DATA LIBRARY Woods He c / of ^ X \ z S < 1 m o H erf pq •V^ H H CO PS M •> O Q CO 04 w CO M z W Pi 1 o o OS H H C3 CO O M 3 CO S3 Ed ej Pi M CO >> X Pn X z O M H < N fcj M o Pi rt U !3 H O O CO 1 \ C_> / co> w fi^l 3 z O CO z > M W H P 5 5C W < « H eS > ►J o 1 [§y 1 o as C o •t-l . > CD C 0) 0) 3 H U-4 CO O > CQ 0> 0) u l-l 3 o w 3 60 12 the decisionmaker for any particular developing pollution situation, steps 2-8 represent generic needs for information and predictive methodologies that will be very similar in most cases. These generic needs are schematicized in figure 2. A major intended implication of the outlined objectives is that research and information products not contributing directly to these objectives are likely to be of little direct value to the decision process leading to the selection of optimal waste management alternatives. The principal processes and relationships that must be modeled to support the successive predictions outlined in figure 2 are discussed below. The source characteristics, which are sometimes controllable decision variables, serve as input data for trajectory and dispersion models. These models should include subroutines that account for physical and chemical transformations and transport of different phases. Relevant processes to be modeled may include: biodegradation, adsorption-desorption, sedimentation, bed- load movement, bioturbation, etc. The resultant time and space scales for physical contaminant distributions can be coupled with information on demography and susceptibilities to generate the measures of outcome for human health and recreational values. Coupled with information on resource distribution and life histories, the physical distribution of contaminants leads through bio accumulation or uptake models to predicted tissue concentrations in key organisms. These tissue concentrations are used in computing exposure for humans and other predators, like birds and mammals. Coupled with information on effects, especially on direct mortality or rep-roductive success, the bio accumulation information provides input for population dynamics models for key species, which lead to predictions of stock size and production. Many of the measures of outcome in table 2 can be predicted or estimated directly through these computational or modeling steps. The specific modeling requirements must be driven, however, by the measures of outcome, which must in turn be convertible to measures of value. Thus, all-encompassing ecosystem modeling is probably not required, and many potential input variables may be suppressed after appro- priate sensitivity analysis. When time scales of concern exceed the predictive capacity of existing models, the decisionmaker must rely more heavily on associated research-monitoring programs to test the adequacy of the predictions. The one class of resource values for which comprehensive ecosystem modeling may be most applicable is wilderness values. In this instance, however, the valuation approaches are not at all defined, and the suggested measures of out- come may ultimately prove to be of little use. In this case, greatest initial emphasis should be placed on development of acceptable and useful valuation approaches . Most of the needs outlined above and in figure 2 have been recognized for some time as essential to understanding and predicting pollutant fates and effects on marine ecosystems (Wolfe and Rice 1972, Wolfe 1975, Warlen et al . 1977). The particular point emphasized here is that these predictions can be much more highly focused on the problem at hand by explicit consideration of the management alternatives and the specific values at risk. Even though full and complete marine ecological understanding Is never achieved, decisions impacting the marine ecosystem will continue to be made. These decisions can be strengthened by careful documentation of a structured decision-oriented analysis that considers alternatives, values, and uncertainties. Formal techniques for such analysis have become well established in operations research 13 and systems engineering (Howard et al . 1977). In the context of environmental assessment, decision analysis assumes the additional dimension of iterative and adaptive application to an evolving problem (pollution) as our basic understanding of ecosystem function grows. This added dimension derives from the stability and resiliency of ecosystems which allow us to research the validity of our understanding and predictions after instituting the policy that resulted from those same predictions. In this adaptive environmental assessment and management approach (Holling 1978), however, the basic decision analysis remains the same. Modeling needs that arise from the decision analysis itself, as opposed to fundamental cause-effect relationships in the marine ecosystem, are outlined in the next section. MODELING REQUIRED FOR DECISION ANALYSIS OF POLLUTION MANAGEMENT PROBLEMS Decision analysis is merely a structured procedure for analyzing the merits of various alternatives in a decision. Decision analysis helps to ensure that essential steps have been consciously considered in the decisionmaking process and also facilitates explicit documentation of the form and content of those considerations. For decision problems as complex as marine pollution management, in which tradeoffs must be balanced among economic costs, human health values, commercial and sports fishery production, recreational and wilderness values, and global human habitat values, over potentially large scales of- space and time, it is important to document the analysis as completely as possible to provide both a logical justification for the chosen policy or alternative and a basis for improved decision making in the future. Careful, complete, and objective documentation of the decision process is particularly important in matters of public policy, such as marine pollution, where numerous different factions, with different perceptions of the problem and different values on the potential outcomes, may be affected by the decision. The documented analysis serves to inform all concerned on the specific consider- ations that entered the decision and provides a basis for iterative objective improvement in successive applications of the analysis. Matheson and Howard (1968) have succinctly outlined a comprehensive approach for the analysis of decisions and the identification of optimal alternatives, which is summarized in the following section. The sequence of steps involved in the analysis (fig. 3) involves the creation and manipulation of a hierarchy of models (fig. 4) through which the relative merits of each available alterna- tive are assessed. These steps are consistent with the elements of a rational decision process discussed earlier (fig. 1). The minis ti c phase of the decision analysis focuses on the development and sensitivity analysis of the structural and value models used in predicting outcomes and values for each identified alternative. For assessment of marine pollution decisions, the structural model would address the environmental processes outlined in table 3 following the structure suggested on pages 13-14 and in figure 2 to generate a set of outcome variables (table 2) for each alternative. The value model Is designed to assign each outcome a value, by translating each member of the set of outcome variables (table 2), as defined and quantified in time and space, into an appropriate measure of value. A time- preference model is also created to translate values that occur over a timestream 14 B O •■-I CO to •H a u QJ CU CO 00 •o o XI c CO c o CO CU J= 4J CO s § En CO I-H CO c CO c O o CU CO CU CO CO X. a CU .c H CO 0) 3 60 •r-l En 15 CERTAIN EQUIVALENT WORTH RISK PREFERENCE MODEL WORTH TIME PREFERENCE MODEL VALUE VARIABLES VALUE MODEL OUTCOME VARIABLES STRUCTURAL MODEL I FIXATED ALEATORY POTENT \ IMPOTENT STATE VARIABLES DECISION VARIABLES SYSTEM VARIABLES Figure 4. Hierarchy of models needed in decision analysis. (From Matheson and Howard 1968. Used with permission.) 16 in the future into present net value or worth. For economic values this time- preference model is used to account for discount rates and inflation. The objectives of the deterministic phase are: 1) to determine whether particular alternatives are deterministically dominated, i.e., they always have a lower worth than some other alternative, regardless of what values are selected for the state variables; and 2) to determine which of the decision variables (table 1) and the state, or environmental, variables are most influential in affecting the worth of each alternative. The probabilistic phase of the decision analysis focuses on resolving the uncertainty in value or worth of any alternative. This uncertainty arises from those (aleatory) state variables to which the worth of the alternatives is found during the deterministic phase to be most sensitive. Probability dis- tributions must be assigned to these aleatory variables over their potential ranges of values. The structural model is expanded to include the probabilistic values for the aleatory variables and the expected outcomes, expected values, and expected worth are then calculated by the structural model, value model, and time-preference model, respectively, based on the assigned probabilities. A probabilistic sensitivity analysis then reveals whether one or more alterna- tives are stochastically dominated, i.e., whether a particular alternative always has a lower probability of achieving any specified worth than another alternative, over the entire range of potential worth. Stochastically dominated alternatives may be dismissed from further consideration because they consistently have lower probable worth and are logically excluded. If certain alternatives prove not to be stochastically dominated, then there is a logical risk in selection of any alternative, and risk preference (Matheson and Howard 1968) must be modeled for the decisionmaker (s ) or his surrogate. The risk-preference model translates the expected worth under uncertainty into a certain worth equivalent. The certain worth equivalent may be viewed as the smallest offered certain worth that would induce a risk-averse decisionmaker to prefer another alternative in place of one with uncertain outcomes and therefore an uncertain worth. At this stage of the analysis, the optimal alternative, the one with the highest certain equivalent worth, has been identified. The informational phase of the decision analysis focuses on determining how much one should pay to upgrade the quality of information available for the analysis prior to making the final decision. In effect the nature of the information to be gained from a particular experimental program is anticipated, with an associated cost, and the decision analysis models are then restructured about that information, with its new (expected) attendant uncertainties. The deterministic and probabilistic phases of the analysis are repeated for this "new" information, and the optimal alternative is again identified with its new certain equivalent worth. The value of the information is exactly that cost of the experimental program which would make the certain equivalent worth of the decision with the improved information just equal to that of the decision without the information. This process is repeated for each of the aleatory variables, separately and in combination, for which improved information might be obtained. Where the value of the information exceeds the cost of the experimental program (including the costs of the delay in the decision), then the decision to gather new information is supported (fig. 3) and the analysis should be reiterated on the basis of the improved information. When further information gathering is projected to be overly costly, then the decisionmaker is ready to act. 17 Computerized modeling techniques have been developed for structuring and analyzing decisions. For example QUICKTREE® is a set of APL functions designed to assist in evaluating decision trees with up to 1000 trajectories (alternative- outcome combinations). This program was developed by the Decision Analysis Group of SRI International, Menlo Park, California. * SUMMARY CONCLUSIONS 1. Marine pollution assessment would serve environmental management decision needs much more effectively if the assessment and decision processes were integrated through a systems analysis of waste disposal management decision problems . 2. In such a decision-analytical approach, the objectives of the environmental assessment are focused by the available alternatives and the values of the decisionmaker, and the modeling needs can be successively refined through sensitivity analysis oriented toward those limited objectives. 3. Generic consideration of marine pollution problems suggests major environ- mental modeling needs in the following areas: A. Transport and Transformations of Pollutants 1. dispersion of dissolved and suspended materials 2. bed load transport 3. adsorption dynamics 4. bioturbation 5. evaporation-dissolution 6. physical-chemical transformations (state changes, chemical reactions, decomposition, etc.). B. Bio accumulation and Effects 1. uptake-depuration kinetics C. Population Dynamics of Key Species D. Ecosystem Modeling 1 . nutrient-planktonic dynamics 2. stability-rr.siliency-recovery from stress. 4. Decision-analytic modeling and systems approaches are currently available and are applicable to marine pollution management problems. Effective application requires close interaction among environmental scientists and modelers, decisionmakers, and systems or decision analysts. ^Mention of tradenames does not imply endorsement by NOAA or the Department of Commerce. 18 REFERENCES Bonnieux, F. , P. Dance, and P. Rainelli, 1980. Impact socio-economlque de la maree noire provenant de 1 'amoco-cadiz . Rapp. Institut National de la Recherche Agronomique. Station d'Economle Rurale, Rennes , France. ISBN 2-85340-322-X. iii+100 pp. Clark, C. , 1976. Mathematical Bioeconomics : the Optimal Management of Natural Resources . Wiley- Inters cience , New York. 352 pp. Engelmann, R. J., 1979. "The Alaskan Outer Continental Shelf Environmental Assessment Program," Environmental Conservation 6(3) : 171-180. Frenkiel, F. N. and D. W. Goodall (eds.), 1978. Simulation Modeling of Environmental Problems , SCOPE 9. John Wiley & Sons, New York, xvi+112 pp. Holling, C. S. (ed.), 1978. Adaptive Environmental Assessment and Management . John Wiley & Sons, New York, xviii+377 pp. Howard, R. A., 1966. "Decision Analysis: Applied Decision Theory," pp. 55-71. In Hertz, D. B. and J. Melese (eds.), Proceedings of the Fourth Interna- tional Conference on Operational Research . Wiley- Interscience , New York. Howard, R. A., 1968. "The foundations of decision analysis," IEEE Trans . Systems Sci. Cybernetics SSC-4 (3) :21 1-219. Howard, R. A., 1979. "Life and death decision analysis," Res. Rpt. No. EES DA-79-2. Dept. Engineering-Economic Systems, Stanford Univ., Stanford, California, ix+145 pp. Howard, R. A., J. E. Matheson, and K. L. Miller (eds.), 1977. Readings in Decision Analysis , (2d Ed.). Stanford Research Institute, Menlo Park, California, vi+613 pp. Kates, R. W. , 1978. Risk Assessment of Environmental Hazard , SCOPE 8. John Wiley & Sons, New York, xvii+112 pp. Keeney, R. L. and H. Raiffa, 1976. Decisions with Multiple Objectives: Preferences and Value Tradeoffs . John Wiley & Sons, New York, xxviii+569 pp. Matheson, J. E. and R. A. Howard, 1968. "An introduction to decision analysis," pp. 5-43. In: Howard, R. A., J. E. Matheson, and K. L. Miller (eds.), Readings in Decision Analysis (2d Ed., 1977), Stanford Research Institute, Menlo Park, California. Mattson, J. S. , 1979. "Compensating states and the Federal Government for damages to natural resources resulting from oil spills," Coastal Zone Mgmt. J. 5(4): 307-332. Munn, R. E. (ed.), 1975. Environmental Impact Assessment: Principles and Procedures , SCOPE 5. John Wiley & Sons, New York, xviii+190 pp. North, D. W., 1968. "A tutorial introduction to decision theory," IEEE Trans . Systems Sci. Cybernetics SSC-4 (3) :200-210. North, D. W. and M. W. Merkhofer, 1976. "A methodology for analyzing emission control strategies," Comput. Ops. Res. 3:185-207. Ward, C. H. , M. E. Bender, and D. J. Reish (eds.), 1979. "The Offshore Ecology Investigation. Effects of Oil Drilling and Production in a Coastal Environment," Rice University Studies 65 (4-5) : 1-589. Ward, D. V., 1978. Biological Environmental Impact Studies: Theory and Methods . Academic Press, New York, viii+157 pp. Warlen, S. M. , D. A. Wolfe, C. W. Lewis, and D. R. Colby, 1977. "Accumulation and retention of diatary ^C-DDT by atlantic menhaden," Trans . Amer . Fisheries Soc. 106( 1) :95-104. 19 Weller, G., D. Norton, and T. Johnson (eds.), 1979. "Environmental stipulations relating to OCS development of the Beaufort Sea: Proceedings of a Synthesis Meeting of OCSEAP investigators, Fairbanks, Alaska, 26-29 July 1979," Special Bulletin No. 25. Arctic Project Office, Univ. of Alaska, Fairbanks. 36 pp. Wilman, E. A., 1978. "Economic values and ecologic impacts associated with oil spills," pp. 119-135. In: Proc. Conf. Assessment of Ecological Impact of Oil Spills, 14-17 June 1978, Keystone, Colorado . Am. Inst. Biol. Sci . Wolfe, D. A., 1975. "Modeling the distribution and cycling of metallic elements in estuarine ecosystems," pp. 635-671. In: E. Cronin (ed.) Estuarine Research, Vol. I: Chemistry, Biology and the Estuarine System . Academic Press , New York. Wolfe, D. A. and T. R. Rice, 1972. "Cycling of elements in estuaries," U.S. National Marine Fisheries Service, Fishery Bulletin 70 (3) : 959-972. 20 COMMENTS ON MODELING FROM THE STANDPOINT OF A RESEARCH ADMINISTRATOR Robert L. Edwards and James E. Kirkley Northeast Fisheries Center National. Marine Fisheries Service National Oceanic and Atmospheric Administration Woods Hole, MA 02543 21 EDIS asked me to comment on modeling from the standpoint of the research administrator. The bulk of the effort in the Fisheries Center is essentially ecological. We have a particular focus on the resources utilized by man. For each of our very large ecosystems we really have only one model and this model often drives our entire program. Each Fishery Research Center is concerned with one or two very large ecosystems of varying complexity and size. These VLE's, as we call them, don't have hard boundaries, of course. For us, modeling is a continual process — a way of life not, unfortunately, marked by continued, steady progress, but rather a process that leads to questions as well as answers and tends to vary in intensity through time. One continuing problem that we have is that of deciding how much centralized effort to put into modeling at any one point in time. In general each scientist models his own specific element of the work. There is a need, however, for the integrative modeling effort. There is a marked tendency to centralize this effort. Unfortunately, all too often such centralization leads to the development of one general model driven by a single dominant personality. While it may seem more efficient to proceed this way, it can lead to serious difficulties because of the limited grasp of the complexity of an ecosystem by any one individual and by the pulsating and varied nature of opportunity to make advances. It is better, in my opinion, to proceed somewhat more slowly by rotating the involvement of individuals dealing with these more general models. This has the benefit of varying the effort commensurate with the opportunity to make specific advances and does not result in a model frozen in the image of one individual. Change is the name of our game. The situation that we are dealing with is dynamic and, to some degree, intractable to deal with because of the logistics of sampling the ocean. Almost all of the decisions that are made with respect to resources events are based on our ability to measure change, not absolutes. Our fundamental problem is one of specifying rates. The first version of our model, possibly it is the first ocean model, was published by George L. Clark in 1948 in a paper entitled "The Dynamics of a Marine Ecosystem."! The present model is about the fifth generation of this model. It has improved with each succeeding year as more and more data accumulates and interactions become clearer. At the present time it is relatively robust at the two ends — that is, we know a great amount about the initial energy inputs and primary production, and the outputs, that is, the fish and other resource populations. The internal anatomy is still unclear although even here rapid advances are being made. The general robustness of the model, even in its third generation, was sufficient to convince us and others that the only solution to the problem created by the massive foreign fishing effort off our shores was to impose an overall biomass quota. We had enough information, for example, to convince others that we had a sausage machine that produced at a reasonably well defined rate, although we couldn't specify precisely the type of sausage that might be produced at any one time. Such a biomass limit was imposed, an impressive 'first' in the history of ocean management. 1 Ecological Monographs, vol. 16, no. 4, pp. 321-335. 23 Many of the changes that are taking place today in the resource populations are not always predictable, but at least many can be rationalized on the basis of the model. It has become very clear, for example, that predation on the smaller size of fishes has a great deal to do with the ultimate size of adult populations. It is also becoming obvious that disease plays a far more significant role than previously acknowledged. In any event, in the Northeast Fisheries Center we have one overall guiding model but many variants, subsets and aspects of this model. There are three general needs which may be categorized as follows : 1. A critically important need is that for a relatively transparent version in popular science terms. This version is one basis for communicating to our constituents. These are usually static versions of the model and contain information about the processes and rates in terms that every man can understand. An example (fig. 1) of one such version — it speaks in terms of dollars rather than grams of carbon per meter squared per year — was developed to illustrate energy flow for an audience of non-biologists. 2. The second need is to clarify the dynamics of the ecosystem and ultimately to predict resources events. This version of the model is characteristically scientific, opaque, and cumbersome to deal with unless you are an expert. The dynamic model required in this instance has a minimum of 50 or more state variables. This number of variables is sufficient to overload most computer main frames and accordingly it is less dynamic than we would like, but it exists and it is used. 3. From my point of view the most significant version of the model is that one we use for communicating with our bosses. It is transparent, perhaps deceptively so, but it does lead to the development of other kinds of information that are necessary. Our simple form of the model can be stated as follows (fig. 2): Change = Fishing (F) + Predation (P) + Disease (D) + Environment (E). Let's have some definitions: Fishing (F). This component refers to all the activities associated with population assessment. It includes survey cruises, age and growth studies, analysis of landings data, and a great deal of analysis. Predation (P). The feeding of one species on another. We now can demonstrate the likelihood that predations can and do structure the fish segment of the ecosystem. The function is felt by man 2 or more years after the fact since much of this predation takes place on the very young fish. Disease (D). Disease is an obvious "player" in the ecosystem. It's full importance has only recently begun to be recognized. Fortunately, disease phenomena are also relatively easy to deal with in the predictive mode once the principles of their involvement are understood. Disease generally has the same "future" significance that predation has. It manifests itself early and usually on younger fish. Its effects peak several years after the initial "insult." Disease is often directly linked to physical environmental change, both natural and man-caused. 24 u 03 O co (0 o o TJ c to .—I 60 c w !Z <3J M-l o c o •H 60 0) u 3 60 26 Environment (E) . The physical environment component refers to the flux of temperature, salinity, and nutrients in the water. The environmental effects are usually subtle, often difficult to predict, and always difficult to sample. This element is subdivided into two aspects — the natural environment and the man-created environment (e.g., pollution). For a moment let's look, at each of these variables and examine how they can be used to evaluate our program and its usefulness to the country. We are often asked by our bosses, "How much is enough?" In figure 3 each of these "system" variables (F, P, D, and E) has been evaluated by the appropriate staff members of the Center. These are performance functions and they are their estimates of the value of information provided by these experts in answering questions asked of them. Program cost is displayed on the X axis and the relative value of information on the Y axis. Perfect information, of course, is represented by 1 and is not attainable. The dots on the curves represent our estimate of our present position. In a way you can call this the "expert's comfort (or discomfort) index." You will note in the performance function for F (Fishing), the experts estimate their present ability to provide information as approximately 0.46. It is costing us approximately $6 million to provide this information. The F_ function has been broken down into two curves. The lower curve is concerned with the data collected from the fishermen themselves. In terms of predicting change, it very quickly flattens out. The reason for this is obvious. Fishermen go to those areas where the fish congre- gate and the data collected from them is accordingly strongly biased when used to estimate population size. This is the same bias that a census taker would experience were he to be confined solely to basing his estimates on the popula- tion of our country on New Year's Eve in Times Square. Catch data has some real value up to a point, but beyond that point it tells you virtually nothing. In order to significantly improve the amount of information from such data, more than the expenditure of money would be involved. One would virtually have to mandate nearly perfect data from each fishing vessel and from a socio- logical point of view that is a virtual impossibility. The other curve is based on data that activities such as research vessel surveys would provide, the more appropriate and unbiased census type data which is necessary to make a prediction. In the case of predation and disease research, the costs for significant information are not as overwhelming as they are for developing appropriate F_ information. In no small part this is because much of the data required can be, and is, collected during cruises carried out for other purposes. Further, such data is used more abstractly - in prediction models - where the principles are as important as the facts. The environmental curves have been subdivided into two components, that which is natural and that which is man-caused. The value of information available today is estimated at about 0.3, mostly because of our inability to deal with the flux of the environment in real time using such sampling tools as vessels. It is in this curve that one sees the opportunity for a small investment in modern technology that would greatly increase the value of information that can be provided. This technology exists - it is called remote sensing. The man-caused environmental changes potentially can reach a higher level of information value than can any other category, simply because the man-caused environmental effects tend to persist for a long time, as for example DDT, which can now be found throughout the world. 27 CO CO h- *r c o CO coZ •a o (A '"CO cnZ 1-4 4-1 CO c (0 O 3 ^2 i—i a mmm cc X w-J O o-J a. T " > 111 1-1 4J CO o o CM ^r CO O OJ c o <8 l S C0 CO 4J O c 3 cm "O coZ coZ CO i *-o O en o 1 10 H • w-J CO "5 1 T--J -j o o I •MS i-i cm5 *•! 4-1 CO i ■ ^ ■ -i ii o CM c 1-1 CO CO cu coo CMO o O o 3 i-i CO *- CO > © CO • • co ^ cn ■ ■ • o o CO CO ^ CM O • en > IL r- > o 01 u 1-4 fa 28 This is only part of the story, however. In figure 4 the relative value of information with respect to time is illustrated. It depends very much on the time horizon of concern how valuable the information from each of these system variables may be. The effects of predation and disease, vis-a-vis the fisherman, both tend to peak 3 to 4 years ahead. In the case of environment, it becomes very important as you look further ahead, and finally the environment concerns dominate when one is looking ahead as much as 10 years. The highest value for the natural environmental information at the zero time horizon, for example, is related to the fact that this information is useful to those pursuing, for example, tunas, or any other resource which tends to congregate along ocean fronts near the surface. Because one of our overriding scientific questions is that of the significance of temporal and spatial variability of environmental parameters, knowing what is going on at the time the research vessels are on the scene is also important. Again that new technology - remote sensing - pops up as a significant and cost effective tool to be invoked as soon as possible. To return now to the more complicated segment of this figure, that referring to fishing, there are three curves. If one wishes to maximize "now," or to put it into other words, discount the future, then fishing data is very important. If, however, one wishes to maximize both revenue and stability (longer term parameters) from existing populations of fish, then the curve is far different. Since most of the resources have moderately long lives in the fishery, a haddock year class, for example, can last several years , then other information becomes important, as for example, predation and disease. Since recruitment to popula- tions of fishes is controlled by many variables, at the present time it is almost impossible to predict recruitment much more than 3 or 4 years ahead. This curve thus tapers off quickly after 6 years. If one desires to maximize ecosystem revenue, the curve falls somewhere in between these last 2 or at about 5 years. Going back to the sausage machine analogy, such activity would require directing the fleet to harvest certain species as opposed to others, with the consequence of somewhat less ability to pass judgments on fishing data alone, but required far more reliance on knowledge of the recent history and future consequences of predation and disease. Figure 5 is simply a single graph combining all the information in figure 4. The most desirable mix of programs is thus indicated in terms of the relative value of information against the time horizon. When one puts this information into the computer and evaluates the overall information return, given various budgetary levels, a new series of curves is generated. These are presented in figure 6. To generate figure 6 we assumed that we had 1.5 full-time research vessels available and a capacity to use satellite derived products. Actually we have the full time use of two vessels and cannot as yet effectively use remotely sensed data. It should be noted that these figures do not precisely reflect our actual budget situation. The computer was asked to evaluate the relative information return from each of the four system variables, given that each time horizon was equally important. We are incidentally, at about the IX level. 29 UJ ID HI c o t4 4J to c 0} <-t D. X > 0) 01 3 00 i-l to 30 c o LU LU o 14-1 o > (U > <1) u T3 0) u c c •H 3 0) IH o O- CO B O (0 0 31 X IT) CO in CO O N CL o CO CD CN UJ-JU-Z 0) CO s-> > m o l y-i o c o o c 3 14-1 . X to o en co .-I C en O -H •H 4J 00 nj C £ •!-( C T> O C 14-1 3 U-) O cu .-l 4) CO i-< > C 9 > O o 60 •H C co -a rH C 4) D OS 100 mg/1 Low Solids Water TSS < 100 mg/ 1 High Solids Water TSS > 100 mg/ 1 Dissolved Oxygen > 5.0 mg/1 Coldwater Gamef ish > 3.0 mg/1 Warmwater gamef ish/ panf ish > 2.0 mg/1 < 2.0 mg/1 Rough f Ish Rough f ish a/ a/ Unfishable Unfishable Roughflsh a/ Unfishable Roughflsh a/ Unfishable Other unfishable conditions include pH greater than 10 or less than 5; temperature greater than 32°C; summer flow zero. PROBLEMS IN APPLICATION TO MARINE RECREATIONAL FISHING Application of the above method to the case of marine recreational fishing requires that we find some index or measure of water or ecosystem quality that can be used to link changes in pollution discharges to changes in individuals' participation decisions. Beyond this, we must have data on pre-policy levels of quality and on the related level of participation. Further, it must be possible to value in a defensible way pre-policy and post-policy participation. Here we can continue to concentrate on the linkage requirement, since this brings us to the core of the paper. First, let us note the several conditions that such a linking index or measure must satisfy. 44 1. The index, be it dominant species type or whatever, must be linked backward to measures of ambient water quality, which in turn are linked to levels of pollution discharge. 2. There must be an a priori reason to expect participation decisions to reflect values of the index. 3. Data must be available to test the hypothesis from 2 and in the process to product projection equations for participation as a function of the index. In moving from freshwater to saltwater recreational fishing, several possibilities for this key index present themselves. An obvious candidate is dominant species, as in freshwater, and there exists some evidence that this effect can be significant, at least for well-defined bodies of water such as enclosed bays and estuaries. Resident populations within semi-enclosed marine waters with serious pollution problems apparently are dominated by such fish species as bay sardines and toadfish. Marine gamefish such as bluefish and striped bass do not seem to frequent dirty areas, nor do they use contaminated estuaries for spawning or nursery areas, but because of the tremendous available dilution and the mobility of major marine species, it seems unlikely that any significant fraction of the nearshore ocean would be found to be dominated by low quality species (e.g., U.S. Department of the Interior 1970, Sindermann 1975, McErlean et al. 1972). A second possibility for the index is some measure of fish population (numbers, or average size, for example) as a proxy for prospective success in angling (bag). This was the basis of the benefit estimation procedure used by Bell and Canterbery in their major study (1976). There seem to be four problems with this measure, however. First, the appropriate variable, standing as a proxy for prospective angler bag, may be a complicated, but is certainly an unknown, combination of the several possible measures; number of fish, average size; and number of very large fish. Second, the measurement and prediction of any indicator of bag raises a difficult question of simultaneity, for participation and bag are probably connected through the sustainable yield function for a species (or group of species) as well as through the participation equation reflecting human decisions. Third, with so many important species highly migratory in habit, tracing any pollution effects on a size variable would require tracing migration routes and understanding effects that may involve only one stage of a fish's life cycle. Fourth, even for resident fishes, the effect of some kinds of pollution, especially that involving nutrients and organic carbon, may be ambiguous. (See, for example, Chittenden 1971, 1976; Roberts et al. 1975; Wise 1974; McHugh 1972, 1975; Hedgpeth 1966; Pararas- Carayannis 1973; Carlisle 1969; Brown and Beck 1972; Bascom et al . 1979; Soule and Oguri 1979; National Marine Fisheries Service 1972; Butler et al. 1972; Smith 1973.) A third possibility might be thought of as an amalgam of the first two — some index of ecosystem health, reflecting species types present and abundance, and indicating the overall quality of the available experience. (See, for example, Hillman et al. 1977, McErlean et al. 1972, Bader et al. 1970, Bechtel and Copeland 1970, Haedrich 1975, Livingstone 1975.) 45 Practical application of any of these possible indexes requires that we be able to make the backward (to discharges) and forward (to angling decisions) links. An initial reconnaissance of the literature on pollution effects on marine fishes has convinced us that the backward link will at least be very difficult. With the exception of laboratory studies of acute toxicity using a variety of elements and compounds and many different fishes , there seems to be remarkably little systematic quantitative understanding of how pollution of natural saltwater effects the availability of fishes of different species, or of how it affects population numbers and size distributions. In table 3 we summarize the findings of our survey to emphasize for the reader the paucity of pollution-related information (that on suspended solids) relative to that on temperature and salinity tolerances. Further, the information we were able to find on dissolved oxygen tolerances — the single most important pollution-related water quality measure for our freshwater work — is confined to a few studies, too varied in methods and results to permit meaningful comparisons .-* Thus, unless further searching can improve our data base in this key area, it appears we cannot make the backward link using models of pollution discharge dispersion, dilution, and transformation as the bases for predicting fish availability or quality measures. This will be especially unfortunate because we do anticipate having available at RFF a coastal county discharge inventory, complete for the lower 48 states, and reflecting point and nonpoint sources, including the inflow of pollution loads in freshwater streams." It will be possible to project how these loads will change with implementation of the clean water legislation of 1972 and 1977. In terms of the link forward to angling decisions, the species type, prospective bag, and ecosystem effects are all defensible as influences on participation decisions. However, to use any of these indices it will be necessary that it match the available data on participation. That is, differential levels of participation across measurement units must be tied to differential values of the chosen index(es). For example, to our knowledge the best available data on participation in saltwater angling identifies individuals no more finely than by state of residence.' In order to test the effect of an 5 We found DO references for Atlantic cod (Wise 1961, and Davis 1975); American eel (Usui 1974); Atlantic herring (Dorfman and Westman 1970); white perch (Dorfman and Westman 1970); shad (Chittenden 1973, Hoff et al. 1966); striped bass (Chittenden 1971, Dorfman and Westman 1970); and tuna (Dizon 1977). " The inventory is being developed as part of the Strategic Assessment Project of the Office of Resource Coordination and Assessment of the National Oceanic and Atmospheric Administration. This project also involves the generation of maps showing areas of critical importance in the life cycles of numerous coastal fishes (fin and shell) and it may be possible that by combining inventory and maps we can produce a useful mixed (quantitative and qualitative) model. See Ehler et al. n.d. ' This is the nature of the data from the 1975 National Survey of Hunting, Fishing, and Wildlife-Associated Recreation, NSHFWR (U.S. Department of the Interior n.d.). 46 index value on participation we would require measures of pre-policy index values appropriate to each state. For example: - If we had species availability data, we might use the analog to our freshwater measures, state acres per capita of water dominated by particular species types. - For abundance, we might use average state saltwater bags. - For ecosystem "health" we might use a state index of marine water quality, perhaps one created out of a combination of species dominance and abundance measures. In any event, the basic problem is finding a measure for which values are available for all coastal states... no trivial requirement. One route to such knowledge that might be explored is via state fish and game departments, coastal zone commissions, sea grant universities, and other state and local institutions, This would involve a mail survey supplemented as necessary with phone calls and even personal visits. Our experience with a similar study in the freshwater context was that a significant fraction of the actual respondents (staff bio- logists) will quarrel with the terms of reference and even with the purpose of the overall project and only grudgingly provide their own versions of local knowledge. CONCLUSION In short we cannot be entirely sanguine about the prospects for successful completion of a study in the marine context analogous to our study of the bene- fits of water pollution control accruing via freshwater recreational fishing. The major reason for this very cautious view is the apparent lack of an index of water or ecosystem quality about which we know enough to provide model links backward to pollution discharges and forward to angler perceptions and tastes. Important questions for this workshop might well be: - To what extent can existing marine ecosystem models provide predictions in terms relevant to pollution control policy assessment — whether strictly benefit estimation or something not involving translation to dollar terms? - If our rather negative answer to this question for marine recreational fishing holds across other benefit categories, such as swimming and commercial fishing, should it matter to NOAA? To the modelers at this workshop? - If it does matter, what might be done to improve the situation? Is it entirely a matter of long-run research, or are there short run improvements that could be made by, for example, clever linking of existing models and isolated research results? 47 B 01 u eg cm •a o k. ■H > c w V V k, 41 y c kl CV i-l 8. CO C O c o o •4-1 c kl CO c II kl cm 41 41 r-t x H CO CO 73 73 41 ■a 8. CO s o. * o o -H /-N V c 41 c •o u- CO CJ X ri c *^ g 01 s^ > •H 60 4-1 B •H i-i S ki 01 41 CO X a a a o o o v| o uo o o o CO ■8 3 ki XI a. tj ki co 41 41 -H 41 a ^ <4-l *■> "H CO ._< •H kl O «0 X 4-1 kl 41 3 91 o 5 a — o o M o o o in O O O ki c CO CJ a O O O o o u-i O -J "3 o •8 CO CO 3 5 kl 8. CO a 4J c CO CO 41 kl 3 -> CO ki 8. § 1 • ki n 41 C • C ki 4) ^N -Ji kl « rH iH CO Q 1 41 kl - — 1 a CO S ^N 41 > UO rH y »r-s o C rH u-i 01 CO X 73 4*% M TJ 0) ki CO CO rH -a- « C t< 41 to CO 4) o CO C u-i m p. ~ X CO V X 3 C CO A •H 4-1 CN 41 3 <4-l Jrf 41 4J CO co co y ih n-' *«• - 41 MH 4J ^H C n M rH iH a "^ 3 O 3 CO kl 41 73 V •H Xl h ao oh u » o UH ■ O 4J >v >v • U-I kl r-J O.^ 41 • O » 3 i-l 73 rH 60 CO ki 3 4-1 y co a. co 4J >-• C 73 CO O iH (0 > w ■H e - a. -h s ki oi ■W 3 41 ki ji y co 4-i c rH 4) • c dowpa rrlng , scu cod, ) , ee grey t, St. s pott 41 ■H X kl 4J _l ■H 60 CO CO rH y cu 3 .-1 c • ki • 41 CO rH O CO cS X k. ^ CO CO 73 "H ki c/> 00 CO X 4) r«N 3 4-1 C 4) /-v • -Si O ■H£ k^ U . Q, • 3 oi mi « x co -a « S C iH 09 C H J I • X ki ki a a 3 >> <4H -H CO 60 w 0) 0) _ S >^ co a a oi .3 4-1 •H 41 « C C 73 4J CO 3 ■H rC * 73 8 iH X iH * 4) c CO C 0. 4-1 3 3 H - kl CO 60 4J >4-l 73 3 Jl! M-l 4J CO •H a 3 oi U-i O CO U co ^^ * rS H **-* j3 C 73 73 4) co co » i— i oi to a 4-1 rH CO no k^ • 60 /~> O -°~ . g. 4-1 en rH 73 4J « Ih x: CO a 4J ki 41 ki i-l CO 3 X • 41 CU c ki 4) CO CO ■ 11 kw o 3 CO O kl Ck CO 41 73 iH •> kl X! » 4j 41 • • ki O •H 4J 41 -H ki Jd C U-I 4-1 01 ox cj^ .* x 4-i ki IU T* M U 41 CO 3 rSl-H H klCO-HUpUojCO 0J-H3cOkicOi-l4) kl Eta vf •-I O 73 4) kl rH CO B y CM rH oo £ CO z a_8 I " eS kl r^ /-N ai e i $ 60 X! 73 i-l Ji * r*N QO • c xi y I 4J kl CO >,-rt ^ 41 3 o 9 — i .* o ki e 60 U o B u 41 C 3 01 k. CO 4J H 60 O- ai ih ^h a. M 4) CO 73 ^v 41 « ■»-' 01 ki u 41 CO rH CO D. ■H ki x >4-i cu cj 3 kl CO iH CJ 41 .-^ WW i ■H « kl 4J CO BO ki X ki 01 CO C 41 CO 4-1 S ^a i ii • 3 4-i oj - c c 41 "S .8.8.1? A 01 CT rH 4-1 '^ CO * CO B> CO 73 iH 3 « a 3 y CWH4JO-H 3 60 JX 00 ke (s , sme , win liver , Atl unner bass Oi C « >, 3 CkiH cj c XtlH 3 01 Q ki 01 U) h O 3 au tlv CO 41 rH rH qq CQ <0 CO C o co .i CO X a 41 M Oi ~§ feS w TO CO CO iH 3 OJ B >4-i X 41 C kl CO < ■ CO CO 4) 41 4J 4-1 •-< 11 73 U-i 01 >4J OJ X w 4J in ,4 j: a CO x y ki CO TSi-HkiXJrfOCUki 0O4)3SICki4U y 6. x ^ x ^ eu co a ki co 4-i ki c y ih 4) iH 60 n CO o CO rH v_/ XI B 73 01 k. ki co kl a HHjojiiH x >4H m Ckaco uo O O uO 3 O u-l O O O H uO CO O CN UO 00 H — I CN CN CM CO CO -* C-l CO — 1 -IN CO CO CO lit 1 iH 1 1 A 1 V O O CO Xi /\ /\ /S V A v 3 O O m u-i 1 60 3 rH —t~* •H 3B CM 0) 73 CO rH 0) 3 X O u 3 CO a 3 o f5 iH 4-1 CO ki • 4-1 0) a kl 4) 3 y CO c o o a y X 01 r • C 41 B O 60 ■H C 3 4J CO O « X X k< y v H 4J CN O O 4J C X 4-J C 4J 3 C 60 iH 73 k< a 0) CU •H 60 ki 73 C o o 0) H 4J rH X a. ki 60 s O "H c- a CO 00 U-I • 4) O CO 00 4-1 y S c 41 i-H OJ u-i y U <4H U 41 C a iH 4) 4-> X o 3 co m y -h CO u-i X y tt -H i 41 -J 4J c 4-1 CO r. B O 41 3 x y CO CO ■H rH H O Q. O 4-1 03 73 S^ C U_| rH CO O c o * C kl O ■H . s «J u- o CO 01 rH k. kl 4-1 X c 01 s i CJ o a c iH o 4-1 41 y CO X ki CO ti o iH C 4-1 4) OO 41 UOC > <-2 o fects discussed by Simons (1976); however in a one-dimensional model, all advtctive and dispersive processes are parameterized as vertical mixing. Although it is clear from the above analysis that nutrient limitation does not solely control phytoplankton dynamics, the role of nutrients, especially phosphorus and silicon, is certainly critical during the period of stratification. Phytoplankton dominant during summer months in Lake Ontario are limited primarily by phosphorus, as demonstrated in simulations discussed above and as shown by recent experimental work (Sridharan and Lee 1977). Therefore, to understand better the control of phytoplankton dynamics in Lake Ontario, one must investigate processes influencing the cycling of phosphorus. Figure 7 illustrates the seasonal changes in simulated concentration of available phosphorus and rate of gross primary production in the epilimnion. As discussed above, after spring, phytoplankton become limited by nutrients and thus the production rate decreases sharply. It is interesting to note, however, that although production decreased considerably, it did not approach low winter values and, in fact, after the initial drop, it increased gradually. This sustained production proceeds at the same time that available phosphorus con- centrations, both actual and simulated (fig. 3, 7a), are extremely low. One might expect that, with phosphorus concentrations this low ( a. O rn JZ. T (1 0) 11 .n rr (0 « as 01 > <> < o Q. -2.5 Excretion Detritus Decay Phytoplankton I ^ Net Rate Uptake^^ J i / J i I i V i i IOVEDTUDHI I L JFMAMJJASOND Figure 7. Top: Simulated concentration of available phosphorus and rate of gross primary production in the epilimnion. Bottom: Rate plots indicating control of available phosphorus dynamics. Stippled area represents net rate of change. From Scavia (1979). :0.069 ■ — U.IU) -— 1 ui -'!__ nPo.02= j External S ources H jj ^ ' ". Hypolimnion Figure 8. Phosphorus flow diagram. Concentration in boxes are in yg P«L _1 «d~ . These values are averaged over the period of July-September for the top 15 m. Phytoplankton, zooplankton, and detritus are evaluated in the model in terms of carbon and are converted to phosphorus for this figure by assuming a Cj^qqPj^ atomic ratio. From Scavia (1979). 68 which eventually adds to the available nutrient supply. Thus, it may well be that the zooplankton is principally responsible for high recycling rates esti- mated by Stadelmann and Fraser in Lake Ontario. Control of spatial gradien ts - While the above analyses provide information regarding the major controls of biological production in Lake Ontario, they were based on intentionally crude physical segmentation. The two-layer representation reduced the complexity of physical interactions and thus allowed focus on bio- logical and chemical interactions. However, there can often be vast differences between lakewide averaged conditions and nearshore conditions, especially during the spring-summer transition period. Because it is nearshore conditions that most often affect people, one must be concerned with spatial variations. The spatial distribution of biological and chemical properties in the Great Lakes is determined by variations of water depth, sunlight, and temperature, by the loctions of rivers, and by currents. Separating these physical factors from in situ transformation is a difficult problem, a problem which numerical simulation can help solve. During the transition period between spring and summer both vertical mixing and large-scale circulation are important. Also, the temperature and current patterns of Lake Ontario are relatively two-dimensional (i.e., small longshore gradients); therefore, variations along the long axis of the lake are negligible. We (Scavia and Bennett 1980) simulated flow, temperature, and biological and chemical processes for a north-south transect of the lake. Our approach was as fo Hows . Temperature calculations of a hydrodynamic model (Bennett 1971, 1974) were compared to observations, and the model was adjusted until computed and observed temperature contours were in general agreement. We then repeated the calcula- tions, this time including the chemical and biological processes from the ecological model decribed above (Scavia et al . 1976, Scavia 1980a), and compared them to observations. During the spring transition period a combination of strong heating and low wind speeds causes the thermocline to form. Because the lake begins the spring colder than 4°C (the temperature of maximum water density), this process starts in shallow water. Thus the lake is divided into two hydrodynamic regions — a deep region where the water is less than 4°C and where surface heating causes vertical mixing, and a shallow region near the coast with temperature greater than 4°C, which may stratify. In figure 9, simulated temperature is compared with observations. The model correctly simulates this general spring temperature pattern and the depth of the thermocline. In addition, thermocline extension further off the north shore than the south is also reproduced. Wind and buoyancy combine to cause simple but interesting flow patterns. The wind tends to drive a one-cell pattern, with upwelling at shore to the left of the wind and downwelling near the opposite shore. Heating drives a two-cell pattern, with warm water rising near both shores and colder water sinking in the deep region. 69 Observed Temperature (°C) Observed Temperature (°C) 30 30 40 50 km from north shore 20 30 40 50 km from north shore J L Simulated Temperature (°C) Simulated Temperature (°C) 20 30 40 50 km from north shore J L 20 30 40 50 km from north shore Figure 9. Comparison of observed (upper) and calculated (lower) temperatures (°C) corresponding to May and June cruises. From Scavia and Bennett (1980). 70 Figure 10 shows mean circulation for three 24-day periods and the mean for the entire 72-day simulation. During the first two 24-day periods the wind came predominantly from the east; because the Coriolis force deflects the water to the right, water near the surface flows north and deep water flows south in compensation. Upwelling near the south shore and downwelling near the north shore closes this circulation. During the last 24-day period the wind and circulation reverse. For the whole 72-day period the wind's effects tend to cancel and circulation looks like the thermally driven pattern. The simple picture of a combination of wind and buoyancy effects should be considered an average circulation pattern. At any given time flow is dominated by wind, and it is only because the thermally driven flow is more persistent that it is as important as the wind-driven flow. In general, nutrient concentrations had slight offshore gradients and were homogeneous vertically in early April. These data were used to initiate the simulation. By the end of May, distinct offshore gradients had developed; nutrients, especially phosphorus and silica, were severely depleted in regions within 10 km of both shores (fig. 11). Also, at this time, no strong vertical gradients were obvious in either the model output or the data. At the end of the period of simulation (corresponding to the June 20-22 cruise) the symmetry of the north and south shore contours was lost. The region of nutrient depletion along the north increased to greater than 25 km and vertical stratification was not as strong there. The spatial and temporal progression of the region of nutrient depletion is demonstrated for phosphorus in figure 11; inorganic nitrogen and soluble reactive silica have similar patterns. The comparison between observed total dissolved phosphorus (TDP) and modeled available phosphorus (AP) is made because previous simulations of phosphorus cycling in Lake Ontario (Scavia 1979) indicate that in spring AP is approximated best by TDP, due, presumably, to production of easily hydro lyzed phosphorus compounds during the previous winter. Biomass parameters (chlorophyll a and particulate organic carbon) in April were relatively homogeneous vertically, with higher values nearshore. Patterns in May generally showed offshore gradients and little vertical structure (fig. 12), except for evidence of nearshore subsurface chlorophyll peaks, which the model did not reproduce. The simulations indicated that between May and June cruises, upwelling moved the higher nearshore concentrations offshore, creating a lens of high biomass about 15 km from the north shore. (See figs. 9 and 10.) By the June cruise, increased nearshore production apparently created higher biomass again close to shore. A similar structure was produced along the south shore with the exception that, like the nutrient contours, the biomass contours were constrained closer to shore by the wind-driven flow. Given this model as an adequate representation of major dynamics in Lake Ontario, we performed numerical experiments to find out which of the physical mechanisms was most important. In these experiments we ran two simulations and compared the results to the original calculations discussed above (henceforth, the normal case). In the first simulation, we eliminated mass transport by advection and diffusion. In the second simulation, only vertical diffusion was included. In all cases in situ biological and chemical processes and sinking were included and temperature distributions were kept the same as in the normal case. The results of these experiments are summarized in figure 13 for available phosphorus and particulate organic carbon. 71 J -. ; ?; Jfii — .--:- ;-- ; Lgure 10. tours :: calculated szreaz : _r : : : - .-' s - eraged . three 2 — z :e::::5 = - z = r ; . : i r e : ; :::: riz.lm: - . Free . . i Bennett (1980). Observed Total Dissolved Phosphorus (ug P L) Observed Total Dissolved Phosphorus (ug P U 20 30 40 50 km from north shore 20 30 40 50 km from north shore 25 50 75 E C 100 a, Simulated Available Phosphorus (ug P L) Simulated Available Phosphorus (ug P L) 24-26 May 20 30 40 50 km from north shore J L 20 30 40 50 km from north shore Figure 11. Comparison of observed (upper) and calculated (lower) concentrations of phosphorus (ug P/L) corresponding to May and June cruises. Observed phosphorus is total dissolved phosphorus and calculated phosphorus is that considered available for phyto- plankton growth. From Scavia and Bennett (1980). 73 Observed Particulate Organic Carbon (mg C L) Observed Particulate Organic Carbon (mg C L) 20 30 40 50 km from north shore J L 20 30 40 50 km from north shore Simulated Particulate Organic Carbon (mgC L) Simulated Particulate Organic Carbon (mg C L) 20 30 40 50 km from north shore 20 30 40 50 km from north shore Figure 12. Comparison of observed (upper) and calculated (lower) concentrations of particulate organic carbon (mg C/L) corresponding to May and June cruises. Calculated carbon is the sum of simulated phytoplankton, zooplankton, and detrital carbon. From Scavia and Bennett (1980). 74 Simulated Available Phosphorus (ug P/L) Simulated Particulate Organic Carbon (mg C L) 25 50 75 E £ 10C a * 125 150 175 20 30 40 50 60 67 km from north shore Simulations of Available Phosphorus (ug P/L) J - " \ 3 >2 2 «^ ^x A J - 20-22 June - I l l I I I i i 20 30 40 90 km from north shore Simulations of Particulate Organic Carbon (mgC/L) 50 60 67 J L km from north shore 10 20 30 40 50 60 67 km from north shore Figure 13. Contours of model output demonstrating the relative effects of physical and biological processes on the distributions of available phosphorus (pg P/L) and particulate organic carbon (mg C/L) . a,b - contours of model output generated with no advection or diffusion included; c,d - broken lines represent normal case (i.e. all processes included); solid lines represent simulation without vertical or horizontal advection. From Scavia and Bennett (1980) . 75 The simulation with no advection or diffusion showed that phytoplankton production in the cold offshore waters, although reduced when compared to inshore waters , is sufficient to deplete nutrients from the entire surface layer within the 72-day simulation (fig. 13a). Therefore, temperature-controlled, in situ production alone was not sufficient to produce the persistent offshore gradients. The second simulation illustrates that vertical mixing, together with in situ production, is sufficient to reproduce the observed biological and chemical patterns. The major effect of vertical and horizontal advection is to smooth the nutrient gradients through increased mixing caused by repeated reversals in flow direction. The same is true for plankton along the relatively dilute, cold-water boundary (fig. 13d). It appears that the distribution of chemical and biological properties in the vicinity of the 4°C isotherm is controlled primarily by the interaction of in situ processes and the differences in vertical mixing on either side of the isotherm. Shoreward, the water mass is weakly stratified vertically. This reduces the mixed depth and allows increased biomass production and subsequent nutrient depletion (Sverdrup 1953, Stadelmann et al. 1974). Lakeward, deep vertical mixing keeps a significant portion of the phytoplankton removed from the sunlit surface layers and therefore inhibits their growth. The region near the 4°C isotherm has been referred to as the "thermal bar." Many notions regarding this region have suggested, as the name implies, that elevated biomass concentrations shoreward of the isotherm are caused by some barrier to their transport offshore. The above analysis demonstrates that in fact no such barrier exists. Biomass, concentrated in surface waters shoreward, are simply diluted vertically throughout the water column when transported offshore. In this fashion, the presence of this convergence zone near the 4°C isotherm in fact enhances offshore transport. LIMITS OF ECOLOGICAL MODELS Compensating errors - Ecosystem models that are faithful to extant theories relating various processes of nature tend to become complicated, nonlinear collections of equations. Often, verification of these more mechanistic models is not possible by usual techniques because it is difficult to obtain complete and independent data sets. This is because sampling all of the properties simulated in more mechanistic models is difficult and expensive (e.g., zooplankton biomass). Even when such data sets are available and these models have been "verified" by usual techniques, one is left with serious questions concerning model reliability because these generally nonlinear models have increased degrees of freedom. Increased degrees of freedom, in this context, means that more than one set of coefficient values will satisfy the usual tests for calibration and verification. The basis for increased degrees of freedom is the cyclic nature of mechanistic models. Since these models generally simulate ecosystem cycles, one would not expect material to accumulate excessively in one particular component but rather to flow among all of the components. Then, because of principles of mass conservation, one could expect that, if flow rates were increased or decreased proportionately, state variable concentrations would not 76 be affected significantly (at least not within the variability usually inherent in field verification data). The following example demonstrates one way in which compensating errors at the process level can lead to erroneous conclusions regarding system controls. After initial calibration to measurements of state variables for the model described above (fig. 1), simulated process rates were compared to measurements. For this comparison, a summer-averaged (July-Sept.) phosphorus flow diagram was constructed as described above from aggregated model output. Flow (or transfer) rates were then compared to measurements and calculations from Lake Ontario and to other, more theoretical estimates. Many of the simulated process rates (fig. 14a) were very low (as much as 3-7 times lower) compared to measured values. Most serious discrepancies in transfers were among available phosphorus, phytoplankton, and zooplankton. I recalibrated the model keeping process rates in mind and most coefficient values still within acceptable ranges. The new calibration, shown in figure 14b, is the same as that discussed above (fig. 8). Here, state variables are close to the originally calibrated values and can still be considered calibrated; however, process rates are much higher and, in fact, much closer to observed values (Scavia 1979). This example demonstrates that if the model were calibrated only against state variables and then used to examine control of phosphorus cycling, then the relative importance of certain processes would be overestimated by almost an order of magnitude. For example, regeneration of available phosphorus from detritus P is relatively more important in figure 14a than in figure 14b and the relative importance of external loads and of transport into and out of the epilimnion is exaggerated in figure 14b. Because of increased degrees of freedom and the usual lack of long-term verification data, mechanistic models need verification tests beyond the standard tests used for state variable simulation. Two general types of verification can be useful additions to the usual tests (Scavia 1980b): 1) comparison of aggregated output from the mechanistic model with output from simpler models and empirical correlations that have been verified or proven to be generally applicable and 2) a comparison of simulated process rates with rates measured in the field or in the laboratory to determine if the model's internal dynamics are consistent with measured and theoretical dynamics. Effects of uncertain inputs : Numerical models have become relatively common tools in lake management. In many cases, they have also become useful for suggesting research needs, synthesizing extant information, and analyzing aquatic ecosystems in ways that are not tractable through field and laboratory studies alone. Models used most often in both contexts have similar attributes; they are generally time-dependent, often nonlinear, ordinary differential equation models based on parameterized physiological processes and mass conservation. These models, whether from the management or the research milieu, have another common thread: they are generally deterministic. That is, although it is often recognized that model initial conditions, parameters, and forcing functions have stochastic components, they are seldom accounted for. Moving beyond acknowledgment of variances of these elements to assessment of their effect is important because these stochastic properties affect the confidence 77 < ce m _i < o CO CO Ul o o cr 0. < UJ _l CD < cr > UJ I- co CQ o 00 ON 5 a CO S o fa < cr CD _l < O Ul _l CD < or Ul CO n M ct) 3 O CO M O a. CO o 4= Pn a) u 3 60 •H fa 78 that can be placed in the model output; that is, confidence is generally inversely related to variance. Analysis of this variability is important in a management context to establish error bounds on predictions. Output from these deterministic models often influences decisions affecting many thousands of people socially and economically (e.g., Vallentyne and Thomas 1978); yet quantitative confidence limits are lacking for these models (Thomann and Barnwell 1980). In particular, only qualitative evaluations of calibration and verification results have been carried out to date, and experience with even these tests is limited. Because eutrophication models are crude representations of highly variable, stochastic systems, ignoring such important attributes often results in naive confidence or unwarranted disbelief in the models' solutions. For these models to become more generally accepted and effectively used, they must be placed in their proper perspective. Evaluating the effects of input (forcing function and parameter) variance on model output provides some of the needed perspective. Analysis of model variability is also important in research contexts where a model's ability to simulate must be evaluated prior to investigation of specific system properties and recognition of actual system variability is important. Output from these models is often used to assess the relative importance of various system compartments or processes and thus to focus additional effort on key problems. Prior to using a model in this context, it is important to evaluate its ability to function as a synthesizer or interpolator. Traditionally, this evaluation is done by comparing model and measurement trajectories, with no quantitative assessment of model or measurement variability. As discussed above and in Scavia (1980b) comparison of modeled and measured state variables alone is not sufficient for this purpose. Calculation of variance associated with state variables and of correlations among state variables and parameters will assist in evaluation of these models for use in research contexts. We (Scavia et al. 1981a, b) have used Monte Carlo and first-order variance propagation analyses to explore impacts of uncertain parameters, loads, and initial conditions on a relatively simple model of plankton seasonal dynamics in Saginaw Bay, Lake Huron. In these analyses, we use estimates of natural variability of the input properties as sources of uncertainty. For Saginaw Bay, natural variability far outweighs uncertainty introduced by inaccurate measure- ments . Treating input errors in that way does not strictly estimate error associ- ated with the ability of the model to predict. To do this, one certainly must examine errors introduced by the equations themselves and perform the analysis over the time frame of the prediction, as has been done for some empirical and simpler lake models (e.g., Reckhow 1979). However, because variance due to measurement errors is small compared to natural variability in this system, these variance estimates measure at least their contribution to prediction variances . Using the input statistics and first-order variance propagation, state- variable variance estimates were made for an annual simulation based on variances of the initial conditions and given parameters. Resulting variances are represented as model output plus or minus its standard deviation in figure 15. 79 1 S I — 06 005 - 004 003 - 002 01 ~S ' 00 5 I 100 Phytoplankton (mgChla/L) 10 Organic -N (mgN/L) i i 06 > I < T 04 ^^r • il It t < "('*■* "^s^ I yi 0.2 0.0 -J i I Hi i j . . [ i i 1 Nitrate (mgN/L) P0 4 (mgP/L) 150 06 04t- 02 00 08 06 04 03 01 00 350 50 Time (days) Herbivores (mgC/L) L-u-i. i 1 1 r 04 Ammonia (mgN/L) 03 02 - 0.1 ^/-" 17TT^ si I T_r \ > | 00 i—S loH — tlT^ I i i I ! I" I Organic- P (mgP/L) Carnivores (mgC/L) Figure 15. Plots of eight state variable trajectories (smooth curve) from Saginaw Bay model. Model error estimates are represented as +1 standard deviation bands (shaded) from first-order analysis. Data (circles with error bars) are represented as baywide mean plus or minus standard deviation of all samples. From Scavia et. al, (1981a). 80 Peaks in variance estimates occurred at times when state variables were changing fastest. Maximum coefficients of variation (CV=standard deviation divided by state variable value) ranged between 148 and 722 percent, however, the average CV during the summer ranged between 33 and 407 percent. While these values are large, they are in many cases comparable to natural variability within the bay (table 1, fig. 15). However, because we have included only a subset of potentially important error sources and because we expect longer term prediction errors to be larger than those estimated here, it is of interest to determine the most significant sources of variability in this model. From the standpoint of model variance the relative effects can be demonstrated easily. In the simulations discussed below, initial condition, parameter, load, and mixing parameter variances were each used singly or in simple combinations. Assuming perfect knowledge of initial conditions (i.e., initial condition errors=0) reduced maximum output variances only slightly. Conversely, assuming uncertain initial conditions and perfect knowledge of parameter values resulted in much lower errors. Thus, parameter variance contributes far more than initial- conditions variance. (See first 3 lines of table 2.) Variance associated with loadings contributed little, even when compared to the low initial-condition contribution (line 4, table 2). None of the CV increased more than 20 percent when loading variances were included. In fact, only ammonia-nitrogen (NH3-N) and nitrate-nitrogen (NO3-N) CV increased more than a few percent. Including uncertainty (CV=10%) in a mixing parameter describing transport between bay and lake also had little effect (line 5, table 2). In fact, even when its assumed variance was doubled, no state-variable maximum CV increased more than 1 percent. These results are consistent with more detailed analyses performed on an ecologically simpler, two-segment model (Scavia 1980c). Input loading variance estimates represent time variability only. It is well known that estimating loads from highly variable, episodic inputs is difficult. To examine the potential influence of these inadequacies, two more cases were run, two and ten times the load variance, respectively. These runs assume that loading standard deviations, other than due to temporal variability, are equal to that due to temporal variability and equal to three times that variability, respectively. Dilution effects of the inner bay (volume 10*0 nH) somewhat mitigated even this variability when compared to variance propagated from initial-condition and parameter sources (lines 6 and 7, table 2). The largest effects were seen in the CV for NH3-N and NO3-N; increasing load variances by a factor of ten, an extreme case, resulted in doubling their model-output standard deviations. These tests of relative effects of different variance sources on propagated variances for a 1-year simulation indicated that parameters were by far the most significant contributors. The effects of initial-condition variance were quickly surpassed by the effects of parameter variance during the simulation, and only when very large loading measurement errors are assumed do load variances contribute significantly. We did not examine results of errors propagated over longer than the 1-year time frame. If we were examining long-term prediction errors, the effects of uncertain load predictions (not measurements) would have to be considered. This would certainly increase the variance contribution of loading estimates. Because the parameter errors had the largest impact on model output errors, we (Scavia et al. 1981a) made use of the propagated covariance matrix to identify 81 TABLE 1. Maximum and Mean of Summer Coefficients of Variation (Percent) Calculated by the First-Order Analysis From Uncertain Initial Conditions and Parameters Compared to Coefficients of Variation From Measured Variables. Variable Maximum Summer Mean* Observationst Phytoplankton 593 Herbivores 772 Organic N 148 Ammonia 201 Nitrate-nitrite 550 Organic P 163 P0 4 552 Carnivores 707 78 206 33 155 407 48 186 266 52 65 48 92 40 96 115 67 * Summer: July-September t Calculated coefficient of variation of spatially averaged values from all sampling dates 82 CO V o u 3 o w cu y c CG •H <2 c cu u cu <4-l <4-l •H Q o hi cu a o CO > CO u c cu <4-l cu o CJ I s cu CO w W -J I •H CO c cu u u CO O u > o •H c CO cu 60 U O en o •H c (0 z 60 o I i-t CO £1 CU M u cu o as > c I o O 4J ■u Jrf >v c •C CO Ph i-H a CO cu y u 3 o c/> (U CJ c CO •H V-. CO > CO -cf m o i-» o O vO vO v© -cr 00 00 CM O o m -* IT) VO p* m u-l CN CN en vO en vO vO CN vO vO CN en vO oo CM 00 u-1 en en VO O 00 O CN CO m *t Z r-l in m i— i m CN .-1 r-» Ov i-i CN VO v© i— 1 m m en vO >» OV O m m vO o i-l CN ■h CN oo oo en oo oo vo CN lO 00 CN en Ov m m CN CN oo oo vO CN CN oo c o c o ■H fs c o T3 a o u CO CO U cu T3 C CO CO C CO o CU 4-1 cu cu "O s s c *-> CO CO O n u u y C co co h cu a. en en oo oo oo a\ 1—1 en n en en CO u cu 4-1 CU CO CO CU co S •d •o s ■O CO CO CO CO CO U T3 o u o lJ O CO C r-l cu T-i CO ■H CX CO 4-1 Q. ft cu A « 60 co CO s CO 60 CO c C c CO c C C -H o o u o •H o K i-l •H CO •H X •H iH 4-) 4-1 a 4J T-l 4J e ■H i-l •H e •H T3 •o 60 •o TJ TJ C c c c •a c c o o •H o B O CO y y y CO y •-I CO r-l s ■H « iH ^-\ CO •o CO CO •-N CO o i-i CO i-i •a i-l CN •H r-l 4-1 O 4-1 c i-l X 4J X ■H r-l •H CO •H \^r 1-1 v-x C c c c 83 which parameters were most important in terms of both model sensitivity and model errors. Identification of those parameters and associated processes suggest areas requiring further research. One final aspect of model uncertainty becomes apparent when viewing the distributional properties of model output generated from Monte Carlo simulations (Scavia et al. 1981b). In this analysis the following procedure was used. The model equations were solved repeatedly. Each model execution was performed with initial conditions and parameter values selected randomly from their individual distributions. From these repeated model executions, state variable means, variances, and other statistics were calculated at 4-week intervals throughout the period of simulation. The analysis was terminated after 1000 simulations, at which time state variable means and variances were converging. Histograms generated from the 1000 simulations for selected state variables at different points in time during the annual cycle are shown in figure 16. It is interesting to note that even though the initial conditions and parameter values are drawn from relatively smooth, symmetric distributions, the resulting model output distributions can be dramatically asymmetric and polymodal. Several implications are suggested by these distributions. Those that are spread out suggest that variability in control parameters (initial conditions and coefficients) has a dramatic effect. That is, a fairly uniform output distribution suggests many possible model outcomes are equally likely as control parameters vary within their confidence ranges. Distributions that are dramati- cally narrow indicate relative insensitivity of that state variable to the uncertainty in control parameters. Bimodal or polymodal distributions suggest that even though control parameter values have probabilities of occurring that vary smoothly through their distributions, the model produces state-variable values that jump from one category of high probability to another with very few outcomes occurring in between. Controls of such threshold behavior both in models and in nature are not well understood, but careful attention must be paid to the possibility of it occurring. SUMMARY AND CONCLUSIONS I have outlined how a particular ecosystem model has been used to better understand the structure and dynamics of Lake Ontario. This is but one example of how integrated modeling and experimental science has advanced our ability to understand and perhaps simulate and predict dynamics of the Great Lakes. This and similar models generally represent collections of process relationships developed through independent empirical studies and as such they merely test those relationships in the context of the whole system. That is, the models test our ability to simulate algal dynamics, for example, by balancing rates of gain and loss calculated from expressions developed independent of the whole system. While certain relationships among processes may be testable, it usually becomes intractable to study such relationships in nature. This is particularly true in systems like the Great Lakes and marine waters where physical processes can be important. In those cases, a model that is firmly based on independent, empirically tested constructs and exercised within the framework of carefully designed field observations may be the only means to improve our understanding of, and thus our ability to predict, responses of the aquatic ecosystem to altering stresses. 84 Phytoplankton ILLInJUJ Nitrate 278 325 278 580 532 246 ULLLL Available Phosphorus Day 112 Day 140 Day 224 Day 280 Day 364 Julian Day Figure 16. Frequency distributions of Monte Carlo output for four state variables at selected time slices during annual simulation. Day 50 distributions are initial distributions. 1000 cases were used. From Scavia et al. (1981b). 85 Because these models are based generally on mechanistic relationships rather than on strict statistical criteria, they can certainly produce reasonable simulations using sets of different coefficient values. While each set of values may result in indistinguishable state variable predictions, they will not likely all give accurate representation of the processes that contribute to those state variable trajectories. Errors in those process rate simulations may lead to erroneous conclusions concerning critical controls of ecosystem dynamics. It is thus important to measure simultaneously both state variables and process in situ when developing or testing these models. Although many field estimates of processes are difficult to obtain, measurement of even some will provide checkpoints of the simulation of internal dynamics. It is less likely that compensating errors in process simulations will remain if the model adequately reproduces both measured states variables and certain processes. Finally, it must be remembered that nature is not predictable in a strictly deterministic fashion given the state of our understanding of its laws. Many events occur on time and space scales smaller than we can attempt to model. These events will add stochastic properties to our measurements. Recognition of these processes and their effects on model prediction confidence must be addressed in any model of nature. REFERENCES Bennett, J. R. , 1971. "Thermally driven lake currents during the spring and fall transition periods," pp. 535-544. In Proc. 14th Conf . Great Lakes Res., Ann Arbor, Mich . Bennett, J. R. , 1974. "On the dynamics of wind-driven lake currents," J . Phys . Oceanogr . 4:400-414. Burns, N. M. , and A. Pashley, 1974. "In situ measurement of the settling velocity profile of particulate organic carbon in Lake Ontario," J. Fish. Res. Board Can. 31:291-297. DePinto , J. V., and F. H. Verhoff, 1977. "Nutrient regeneration from aerobic decomposition of green algae," Environ. Sci. Tech. 11:371-377. Ganf, G. G., and P. Blazka, 1974. "Oxygen uptake, ammonia and phosphate excretion by zooplankton in a shallow equatorial lake (Lake George, Uganda)," Limno 1 . Oceanogr. 19:313-325. Glooschenko, W. A., J. E. Moore, and R. A. Vollenweider , 1972. "The seasonal cycle of pheo-pigments in Lake Ontario with particular emphasis on the role of zooplankton grazing," Limno 1. Oceanogr. 17:597-605. Golterman, H. L. , 1973. "Vertical movement of phosphate in freshwater," pp. 509- 538. In E. J. Griffiths, A. Beeton, I. M. Spencer, and D. Mitchell (eds.), Environmental phosphorus handbook . Wiley, New York, N.Y. Heidke, T. M. , 1979. Modeling the Great Lakes system; Update of existing models . Great Lakes Basin Commission, Ann Arbor, Mich., p. 79. McNaught, D. C. , M. Buzzard, and S. Levine, 1975. "Zooplankton production in Lake Michigan as influenced by environmental perturbations," U.S. Parsons, T. , and M. Takahaski , 1973. Biological oceanographic processes . Pergamon Press, New York, N.Y. PTSTF, 1980. "Phosphorus management for the Great Lakes," Final report , the phosphorus management strategies task force. International Joint Commission, Windsor, Ontario, p. 129. 86 Reckhow, K. H. , 1979. "Empirical lake models for phosphorus development: Applications, limitations, uncertainty," pp. 183-222. In D. Scavia and A. Robertson (eds.), Perspectives on Lake Ecosystem Modeling , Ann Arbor Science, Ann Arbor, Mich. Rigler, F. H. , 1973. "A dynamic view of the phosphorus cycle in lakes," pp. 539-572. In E. J. Griffiths, A. Beeton, J. M. Spencer, and D. T. Mitchell (eds.), Environmental phosphorus handbook , Wiley, New York, N.Y. Riley, G. A., 1942. "The relationship of vertical turbulence and spring diatom flowerings," J. Mar. Res. 5:67-87. Riley, G. A., 1946. "Factors controlling phytoplankton populations on Georges Bank," J. Mar. Res. 6:54-73. Riley, G. A., 1963. "Theory of food-chain relations in the ocean," pp. 438- 463. In M. N. Hill (ed.), The sea , Interscience, New York, N.Y. Riley, G. A., H. Stommel, and D. F. Bumpus , 1949. "Quantitative ecology of the plankton of the western North Atlantic," Bull. Bingham Oceanogr. Coll. 12:1-169. Scavia, D. , 1979. "Examination of phosphorus cycling and control of phytoplankton dynamics in Lake Ontario with ecological model," J. Fish. Res. Board Can. 36:1336-1346. Scavia, D. , 1980a. "An ecological model of Lake Ontario," Ecol. Model. 8:49-78. Scavia, D. , 1980b. "The need for innovative verification of eutrophication models," pp. 214-225. In R. V. Thomann and T. V. Barnwell, Jr. (eds.), Proceedings of National Workshop on Verification of Water Quality Models . U.S. Environmental Protection Agency, Athens, Ga. Scavia, D. , 1980c. Uncertainty analysis of a lake eutrophication model , Ph. D. dissertation, University of Michigan, Ann Arbor, Mich, 159 pp. Scavia, D. , and J. R. Bennett, 1980. "The spring transition period in Lake Ontario - A numerical study of the causes of the large biological and chemical gradients," Can. J. Fish. Aquat. Sci. 37:823-833. Scavia, D., R. P. Canale , W. F. Powers, and J. L. Moody, 1981a. "Variance estimates for a dynamic eutrophication model of Saginaw Bay, Lake Huron," Water Res our. Res. 17:1051-1059. Scavia, D. , B. J. Eadie, and A. Robertson, 1976. "An ecological model for Lake Ontario - model formulation, calibration and preliminary evaluation," NOAA Tech. Rep. ERL 371-GLERL-12, 63 pp. Scavia, D. , W. F. Powers, R. P. Canale, and J. L. Moody, 1981b. "Comparison of first-order error analysis and Monte Carlo simulation in time-dependent lake eutrophication models," Water Resour. Res. 17:1115-1124. Simons, T. J., 1976. "Analysis and simulation of spatial variations of physical and biochemical processes in Lake Ontario," J. Great Lakes Res. 2:215-233. Sridharan, N. , and G. F. Lee, 1977. "Algal nutrient availability and limitation in Lake Ontario during IFYGL," U.S. Environ. Prot. Agency Rep. EPA-600/3-77- 046a. Stadelmann, P., and A. Fraser, 1974. "Phosphorus and nitrogen cycle on a transect in Lake Ontario during the International Field Year 1972-1973 (IFYGL)," pp. 92-107. In Proc. 17th Conf. Great Lakes Res. Int. Assoc, for Great Lakes Res. Stadelmann, P., J. E. Moore, and E. Pickett, 1974. "Primary productivity in relation to temperature structure, biomass concentration, and light condi- tions at an inshore and offshore station in Lake Ontario," J. Fish. Res. Board Can. , 31:1215-1232. Steele, J. H. , 1965. "Notes on some theoretical problems in production ecology," pp. 383-398. In C. R. Goldman (ed.), Primary Production in Aquatic Environ- ments . University of California Press, Berkeley, Calif. 87 Stoermer, E. F., and T. B. Ladewski, 1978. "Phytoplankton associations in Lake Ontario during IFYGL," Great Lakes Res. Div. Univ. Mich. Spec. Rep. 62: 106, University of Michigan. Sverdrup, H. U. , 1953. "On conditions for the vernal blooming of phytoplankton," J. Cons. Perm. Int. Exp. Mer. 18:278-295. Thomann, R. V., and T. 0. Barnwell, Jr., 1980. "Workshop on verification of water quality models," U.S. Environ. Protection Agency Report , EPA-600/9- 80-016, p. 258. Vallentyne, J. R. , and N. A. Thomas, 1978. Fifth year review of Canada-United States Great Lakes Water Quality Agreement , Report of Task Group III, a technical group to review phosphorus loadings, 86 pp., U.S. Dept. of State, Washington, D.C. 88 OIL SPILL MODELING FOR OCS ENVIRONMENTAL ASSESSMENT AND DECISIONMAKING David E. Amstutz Bureau of Land Management Department of the Interior Washington, D.C. 89 ABSTRACT The Outer Continental Shelf (OCS) Lands Act, as amended, requires the Secretary of the Interior to make publicly owned oil and gas resources available to help meet our Nation's energy needs. Within the Department of the Interior, the Bureau of Land Management (BLM) carries out the leasing process; the Minerals Management Service supervises leases once they have been sold (until early 1982 this function was performed by USGS). Since OCS leasing involves actions which may impact the environment, the leasing process is subject to the National Environmental Policy Act (NEPA) . NEPA in turn requires that BLM produce environ- mental statements (ES) for each scheduled lease sale. A significant portion of each ES deals with accidental spills related to production and transportation of offshore oil. Oil spill risk analysis modeling is performed in DOI jointly by BLM and USGS. Input data are provided by BLM; the modeling work itself is performed and formally reported by USGS; and the model results are used by BLM. The modeling work is undertaken from a cumulative perspective and includes spill simulations from existing transportation routes and, where applicable, from existing Federal leases. Model results are analyzed for use by decisionmakers on matters relating to lease tract deletion and transportation alternatives and lease stipulations. The modeling work is predictive (for the next two to three decades) and couched in probabilistic terms, to account for the numerous uncertainties existing at the prelease sale stage. Thus, the model is unlike deterministic or real time models. INTRODUCTION 1. Background The Outer Continental Shelf Lands Act of 1953 charges the Department of the Interior (DOI) with carrying out a national offshore oil and gas leasing program. Within the DOI, the Bureau of Land Management (BLM) is responsible for leasing offshore resources and the Minerals Management Service is responsible for supervising offshore leases. (Lease supervision and enforcement activities were transferred from the U.S. Geological Survey (USGS) to the Minerals Management Service in early 1982.) Since its first offshore lease sale in 1954, BLM has sold more than 4,000 leases for a total in bonus bids of $36 billion; lease rental and production royalties have exceeded $10 billion (BLM 1981). Through 1980 a total of 5.4 billion barrels of crude oil has been produced (USGS 1980). The Outer Continental Shelf (OCS) program is carried out in compliance with numerous laws. A compilation of laws governing resources and resource management on the OCS is presented in DOI (1981). The National Environmental Policy Act of 1969 and the OCS Lands Acts Amendments of 1978 provide for the environmental modeling work reported on in this paper. 91 The Federal OCS includes all areas seaward of a line drawn 3 geographical mi from the shoreline. Off Texas and western Florida the line is drawn 3 marine leagues from the shoreline. Legislation defining the OCS is contained in the Submerged Lands Act of 1953 and the Outer Continental Shelf Lands Act of 1953. Seaward limits to America's OCS are either not defined or are in negotiation with adjacent nations. OCS lease sales have been held within all regions of the OCS except for portions of the Bering and Chukchi Seas and areas with water depths greater than 2 to 3 km. Oil spills are a major environmental issue identified by parties associated with OCS leasing (Federal and State governments, industry, conservation groups, and the general public). Expressions of concern and requests for information for use in analysis have led to the development of the DOI oil spill risk analysis model. Model outputs address three topics: the likelihood of spill occurrence, likely pathways that spills might follow, and the risks that spills will occur and will contact various resources. The model treats uncertainties inherent with OCS related spills as well as spills from other sources. The model serves as a quantitative framework for synthesizing enormous amounts of environmental information. Model outputs are intended for use by environmental analysts and program decisionmakers. 2. Scope Oil spill risk analysis modeling is carried out as a joint effort of BLM and USGS. Because of this and the wide collection of users of the modeling results, the work may be considered as a DOI project. Input data to the model are provided by BLM. These data includes all environmental information as well as definition of leases to be offered and transportation senerios to be used if oil is found. USGS provides estimates of oil and gas resources within the lease sale area and carries out all of the model computations. Reports of model output are provided to BLM for use and analysis in writing environmental statements and drafting DOI decision documents. Decision documents are finalized within the Department. Input data reflect the best available information for each of the OCS lease sale areas. The existing information base is supplemented through studies sponsored by BLM's Offshore Environmental Studies Program. The Studies Program began in the mid-seventies and continues to date. Approximately half of the BLM sponsored studies have been conducted off Alaska. Data on ocean circulation, local winds, and locations of various marine resources comprise the bulk of model inputs . The spatial extent of each model accounts for spills originating from pro- duction sites as well as along transportation routes. Spill trajectories are analyzed for up to 30 days. For example, a recent model for the area off the southeastern coast extended from Miami to Norfolk and from the shore to well over the Blake Plateau. The oil spill risk model is predictive in that it treats future events. The future, extending 2 to 3 decades, is defined by the estimated time to complete production from an offshore lease. Because the model deals with uncertain events and environmental circumstances, it is couched in probabilistic terms. This treatment of uncertainty distinguishes the DOI model from deterministic 92 oil spill models, such as those developed and run by the National Oceanic and Atmospheric Administration and the U.S. Coast Guard for response to real time spills (Wallops Workshop 1980). OIL SPILL RISK MODEL 1. Overview The DOI oil spill risk model is one of the largest environmental models in current use; larger models are used by the National Weather Service. The overall workings of the model are discussed by Smith et al. (1982). Detailed documentation of the model is presented by Lanfear and Nakassis (1980) and Lanfear and Samuels (1981). A brief overview of the model with examples of lease sale applications is presented by Lanfear et al. (1970). Model runs for each OCS lease sale, beginning in 1976, are described and results presented in the USGS Open File Report series; recent examples, for Southern California and the Gulf of Mexico, are presented by Samuels et al . (1981a) and LaBelle and Samuels (1981), respectively. For a specific application the region to be modeled is defined based upon the locations of potential spill sites (production and transportation), locations of potentially vulnerable resources, and knowledge of winds and currents which characterize the region. Typically the modeled region includes 600 to 800 nmi of coastline and extends seaward about 400 nmi. The model is structured on a grid (480 x 480). All spatial information are digitized and portrayed on the model grid. Computer programs are incorporated to allow input data to be on virtually any map projection and map scale. Digitized inputs include the shoreline, ocean currents, and locations of biological resources. Hypothetical oil spills are launched from potential spill sites and advected within the model grid by wind and current. Spill contacts with various "targets" are recorded along with the time between spill launch and contact. Spills are launched throughout the year and in sufficient numbers to establish statistical significance to the contact probabilities. Contact probabilities derived in this fashion are conditional in that spill occurrence is assumed. The conditional probabilities are then combined with the likelihood of spill occurrence to yield the final (joint) probabilities that spills will occur and will contact specific targets. 2. Salient Points The oil spill risk model is well documented in the published literature. In addition, as noted above, there are published reports describing each sale specific model run. Thus, I attempt here to summarize selected topics. The topics were chosen to reflect the capabilities of the model. The topics discussed below are intended to also represent the collection of issues most frequently raised by the public who interact with the OCS Program. The model deals with oil only on the ocean surface. Spills, represented by points (hypothetical center of mass of a surface slick), are advected by surface currents and winds. The surface currents are climato logical (generally average 93 ninthly) and do not contain an accounting of local wind effects (geos trophic flow). Local wind-induced drift is computed to be 3 1/2 percent of the local wind speed with a 20° clockwise rotation from the wind direction. Following the work of Samuels et al. (1981b), recent model runs incorporate a variable deflection angle which is an exponential function of wind speed. The oil advection algorithm is taken as the vector sum of the surface current vector and local wind-induced drift. Seasonal portrayals of surface currents vary spatially as dictated by available data; examples are contained in Samuels et al. (1982a). Consequences of using various hypothesized versions of surface currents have been examined (Lanfear and Amstutz 1981). Local winds are sampled at 3-h intervals, in Monte Carlo fashion, from seasonal wind transition matrices. The matrices are constructed from time series observations, generally measured at coastal stations. The 41 x 41 matrices represent eight directions with five speed classes each, and the calm condition. Winds are thus treated as a first-order Markov process. Wind zones are assigned over the modeled area to dictate where each station time series applies. Wind zone definition is determined through comparison of wind roses at sea, derived from ship observations, with those constructed from the coastal station time series. Spill advection continues until the spill contacts land, encounters a model boundary, or remains at sea for more than 30 days. Spills are launched through- out the year (500 per season or 2,000 per year) from each launch site; thus the Monte Carlo sampling error does not exceed 2 to 3 percent. Launch sites may be: single points, to simulate platform locations; along lines, to simulate pipelines and tanker routes; or collections of uniformly distributed points, to simulate several platforms in a small area. Biological resources (commercial fishing areas, whale migration routes, sea otter ranges, pelagic sea bird feeding grounds, etc.) are represented spatially for times (months) they are considered sensitive to oil spills. Shorelines may be subdivided by type or use (e.g., rocky shores, salt marsh, high- intensity use beaches) and by land segments. Land segments may be designed to be of equal length (typically 20 to 30 nmi) and also by arbitrary criteria such as to be coincident with political subdivisions. Conditional probabilities are tabulated annually by launch site identifica- tion and target name, using transit intervals of 3, 10, and 30 days. These conditional probabilities (numerically determined by winds, currents, locations, and temporal sensitivities of resources and locations of spill sites) portray risks from oil spills without consideration of the likelihood of oil spill occurrence. Conditional probabilities, though useful to decisionmakers, are beyond their control. Determination of future spill incidence is a complex task. The DOI oil spill risk model projects future spill incidence upon past experience using volume of oil as an exposure variable. Predicted probability distributions for spill incidence are calculated separately for platforms, pipelines, and tankers. Spill occurrence is calculated for size categories equal to and greater than 1,000 barrels and 10,000 barrels. The predictive procedure used in the model was initially developed by Devanney and Stewart (1974) using Bayesian techniques, Additional analyses have addressed the data bases and alternative exposure 94 variables (Stewart 1976, Stewart and Kennedy 1978). Devanney and Stewart (1974) estimated the number of spills using the negative binomial distribution. Since the exposure variable (volume of oil) associated with an OCS sale is less than the historical exposure, the Poisson distribution serves as an excellent approximation to the negative binomial distribution (Smith et al. 1982). The Poisson distribution is defined by its single parameter, the expected number of spills . The oil spill data bases used in the DOI model were initially assembled by Devanney and Stewart (1974) and Stewart (1975, 1976). OCS production records are maintained by USGS, as are records of platform and pipeline spills. Oil spill incidences are currently under review by the Futures Group (1982) under contract with BLM. There have been 10 OCS platform spills greater than or equal to 1,000 barrels since 1964; eight of these occurred prior to 1974. A critical analysis of these platform spills and production since 1964 has quantified a decrease in platform spill rate (Nakassis 1982). Pipeline and tanker spill rates are also under review by personnel in BLM and USGS. Spill incidence rates used in the DOI model receive considerable criticism from numerous groups. Generally the criticisms suggest that the spill rates are either too high or too low; that the rates are not applicable to some OCS regions (recall that all of America's OCS production to date has been from the western and central Gulf of Mexico and southern California); and that there must be better exposure variables. Those of us responsible for the oil spill model have adopted the following policy: 1) the past OCS experience is the best available quantitative basis from which to predict future events; 2) the number of spill incidences is of record and will be made available to those who desire testing their hypotheses regarding spill incidence; 3) if alternative exposure variables can be demonstrated to be more precise and of more utility to decisionmakers they will be incorporated; 4) it is reasonable to use a length of past record comparable with the length of forecast (this, incidently, is necessary if we are to examine the record for changing rates); 5) as for regional applicability, the spill rates used in the model have been examined against the records of experience in Cook Inlet (pro- duction from State leases) and in Prohdue Bay (onshore production) - the experiences in these two locations are not different, in a sta- tistically significant sense, than our experiences in the Gulf of Mexico ; and 6) lastly, we are able to define the circumstances (required production without spills) needed to achieve a difference from our past experience. 95 The spill incidence record on America's OCS is in fact remarkedly low. Although we cannot quantify specific reasons, the consequences are a tribute to all from our society who participate in offshore oil and gas activities. The final stage of oil spll risk analysis combines the conditional probabilities and spill likelihood to yield final (joint) probabilities — the probabilities that spills will occur and will contact various resources. The calculation of joint probabilities involves use of hydrocarbon resources expected to be produced and transported. The volumes used in the model are provided by USGS and are the mean economically recoverable resources. Joint probabilities are calculated as follows: (a) Form matrix [c], where elements c^ i are the conditional probabilities that a spill from site j will contact target i. (b) Form matrix [s], where elements Sj ^ are the expected number of spills from site j due to unit volume produced at site k. (c) Form matrix [u] « [c] x [s], where elements u^ -s are the expected number of spills occurring and contacting target i due to production of a unit volume of oil at site k. (d) Form vector [v], where element v(k) represents the volume of oil expected to be produced at site k. (e) Form vector [L] » [u] x [v], where element l(i) is the expected number of contacts to target i. (f) Using the Poisson distribution discussed earlier and the vector [L] , the probability of exactly n contacts to target i may be calculated by: P(n,i) = [l n (i) exp(-l(i))]/n! The joint probability that one or more spills will occur and will contact target i is obtained by summing P(n,i) over n > 0. More detailed discussion of the above procedure is contained in recent lease sale specific oil spill reports published by USGS, such as Samuels et al. (1982b). The steps outlined above illustrate the combining of OCS production and transportation spills. This is an important factor — one cannot decide to produce at a specific site without concomitant consideration of oil transport. USE OF MODEL RESULTS 1. Applications Model results are presented and analyzed in Environmental Statements (ES) prepared in BLM's OCS Field Offices. The content and analyses of an ES are defined by the National Environmental Policy Act and its implementing regulations In compliance with these, the model addresses each proposed lease sale from three perspectives. First, a portrayal of the proposed sale by itself, as though there were no other activities taking place in the sale area. Second, a 96 portrayal of the sale area environment as it would be without the proposed sale. Third, the cumulative circumstances, a portrayal of the sale area environment as it would be if the lease sale were held and all other activities continued. Sources of accidental spills considered in the model framework for the cumulative circumstance include: production and transportation of petroleum hydrocarbons into and within the area. For each of these portrayals the analysts must consider the proposal in its entirety as well as various alternative sale configurations. Oil volumes, used as exposure variables in the oil spill risk model, are also used in sale specific economic analyses thereby affording a common basis for cost/benefit analysis. Analytical attempts have been made to maximize oil production while minimizing oil spill risk (Smith et al. 1979). Although this can be readily accomplished, numerical solution would presumably require assignment of weights to the targets. Without weights the targets would be treated with equivalence, which implies that spill contacts yield equivalent consequences. Weights can be assigned in terms of acceptable numbers of contacts. It seems reasonable to assume that decisionmakers make such judgments during their personal deliberations. At the final stages of a proposed OCS sale process, the model results are incorporated into decision documents. Decision documents are used by agencies within DOI to make recommendations, concerning the proposed sale and its alternatives, to the Secretary. 2. Complications The oil volumes used in the model analysis are mean estimates of economically recoverable amounts. In analyzing sale alternatives (tract deletions for example), one often encounters circumstances where volume dependencies (for economic and geologic reasons) exist. The analysis thus goes well beyond simply deleting an element from the volume vector. Quite often we observe that oil spills from transportation routes present greater threat to resources than spills from production sites. This is expected based upon proximity. The DOI is committed to write an ES in frontier areas which addresses the development stage of the offshore leasing process. (Frontier areas are those in which there has been no previous production; the development stage follows leasing but precedes actual production.) A development ES will be greatly enhanced by the capabilities of the oil spill risk model to evaluate alternative transportation routes. Model outputs consist of thousands of numbers. The majority of the proba- bilities is negligible; nevertheless, analysis of the remaining values is a difficult and complex task. Not all of the values can be analyzed. All of the probabilities are presented in the ES, however, to enable readers to apply their own judgments and analyses. The model deals with contacts; impacts are quite another matter. Impacts from oil spills are very much dependent upon how and where a spill occurs (underwater blowout versus blowout into the atmosphere; pipeline failure at sea versus tanker spill near shore, etc.). Impacts are also very much dependent upon the physical-chemical properties of the spilled oil. These properties 97 are, in almost all cases, unknown at the time of a lease sale analysis. Oil properties will be far better understood at the time a development ES is written. Impact assessment requires knowledge of oil properties not only at the location of the spill but at the time and location of contact with sensitive resources. Thus, impact analysts desire quantification of oil weathering. Examination of oil weathering studies reveals that time is the single most important variable. Study of weathering algorithms reveals near linear dependencies on time. As a first approximation then, the model retains measure of the time between spill occurrence and target contact. The times for which conditional and joint probabilities are accumulated (3, 10, and 30 days) were chosen for their use as implicit measures of oil weathering — as well as for matters relating to contain- ment and clean-up. Spill sizes range over seven to eight orders of magnitude. In general, given such a range, the average size has little meaning. Observed OCS platform spills have the least range in sizes, with pipelines second; tanker spills account for the widest range in sizes. The OCS platform spills (1964 to present) greater than or equal to 1,000 barrels range from 1,500 to 77,000 barrels; the average is approximately 19,000 barrels. 3. The Future Efforts have been underway for some time now to apply three-dimensional ocean circulation models on the OCS. Results have been substantial: to our understanding of coastal circulation, to our abilities to quantify stochastic elements of the circulation, and to supply circulation data beneath the sea surface. Meteorological studies have successfully established quantitative means to construct at-sea time series winds from time series measured along the coast. Several years of direct measurement of time series winds are being acquired with anchored meteorological buoys. Studies of the distribution of sea ice, its properties, and oil/ice interaction are yielding results applicable to oil spill risk analyses. ACKNOWLEDGMENTS I wish to acknowledge James Slack, USGS, who constructed the initial DOI oil spill model; Kenneth Lanfear and the Environmental Modeling Group, USGS, who so ably perform the modeling work; the several BLM field office personnel who have suggested improvements to the model; and the several marine scientists who have contributed data used in the model runs. REFERENCES BLM, 1981. OCS Statistical summary . U.S. Department of the Interior, Bureau of Land Management (4 volumes). Devanney, J.W., III, and R.J. Stewart, 1976. Analysis of oil spill statistics . Report to Council on Environmental Quality, Washington, D.C. 98 DOI, 1981. Compilation of laws related to mineral resource activities on the outer continental shelf . U.S. Department of the Interior, January 1981. ( 2 vo lumes ) . Futures Group, 1982. An analysis of oil spill statistics . Reports in preparation by the Futures Group, under contract to BLM. BLM Contract No. AA550-CTO-69. LaBelle, Robert P. and Kenneth J. Lanfear, 1981. "An oilspill risk analysis for the Gulf of Mexico (Proposed Sale 67 and 69) outer continental shelf lease area," USGS Open File Report 81-806. Lanfear, K.J. , R.A. Smith, and J.R. Slack, 1979. "An introduction to the oil spill risk analysis model," Presented at the 11th Annual Offshore Technology Conference in Houston, Texas, April 30-May 3, 1979. OTC Paper No. 3607. Lanfear, Kenneth J. and David E. Amstutz, 1981. "Environmental studies for oilspill trajectory modeling in the southeastern U.S. outer continental shelf lease area," Presented at the 13th Annual Offshore Technology Conference, Houston, Texas, May 4-7, 1981 . Lanfear, Kenneth J. and Anastase Nakassis, 1980. "Documentation and users guide to the spatial environmental data digitizing systems, SEDDS," USGS Open File Report 80-871. Lanfear, Kenneth J. and William B. Samuels, 1981. "Documentation and user's guide to the U.S. Geological Survey oilspill risk analysis model: oilspill trajectories and calculation of conditional probabilities," USGS Open File Report 81-316. Nakassis, Anastase, 1982. "Offshore oil production becomes safer?," USGS Open File Report 82-232. Samuels, William B., Dorothy Hopkins, and Kenneth J. Lanfear, 1981a. "An oilspill risk analysis for the Southern California (Proposed Sale 68) outer continental shelf lease area," USGS Open File Report 81-605. Samuels, William B. , Norden E. Huang, and David Amstutz, 1981b. "The sensitivity of an oilspill trajectory analysis model to variation in wind deflection angle," Presented to 44th Annual Meeting, ASLO, Milwaukee, Wisconsin, 15-18 June 1981 ; accepted for publication, International Journal of Ocean Engineering (1982). Samuels, W.B., LaBelle, R.P., and D.E. Amstutz, 1982a. "Oilspill trajectory modelling in Alaskan waters," Presented to the American Geophysical Union - American Society of Limnology and Oceanography joint meeting, February 1982, San Antonio, Texas . Samuels, William B., Dorothy Hopkins, and Kenneth J. Lanfear, 1982b. "An oil spill risk analysis for the Beaufort Sea (Proposed Sale 71) outer continental shelf lease sale area," USGS Open File Report 82-13. Smith, Richard A., Kenneth J. Lanfear, and Ivan C. James, III, 1979. "Oilspill risk minimization through optimal tract selection," Presented at the National Weather Service Conference, "Physical Behavior of Oil in the Marine Environment," at Princeton University, May 8-9, 1979 . Smith, R.A. , James R. Slack, Timothy Wyant, and Kenneth J. Lanfear, 1982. "The oilspill risk analysis model of the U.S. Geological Survey," USGS Professional Paper 1227. Stewart, R.J., 1976. A survey and critical review of U.S. oil spill data resources with application to the tanker/pipeline controversy ^ Report to Office of Policy Analysis, U.S. Department of the Interior, Contract No. 14-01-0001-2193. USGS, 1980. PCS Statistics, 1953-1980 . U.S. Department of the Interior, U.S. Geological Survey, Conservation Division, June 1981. Wallops Workshop, 1980. Proceedings of the workshop on government oil spill modeling, November 7-9, 1979, Wallops Island, Va . U.S. Department of Commerce, NOAA/EDIS, February 1980. 99 OIL SPILL-FISHERY IMPACT ASSESSMENT MODELING: APPLICATION TO GEORGES BANK Malcolm L. Spaulding Department of Ocean Engineering University of Rhode Island Kingston, Rhode Island 02881 and Mark Reed Applied Science Associates, Inc. 529 Main Street Wakefield, Rhode Island 02879 and Saul B. Saila, Ernesto Lorda, and Henry A. Walker Graduate School of Oceanography University of Rhode Island Narragansett, Rhode Island 02882 101 ABSTRACT An oil spill fishery impact assessment model composed of an oil spill fates model, a continental shelf hydrodynamics model, an ichthyoplankton transport and fate model, and a fishery population model has been applied to the Georges Bank — Gulf of Maine region to assess the probable impact of oil spills on several im- portant commercial fisheries. The model addresses direct impacts of oil on a commercial fishery through hydrocarbon-induced egg and larval mortality. This early life stage mortality is estimated by mapping the dynamic spatial/temporal intersection of the surface and subsurface oil concentrations resulting from the spill, and juxtaposing dynamic maps of the developing eggs and larvae. Ichthyoplankton entering an area with hydrocarbon concentrations in excess of a specified threshold are assumed lost. Model output is given in terms of differential catch, comparing the non-impacted and the hydrocarbon impacted fisheries. Simulations of the impacts of monthly oil well blowouts at a site in the North Atlantic Outer Continental Shelf (OCS) lease area have been completed for Atlantic herring and Atlantic cod. Results of these case studies clearly show the importance of spill timing and location, spatial and temporal spawning patterns, and details in the hydrodynamic transport field as critical factors in determining spill impact. Model system sensitivity studies to incremental losses of a given year class and the influence of the Georges Bank gyre on impact are also presented. INTRODUCTION One of the major concerns associated with Outer Continental Shelf (OCS) hydrocarbon exploration and development, is the release of oil into the marine environment, and the resulting short- and long-term effects on the ecosystem. Of particular concern in productive fishing areas such as Georges Bank, is the impact of spills on the higher trophic levels, and more specifically on com- mercially important fish species. Realistic assessments of the impact of spills on these species are essential if the fish and mineral resources of these shelf regions are to be managed rationally. An objective assessment methodology is clearly needed to support a rational resource management policy. The methodology selected should be able to (1) reliably quantify impacts, (2) appropriately represent the space and time scales of the pollutant events of interest, (3) take advantage of existing environmental data and therefore minimize the need for additional data, (4) be formulated within the framework of management needs, (5) address a well-defined ecological unit, (6) limit the number of empirical formulations and, (7) be trans ferrable to other geographical areas and species. These requirements are intended to assure that the approach to impact assessment is credible, useful, and affordable. Scientists addressing biological impact assessment problems have developed three major model methodologies, characterized here as the statistical, the indicator species, and the ecosystem dynamics approaches. 103 The statistical approach is the oldest, most common, and potentially the simplest form of impact estimated. The basis of the approach is the derivation of an empirical transfer function relating some ecosystem metric to impact magnitude. Typical metric constructs include matrices, perhaps with weighting functions (Cantor 1977), species diversity indices, and correlation procedures (Pielou 1977). Field sampling and monitoring programs, coupled with a variety of statistical methods, represent the standard format for statistical environ- mental impact analyses. Major drawbacks of the approach arise from its sensi- tivity to the metric selected, and the high degree of natural variability usually reflected in the underlying data. This latter problem results in a common inability to answer the question, "When is an impact (i.e., a change in the system metric) significant?" In addition, questions of which variables to measure, over what time span, and at what sampling intervals and locations are non-trivial, but must be answered. Because of the high degree of agglomeration inherent in statistical methods, the dynamic complexities of the "real" system become obscured. Thus, little insight into the processes controlling system response is gained, and the ability to address a variety of realistic management options is usually lost. At the other end of the complexity spectrum lies the full ecosystem modeling approach. In its purest form, this methodology incorporates all our knowledge of environmental functioning at the process level. In practical applications some agglomeration at the biomass level (Laevastu and Larkins 1981) or at the level of numbers (Anderson and Ursin 1977; Reed and Balchen 1982) is necessary to achieve completion of a project within prescribed temporal, economic and computational constraints. In the past, this agglomeration commonly led to a "box model" representation, in which species or species types and their inter- actions were represented through a set of highly parameterized differential equations (e.g., Chen 1975; Kelly 1975). Increasing accessibility of powerful computational facilities has made feasible more detailed, pro cess -specific representations. Such detail is attractive because of increased "realism". The construction of such an ecosystem model is based on some form of conservation laws, including appropriate source, sink, and interaction processes, to relate ecologically critical components. The more detailed and "realistic" these models are, the more variables and parameters they require. Such efforts are therefore sometimes characterized as "data hungry", and may require relatively arbitrary assignment of values to many parameters. Conclusions drawn from simulation outputs may therefore be subject to considerable uncertainty. The data necessary to support this modeling approach, in terms of input as well as validation, is rarely available at present. On the positive side, an ecosystem model with appropriate spatial, temporal, and biological resolution makes better use of available data than simple statistical analyses, in that information inherent in the data is conserved in the model. Spatial species dynamics, for example, are well documented in most fisheries data, but traditional fish population models (e.g., Beverton and Holt 1954) have essentially neglected this aspect. More recent ecosystems models for fisheries management (Laevastu and Larkins 1981; Reed and Balchen 1982) have demonstrated the importance of spatial dimensions in arriving at an under- standing of the governing processes. 104 These process-explicit ecosystem models with spatial representation are also useful for investigating the impacts of various alternative management strategies. Improved data gathering technologies (e.g., satellite remote sensing), combined with improved computer capabilities will render these complex models more credible, giving them wider acceptance, and they will probably become the preferred management tool of the future. The fact remains that rational decisions must be made now, in the 1980s, concerning relatively complicated and important resource management alterna- tives. The indicator species, process-explicit modeling approach to environmental impact assessment provides a workable compromise between the statistical and full ecosystem approaches. Single species indicator models have seen extensive use in assessing the impact of power plant location and operation in riverine, estuarine, and coastal areas on commercial and sport fisheries (Lawler 1972; Hess et al. 1975; Van Winkle 1977; Spaulding and Isaji 1979). Narrowing the scope of an ecosystem model to address one important species, retains process- explicit capabilities (e.g., physical transport, migration, or feeding), increases validation potential through parameter identif lability (Gentil and Blake 1981), and therefore model credibility, while reducing time, expense, and data demands in development. In short, a single species modeling approach, augmented by a judicious choice of processes for inclusion in the model, represents a utilitarian compromise between the two other alternatives discussed above, and provides a useful methodology for estimating environmental impacts and investigating management alternatives. The single species modeling approach has been used to estimate the impact of oil spills on the Georges Bank cod fishery (Reed 1980; Reed et al. 1981). This impact assessment methodology is currently being extended to additional species (Spaulding, Saila et al. 1981; 1982). A summary of some recent results from this work follows. OVERVIEW OF OIL SPILL FISHERY IMPACT ASSESSMENT MODEL SYSTEM The oil spill fishery impact assessment model system addresses first order direct impacts of oil on a commercial fishery through hydrocarbon-induced egg and larval mortality. The model system components follow the approach given by Reed (1980), Reed and Spaulding (1978, 1979), and Reed et al . 1981 for the hydrodynamic, ichthyoplankton transport and fates, and oil spill fates components, and the approach given by Lorda and Saila (1980) and Lorda et al. (1982) for the fish population dynamics component. The simulated processes within an relationships among these four models are shown in figure 1. In operation, the ichthyoplankton transport and fates model, using a toxicity threshold assumption, output from the oil fates model on the distribu- tion of spilled oil, and a definition of the spatial and temporal spawning patterns of the species of interest, estimates the oil-induced mortality of eggs and larvae. Surviving eggs and larvae undergo transport and mortality as usual, except that density-independent mortality is increased in proportion to the number lost due to the oil spill. An impact analysis is achieved by running the fishery model to determine the perturbations caused by the spill event on the baseline equilibrium catch. 105 < >■ H t- < Q > S Ul X to ENT A ITY TION ORTALI O _i< 2 2 < J Ul K? , O u. ftCC* uj <* uiqO a) _iO ITY NT M H. « P TERS TERN PORA INOU -1 < a zo;3 u. Z < to Z / CO (0 CU O O CD J3 C to CO o c o to u o CD x. 60 C •r-l o X • CO CN t— I •H TD CO C 4-1 CO CU CO cu co u CD CU CO 3 !l J-i O C/3 <4-l CU 3 00 PK 109 N SUMMARY OF RESULTS Oil Spill Fates Model The oil spill fates model was used to simulate the 12 monthly spill cases noted in table 1. Figures 3 and 4 show the trajectories of the surface slick and subsurface hydrocarbon concentrations with time, noted in 5 day increments from the start of the spill event, for the May and December spills respectively. The other 10 monthly cases are documented in Spaulding, Saila et al . (1982). The centers of mass of the surface slick and the 50 part per billion (ppb) subsurface oil concentration contour are used to define the trajectories. An outline of the coast and the 100 m and 1,000 m bathymetric contour lines have been included to assist the reader in orienting the spill location to both land and important shelf and basin bathymetry. Noted at the bottom of each figure is detailed information on key spill parameters. An analysis of the 12 simulations shows differences in the response of the surface and subsurface oil to the combined effects of the wind-induced and long- term residual flow patterns drawn from a summary of drifter studies (Bumpus and Lauzier, 1965). The surface spillets readily respond to wind forcing, while the subsurface oil is more markedly influenced by the long-term advective field. In overview, the subsurface oil trajectories are toward the southwest and west in the fall and winter months, and become strongly influenced by the Georges Bank gyre in the spring and summer. The trajectories of the surface spillets are more convoluted than the subsurface trajectories because of their strong response to the passage of weather events. Generally, surface spillets on Georges Bank are transported to the southwest or southeast during the fall and winter and toward the northeast from spring through early summer, following the seasonal mean wind conditions. Stronger winter winds carry the surface oil away from the spill site more rapidly than do the lower velocity late spring and summer winds. It is interesting to note that when the wind speeds become more moderate, such as in August, the surface spillet trajectories are affected by the transient wind and residual flow in about equal magnitude. This observation suggests that knowledge of the offshore residual current patterns is critical in determining the trajectory of spilled oil in moderate weather, and indicates the need for a proper hydrodynamic modeling effort to produce improved impact estimates. Table 2 presents the final mass balance in percent for each monthly spill simulation, with the environment partitioned into atmosphere, sea surface, and subsurface (water column) components. The first two columns show the mass and volume of oil spilled. The last column, labeled "outside domain", indicates that this percentage of the oil has left the study area by exiting through one of the model boundaries. A review of the monthly oil spill final mass balances shows that 36.5 +0.5 percent is evaporated, 6+3 percent is in the water column and the remaining 57+3 percent of the spilled oil is found on the sea surface. Based on previous studies of mass balance for other oil spills (Spaulding, Saila, et al . 1981) it is clear that the major structure of the final mass balance relationships is determined by the oil type, with environmental factors such as wind and temperature exhibiting only secondary roles. 110 39- 38' £ t SPILL LOCATION D i SURFACE TRACK 15 DAT INTERVALS! A t SUBSURFACE TRACK IS OAT INTERVALS) 50 100 KM r46 45 44 43 -42 -41 40 -39 T 38 72 71 70 69 68 67 66 65 64 MOVEMENT OF SPILLED OIL FOR 90 OATS AFTER THE RELEASE SPILL PARAMETERS OIL- STATFJORO NORMAT CRUDE SITE- 68 DEG 12 NIN M. 40 DEC 37 HIN N TTPE- MELL BLOMOUT AMOUNT- 68 MILLION CAL OVER SO OATS START- MAT . JULIAN OAT 121 Figure 3. Trajectories of the surface and subsurface hydrocarbon centers of mass for the May spill. Ill 71 70 69 68 67 66 65 64 63 45- N -Vk-*^ MAINE .^'ifW^r^-^ Jf»™ i'^^l 44- jV.SCOTIA :*Jr J^J . Iff >\\jf rf 43- n.h. -iy /d t> 42- 41- 40- 39- * t STILL LOCATION • SURFACE TMCK IS MT INTERVALS! A • SUBSURFACE TfWCK IS OUT INTERVALS) 38- 1 1 1 1 d 5b ido km 1 II 1* 46 45 -44 43 42 41 40 39 38 72 71 70 69 68 67 66 65 64 MOVEMENT OF SPILLED SIL FOR 70 OflTS AFTER THE RELEASE SPILL PARAMETERS OIL- 3TATFJ0R0 NORNAT CRUDE SITE- 60 0EG 12 NIN N. 40 DEC 37 MIN N TTPE- HELL BL0HOUT AMOUNT- 68 MILLION GAL OVER 30 OATS 3TART- OECEMBER. JULIAN OAT 335 Figure 4. Trajectories of the surface and subsurface hydrocarbon centers of mass for the December spill. 112 1 a Id u o. to •« *P * » ^ » < < Q ■ < H Z id Z z O DC > Z u * *r* l" ao co C o s -1 < u o z i •u c 8 at u o (0 OJ o c CO I-l CO .£> CO CD I a CO CO c CM (U t— I .£> CO H u u z < ■ < m CO to < z o a u Q h D >> U X H Id Q » ^ *m m irt un ♦ Nir in ^ « N-fic » r- | «•" o m ** *n pa 'O r* «0 *© *o - * *> ,~ ui ir. >esS ZCQ* *E3 as>-Z o.<5 < in as u w uaa O Id 3 H> o UOid O Z Q a3itfiM SMiajgagwrtns — ttvi Id a 3 OS u >■ u O w a* oc . °* (0 -o en z B u b. I a H z I a *- ft s ° I 2 u.< Z Q OS H J Q 5<«iu - nO>-HQ J, a S Si Ho Q W z < H < H to -J J a 10 < a z < 3 3 z to to to Id s < J u. J td J T a. H co « - O O u. u X U Z Ha td > td CD * a Id a 8 m > z ae a. co IT as < O a. oc 10 td Id H u. < O 2 a H z (0 _) Id O Id a. uo z H < -1 Ld < _> 03 a CO < — 1 Z Id H CO Id z X3 H to 113 A comparison of the amount of oil in the water column with the mean wind speed over the first 30 days of the simulation period, for each spill event, shows a strong correlation. The amount of oil in the water column under a spillet increases approximately as the square of the mean wind speed (Audunson el ai . 1979). The windier winter months show double the oil mass in the water column compared to the quiet summer months. It is also seen that the source for this additional subsurface oil is the sea surface. Water temperature has no effect on the oil mass entrainment as presently formulated, and therefore no correlation between subsurface oil and water or air temperature is noted. The slight variations in the amount of oil found in the atmosphere represent the combined effects of temperature and wind with higher temperatures and faster winds giving increased evaporative losses. The summer spills appear to be dominated by the temperature effect, while the winter spill mass balances are largely governed by the wind. Fishery model Figures 5 and 6 show the temporal distribution of spawning activity as simulated for herring and cod, and the reductions of incoming year classes due to each of the 12 spills investigated. These figures demonstrate the critical importance of spill timing relative to the temporal distributions of spawning. The results shown are based on egg and larval oil-induced mortalities estimated with the ichthyoplankton transport and fate model. The ensuing year class reductions were then incorporated into the adult fish population model, a non-linear, non-spatial matrix formulation (Lorda and Saila 1980), to project variations in catch. It is clear also that the size of the spawning grounds and their locations relative to the spill origin (fig. 7 and 8) play a major role in determining the extent of the cohort reduction in each case. The two cod spawning grounds (fig. 7) are well defined, and the spill site is located on the edge of the largest one (on Georges Bank). The result is that the largest cod cohort reduction (77.5%) occurs when the spill starting time (day 60) matches the peak of the combined Georges Bank and Nantucket Shoals spawning. This indicates a concentration of the oil-induced mortality in the earlier life stages (eggs and yolk-sac) before significant larval dispersion occurs. Unlike cod, the herring spawn over a large area of Georges Bank (fig. 8). This initially higher dispersion of the herring eggs results in consistently lower oil-induced mortalities. The largest oil impact for the herring (17.6% cohort reduction) does not match its spawning peak as in the case of cod, but occurs about 45 days later. Although a mismatch of 12 days can be explained by the initially demersal herring eggs (yolk-sac is the first pelagic stage), the remaining 30 days of difference between the spawning peak and the largest oil-induced mortality seem to indicate that the spreading of the spill must proceed for that number of days before the largest possible number of herring larvae can be oiled at a lethal concentration level. The effective reductions of the initial cohort sizes caused by the oil were implemented in the fishery model by specifying equivalent one-time reductions in the probability of survival through year-0. The effects of this reduced cohort survi al on the expected catches over the 50 years following the occurrence of the oil spill are summarized in tables 3 and 4 for the herring and cod fisheries, 114 HERRING STOCK IGE0RGE3 BANK) - TEHPOARL SPANNING DISTRIBUTION TIME (JULIAN DATS) HERRING PISHERT - IMPACT OP NORTH BLONOUT (TIMING EFFECT) *^>. 00 SO. 00 100.00 ISO. 00 200.00 2S0.00 300.00 350.00 SIMULATED SPILL STARTING DATE (JULIAN DATS) Figure 5. Herring population. Temporal spawning distribution and oil-induced mortalities in Year-0 class for the twelve monthly blowout spills. 115 COO STOCK I6E0AGES BANK) - TEMPORAL SPANNING OISTMIBUTION a t. 00 C«org«t Bonk SO. 00 100.00 150.00 200.00 250.00 300.00 350.00 TIME (JULIAN DATS COO FI3HEP.T - IMPACT OF N0P.TM BLONOUT (TIMING EFFECT) %. 00 SO. 00 100.00 ISO. 00 200.00 250.00 300.00 850. 00 SIMULATED SPILL STARTING DATE (JULIAN DATS) Figure 6. Cod population. Temporal spawning distribution and oil- induced mortalities in Year-0 for twelve monthly blowout spills. 116 39- 38 I *SPILL SITE 72 — T- 71 — r- 70 T 69 68 SCALE I 1 1 50 100 KM l I 67 66 65 -39 38 64 Figure 7. Spawning locations for Atlantic cod, 117 39 71 70 69 68 67 66 ■ ■ ■ i ■ * SPILL SITE 46 38-J 72 71 70 69 SCRLE I 1 1 50 100 KM 68 T -45 -44 43 -42 41 40 39 r*-38 67 66 65 64 Figure 8. Spawning locations for Atlantic herring. 118 respectively. These tables show the values of three selected model output variables for each of the 12 spills simulated. The interpretation of these output variables, largest annual catch reduction A, largest cumulative reduction B, and final long-term loss C, should be clear from the headings, and from figures 9 and 10. Variable C, the sum of the catch losses, or ultimate loss to the fishery for each spill, is discussed below. The January (Julian day 1) oil spill for herring and the March (Julian day 60) spill for cod are used for this discussion since they resulted in the largest percent cohort reductions in each of the two species. The behavior of the three fishery model variables for these two cases over the 50 years following the spill is shown in figure 9 for the herring and figure 10 for the cod. It is clear from these plots that the cod compensatory mechanism over compensates for the added oil-induced mortality, while the herring population, with a weaker compensatory mechanism, is only able to return to its initial equilibrium size. Although A, the largest one-year catch loss, would be similar for comparable oil-induced mortalities in both species, B and C, the largest cumulative catch loss over 50 years and the ultimate loss, respectively, are quite different. In the case of the herring, both B and C are virtually the same variable because the largest cumulative loss always corresponds to the last year in the series of catch projections. For the cod fishery, B is maximum in the 7th year after the spill, decreasing thereafter and converging to an ultimate loss C, which is less than half of the maximum value of B (fig. 10). It should be noted that C, the ultimate loss to the fishery, of roughly 21 percent of a single year's equilibrium catch in the case of cod for the March spill, cannot be shown in figure 10 except as the value to which B ultimately converges. The difference in the dynamics of the impacted herring and cod fisheries results from the different formulations of compensatory mortality used to model these two species. It is interesting to note that for the most damaging time of the blowout, the ultimate catch loss for both fisheries is roughly 21 percent of the one-year equilibrium catch, despite the fact that the actual oil-induced ichthyoplankton mortality for the herring was only 17.6 percent versus 77.5 percent for the cod. MODEL SYSTEM SENSITIVITY STUDIES When applying a system of models as complex as that outlined here, the interactions between the submodels and the sensitivity of model system predictions to various assumptions and parameterizations are critical to determining model validity, and in understanding what processes are most important in controlling system behavior. Sensitivity studies to percent loss of a year class and the effect of the Georges Bank gyre on impact predictions are discussed below. These two topics have been selected for presentation since they are the key to answering some common questions related to the impact of oil spills on commercial fisheries for the Georges Bank Region: What if a spill eliminates an entire year class? Does the Georges Bank gyre trap pollutants and cause increased impacts? Additional sensitivity studies on toxicity threshold levels, compensatory mortality regimes, and all major fishery model parameters are documented in Reed et al. (1979, 1981), and Spaulding, Saila et al. (1981, 1982) and are currently an area of active investigation. 119 Table 3. Herring fishery (Georges Bank): North Blowout spill impact simulations. Summary of simulation runs for different spill starting dates. MODEL Output Variables: (A) (B) (C) CUMULATIVE FINAL SUM JULIAN Percent TIME CATCH % LOSSES CATCH % 1 LOSSES CATCH f DATE OIL-M0RT STEPS Largest YR. Largest YR. LOSSES 0001 17.61 50 3.99 4 th 21.30 50 th 21.209 0032 6.69 50 1.51 4 th 8.07 50th 8.063 0060 3.34 50 0.76 4th 4.03 50th 4.023 0091 0.14 50 0.03 4 th 0.17 50 th 0.1b7 0121 0.03 50 0.01 4 th 0.03 50 th 0.034 0152 0.00 — 0.00 ... 0.00 ..— 0.000 0182 0.00 — 0.00 ... 0.00 .... 0.000 0213 0.02 50 0.01 4th 0.03 50th 0.027 0244 0.04 50 0.01 4 th 0.05 50 th 0.048 0274 0.07 50 0.02 4 th 0.09 50 th 0.067 0305 6.40 50 1.45 4th 7.73 50th 7.721 0335 12.33 50 2.79 4 th 14.89 50th 14.884 Table 4. Cod fishery (Georges Bank): North Blowout spill impact simulations. Summary of simulation runs for different spill starting dates. MODEL Output Variables: (A) (B) (C) CUMULATIVE FINAL SUM JULIAN Percent TIME CATCH f LOSSES CATCH % 1 .OSSES CATCH % DATE 0IL-M0RT STEPS Largest IR. Largest YR. LOSSES 0001 3.9* 50 0.68 4th 2.82 7 th 1.342 0032 34.50 50 5.99 4th 24.80 7 th 10.744 0060 77.52 50 13.46 4th 55.68 7th 20.733 0091 54.62 50 9.49 4 th 39.25 7 th 15.889 0121 37.09 50 6.41 4 th 26.67 7th 11.454 0152 14.24 50 2.47 4th 10.24 7 th 4.725 0182 1.10 50 0.19 4 th 0.79 7 th 0.380 0213 0.00 — 0.00 0.00 0.000 0244 0.00 — 0.00 — 0.00 — 0.000 0274 0.00 — 0.00 0.00 0.000 0305 0.00 — 0.00 0.00 0.000 0335 0.11 50 0.02 4th 0.00 7th 0.037 120 HERRING FISHERY (Georges Bonk) - CATCH PROJECTIONS AFTER NORTH BLOWOUT (Jon ) 20. IS. M. TIME IN YEARS ss. Figure 9. Impacted herring fishery. Catch projections after the January (Day 1) North Blowout spill simulation. COO FISHERY (Georges Bonk) - CATCH PROJECTIONS AFTER NORTH BLOWOUT (March) 10 is *0. 2S. JO TIME IN YEARS ». so. Figure 10. Impacted cod fishery. Catch projections after the March (Day 60) North Blowout spill simulation. 121 Sensitivity to percent loss of a year class The sensitivity of the two selected species to percent loss of a year class has been investigated by projecting the age-specific density vector of each equilibrium population for 50 years after simulating 10 percent incremental reductions in the probability of survival of the first cohort through year-0. The results of these simulations for the herring and cod fisheries are shown in tables 5 and 6, respectively. The interest of these simulations comes from the fact that within the limitations of the fishery model itself they cover the full range of impacts resulting from any oil spill event that might occur on Georges Bank regardless of size, type, timing, and duration, as long as two or more successive cohorts of the same species are not impacted. It is interesting to note that even the elimination of an entire year class (i.e., 100% cohort reduction) results in rather moderate losses to the fishery in the case of cod. On the other hand, the loss of an entire year class in the herring population results in ultimate losses to the fishery equivalent to more than a one-year equilibrium catch (table 5). These differences in sensitivities of the cod and herring are the result of the different population structures and fecundities of each species, together with the different compensatory formulations used to model them. The strongest compensation occurs in the cod, which typically follows a dome-shaped (Ricker 1958) form of stock and recruitment relationship. The herring compensatory ratio is about 25 percent less than for the cod. In addition, the herring com- pensatory mortality follows the Beverton and Holt (1954) form of stock recruit relationship, which weakens even further the ability of this species to compen- sate for year class reductions. The net result is that the herring stock demonstrates less dynamic response to large cohort reductions than the cod stock. Finally, the largest one-year catch loss (model output variable A), for the two species can be understood by examining the age structure of the total catch. It is clear that the value of A cannot be larger than the largest percent contribution of a single age-class to the total catch. For instance, A is 17.4 percent for the cod for the loss of an entire year class, which is the contribu- tion of the four year olds to the total catch. Thus, a fishery based on a long- lived species such as cod, in which the catch is typically composed of many year classes, is much less susceptible to large catch losses from single event, acute oil spills. A fishery based on the current herring stocks on the other hand, will tend to rely upon only one or two age groups to supply most of the catch, so that recruitment fluctuations caused by oil spills (or variations in natural mortality) will be more strongly reflected in the catch on a percentage bas is . Effect of Georges Bank gyre on impact predictions An issue which has consistently been raised at the public hearings on the lease sales of tracts off the New England coast concerns the possible formation of an anticyclonic gyre on Georges Bank. There is a variety of oceanographic data supportive of such a gyre during the spring and early summer (Bigelow 1927; Bumpus and Lauzier 1965; EG&G 1979). The existence of such a feature makes sense biologically in that ichthyoplankton would be retained in the Bank's 122 Table 5. Herring fishery (Georges Bank) : Sensitivity to percent loss of a year-class. Summary of simulation runs with 10 percent incremental losses. MODEL Output Variables: (A) (B) (O CUMULATIVE FINAL SUM SIMUL Percent TIME CATCH f LOSSES CATCH % LOSSES CATCH % No. 0IL-M0RT STEPS Largest YR. Largest YR. LOSSES 1 10. OU 50 2.2b 4 th 12.08 50th 12.0b7 2 20.00 50 4.53 4th 24.21 50 th 24.194 3 30.00 50 6.79 4th 36.41 50 th 36.382 4 U0. 00 50 9.0b 4th 48.67 50 th 48.636 5 50.00 50 11.32 Hth 61.00 50th 60.959 6 60.00 50 13.59 nth 73.41 50 th 73-356 7 70.00 50 15.85 nth 85.89 50 th 85.830 8 80.00 50 18.11 4th 98.46 50th 98.387 9 90.00 50 20.38 4 th 111.12 50th 111.031 10 100.00 50 22.64 4 th 123.86 50 th 123.768 Table 6. Cod fishery (Georges Bank): Sensitivity to percent loss of a year-class. Summary of simulation runs with 10 percent incremental losses. MODEL Output Variables: (A) (B) (C) CUMULATIVE FINAL SUM SIMUL Percent TIME CATCH I ] .OSSES CATCH t LOSSES CATCH % No. 0IL-M0RT STEPS Largest YR. Largest YR. LOSSES 1 10.00 50 1.74 4th 7.19 7 th 3-361 2 20.00 50 3.47 4th 14.39 7 th 6.522 3 30.00 50 5.21 4th 21.57 7th 9.480 4 40.00 50 6.95 4th 28.76 7 th 12.234 5 50.00 50 8.68 4th 35.94 7 th 14.782 6 60.00 50 10.42 4th ■43.11 7 th 17.124 7 70.00 50 12.1b 4th 50.29 7 th 19-261 8 80.00 50 13.89 4 th 57.46 7 th 21.193 9 90.00 50 15.63 4th 64.63 7 th 22.925 10 100.00 50 17.37 4th 71.79 7 th 24.4b1 123 plankton-rich water, explaining the intense productivity of the area (EG&G 1981). The problem is that an oil spill might be similarly retained, and its impact on the ichthyoplankton thereby increased. The oil spill - fishery impact assessment model system has been used to investigate this question for winter (Julian day 32) and spring (Julian day 121) spills. The residual advective field underlying these two test simulations is the summer data set shown in figure 11 derived from charts compiled by Bumpus and Lauzier (1965). This residual current field, which was held constant through the two simulations, was selected because it showed the strongest gyre-like configuration over the banks. Random walk diffusion and the definition of the wind driven hydrodynamics remain as in prior simulations. The impacts on cod of the 68 million gallon north blowout spill scenario under this altered current pattern are summarized in table 7. Not surprisingly, the replacement of the winter residual current field, which is essentially unidirectional to the south and southwest, with the strong summer gyre results in greatly increased impacts. The impacts for spills on days 32 and 121 are nearly the same, subject to this constant residual advective field. This is due firstly to the fact that dispersion of oil into the water column is relatively complete after a few days. The effect of wind on the subsurface distribution is therefore limited primarily to this initial period. Secondly, the two spills occur on either side of the cod spawning peak near Julian day 90 (fig. 6). Because the day 32 subsurface oil remains in the general vicinity, it affects the eggs spawned during this peak time. Conversely, in the case of an oil spill occurring on day 121, the eggs stay in the area longer, and therefore remain subject to the toxic action of hydrocarbons released after the time of maximum spawning activity. These observations simply reflect the expected tendency of a gyre-like formation to retain transported constituents. Although this model experiment is relatively artificial, it demonstrates the importance of the inclusion of subsurface representations in oil spill impact prediction work, and strongly underscores the importance of determining the correct structure of the advective field for correct estimation of impact magnitudes. CONCLUSIONS It is clear from the above discussion that spill timing, spatial and temporal spawning distributions, and population dynamics of the species of concern are critical factors in determining the impact of spill events on the Georges Bank herring and cod fisheries. In the case of cod, compensatory mechanisms within the fishery, combined with the existence of several year classes of mature fish, will tend to attenuate or reduce large losses in one year class caused by the proximinity of spawning site to spill location. For herring, diffuse spawning locations protect the population from localized oil spill pollution events, although internal population dynamics result in relatively large catch losses in the long term due to small numbers of year classes. Sensitivity studies on percent loss of a year class further display the importance of compensation in determining the impacts of a pollutant event on a fishery. A fishery such as cod, with relatively strong compensatory behavior, 124 39' -44 -43 38+ -42 -41 -40 -39 i i 50 100 KM i i r 38 72 71 70 69 68 67 66 65 64 Figure 11. Current field used for gyre transport investigations, 125 Table 7. Effect of gyre on impact predictions for Georges Bank cod eggs and larvae. Entries are percent of one year's ichthyoplankton oiled. SPILL DAY 32 (WINTER) 121 (SPRING) RESIDUAL ADVECTIVE FIELD NORMAL DYNAMICS 34. 57. 36. 97. GYRE 60. 17. 60. 57. 126 can undergo a single pollutant event with relatively small losses in projected catch. Herring, on the other hand, with its weak compensation shows a much greater sensitivity to losses of new recruits. The presence and duration of the Georges Bank circulation gyre, as represented by the simple parameterization used here, are important factors in determining impact magnitudes. Significantly higher egg and larval mortality are observed if the gyre is present, because this circulation feature increases the exposure time of spawned products to toxic levels of the pollutant. This simple simulation demonstrates that the circulation features of this area must be carefully considered in order to make realistic assessments of impacts. While significant progress has been made through this modeling approach in understanding the impact of oil spills on a commercial fishery, we have only taken the first step. The most important lesson of the research to date is that realistic impact assessment procedures need to take an integrated view of the environment. The interrelationships among the components of the physical and biological systems under study must be correctly represented within the model to provide a sound basis for rational resource management decisions. As with all productive modeling studies the present work has helped focus on the changes necessary to improve the credibility and accuracy of the model system. Efforts are currently in progress to further validate and upgrade each component of the impact assessment methodology. ACKNOWLEDGMENTS This work was funded by the United States Department of the Interior, Minerals Management Service (MMS) under contract AA851-CTO-75, with Dr. William Lang of the MMS New York Outer Continental Shelf Office serving as the technical contract monitor. To complete a modeling project as large and comprehensive as that outlined here has required the combined talents of a large integrated multi-disciplinary team. Key team members and their areas of contribution are as follows: M.L. Spaulding, Department of Ocean Engineering, and S.B. Saila, Graduate School of Oceanography, University of Rhode Island - principle investigators; E. Lorda, H. Walker and V. Pigoga, University of Rhode Island, Graduate School of Oceanography - fishery modeling; C. Swanson and T. Isaji, Applied Science Associates, Inc. - hydrodynamic modeling; M. Reed, E. Anderson, Applied Science Associates, Inc. - oil spill fates and ichthyoplankton transport modeling. The typing of this paper in its numerous versions was performed with admirable good cheer by Ms. Teri Highling of Applied Science Associates, Inc. 127 REFERENCES Anderson, E. and M.L. Spaulding, 1982. "Application of an Oil Spill Fates Model to Environmental Management on Georges Bank," Env. Prof . 3:119-132. Anderson, K.P. and E. Ursin, 1977. "A Multi-species Extension of the Beverton and Holt Theory of Fishing, with Accounts of Phosphorous Circulation and Primary Production," Medd. Denmarks Fiskeri og Havundersogelser , Vol. 7: 319-435. Audunson, T. , P. Steinbakke , and F. Krogh, 1979. "Fate of Oil Spills on the Norweigian Continental Shelf," Presented at the 1979 Oilspill Conference, Los Angeles, Calif., unpublished memo. Balchen, J.G. , 1980. "Modeling and Identification of Marine Ecological Systems with Application in Management of Fish Resources and Planning of Fisheries Operations," Norwegian Journal of Modeling, Identification, and Control 1(2): 67-68 Beverton, R.E. , and S.J. Holt, 1954. "Notes on the Use of Theoretical Models in the Study of the Dynamics of Exploited Fish Populations," Misc. Contrib. #2, U.S. Fish Laboratory, Beaufort, N.C., 181 p. Bigelow, H.B., 1927. "Physical Oceanography of the Gulf of Maine," Bull. U.S. Bureau of Fish ., 40(2): 511-1027. Bumpus , D.F. and L.M. Lauzier, 1965. "Serial Atlas of the Continental Shelf Off Eastern North America between Newfoundland and Florida." American Geographical Society Serial Atlas , Folio 7. Cantor, L., 1977. Methods for Environmental Impact Assessment , University of Oklahoma Press. Chen, C.W., 1975. "A Discussion of CLEAN, The Aquatic Model of the Eastern Deciduous Forest Biome," p. 83-102 in Russell, C.S. (ed) Ecological Modeling , Resources for the Future, Washington, D.C., 394 p. Danenberger, E.P. , 1977. "Oil Spills 1971-75, Gulf of Mexico, Outer Continental Shelf," U.S. Geological Survey Circular #741, 47 p. Devanney III, J.W. and R.J. Stewart, 1974. "Analysis of Oil Spill Statistics," In: Primary Physical Impacts of Offshore Petroleum Developments, Report No. MITSG 74-20, Massachusetts Institute of Technology, Cambridge, Mass. 126 p. EG&G, 1979. "Analysis Report," Appendix F, 10th Quarterly Progress Report , New England Outer Continental Shelf Physical Oceanography Program, EG&G Environmental Consultants, Waltham, Mass. Gentil, S. and G. Blake, 1981. "Validation of Complex Ecosystem Models," Ecological Modeling (14): 21-38. Hess, K.W., M.P. Sissenwine, and S.B. Saila, 1975. "Simulating the Impact of the Entrainment of Winter Flounder Larvae," p. 1-30 in Saila, S.B. (ed.) Fisheries and Energy Production, A Symposium , Lexington Book, Mass., 300 p. Kelly, R.A., 1975. "The Delaware Estuary Model," p. 103-134 in Russell, C.S. (ed.) Ecological Modeling , Resources for the Future, Washington, D.C., 394 p. Laevastu, T. and H.A. Larkins , 1981. Marine Fisheries Ecosystem, Its Quanti- tative Evaluation and Management , Fishing News Books Ltd., Surrey, England. Lawler, J. P., 1972. Testimony of October 2, 1972, on Indian Point 2 Licensing Hearings, before U.S. Federal Power Commission, Albany, N.Y. 128 Lorda, E. and S.B. Saila, 1980. "A Stochastic Nonlinear Leslie Matrix Model for Simulating the Effect of Increased, Age-specific, Density-independent Mortality of Fish Populations," Paper presented at the Symposium on Age- Structured Models in Fishery Science, 110th Annual Meeting of the Amer. Fisheries Soc. , Sept. 22-24, 1980. Louisville, Ky. Lorda, E. , S.B. Saila, and H.A. Walker, 1982. "The Stochastic Modeling of the Dynamics of Fish Populations Using Nonlinear Leslie Matrices: Methodology, Probabilistic Analysis, and Application to the Georges Bank Cod Fishery," (in preparation). Moore, S.F., R.L. Dwyer, and A.M. Katz, 1973. "A Preliminary Assessment of the Environmental Vulnerability of Machias Bay, Maine to Oil Supertankers," T.R. No. 162, Ralph M. Parson Laboratory for Water Resources and Hydro- dynamics, MIT, Cambridge, Mass. 162 p. Pielou, E.C., 1977. Mathematical Ecology , Wiley International Pub., New York, N.Y. Reed, M. , 1980. "An Oil Spill Fishery Interaction Model — Development and Applications." Ph.D. Thesis, Department of Ocean Engineering, University of Rhode Island, Kingston, Rhode Island. Reed, M. and J.G. Balchen, 1982. "A Continuum Model of Fish Population Dynamics and Behavior," Norwegian Journal of Modeling, Identification, and Control , April 1982 (in press). and M.L. Spaulding, 1978. "An Oil Spill-Fishery Interaction Model." Part X in Environmental Assessment of Treated Versus Untreated Oil Spills: Second Interim Progress Report , U.S.D.O.E. Contract No. E(ll-l) 4047. and M.L. Spaulding, 1979. "An Oil Spill-Fishery Interaction Model: Comparison of Trested and Untreated Spill Impacts." Proc. of 1979 Oil Spill Conference, EPA-API, L.A. , p. 63-73. , M.L. Spaulding, and P. Cornillon, 1981. "A Fishery-Oil Spill Interaction Model: Simulated Consequences of a Blowout," Proc. of NATO Symposium on Applied Operations Research in Fishing , Plenum Press, New York, p. 99-114. Reed, M. , M.L. Spaulding, E. Lorda, H.A. Walker, 1982. "An Oil Spill Fishery Interaction Model: The Fisheries Recruitment Problem." Paper presented at the Technical Conference on Hydrocarbon Exploration and Development on Georges Bank, April 27-30, 1982, Nantucket, Ma. Ricker, W.E. , 1958. "Handbook of Computations for Biological Statistics of Fish Populations," Bull. Fish. Res. Bd. Can. #119, 300 p. Spaulding, M.L. and T. Isaji, 1979. "Numerical Modeling of Entrainment and Far Field Thermal Dispersion for NEP 1 and 2," ORNL/TM- 7590, Oak Ridge National Laboratory, Charlestown, Rhode Island. Spaulding, M.L. , S.B. Saila, M. Reed, J.C. Swanson, T. Isaji, E. Anderson, E. Lorda, V. Pigoga, K. Marti, J. Hoenig, H. Walker, C. Jones, F. White, R. Glazman, and K. Jayko , 1981. "Assessing the Impact of Oil Spills on Commercial Fishery." First Interim Report , prepared for the Bureau of Land Management NYOCS Office under Contract No. AA851-CTO-75. , 1982. "Assessing the Impact of Oil Spills on a Commercial Fishery," Final Report , prepared for the Bureau of Land Management NYOCS Office under Contract No. AA851-CTO-75, November 1982. , K.B. Jayko, and E.L. Anderson, 1982. "Hindcast of the ARGO MERCHANT Spill Using the URI Oil Spill Fates Model," Ocean Engineering (in press) Van Winkle, W. , 1977. Proceedings of the Conference on Assessing the Effects of Power Plant-Induced Mortality on Fish Populations , Pergamon Press, N.Y., 379 p. 129 A SIMULATION MODELING FRAMEWORK FOR ECOLOGICAL RESEARCH IN COMPLEX SYSTEMS: THE CASE OF SUBMERGED VEGETATION IN UPPER CHESAPEAKE BAY 1 W. Michael Kemp 2 Walter R. Boynton 3 Albert J. Hermann 2 ,^ ^University of Maryland Center for Environmental and Estuarine Studies Contribution No. HPEL-83-1429 2 Horn Point Environmental Laboratories, P.O. Box 775, Cambridge, MD 21613 3 Chesapeake Biological Laboratory, Box 38, Solomons, MD 20688 ^Department of Oceanography, University of Washington, Seattle, WA 131 ABSTRACT This paper provides a broad overview of a system of simulation models developed in conjunction with empirical research programs to study the de- clining abundance of submerged aquatic vegetation (SAV) in Chesapeake Bay and the effects of this decline on ecological and socio-economic processes. A hierarchical organization of models and submodels allowed the simplification needed for tractability while maintaining sufficient detail for examining mechanisms of ecological interaction. Two models of the SAV ecological subsystem are presented. First, the Autotroph Model was used to investigate consequences of shifting competition for light and nutrients among 4 groups of primary producers (SAV, phytoplankton, epiphytes and benthic microalgae). This model, which has been calibrated and verified against independent data sets, was used to extrapolate from controlled experiments to consider effects of nutrient enrichment. The second of these, the Nekton Model, was developed to test possible effects of declining SAV on the trophic structure and relative abundance of 3 fish groups. The model's design utilizes certain elements of traditional fish population models within the generic structure of an ecosystem model. An SAV resource management model was developed by aggregating the details of these and other ecological submodels and is linked to a suite of simulation models which relate human activities to estuarine processes and societal values in the Bay region. We argue here that this management modeling framework allows results of scientific research to be integrated into political and socio-economic networks toward balanced uses of those estuarine resources related to SAV. INTRODUCTION Estuaries such as Chesapeake Bay are complex and dynamic ecological systems which benefit human societies in many ways. On one hand, these coastal ecosystems provide a bountiful source of fisheries production and diverse recreational ooportunities . On the other hand, natural biogeochemical processes witl#n these systems are capable of transforming many wastes emanat- ing from human activities into useful components of regional and global cycles. In some cases low levels of waste inputs (such as nutrients and organics) can, in fact, enhance estuarine productivity. However, in many of these environ- ments waste loading rates are such that they detract significantly from the estuary's value as a source of fisheries and recreation. Hence, a serious problem evolves wherein legitimate but competitive uses of the natural resource are in direct conflict with one another. In the last two decades, Chesapeake Bay has undergone some documented changes. One such change has been the drastic decline of submerged aquatic vegetation (SAV) which once dominated littoral zones throughout the estuary. Coincident with this loss of aquatic plants, there have been significant changes in water quality (including increased levels of turbidity, nutrients and agricultural herbicides), as well as declines and shifts as in various fisheries (Boynton et al. , 1979). Stevenson and Confer (1978) postulated that many of these alterations in water quality are attributable to increased waste loadings to the Bay from both sewage outfalls and diffuse sources, and that such deteri- orating conditions have led to the loss of ecosystems associated with these submerged plants. Moreover, it was hypothesized that the decline of SAV has contributed to detrimental changes in fisheries production (Boynton et al . , 1979). 133 Hence, this loss of SAV communities represents a convenient case study for examining the emerging problem of managing and balancing the conflicting uses of estuarine resources. The purpose of this paper is to describe a simulation modeling framework which has served to organize, focus and elaborate a broad empirical research program for investigating this problem. RESEARCH ORGANIZATION AND DESIGN Peception of Problem In figure 1 we have illustrated our perception of the important inter- actions related to the decline of SAV, including (1) factors contributing to the decline, (2) ecological consequences of the decline, and (3) socio-economic ramifications of this ecosystem modification (Boynton et al., 1981). In this cartoon SAV are shown to act as natural nutrient sinks and sediment traps, both processes having economic analogs in terms of sewage treatment plants and channel dredging operations, respectively. Furthermore, SAV communities are suggested to be important sources of food and habitat promoting growth of fish, shellfish, and waterfowl stocks which are harvested in commercial and recreational endeavors. Various watershed activities are shown to influence estuarine water quality (nutrient, sediment, and herbicide additions) via direct discharges and runoff which are, In turn, regulated by rainfall and other factors. Throughout this cycle some economic enterprises (e.g., agriculture) may detrimentally influence SAV while others (e.g., fishing and dredging) are affected by plant losses. While this presentation may be useful as an over- view of the basic relationships involved in the problem, it does not indicate the nature of such relationships. Hence, we need a more explicit framework within which mechanistic connections are embodied. We recognized in this research project a rare opportunity to address several scientific hypotheses of theoretical and empirical interest within a broad context of resource management questions. However, to do so effectively it was necessary to use a scheme whereby the complexity of this problem could be dealt with in an organized, piece-wise simplified fashion. We, therefore, developed a hierarchical approach for the overall research program which enabled us to integrate highly controlled experiments (testing mechanistic hypotheses) together with descriptive field measurements (characterizing the structure and function of these SAV ecosystems). This allowed us to combine a spectrum of research methods and scales of interest into a unified effort. We have discussed the relative merits and philosophical underpinnings of this scheme elsewhere at length (Kemp et al. , 1980). A variety of conceptual and simulation models were utilized to integrate this research program. We reasoned that models could facilitate the coupling of experimental findings on "causality" (i.e., influence) with the inherently holistic perspective of descriptive in situ observations. Furthermore, simulation models could be used to confer generality upon specific results at either end of the controllability-realism spectrum (Kemp et al., 1980). This would be done by constructing, calibrating, and verifying models with data from a variety of systems. Thus, we concluded that simulation models could be used to examine the possibility that altered water quality conditions contri- buted to the decline of SAV in various regions of Chesapeake Bay. Such models would help to Interpolate and extrapolate the results of experimentally inferred 134 c (D co CD ■r-l > •H 4J U CO 3 to o c o •l-l 0) c o •H 4J CO ■u CI) 3 CJ r-J •r-l CO 4-1 > CO 3 CD cr o co u 3 T3 O cu co 00 CD M U CD g ' J XI CD 3 .G CO 4J o 00 C £1 O S •H O. 00 CD 3 •O O r-l cw co O C o o •.-I -H 4-1 4J CO O e s CD T3 x: o u vj co a. CO CO 3 O 4-1 -r-l a oo CD o O i-l c o o o CJ CD CD u 3 60 •■-t 135 relations for any combination of water quality factors observed (past or present) in nature. Simulation Modeling Structure We employed two distinctly different strategies for simulation modeling which were central to thee overall SAV research program. One strategy was directed primarily toward understanding the dynamic behavior of the seagrass ecosystem including energy flux, predator-prey interactions, nutrient cycling and trophic structure. As before (in the broad design of our research program), we utilized a hierarchical perception to decompose a detailed SAV ecosystem model into a cluster of subsystem models. This allowed us to maintain sufficient ecological detail against the limits of conceptual and computational tract- ability. The other approach in our modeling program emphasized the role of these plant communities in a larger context of the entire estuarine system in- cluding socio-economic considerations. Here, we developed an aggregated version of the SAV ecosystem model (i.e., combined submodels) and placed it into a sequence of cascading connections of influence, which lead from human uses of the estuary for waste disposal, through the SAV ecosystems, to human uses of the estuary as a source of fisheries harvest and other recreational activities. In this paper we describe the structure and the logic behind this dual modeling framework, and we provide a few selected results from these models to indicate briefly the breadth of research questions which were addressed. SAV ECOSYSTEM MODEL Ecosystem Modeling Framework The initial step in developing a simulation model of the SAV ecosystem involved identification of the level of aggregation and essential state variables. Obviously, there are certain misleading consequences of reduced dimensionality such as artifically conferred stability (e.g., Schaffer 1981). However, we have taken cognizance of population time-constants (Goodall 1974, Schaffer 1981), as well as life histories, trophic relations and habitats (Boling et al., 1975) in defining aggregated biological state variables. We have reduced the number of chemical variables (e.g., plant nutrients) by recognizing basic principles of chemical kinetics whereby biochemical rates are determined by a single rate- limiting step or substrate (e.g., Brezonik 1972). In all we defined 37 state variables to be included in this model. There are a few published examples of analytical or simulation models for seagrasses or other submerged macrophytes (Titus et al., 1975, Belyaev et al., 1977, Short 1980, Weber et al . , 1981, Verhagen and Nienhuis 1983, Adams et al. , 1979). However, all but one of these dealt with plant production only, and none contained more than 8 state variables. It was decided that this many (37) variables in one model would produce a virtually unmanagable system of equations, particularly given the necessary high degree of connectivity. A hierarchical scheme of six subsystem models was used to define the SAV ecosystem (fig. 2). Other modelers have similarly utilized hierarchical approaches (Goodall 1974, Overton 1975, Mclntire and Colby 1978), and various methods have been suggested for interconnecting subsystem models. We elected to simulate subsystem models independently and then to use outputs of each as 136 CO Ui H < OC i_ CO CO ■'■w •• Numbe Number UI H oc HI CO CO (0 CO E CO 1- 0) E z CO CO co > z CD: •co : " o m E o W HI :C; a a m o o _l m :C0 : ;i*-'. E E co ■a *- O 2 a. c £ i_ co o .ui CO CO O X CO :•'•'•• • • • • • >. T> O) C o CO o CO o 0) u *■» 2 co c OC CO 3 111 c co 1- 3»- < 2 a. § cE^ 1 "O c Sesto rabs; hrimp; ■■••■'-•■.■■ ■-':- c '.'C--.'-.:-. o co arikto s DIN c o CO <1> z O XL c CO CO O oco H Phytopl Aqueou CO Inorgan Larvae Z < _l O O Q a o o N D a. **&:•'! • • • • • >« CO CD w *- CD CO CO § CD 1 a r CO en CO a CD cu ! c '• '.••;•'. O) :•.:•:■.-. Zj CO x o ..-.* < t». :: ."-.3 q; a. c ■•..•• co • u i2 a ■=. 1 s « 5> O r n* t; co *- J= O 3 •-: o Q. k_ :0 o: £ a c CD .Q O AVS AV R queo terst 3 < -Q. HI 111 • co..co< i; • Resident Fish Biomass • Resident Fish Numbers • Predator Fish Biomass • Predator Fish Number: f^\i* •Schooling Fish Biomass • Schooling Fish Number: A^uit z o :*: ui z "co'; co: CO :E O to; CO: 'C- 3: CO »♦— ■ '.£" • i (0 +rt c CD CO co E t- C T3 £ CD lli E w ^ CO ** TJ O Q c CD E ■5 CD CO CD "? HI CO ■o c CO CO CO 1_ CD XI E 3 ganic S y Orga Z: Q ,co _ CO co E o CO o X 1- V Rool erstitia CD CO c 3 CO z CO c 3 CO CJ 'c CO bile Or tractor Z < «-■■ •^ M— O CO a> HI CO h: c c c -1 OC CD •- ..•*■ • • • • • w: CO. co: E: ; .co O o CO ;-o co X: co CO > o 3: co < Cl Q. :C0 HI HI ■>•.-: .;.•:-; • a o Q. a HI c CD E ■a CD CO Q. a. HI CO CO c (0 X u. HI ■o (0 *- *- CO c •o ™ F n Wl ■a c oco O co <4l u ^ o M-l .— i en c • V-i co 0) .-1 4-1 CD X •o (i) O e -a Xi c 3 TO CO ^ 1-1 A3 o s: (0 4-1 01 o c •rJ 1-4 o "O U to 0) a. J-l a. 03 CO 4-1 m i-i l-W CO O X w O H Xi 0) T3 4-1 XI c CO •H x; CO ai ^-i C 1) ■H "O n T1 B ja 4-1 D CQ CO •i-l i-H e 0) CQ 4-) CO ^-1 >,X3 (0 X •H an • e 14-4 o x; 4J c o c •H o 4-1 •r4 on c c ■<-< •M 14-1 U CU O p CO u D 137 inputs to the others. This procedure is necessarily iterative, where the modeler serves as an interfacing mechanism. While it can be tedious, this approach has the flexibility to allow the modeler's intuition to function freely. Theoretically, if each subsystem model is well-calibrated, the interconnections among them would match. Subsystems were defined so as to maximize internal interactions and minimize connections with external variables (Simon 1973). The resulting Sub- systems are (fig. 2): (1) the Autotrophs which compete for light and nutrients, (2) the Epibiota which inhabit leaf surfaces of the dominant autotroph (SAV), (3) the Water, with its suspended and dissolved substances, (4) the Benthos and the sediments supporting them, (5) the Large Mobile Invertebrates, and (6) the Nekton which feed on production from other subsystems. The sum of the state variables contained in all 6 subsystem models is 45; however, 8 of these occur in more than one subsystem. This redundancy of variables means that the state spaces overlap, and it further insures consistency in the overall behavior of the SAV ecosystem model and its subsystem simulations. It is apparent in figure 2 that the number of common variables (in shaded boxes) decrease away from the Autotrophs, suggesting a reduction in the number of direct interactions among variables at higher trophic levels. These models were designed to represent a unit area of water and sediment in an SAV ecosystem with spatial averaging implied. Both carbon (C) and nitrogen (N) are modeled in this scheme, where N is conserved within the model during all transactions while C is transformed (with CO2 making the difference) as needed according to prescribed C:N ratios for all biological state variables. Flows of both C (and associated free energy) and N are crucial to the behavior of this ecosystem. However, to include both with completely conserved materials would require nearly twice the number of variables. Other chemical factors such as oxygen and phosphorus are assumed to be nonlimiting to the ecosystem's behavior, and those are thus omitted. Several previous modeling studies have explicitly considered both C and N (e.g., Walsh 1975 a,b, Kremer and Nixon 1977, Hopkinson and Day 1977, Najarian and Taft 1981). However, most ecosystem models have been confined to tracing the flows of either carbon (energy) or nutrients but not both (Najarian and Harleman 1977, Wetzel and Wiegert 1983). The mathematical structure of this model uses nonlinear, first-order differential equations simulated by finite difference techniques. There is one equation for each state variable, and each term in an equation represents an interaction between variables. In the following two sections of this paper, we report some salient aspects of two of these subsystem models, the Autotrophs and the Nekton. These subsystems are at opposite ends of the ecological trophic chain, one (Autotrophs) being more externally regulated (by Sunlight, nutrient inputs, etc.), while nekton dynamics result more directly from production at lower trophic levels. The Autotroph Subsystem Mode l A major objective in developing the Autotroph subsystem model was to examine the consequence of changing patterns of turbidity, nutrients and grazing on the competitive balance among the primary producers in an SAV community. This model is depicted in figure 3, where phytoplankton, epiflora, SAV and benthic micro-algae all compete for limited availabilities of light and nutrients. Competition for light is direct via shading, while competition 138 AUTOTROPHS EXTERNAL STOCKS Figure 3. Autotroph ecosystem submodel presented in terms of state variables (shaded symbols) and interactions (lines with arrows) among variables and with external forcing functions (circles). Symbols are based on Odum (1971). 139 occurs for two sources of dissolved nutrients through periodic depletion of supplies, and only the rooted vascular plants have direct access to both nutrient sources. The 7 state variables here are connected to numerous external factors, both those in another subsystem and those entirely external to the SAV community. The nature of mathematical formulations used can be illustrated with the primary production term in the SAV growth equation: P= [C/N] [ATTEN] [LKIN] [TEMP] [NKIN] [LAI]. (1) Here, SAV production (P) is a multiplicative function of 6 auxilliary variables: [C/N], the nitrogen-to-carbon conversion; [ATTEN], the light attenuation relation; [LKIN], the photosynthesis-irradiance function; [TEMP], the temper- ature kinetics; [NKIN], the nitrogen uptake relation; and [LAI], an index of leaf area representing the ability to absorb photons. Light attenuation follows a simple Beers-Lambert relation with various materials contributing to the effect (e.g., Parsons et al., 1979): I z = I e~ kz (2) where I z and I Q are light levels at depth, z, and at water surface, respectively. The attenuation coefficient k is taken as the sum of individual k's for seston, epiphytic material and SAV leaves, where each k is a linear function of the amount of material per m^ with the overall intercept attributable to dissolved substances and the water itself. The photosynthesis-irradiance relation is approximated by a rectangular hyperbola (Parsons et al., 1979): P " P m C K-fT ] ' (3) l z where P m is the maximum photosynthesis possible, and K^ is the light level at 0.5 P m . Data for all of the light relations were obtained from experiments in our laboratory (Kemp et al., 1981). The temperature (T) function used is a simple Arrhenius relation, -(K t /T) TEMP = e (4) Values for K t were obtained from the literature for related species (Titus and Adams 1979, Barko and Smart 1981). A higher order equation (Johnson et al . , 1974) which accounts for temperature stress via protein denaturation at elevated T was used in some versions of the model. Little information was available concerning the appropriate algebraic ex- pression for describing SAV nitrogen uptake (V) from two sources (water column and sediment pore-water). We chose a formulation analogous to the Michaelis-Menten relation, and assuming a single maximum uptake rate (V m = f(P m )) but differing half-saturation constants 140 V = V N +k*N, r a b -I L K +(N +k*N u ) J ' (5) s a b where N a and N^ are aqueous and benthic nitrogen concentrations of dissolved inorganic nitrogen (mostly NHt) , and K is the half-saturation constant for uptake of N fl and (K g /k ) is the half-saturation for N^. Again, these coefficients were calculated from our own experimental data, primarily for the SAV species, Potamogeton perfoliatus (Kemp et al., 1981). Similar expressions were used to describe light, nutrient and temperature interactions in primary production of other autotrophic groups. The basic behavior of this model is illustrated in the calibration output (fig. 4). The close correspondence between model and field data is also apparent here. For clarity the variances associated with these data are not given. However, the model trace is generally well within the 95 percent confi- dence interval for field observations. Subsequently, the veracity of this model was compared to a second independent data set, and again good agreement was obtained between model and measurements (Kemp et al., 1983b). The effects of nutrient additions to this model system were also very similar to those observed in large experimental ponds, and the model was used to extrapolate results from these systems to actual estuarine conditions (Kemp et al., 1983 a,b). It is interesting to note the slight asynchrony of peak summer abundance for these 4 components, indicating some temporal separation of niches to minimize competition toward system homeostasis (Lewis 1980). The Nekton Subsystem Model The hypothesis to be investigated with the Nekton model was that changes in SAV abundance would influence total fish abundance and would shift the balance among various trophic and habitat fish groups. This model is important in the overall simulation framework because nekton provide a crucial feedback control for the other ecosystem submodels (fig. 2) and because its output provides a principal linkage to management concerns. The general organization of the Nekton submodel is described in figure 5, where categories of fish (including total biomass, and adult or juvenile numerical abundance for each) compete for various food items, an important one of which, benthic infauna, is explicitly included in this model. Other food sources are external to the model, and most are variables in other subsystems. The nekton system here is defined by 9 state variables in 4 categories. There are 3 variables within 2 fish groups, and 2 within the third, "Resident Fish". There is some direct predator-prey interaction among the 3 fish groups; however, competition for limited foods also represents an indirect mechanism of inter- action. Model fish groups are connected to external fish populations, with immigration and emigration controlled by temperature cues and density dependent factors. The 3 categories of fish are functional classifications defined on the basis of similar habitat, trophic relations and life histories. The ecological units were developed as a compromise allowing aggregation but retaining some of the mechanistic relationships which characterize populations in nature (namely where they live, what they eat, and how and when they reproduce). This condensation 141 10 . b) Epibenthic Algae CM I E ° 5 CO CO o z < i- co 20 10 J I L. J L. J I L J i_ c) Epiphytes d) Submerged Aquatic Vegetation JFMAMJJASONDJ Figure 4. Calibration simulations from autotroph ecosystem submodel (fig. 3) for standing stocks of a) phytoplankton, b) epibenthic algae, c) epiphytes, and d) submerged aquatic vegetation. Data are taken from Kemp et al. (1981). 142 NEKTON («*) J (h M a E d N en RESIDENT RSH © /" > \/' X EXTERNAL *- WC—) STOCKS PRGRM^ V_^< "^ / \MIGRA-A/ "\ /^ N TION I ADULTS Y ^VE- 1 /PRGRMi I ADULTS I NILES J SCHOOLING RSH EXTERNAL PREDATOR EXTERNAL STOCK Figure 5. Nekton ecosystem submodel presented in terms of state variables (shaded symbols) and interactions (lines with arrows) among variables and with external forcing functions (circles). Symbols are based on Odum (1971). 143 process is a necessary abstraction of ecological modeling, representing an attempt to balance among criteria of realism, precision and generality (Levins 1966). These fish groups are what Boling et al., (1975) referred to as "para- species", defined consistent with modeling objectives. It is fortunate that the fish assemblages in Chesapeake Bay's brackish SAV ecosystems are of relative- ly low species diversity. In fact, 80-95 percent of the fish biomass in each of the 3 categories (defined above) is comprised by 1-3 principal species (Lubbers et al., 1981) with similar functional characteristics. The major "Resident Fish" are Fundulus spp., Lucania parva and Apeltes quadracus ; the most important "Schooling Fish" are Anchoa mitchilli and Menidia spp.; "Predatory Fish" are dominated by Pomatomis saltatrix and Mo rone american a. The elements of nekton life cycles are included in the model by utilizing special subroutines for spawning, recruitment and migration. Recruitment from juvenile to adult age (size) classes is also represented in the model structure for Schooling and Predatory Fish in terms of juvenile and adult numerical abundance. Thus, issues of stock-recruitment and density-dependence can be treated in the model, albeit at a coarse-grained level. The use of numbers and biomass as distinct, but coupled, state variables allows considerable flexibility and structural condensation while maintaining realistic model behavior. This approach, which was used by Steele (1974) for zooplankton in his model of the North Sea pelagic ecosystem, provides a means for tracking both energy flow (as biomass) and population information (as numbers). Predator-prey relations are often best described in terms of numerical abundance, while metabolic pro- cesses are more a function of biomass. Traditional population models consider numbers only (in separate age groups), while most ecosystem models utilize biomass only. This model attempts to combine the strengths of both. The mathematical form of equations used in the model can be illustrated in terms of Schooling Fish biomass and adult numbers. The temporal rat e-of -change for biomass (Q35) is Q35 = assimilation - predation mortality - fishing mortality - spawning effort - respiration + immigration -emigration, (6) while for adult numbers (Q36) the rat e-of -change is Q35 = recruitment from juveniles - predation mortality - fishing mortality + immigration - emigration. (7) Overall, the terms in the biomass equation utilize an interplay between variables of biomass and number, where, for example, assimilation (a fixed fraction of consumption) and mortality involve biomass and numbers for both prey and predator, while the respiration term Involves only biomass (Q35). The terms in Eq. 7 are (with the exception of recruitment) derived from those in Eq. 6, with the reciprocal of average size used to convert from biomass units to numbers. The formulation for predation utilizes concepts of threshold densities (e.g., Wiegert 1975), size-selective feeding (e.g., Brooks and Dodson 1965), other criteria of selectivity (e.g., Ivlev 1961) and refuge provided by SAV structure (Heck and Orth 1980). Predation is taken as the product of the predator activ- ity (PRED) times prey availability (PREY). In the case of adult Schooling Fish, 144 PRED = k x Q 35 [log(L 1 + k 2 (Q3 5 /Q3 6 )] [exp k 3 T] (8) where T is temperature, L\ is related to minimal feeding rate for small organisms, (Q35/Q35) is average size of predator, and k's are empirical coefficients. Similar expressions are used for predation by juveniles but different prey items are involved. Thus, both juvenile and adult feeding contribute to biomass (Q35), allowing for ontogenic changes in diets (e.g., Carr and Adams 1973). Prey availability is defined as the product of prey biomas (Qfc), a poly- nominal function of average prey size, f(Q b /Q n ), and a prey refuge function created by SAV, PREY = k 4 Q b [f(Q b /Q n )] [L 2 + exp(-k 5 (Q p -L 3 ) ) ] (9) where L 2 is the maximum refuge offered, L3 is the lower threshold of plant bio- mass (Qp) for incipient refuge effect, and where the availability function can- not exceed unity. The polynominal function of average prey size exhibits a broad central region (20-180% of mean prey size) with reduced availability when prey become very small or very large. Other details of model formulation are described in Kemp et al., (1981). The general behavior of this model is indicated in the calibration simulation presented in figure 6. Simulated time-course of benthic infaunal biomass follows field observations reasonably close, both in magnitude and timing, although the model shows a slower winter-spring growth in the community than the data would indicate. At this preliminary stage of model development, we can only say that model output is in the right order-of-magnitude , and that certain temporal trends such as abundance of Resident and juvenile Schooling Fish are reasonably consis- tent with data. Seasonal patterns of biomass are generally skewed too far into the autumn, probably due to problems in the emigration subroutines. Ultimately, it is hoped that this model will help us to understand the way in which compet- itive shifts among the autotrophic groups influence the relative balance in fish abundance among the 3 groups (fig. 5) which are well down the trophic chain from those primary producers. Model simulations can be used to distin- guish the relative importance of habitat (e.g., predatory refuge) versus primary food production in leading to these effects, while such a distinction could not easily be made through field experimentation. RESOURCE MANAGEMENT MODELING Management Modeling Framework Parallel to the detailed ecosystem modeling, we developed a system of resource management models for focusing on the multiple interactions of human activities with resource ecosystems. In general this modeling effort was de- signed to assist in utilizing scientific knowledge towards balanced and productive management of Chesapeake Bay resources. In contrast to the detailed ecosystem models, this research was intended to assess both the relative importance of factors contributing to the decline in SAV abundance, and the consequences of this decline (in terms of such factors as fish production). The modeling 145 o CO < z < 12. 00^ 8.00 4.00 0.50- 0.25- CM I E o 3 1.50 CO CO < 1.00 0.50- 1.00 0.50 BENTHIC INFAUNA *f Numbers RESIDENT FISH -''"~\ - t t 1 1 7r \\ / O v \ 1 / Biomass \ \ _ / ^y \ \ 4 t^r V \ /. \V ^z^s o ** ^r 1 ■ ^^--~ 1 i 1 1 1 1 1 1 PREDATORY FISH 4 3 2 ^ CVJ I o I E X 1 *^ k_ CVJ _l N»» z > CO 1 _) -> ~3 Q 1 < SCHOOLING FISH Biomass Juvenile- M 1X1 CM it I! CO 5 < c N D Figure 6. Calibration baseline simulation from nekton ecosystem submodel (fig. 5) for biomass and/or numerical abundance of a) benthic macro- infauna, b) predatory fish, c) schooling fish, and d) resident fish. Data are taken from Kemp et al. (1981). 146 framework explicitly establishes the interactions between SAV ecosystems and human economic systems. Direct and indirect effects of alternative management scenarios were assessed in terms of economic values and ecological processes . Finally, this framework, provided a heuristic format for understanding some principles of resource management. This scheme is illustrated in figure 7 as a cluster of interconnected models representing the influence of human activities, as modified by physical forces (e.g., rain, sunlight, tides), on SAV ecosystem dynamics which in turn affect resources valued by society. Briefly, meteorological conditions coupled with agricultural practices are shown as inputs to the Watershed Runoff Model (Holton and Yaramanglou 1979) which links the Universal Soil Loss Equation to hydrologic and chemical process models, thereby routing water, nutrients, sediments and herbicides from fields to estuary. A simplified 2-layer "box model" of Estuarine Circulation, based largely on continuity at steady-state (Officer 1980), receives agricultural runoff and sewage nutrient wastes and transports water and materials through the estuary providing an ambient water quality field to which SAV are exposed. These materials, along with direct agricultural run- off, provide inputs to the SAV Ecosystem Management Model (SEMM), the details of which will be discussed in the next section. Outputs from SEMM, including fisheries and recreational activities are input functions to the Resource Economics Model which estimates equivalent economic values associated with these features (Kahn 1981, Boynton et al., 1981). Depicted on the left side of figure 7, the marginal costs and benefits associated with various economic activities (Land Development, Agricultural Pro- duction, and Sewage Treatment) are calculated (Boynton et al., 1981). Also asso- ciated with these economic processes are direct or indirect effects on waste loading to the estuary. Resource values are combined with costs and benefits of watershed activities to establish viable Resource and Waste Control Options which managers and citizens consider towards developing resource policies. In general, connections between submodels are undirectional , with feedback occurring only indirectly through the management decision process. For example, materials enter the estuary from the watershed, while the estuary, per se, has little direct influence on watershed activities. In this scheme the modeler serves as the interface between connected submodels, and piece-wise simulations can be per- formed with no loss of information since there are no direct feedbacks. In other words, the output information from simulations in one submodel are used by the modeler to define input conditions for the next sub-model in the sequence. SAV Ecosystem Management Model At the focal point of this resource management framework is the SAV Eco- system Management Model (SEMM). The SEMM was designed to emphasize interactions between SAV ecosystems and human systems (fig. 8), specifically water quality effects on SAV production and abundance, and the habitat and food-chain factors whereby SAV enhances fish production. The structure of SEMM aggregates much of the complexity which had been emphasized in the SAV ecosystem submodels (e.g., Autotroph and Nekton Models in fig. 3 and 5). Our intention here was to preserve sufficient detail in ecological function so that relevant interactions with socio-economic activities could also be included without losing conceptual and computational control. Sensitivity analyses performed for the ecosystem submodels provided some guidance on strategies of aggregation, wherein crucial 147 CO _l LU Q O LU LU o < LU O DC o CO LU CC > < CO s9!J!A|jov ojujouoog pajBpossv 6C C •H a o .— 1 Q> > -l to 3 • o >. 4.) CO PQ T3 HI 0> N X. •H CO c o> CO a. oc CO r-l co o o> JC CO CJ S CO c Ul •1-1 60 O c M n O. •r-l •u oo CO c 4-1 •r-l 01 >— 1 60 T3 > O e u •H en 4-1 3 CO O 3 •r-l IT r-l CO CO > T> CI) 60 60 C U O CI) s B CO X) 3 CO co G o -j •H O 4-J <-t_i CO r-l co 01 0) U •r^ u 60 a> 0) J-l 4-) c CO 1-1 (J 4J IU CO o 4-1 B C CO 0) u B 60 0) CO 60 •r-l CO T3 c CO U B •H 4-) r-l CO r— 1 6 CO o< u A 0> CJ > C/3 o • r~- 200 h CM I E O o) 150 > < CO o Z> 100 Q O DC Q. y 50" EFFECTS of CHANGING LOADING RATES - on SAV PRODUCTION N S. S Herbicides \ «v \ ^ \ \ "**^ Sediments >^ ^ ^ •■» I ^^^IMutrients i i i 100 200 300 400 PERCENT of 1960 LOADING Figure 9. Summary of the effects of changing inputs of a) herbicides, b) sediments, and c) nutrients on annual net production of submerged aquatic vegetation (SAV). Lines represent the distillation of several simulation scenarios from SAV Ecosystem Management Model (fig. 8). 151 4- o CO o 03 CU o co cu > cu +J cu 3 03 > CD O [_ 3 o t/l a> cz 03 CO CU T3 tz 03 3 u • ^ t_ CD • fO ■!-> C c CD CU e CD CU 3 en ■•-> 03 CD C -Q T3 E l/) 4- ■a <4- cu O JCT 1 CO CU L. TD CU C- t) 1— 3 CO o —I i-i ct c_> zzz> Q — I -Z. cc <_> — J I— I *e cc o cd ,-i • O r^ LO oo C\J O UJ *« —• CD o o o o o co o o en i *>o- CO o o o «* o LO O --H + *>o- o o O CO * • o o oo I o o o .-• CD CO CNJ I to- o o •a o IB tO 03 • r— ■X ■p. ■1- — 5«t t_ * oo=*fc L. co fa<* +J o CO rD +> O O +-> o +-> -l-> o +-> • r— CO oo 03 o -f— CO 03 o +-> o ro a. ai 4- r-. a. ai 03 CJ E co •"-" 00 03 cn cu CO CO cu CO CO u •— ' > O +-> > O O 3 ^- — "i— t_ CO •r— T— t_ CJ -t-> 4-> CD CU 4- +-> CD . — +-> 03 •r~ CU 03 4-> 3 CO i *« L. C i — 5*« CJ O CU *^^ CD qi cu CO cc v^^ 4- ■r— cj cc JE cu t_ • • CO • • d CD •r— CD <£. U- CO CO cu CD cu U u 03 ro CO CO CU e 3 o ^~ • ^ fU +-> > fO > CO t_ r>- te en 3 CO CU E > CU cc +-> J^ c c •i- x: o. -i-> o o xz JO CJ cu •<- > . •I- +J — - CD O ■— I 13 CO CU O-CTi t- E i-l ro t- - — — s-o • + 0) -— +-> (D m co E +J +J •!- CU •I- -(-> m- co c: cu cu o C +J 0) O c JD +-> >, o ■o t. CO fz cu (t3 "+- c cu ■<- *~^ L. 1 -o CO CU t- -a co CU ■!- +-> jQ > co E O O 3 t- CJ> ^ CL ■a cu .c: CO (_ CU +J 03 3 CU > cu CO cu jr o CD 3 CO CU >■> ■4-> > ■a o L. a. CU C O CU CO CO CU ■!-> t- c Q. CD CU E t_ CU CD CO 03 •i- C -C 03 O CO t_ o +-> O a> CO CU *-~ > CL. •r- O0 +J CD o — - cu a: CO +J a> (j S- 3 — -a o co t_ CU CL CU 03 JC -r- co t_ •^ o 4- +-> O ■a cu c co 03 CO CU CO t- o 3 t- +-> CD 3 O 4- •-- O t_ CD CU 03 CD CU O O t- 4_) CU Q. +J •f- CU 4- SZ CU +-> CU co J2 03 C- T3 o CU +-> +-> 03 CO i — O 3 O O -C 03 O O 03 CU CO 03 •>- C -•-> o 03 -i- I— CD CU cu t_ t_ cu cu J= JT cu > 03 cu cu CO cu E o CO cu o T3 O 03 £Z o +J 03 CU t_ CJ cu ■o 03 03 CJ t_ CU O co O O •i- 03 c cu •r- O 03 CO CD +-> cu cz J= cu +-> E cu CO CD +-> 03 C C CU 03 co E cu t- X) Q. CU CU -C U CO t_ co CU •1- ■(-> -C 03 H- 3 ■K O (_ CU rsj ■(-> 03 CO o O 0) c cu J3 CU J3 o co 03 cu o co CU -o cc 3 CO c_ O +J o cu CO o JO T3 CU t_ t_ 3 o t_ cu 3 co O CJ 152 advantage (because of the benefits accrued to the farmer) to allow agricultural wastes to diminish the value of estuarine resources, in the long run it may be considered unfair for the smaller, es tuarine-based activities to bear the burden of associated costs. MODELING IN SCIENCE AND RESOURCE MANAGEMENT Models in Scientific Research The role of ecological modeling in scientific research has been much debated over the last decade or so (e.g., Watt 1975, Wiegert 1975). Some of the often mentioned utilities of modeling in environmental research include 1) organizing research objectives and methods (also identifying missing information or poorly understood relationships and formalizing scientific hypotheses), 2) interpolating and extrapolating from a given data base, and 3) testing sensitivities of model variables in relation to their real-world counterparts. Pielou (1981) has recently reviewed and critiqued these and other aspects of ecological modeling, concluding that many of the various uses of models are reasonable but often overstated. In the course of our SAV research program, we have attempted to utilize models toward a number of these objectives. While the ultimate success in meet- ing such objectives will necessarily remain unclear, several more pragmatic benefits have emerged from the modeling effort. For example, the conceptual exercise associated with model development has provided a means for a productive dialogue among diverse research specialists to integrate varied information. As such, a model can serve as a format for discussion. Furthermore, models at the conceptual level can help to bridge the dichotomy between descriptive and experi- mental research objectives. Within the conceptual framework of ecological models the linkage between mechanistic processes and overall ecological structure can be made explicit. Models in Resource Management While ecological models have been used effectively in various resource management programs (e.g., Jeffers 1973, Cooper 1975), logistic and personnel problems can hinder this relationship (Mar 1974). We suggest in schematic form (fig. 10) that balanced use of natural resources, such as Chesapeake Bay, might best be achieved through interaction among resource managers, scientists and users, with ecological models serving as an interface between these diverse groups. In this view, management agencies develop alternative strategies for resource allocation (1, numbers refer to figures) by interacting with user groups and by following traditional management practices. While it is generally difficult to project the potential impact of new management scenarios (2) on natural resources or on regional socio-economic interests, managers attempt to select policies (3) which lead to balanced resource uses (10). Various user groups utilize the Bay in response to needs and desires (7), some of which conflict with one another (8), resulting in an identifiable management problem (9). Scientific groups study the Bay to catalogue and understand the dynamic behavior of the ecosystem and its component parts (4). Such research generates scientific data (5), which in itself is of some use to resource managers, but which finds more utility as it emerges into a conceptual framework (models). Such models can be formalized into a simulation modeling scheme (6), which can be used to forecast potential Impacts of various management options. In addition, 153 CHESAPEAKE BAY MANA GERS SCIENTISTS USERS (1 ) (4) (7) Altern. Mgmt. Scientific Resource / Strategies W Research Uses \ 1 I ' r i (2) (5) C o (8) Ecol. Effects CO CO Scientific ** Z3 Conflicting of Mgmt. Data ° * Uses /DC ' t Trade-Offs \ i / i • (3) Mgmt. Policy Selection (6) SAV ECOSYSTEM MODEL T Sensitivity (9) Problems Identified ^ / '/ \ 1 ' / (10) BALANCED USE OF RESOURCES Figure 10. Conceptual flow chart illustrating potential interactions among resource managers, estuarine scientists, and resource users relating to submerged aquatic vegetation (SAV) in Chesapeake Bay. In this hypothetical framework ecosystem modeling plays a pivotal role integrating the activities of all three groups toward balanced uses of SAV and other natural resources. 154 these models can aid the manager in understanding the explicit and implicit trade-offs involved in each scenario towards judicious selection of an appropriate management policy. The model or suite of models can also be used to aid in identification of conflicting resource uses, thus facilitating the development of a program of balanced resource use. While this scheme suggests modeling to be a central element in unifying various aspects of resource management, there are limitations which must be recognized as well. In some instances, a mathematical simulation modeling approach, with attendent costs and investments of time, may not be required. Simple calculations may be adequate in some instances, while scientific intuition (which in itself represents a qualitative model) may be sufficient in others. Great care must be taken in developing modeling programs appropriate to the resource questions being addressed. For example, models are sometimes developed at the wrong spatial and temporal scales, or models may be so complex that they tend to obfuscate rather than clarify and objectify resource management options. Nonetheless, if such potential problems are avoided, models can play a very useful role in resource management, particularly if they are viewed as one of many tools which can be utilized in this area of research. ACKNOWLEDGMENTS Numerous individuals have contributed directly and indirectly to the research summarized here. Much of the analysis and computer programming was done by S. Bollinger, while additional computation and interaction was provided by T. Schueler, R. Walker, R. Thomann and K. Lezon. Assis- tance in the relevant data collection and experimentation was provided by S. Bunker, J. Cunningham, K. Kaumeyer, L. Lubbers, D. Marbury, M. Meteyer, J. Metz, K. Staver, and R. Twilley. We are indebted to R. Costanza, T. Do Ian, D. Flemer, J. Kremer, J. Smullen, S. Taylor, R. Walker, R. Wetzel, M. Yaramanglou, H. Odum, J. Zucchetto, and many others for stimulating interactions on the conceptual aspects of this work. This research was supported by grants from the U.S. Environmental Protection Agency No. R805932010 & X003248010 and from Maryland Dept. Natural Resources No. C18-80-430 (82). Financial support for digital computation was provided by University of Maryland Computer Science Center. Typing was done by J. Gilliard and drafting by F. Younger and D. Kennedy. 155 REFERENCES Adams, S.M. and others. 1979. "An ecosystem model of fishes associated with the seagrass, Zostera marina ." In R.F. Dame (ed.) Marsh-estuarine systems simulation. Univ. S. Carolina Press, Columbia, S.C. Barko , J. W. and R.M. Smart. 1981. "Comparative influences of light and temperature on the growth and metabolism of selected submerged freshwater macrophytes ." Ecol . Monogr . 51:219-235. Belyaev, V.I. , K.M. Khailov and I.N. Okhotnikov. 1977. "Mathematical simulation of a marine coastal ecosystem containing macrophytes." Aquat . Bot . 3:315- 328. Boling, R.H. , E.D. Goodman, J. A. Van Sickle, J.O. Zimmer, K.W. Cumming , R.C. Peterson and S.R. Reice. 1975. "Toward a model of detritus processing in a woodland stream." Ecology 56:141-151. Boynton, W.R. , W.M. Kemp and J.C. Stevenson. 1979. " A research design for understanding and managing complex environmental systems," p. 168-177, In: R.R. Johnson and J.F. McCormick (eds.) Strategies for protection and management of f loodplain wetlands and other riparian ecosystems . GTR-WO- 12. U.S. Dept. Agric, Forest Service, Washington, D.C. Boynton, W.R. , W.M. Kemp, A.J. Hermann, J.R. Kahn, T.R. Schueler, S. Bollinger, S.C. Lonergan, J.C. Stevenson, R. Twilley, K. Staver, J. Zucchetto. 1981. "An analysis of the energetic and surplus economic values associated with the decline of submerged macrophytic communities in Chesapeake Bay," p. 441-454. In: W.J. Mitsch, R.W. Bosserman and J. Klopatek (eds.) Energy and ecological modeling . Elsevier, Amsterdam. Brezonik, P.L. 1972. "Chemical kinetics and dynamics in natural water systems,' p. 832-913. In: L.L. Ciaccio (ed.) Water and water pollution handbook , Vol . III . Marcel Dekkar, New York. Brooks, J.L. and S.L. Dodson. 1965. "Predation, body size and composition of plankton." Science 150:28-35. Carr, W.S. and C.A. Adams. 1973. "Food habits of juvenile marine fishes occupy ing seagrass beds in the estuarine zone near Crystal River, Florida." Trans . Amer. Fish . Soc . 102:511-540. Cooper, C.F. 1975. "Ecosystem models and environmental policy." Simulation 27: 133-138. Goodall, D.W. 1974. "Problems of scale and detail in ecological modeling." £. Environ . Mgmt . 2:149-157. Heck, K.L. and R.J. Orth. 1980. "Seagrass habitats: The roles of habitat com- plexity, competition and predation in structuring associated fish and mobile invertebrate assemblages," p. 449-464. In: V.S. Kennedy (ed.) Estuarine perspectives . Academic, New York. Holtan, H.N. and M. Yaramanglou. 1979. Procedures manual for sediment , phosphorus and nitrogen transport computations with USDAHL . Maryland Agric. Exp. Sta. , Univ. of Maryland, College Park, Md. Hopkinson, C.S. and J.W. Day. 1977. "A model of the Barataria Bay salt marsh ecosystem," p. 236-265. In: G.A.S. Hall and J.W. Day (eds.) Ecosystem modeling in theory and practice . John Wiley, New York. Ivlev, V. 1961. "Experimental ecology of the feeding of fishes." Yale Univ. Press, New Haven. Jeffers, J.N.R. 1973. "Systems modeling and analysis in resource management." J. Environ . Mgmt . 1:13-28. Johnson, F.H. , H. Eyring and B.J. Stover. 1974. The theory of rate process in biology and medicine . John Wiley, New York. 156 Kahn, J.R. 1981. The economic losses associated with depletion of submerged aquatic vegetation in Chesapeake Bay . Ph.D. Thesis, Univ. of Maryland, College Park, Md. Kemp, W.M., M.R. Lewis, J.J. Cunningham, J.C. Stevenson and W.R. Boynton. 1980. "Microcosms, macrophytes, and hierarchies: Environmental research in the Chesapeake Bay," P. 911-936. In: J. P. Giesy (ed.) Microcosms in ecological research . Symp. Ser. 52 (CONF-781101) NTIS, Springfield, Va. Kemp, W.M. , W.R. Boynton, J.C. Stevenson, and J.C. Means (eds.). 1981. "Submerged aquatic vegetation in Chesapeake Bay." Univ. Maryland Center Environ . Estuar . Stud . Ref . No . 80-168 , Cambridge, Md. Kemp, W.M., A.J. Hermann, and W.R. Boynton. 1983a. "Ecosystem simulation models for submerged vegetation in upper Chesapeake Bay: A tool for resource management." Univ. of Maryland Center Environ . Estuar . Stud . , Cambridge, Md. (In press). Kemp, W.M., W.R. Boynton, J.C. Stevenson, R.R. Twilley and J.C. Means. 1983b. "The decline of submerged vascular plants in upper Chesapeake Bay: Summary of results concerning possible causes." Mar . Technol . Soc . J_. (In press). Kremer, J.N. and S.W. Nixon. 1978. A coastal marine ecosystem : Simulation and analysis . Springer-Verlag , New York. Levins, R. 1966. "The strategy of model building in population biology." Amer . Scientist 54:421-431. Lewis, M.R. 1980. An investigation of some homeostatic properties of model ecosystems in terms of community metabolism and component interactions . MS Thesis, Univ. of Maryland, College Park, Md. 150 pp. Lubbers, L. and others. 1981. "Comparative abundance and structure of littoral nekton communities at vegetated and non-vegetated sites in Chesapeake Bay," p. 461-574. In: W.M. Kemp et al. (eds.). Submerged aquatic vegetation in upper Chesapeake Bay . Univ. of Maryland CEES, No. 80-168, Cambridge, Md. Mar, B.W. 1974. "Problems encountered in multi-disciplinary resources and envi- ronmental simulation models development." J_. Environ . Mgmt. 2:83-100. Mclntire, CD. and J. A. Colby. 1978. "A hierarchical model of lotic ecosystems." Ecol . Monogr . 48:167-190. Najarian, T.O. and D.R.F. Harleman. 1977. "A real time model of nitrogen-cycle dynamics in an estuarine system." Prog . Water Technol . 84:323-345. Najarian, T.O. and J.L. Taft. 1981. "Nitrogen-cycle model for aquatic systems: Analysis." J. Env . Eng . Div. (Am. Soc. Civil Eng.) 107(EE6): 1141-1156. Odum, H.T. 1971. Environment , power and society . John Wiley, New York. Odum, H.T. 1983. Systems ecology : An introduction . John Wiley, New York. Officer, C.B. 1980. "Box models revisited," p. 65-114. In: P. Hamilton and K.B. McDonald (eds.) Estuarine and wetland processes . Plenum, New York. Overton, W.S. 1975. "The ecosystem modeling approach in the coniferous forest biome," p. 117-138. In: B.C. Patten (ed.). Systems analysis and simulation in eco lo gy , Vol. 3. Academic Press, New York. Parsons, T.R. , M. Takahashi and B. Hargrave. 1979. Biological oceanographic processes . (Second ed., reprinted) Pergamon, Oxford. Patten, B.C. 1975. "Ecosystem linearization: An evolutionary problem." Amer . Natur. 109:529-537. Pielou, E.C. 1981. "The usefulness of ecological models: A stock-taking." Quart . Rev . Biol . 56:17-31. Schaffer, W.M. 1981. "Ecological abstraction: The consequences of reduced dimen- sionality in ecological models." Ecol . Monogr . 51:383-401. Short, F.T. 1980. "A simulation model of the seagrass production system," p. 277-295. In: C.P. McRoy and R. Phillips (eds.) Handbook of seagrass biology : An ecosystem perspective . Garland STPM Press, New York. 157 Simon, H.A. 1973. "The organization of complex systems," p. 1-27. In: H.H. Pattee (ed.) Hierarchy theory . The challenge of complex systems . Geo. Braziller New York. Steele, J.H. 1974. The structure of marine ecosystems . Harvard Univ. Press, Cambridge. Stevenson, J.C. and N.M. Confer (eds.). 1978. "Summary of available information on Chesapeake Bay submerged vegetation." FWS/OBS-78/66, Fish and Wildlife Service , U.S. Dept. of Interior, Washington, D.C. Titus, J.E. and M.S. Adams. 1979. "Coexistence and the comparitive light relations of the submerged macrophytes , Myriophyllum spicatum and Vallisneria americana . " Oecologia 40: 273-286. Titus, J., R.A. Goldstein, M.S. Adams, J.B. Mankin, R.V. O'Neill, P.R. Weiler, H.H. Shugart and R.S. Booth. 1975. "A production model for Myriophyllum spicatum L." Ecology 56:1129-1138. Verhagen, J.H.G. and P.H. Nienhuis . 1983. "A simulation model of production, seasonal changes in biomass and distribution of eelgrass ( Zostera marina ) in Lake Grevelingen." Mar . Ecol . Progr . Ser . (In press). Walsh, J.J. 1975a. "A spatial simulation model of the Peruvian upwelling ecosystem." Deep-Sea Res . 22:201-236. Walsh, J.J. 1975b. "Utility of systems models: A consideration of some possible feedback loops of the Peruvian upwelling ecosystem," p. 617-633. In: L.E. Cronin (ed.) Estuarine research . Academic, New York. Watt, K.E.F. 1975. "Critique and comparison of biome ecosystem modeling," p. 139- 152. In: B.C. Patten (ed.) Systems analysis and simulation in ecology , Vol. 3. Academic Press, New York. Weber, J. A. , J.D. Tenhuren, S.S. Westrin, C.S. Yo cum and D.M. Gates. 1981. "An analytical model of photosynthetic response of aquatic plants to inorganic carbon and pH. " Ecology 62:697-705. Wetzel, R.L. and R.G. Wiegert. 1983. "Ecosystem simulation models: Tools for the investigation and analysis of nitrogen dynamics in coastal and marine ecosystems." In: E.J. Carpenter and D.G. Capone (eds.) Nitrogen in the marine environment . Academic, New York. (In press). Wiegert, R.G. 1975. "Simulation models of ecosystems." Ann . Rev . Ecol . Systematics 6:311-338. 158 OPTIMAL EXPLOITATION, BY MUSSEL RAFTS, OF THE RIA DE AROSA, SPAIN: PREDICTIONS OF A FIRST-GENERATION MODEL Richard G. Wiegert Department of Zoology University of Georgia Athens , Georgia 30602 USA and Ernesto Penas-Lado Instituto Espanol de Oceanografia Atpdo. 130 La Coruna, Spain 159 INTRODUCTION The Ria de Arosa is an estuary on the northwest Atlantic coast of Spain which has been intensively exploited by raft culture of mussels ( Mytilus edulis ) since the late 1940's (Tenore et al . in press). The Ria Arosa has an area of 250 km^ with an average depth of 19 m. The yield of mussels averages 86,000 Tm (wet weight) yr - * from approximately 2,000 rafts (Perez and Roman 1979). Each of the rafts is about 19 m square and supports 700-800 ropes, some 8-9 m long. The mussels and associated organisms attach and grow on these ropes. An additional 200 oyster rafts produce currently about 2,500 Tm yr - * (wet weight) of Ostrea edulis . For the past several years the Ria de Arosa has been the site of an intensive ecological study, performed by a joint team of Spanish and American scientists. The objectives were: 1 ) to document the major processes occurring in the estuary, 2) to ascertain the factors responsible for the high productivity, 3) to compare the effect of the mussel rafts in Ria de Arosa with a relatively unmodified estuary immediately to the north (Ria de Muros ) , and 4) to obtain the data necessary to construct a simulation model capable of predicting the optimal number of mussel rafts. Two years ago we began building a multicompartment ecosystem simulation model of the Ria de Arosa, one which employed realistic (= definable) parameters and functional feedback relationships which included provision for threshold effects (see Wiegert et al. 1975; Wiegert et al. 1981). Thus the model is 'explanatory' in the sense of postulating mechanistic causes for the observed sets of field data. The model is a set of hypotheses, arrived at by induction and utilizing wherever possible ecological, physiological and physical data. The predictions by these (model) hypotheses can be tested by the independent seasonal standing crop measurements from the estuary. This kind of model, although required if any advance in theory is desired, is far too demanding of data and time to be preferred when only simple prediction is required. Unfortunately, in the present instance, no other course was possible because the data requisite for construction of a multiple-regression type "predictive" model were not available. There is only one estuary being heavily exploited. Furthermore, experimental manipulation of the Ria de Arosa in the sense of increasing or decreasing numbers of rafts and observing the results is not possible because the rafts are individual family enterprises. The conserved or bookkeeping unit in this first-generation model is elemental N, because this is a major factor limiting primary production in these Spanish estuaries. The overall high productivity of the Ria de Arosa Is maint- ained by periodic upwelling which results in a large volume of N-rich water being moved into the estuary (Tenore et al . in press). The structure of this model plus associated submodels and details of many simulation experiments performed with them will be published elsewhere (Penas and Wiegert, in prepara- tion). Here we present only a brief report of the salient features of this first-generation model and the effect of varying the number of rafts in the estuary. 161 METHODS This initial nitrogen flux model AR0SAN1 is constructed as a set of non- linear, discontinuous differential equations representing the fluxes. For all fluxes into a biotic compartment, (e.g., ingestion) feedback control is effected by means of functions which provide for representing 1) a maximum rate of ingestion of N, 2) a refuge of available N, 3) an upper satiation level for available N, 4) a maximum density or carrying capacity for the feeding compart- ment (this is based solely on space, assuming available N levels are optimal), and 5) a lower response threshold which specifies that density at which negative effects of crowding are first manifested. In linear fluxes (such as excretion, detritus production, nonpredatory mortality plus egestion, and fluxes between abiotic state variables) simple donor-dependent equations have been used. In the case of the fluxes representing exploitation by man, we have assumed, that the fishing effort (and the harvest) is directly proportional to the density of a resource. Thus a linear donor- dependent equation is reasonable. At least this is so in mussel and oyster fisheries although it is arguable in fishes and crustaceans. Input of inorganic nitrogen by upwelling is represented by a rate of intrusion and is given a different value for each of the four seasons. Thus, upwelling is averaged over seasonal (3-mo) periods and is not simulated as separate events. This is a serious constraint in this initial model although in general its effect is to exaggerate the potential shellfish production and thus reinforce the conclusions about the potential effect of additional rafts. The resource feedback terms included in all nonlinear fluxes are, in most cases, of the linear form given in Wiegert (1974). Only for microbial action on the three particulate organic nitrogen state variables did we employ the resource feedback based on the ratio of microbial biomass to substrate, given by Christian and Wetzel (1978). Crowding feedback functions are linear in all cases. The Program AR0SAN1 . The FORTRAN program AR0SAN1 simulates the dynamics of this system through time. Numerical integration is done with a simple Euler routine and the iteration interval is 0.1 day. By means of an arithmetic IF statement, in one out of 37 iteration loops, the standing stocks of the 22 state variables are stored in a data array. At the end of the simulation, this data array is transferred to an external data file to be plotted. This array then contains the variation of standing stocks and fluxes by means of 99 values through the year of simulation. THE MODEL Twenty-two state variables, five abiotic and 17 biotic, were defined (table 1) based on their position in the trophic web of the ecosystem. In some cases, the state variables represent taxonomic units rather than trophic ones. This is so in epibenthic fauna (Echinoderms , Crustaceans, Fishes), where trophic differentiation is quite troublesome, for most important groups are omnivorous and, in some cases, the data were available only for taxonomic units. Two species cultured on rafts (mussel and oyster) are represented as separate state variables, due to their high biomass, their importance in the trophic web, and our interest in simulating the variations of their yields, and the effects 162 u PS HI •rH C 00 3 i- o TO C S-i OO r- O 4-J >H a •r-l o 4J /-s 4-1 CO a "cu > X> XI CO CU o •r-l CU 0) E o •■-I 4-1 > TD 01 c O r-H c O M-l CO CO O CU •r-l oo co Cu C r-l u CO CO CO CU a 14-1 ■u 3 00 CO o X> CO U CO O 3 • B . CU i-H C C • B CO o 4-1 X ■H •.h 1-1 CU . 6 H E5 vO CU CO •H 3 C 4-1 c/j O r-4 6 S-l a • CU CU 4-1 [d >J s •■-I CO CO CO oc o Sh 00 o 1-1 CO e CU ct-i r-l P W M-l O x> CU co > CJ w w 4J 4-1 o i-i CO -C •iH p cj •H ■r-l •H hJ 1 4-1 /— \ 4J o < C 3 C W -H Z Cu 00 O 53 H • CO C/J CO M CU c •rH M OO OT o 00 en £ XI •I-l X sa 33 OS S -.H c •rH c 1-1 o t v "' OS w CU i-H o CO w CJ CO co CO en H oo >» C 1-4 o£ 00 (11 53 : H CO o CO o cj fa Sn S-l 4-' O u> • ^ 5-4 M 1-1 O o CO •■-I co g CU c CO 33 PC 3 i—i C/J 1-1 cu > CU o h-l o CO W w OS X a S 1-1 M T3 1-1 ^C c u CJ a pq m r T -| 4- 4-1 OS <1^ CO 4-1 •iH CO i-i M *g o 4- c 1-1 H 4-1 o 4-> >*• H Sh * -J; * * * < . B m 0) vO 1^ 00 ON o i-l CM -o .-1 i-H ■-I I-H CNl CN CN CO 1-1 ,_, PS CO 4-1 r-l CO 1 CJ •s en • CO XI e CO iH I— I ■a 4-J CO • CO ■r^ o cu •iH CO X # c UH 4-J co 3 4-1 4-1 J= i-H 3 O 3 CU CO CO 14-1 4-1 o S-I -M CO u CO co >, CO Sh • C/D CU 4-1 1-4 CO U-l CU CU C S-i CO c X 0) CO 1 0) CO — 1 1-1 CO T3 4-1 CO a> p •H •H >* XI CO 3. i_l u CO r-l rJ a i-H X) o 1h 4-1 X XI C 3 X! 4-1 "O Sh i— i >n c 4-J co o D CU o UH r-H 3 4J a> CO CO O 00 «• Sh o 4-1 B C 4-J 1-1 CU N_/ CU cu co O B B Sh O 6 •rl CO a x> 3 1-1 C XI 4-1 3 ^— ' CO iH o O CO S CO u •r-l c 4-1 3 CO •r-l 4-1 i-H £ C s P CJ 6 •i-l • c CU Cu CU TO r-l 4-1 e 1-1 ^ o O -rH • i-i B CO 1-1 ■o 4J 3 3 i-l CO t~S >N tu X CU 1-1 P Of CO o cu 3 >n -H •rH •> O 3 I— 1 CU ^3 • 4-1 •2 o • • c 4-1 UH X) 4-1 co .-1 X Sh CO O o r-4 • 1H •■H 6 ^ • CO < o CO 3 CU 3 CO XJ cu • 3 3 /-N • 3 z r-H XI CJ CO CO , — i p CX -H co 4-1 CO »> >~ XI 4-1 CU 4J , 6 Sh M 4J CU iH •rH M CO fa CO CO 4-1 o Oh > e CO C o 3 0) XI XI o a Ph CU CO C B Ph CO •rH r-H O 4-1 4-1 O ■rH CO o CO S-i c p W Cu 3 CO fa •iH H 04 Sh CO >N cu o S ►j 3 CO • CO CO co e CO 4-1 e Sh (JH CJ CU • B i-H 4-1 £ CO § . >, XI r-l CU w o CO c H o H CU 3 " •rH 4-1 •— V .XJ CO Pxj 4-1 OS x> 3 CD CJ 52 iH 4-1 M c o • o pg 4J 4-1 T" • re U « r-l C •rH ■<4 CO o a) 5C •rH co CO z 3 O •• u CO O u cj h-l fi 1 CO O || CO r-l •r-l 4-1 B lu < § CO CO s-^ 8 T4 i-H i-H s X 15 u cu . CO S-i cu <; . 4-1 o u r-l rJ • OS PQ CJ O 3 s CJ I-H U 15 oo iH - C CO X ts ■rH o M H XI a fa PQ CO 3 -■H OS 3 -o OS 4-1 H C/j g 04 < X) s •H cu e 2 CU OS i-H 3 oS o <4 ,J o i-l o ^H o x 53 CO oo X) XI O rJ a. C XI s 4-1 H CO 00 CJ X) M fa S3 I-H CO Sh fa 3 CO w PQ g >> o H O c CO a CO O Nl w Oj CU C4H H i-H o o rJ c 3 CO CJ CO CO W P B CO 3 r-H r-l 3 CO § f5 • * .-i • • • • • • • • • oo c^ X CO H 163 of their cultivation on the whole ecosystem. The standing stocks are expressed in grams of nitrogen m - ^ , as an average for the whole system. Besides the state variables listed on the next page, some other nitrogen pools are considered only as sources or sinks. These are oceanic water (source of inorganic nitrogen, sink for dissolved and particulate organic nitrogen), dissolved N2 (source/sink by f ixation/denitrif ication) , and man (sink by fishing). The flow diagram (fig. 1) shows the main pathways considered in nitrogen transfer and transformation. Fluxes are expressed in grams nitrogen m - ^ day - * . Most excretion and detritus production (egestion plus nonpredatory mortality) fluxes are omitted to make the diagram clearer. Because of their quantitative importance, fluxes of excretion and detritus production by mussels have been retained in the diagram. The main pathways of nitrogen transfer are: the inorganic N rate of supply (by upwelling, mineralization, and mussel excretion), inorganic N uptake by seaweeds and phytoplankton, and secondary production by zooplankton and mussels. This secondary production has been strongly affected by extensive mussel culture. Formerly, phytoplankton production gave rise to a relatively long and complex pelagic food chain. This has been "short circuited" by the mussels. This fact, along with the other natural short circuit, that of the seaweed production (larger than that of phytoplankton and very lightly grazed) produced a shortening of the initial food chain to one with a large production of detritus. From this arise several questions that the model was constructed to answer. 1) Is the system capable of recycling this much detritus or will it accumulate in the sediments? 2) Can the system support a much larger mussel and oyster culture taking into account the food available and question 1? The important factors affecting the answers to these two questions are, besides nitrogen itself: 1) Light limits production in winter; its influence is simulated by seasonal changes in the maximum rates of uptake by the photosynthetic state variables. 2) Oxygen concentration affects sediment transfers by limiting production by meiofauna and infauna. These bioturbators enhance microbial colonization of organic matter and, thus, organic nitrogen recycling. 3) The saturation threshold of microbial uptake of organic nitrogen in sediments is low because once a certain level of O2 depletion is achieved (in relation to organic matter supply), the realized rate of uptake of organic nitrogen by microbes cannot increase, substrate availability notwithstanding. 4) The carrying capacity of mussels determines the number of mussel rafts. Because most of the data available on the ecology of the Ria de Arosa were in units of biomass or carbon, conversion factors have had to be used. For animals, Vinogradov's (1953) average of 7.6 percent nitrogen has been used. In the case of phytoplankton, two different C/N ratios were used: 6.0 for blooming phytoplankton, and 10.0 for slow-growing phytoplankton (estimated from Donaghay et al. 1977). 164 to e CO >, CO o o A3 ■u ■g CO CO CO o u < 0) M-l o 00 CO 3 o c CU 00 o u u o 00 •H 165 RESULTS AND DISCUSSION Simulations with AROSAN1 produced the following results when the carrying capabity for mussels was set to equal the current number (2,000) of rafts. Phytoplankton (fig. 2) and zooplankton (fig. 3) show standing stock values and an annual pattern of variation in agreement with data reported by Tenore and Gonzales (1975). The simulated primary production by macrophytes (fig. 4) tracks the annual pattern found by Fuentes (unpublished). Increasing the number of mussel rafts depresses the standing stock of phytoplankton, and because of both direct ingestion and reduction of their food, that of zooplankton as well. Macrophytes, because they grow on the ropes of mussels, increase directly with increase in number of rafts. We emphasize that the model was constructed from experimental and literature data on rates. Thus the measured standing stocks cited above are independent of the construction of the model and are available as a first test of the models' predictive ability. Varying the carrying capacity threshold for mussels in order to simulate a change in number of rafts produces a decreasing yield of mussels per raft as the raft number increased (fig. 5). This is due to the reduction in standing stock of phytoplankton and zooplankton. Mussel production is predicted to reach an asymptote of approximately 105,000 Tm (wet weight) yr -1 . The increased standing stock of mussels does, however, have an impact on the recycling of N between the sediments and the water. The model predicts (fig. 6) an accumulation of N in the sediments as the number of rafts increases. This is a consequence of the assumption that the saturation threshold for the microbial uptake of the substrate is readily reached and surpassed by the tremendous amount of detritus deposed by the increasing numbers of mussels. The value of that threshold also takes into account the influence of bioturba- tion by meiofauna and infauna that are strongly limited by the oxygen depletion accompanying the increased rain of detritus onto the bottom. The model also predicts a pronounced negative impact of mussels on the productivity of oysters. This seems to be due to a difference in the refuge response thresholds for these two species with respect to a shared food. This is an intriguing and wholly unexpected result from the model. If it is supported by further field and modeling work, it could form the basis for some very interesting simulations involving the socio-economics of oyster and mussel production and pricing and lead to a better monetary return from this valuable marine resource, the productive Ria de Arosa. This is the focus of our continuing work with this model. CONCLUSIONS A crude first-generation ecosystem model of the Ria de Arosa simulates annual patterns for the major biotic components that are in agreement with field data not used in the construction of the model. The predicted response to an increase of mussel rafts over the current 2,000 or so in the estuary was 1) a decreased return of mussels per raft 2) an increased and accelerating retention of N in the sediments 166 > a ■o E c .uj n Btu c o 1-1 4-1 to r-l CO 2 ffl S > 1-1 u CO 3 o U-l o ^ ^-v (0 CO s-j >•> s i—i O r-t X co 1-1 60 CO 4-> 3 3 u O b 01 K a. J3 O co •u O c G CO CO CO D ■-I 4-1 o a. i-i o O co vO 4-1 •* ^ U CN x: c a. CO CO H ^ Sj ^ co 3 CNI > O =s= ^ > u CO > M O x; 60 1-1 •U U-l o e H C •i-l 3 CD CO 3 g 0) co en i o •D i-l 0) •i-l CO 3 C C 1-4 3 60 169 CO 3 cr , y-i CO -* o o • 4-1 (0 w 4-1 14-1 60 CO C U •H ■o 1-4 c en C/l 3 e . M-l to o 4-1 C u CLI o u CO u •l-l U c o 14-1 o 1-1 •"N c c CO 60 •H Ij 4J O •r-l T3 CU C 4J o cfl o ?— I 3 |— 1 O TO •r4 •H 4J 4-1 U •H CO C CM •H VO - < a. o to o \a o N3QVHN3W sjojlsao s SWV1D 4-1 o 6 CO a 4-1 a cu o C o o cu e cu u 3 60 •l-l 177 A SIMULATION MODEL OF A NEAR-SHORE MARINE ECOSYSTEM OF THE NORTH-CENTRAL GULF OF MEXICO Joan A. Browder Southeast Fisheries Center National Marine Fisheries Service, NOAA Miami, Florida 33149-1099 179 ABSTRACT Possible interactions between shrimp and bottomfish were incorporated into a theoretical model of the near-shore ecosystem of the north-central Gulf of Mexico. The model was used to simulate the changes in the standing stock and harvests of shrimp that might be side effects of reducing the unwanted fish caught in shrimp trawls by either of two methods: (1) decreasing the proportion of the bycatch that is discarded (but keeping the total quantity caught the same) and (2) reducing the total quantity caught (but discarding the same proportion of the bycatch). Possible direct and indirect influences of each change in harvesting procedure on food availability to shrimp, predation on shrimp, and competition with shrimp were explored. Although discards are a minor portion of the dead organic material used by shrimp for food, decreasing the discard rate by utilizing the bycatch decreased the availability of food to shrimp and its competitors in the model system. Using shrimp trawls with reduced catch efficiency for fish relative to that for shrimp resulted in increased total availability of food for shrimp but increased competition from bottomfish for that food. The increased supply of the common food ultimately outweighed increased competition from bottomfish, and the shrimp standing stock and shrimp yield recovered in spite of a higher standing stock of bottomfish. Predation on shrimp by bottomfish was a relatively minor influence compared to the other two effects. Changes in standing stocks not directly connected to shrimp were as important in determining the response of shrimp stocks to different strategies for reducing discards as changes in standing stocks that interacted directly with shrimp stocks as prey, predators, or competitors. INTRODUCTION A simulation model of the near-shore marine ecosystem of the north-central Gulf of Mexico is under development at the Southeast Fisheries Center of the National Marine Fisheries Service, Miami, Fla. The modeling effort is being used to investigate the dynamics of interactions among economically important species such as shrimp ( Penaeus spp.) and menhaden ( Brevoortia patronus ) that are harvested from this highly productive area, which has been called "the fertile crescent" (Gunter 1963). An immediate need of fishery managers is to know the possible effects on shrimp of reducing the discarded bycatch of the shrimp fishery. Fishing operations that use relatively nonselective gear such as bottom trawls usually catch species other than those at which they direct their effort. Incidentally caught, or "bycatch," species are sometimes sold, but they are often dumped overboard when the amount is extremely large or when their value is low relative to that of the target species; such is the case in the Gulf of Mexico shrimp fishery. The fish discarded by the shrimp fishery are almost always killed by the trawl or by handling on deck and are dead when returned to the sea. The total ex-vessel value of shrimp is greater than that of any other U.S. fishery species. More than half of the U.S. shrimp harvest comes from the northern Gulf of Mexico. The weight of "discards" averages approximately 14 times the weight of the shrimp landed and amounts to about 400,000 metric tons in the offshore area from Perdido Bay, Fla., to Point Au Fer, La., which has been delineated as the "primary area" of concentration of these species for purposes of making resource surveys. Atlantic croaker ( Micropogonias undulatus ) , a sciaenid fish, makes up over 50 percent of the bycatch by weight. Five other 181 species, three of which are sciaenids , make up about another 30 percent (GMFMC 1980). A directed fishery harvests about 50,000 metric tons of these species from the north-central Gulf of Mexico. They are used in the manufacture of pet food and oriental fish paste (surimi), as well as being marketed fresh. Catch rates and total landings of the bottomfish fleet have declined in recent years, and some state fishery biologists and participants in the groundfish industry feel that the decline is due to competition from an increasing number of shrimp trawlers. Fishery managers charged with "optimizing yield" from our marine resources need to know whether the discards are a necessary consequence of harvesting shrimp, an economically valuable fishery product, or represent a waste that should be eliminated, preferably without economic hardship to the shrimping industry. A key question is whether killing fish in the shrimping operation and throwing them back to sea may actually enhance the yield of the target species by (1) reducing predation or competition, (2) providing a supplemental food source, or (3) increasing nutrient regeneration and thereby stimulating primary productivity, leading to a greater availability of shrimp food. If the present way of handling discards enhances shrimp production, then reducing discards could affect shrimp harvests detrimentally. There are two possible ways to reduce discards without reducing the level of shrimping effort. One way is to catch fewer fish by using shrimp trawls with a reduced efficiency for catching fish relative to that for catching shrimp. This might increase the standing stock of living fish in the system and decrease the standing stock of dead fish returned to the system. Another way to reduce discards is to land a larger proportion of the bycatch. This would reduce the quantity of dead fish returned to the system without increasing the standing stock of living fish. These alternative approaches to reducing discards might affect shrimp harvests differently because one approach increases the standing stock of living fish and the other does not. The effects depend on interspecies dynamics and nutrient cycling in the system. The ecosystem model integrates qualitative and quantitative information about the system into a mathematical characterization that simulates interspecies dynamics and nutrient cycling. Some simulations, using preliminary data, were performed to determine the possible effect on shrimp of alternative strategies to reduce discards. Modeling results indicate that the shrimp harvest could be affected by either method of reducing discards. If shrimp trawls to reduce fish catchability relative to that of shrimp were used, however, then an initial decline in the shrimp harvest might be followed by a recovery to present levels or slightly higher. In the simulation, the extent to which bottomfish selected against shrimp in favor of alternative prey affected the recovery of shrimp stocks. Indirect interactions among species were as important as direct interactions in determining initial and long-term responses of shrimp stocks to different strategies. Model results may be dependent upon certain parameters in the model that determine the rate of nitrogen remineralization by living bottomfish. They may also depend on the Michaelis-Menten half-saturation constant, which determines the rate of phytoplankton production as influenced by the concentration of nitrogen in the water. 182 It is too early in the development of the model to state with any assurance that model results apply to the real world. The model must undergo further testing and evaluation before specific management recommendations can be made on the basis of simulations. At this point the modeling effort suggests that changes in fishing pressure can alter the balance between species in both competitive and prey-predator interactions. Work with the model suggests that standing stocks that are large relative to other standing stocks in the system can be remarkably stable in the face of heavy fishing pressure. At the same time, fishing pressure on such stocks can influence other stocks, even those that are neither prey nor predator of the harvested species. The model demonstrated that animals at higher trophic levels, such as marine mammals, can have large impacts on lower trophic groups, such as zooplankton. Both the stability and instability exhibited within the model system in the face of fishing pressure were due to the complex interconnections of system structure. In this paper I describe the mathematical structure of the model and present and discuss results of simulations. A more detailed description of the north-central Gulf of Mexico near-shore marine ecosystem and the conceptual basis for the model design were presented in a previous paper (Browder 1981). MODEL STRUCTURE The model is diagramed in the energy flow language of H.T. Odum (1982) in figure 1. It is a food web of 12 compartments connected by flows of energy in the form of organic matter and by flows of nutrients. The compartments represent inorganic nitrogen; standing stocks of phytoplankton and animals in eight trophic groups; and organic material of two types, high nitrogen and low nitrogen. The model receives three inputs: solar radiation (as gross primary productivity), inorganic nitrogen, and low-nitrogen organic material, the latter two of which come from river inflow. The types of exchanges between compartments of the model are nutrient uptake, phytoplankton rain to the bottom, feeding by animals, release of unassimilated organic material by animals, and the release of mineralized nitrogen through the decomposition of organic material and in the elimination of metabolic waste products from animals. Energy leaves the system as carbon dioxide (through the respiration of plants, animals, and decomposers) and as harvests. The basic mathematical structure of the model is a set of 12 differential equations : Ql = Jl +1% - L - P 1>2 , Qi Q2 = J 2 " p 2,3 -£ p 2,j " R 2. Q2 q 3 = j 3 + p 2>3 +2fj -£p 3 ,j. Q3 Q 4 = F 5 + (B - P 4>12 ) -£P4,j> Q4 Q5 ^P'i.5 -£P 5 ,j " R5. Q5 Q6 =£P*i,6 "2 p 6,j " H ~ H 6> Q6 Q7 =2P'i,7 -£ p 7,j - R 7 , Q7 183 INORGANIC NITROGEN PHYTOPLANKTON LOW-NITROGEN ORGANIC HIGH-NITROGEN ORGANIC ZOOPLANKTON PELAGIC FISH BENTHOS 184 Q8 = ZP'i.8 ~^8,3 " R 8 " H 8» Q9 : SHRIMP Q9 = SP'1,9 "^ p 9,j - R9 " H9 - B, Q 9 : GROUNDFISH QlO ■ ^ p 'i,10 -^ P 10,j " R 10 ~ H 10» QlO : MIGRATORY PELAGICS Qll = SP'i.H -^ p ll,j " R ll» Qn : MARINE MAMMALS Q 12 = lP' ± 12 + cB -IP12 i " r 12> Ql2 : LARGE SCAVENGERS where i is a number from 2 to 12 representing food compartments, j is a number from 5 to 12 representing predator compartments, Jk are flows into the system from outside, Hi are nitrogen release and remineralization rates, L is the rate of loss of inorganic nitrogen in currents, Pj a are flows of energy between compartments, most of which represent feeding , P'i a are the assimilated portion of food intake, F^ are flows of unassimilated food to organic compartments from animal compartments , B is the bycatch rate, R.» are respiration rates, Hj are harvest rates. Flows indicated in the equations but not represented by connecting lines in the model diagram can be assumed to be null flows. J I and J3 are constants representing nitrogen and low-nitrogen organic matter flows into the system. J 2 , gross primary productivity (GPP), is a variable that is a product of S2, GPP at saturation, multiplied by a Michaelis- Menten equation, with nitrogen as the substrate (controlling factor): J 2 = 82(01/1011! + Qi). ?l 2 is the rate of uptake of organic nitrogen by phytoplankton: p l,2 " n 2< J 2 " R 2>» where n2 is the nitrogen concentration in phytoplankton. 185 L, the loss of nitrogen from the system, and ?2 3, phytoplankton rain, are simple functions of standing stock: L = c^Qi and p 2,3 = c i,3Q2- Rates of predation are donor and recipient controlled and governed by the simple equation where p i,j = c i,j w i,jQiQj> Q^ is predator standing stock, Q^ is prey standing stock, W^ 4 is the selectivity factor of j for each i, and c± a is a rate-coefficient determined by balancing the model for steady- state conditions by a method to be described. The material assimilated by the predator is determined by the equation P'i,j = p i,j A i,j. where A^ is a matrix of assimilation coefficients, ranging from to 1.0. The assimilation coefficient is set at in cases where there is no exchange between compartments. Unassimilated food, which is deposited in the low-nitrogen organic compartment, is a function of the rate of feeding on each prey and the assimilation coefficient specific to that prey: Fj -S(P lfj (l-A itj )). Respiration and harvest rates are simple functions of standing stock: where r^ is the respiration rate-coefficient specific to each compartment and e.s is the harvesting rate-coefficient specific to each compartment. The coefficient e is equivalent to the fishing mortality (F) of mathematical fishery biology (Ricker 1975), which has two components, effort (f) and catchability (q). 186 Discard rate is a function of the rate of shrimping effort (eg), bottomfish standing stock (Q9), and the ratio of bottomfish catchability to shrimp catchability (b): B = be 8 Q 9 . Remineralization is treated as a simple function of respiration rate: N j = n j R j' where nj is the ratio of nitrogen released in excrement (or bacterial excretions, in the case of the detritus compartments) to carbon dioxide released in respiration from each organic compartment except phytoplankton. FLOW-BALANCING PROCEDURE Flows and rate-coefficients were calculated by an iterative top-down flow- balancing procedure that was based on three assumptions : 1) The system at present is in steady state (over the long term, standing stocks are neither growing or declining). 2) Animals with several food sources feed no ns elect ively on these sources in a proportion that is equal to the proportion of the standing stocks of these food sources in the environment. 3) Selectivity can be approximated by differential "weighting" of two or more feeding flows to the predator. In the steady-state situation, the inflows to a compartment equal the outflows. Total inflows to a compartment can, therefore, be determined if the total outflows are known. The assumption that relative feeding rates on alternative food sources are equal to relative standing stocks of these food sources allows the inflows to a compartment to be apportioned among food sources. If selectivity is considered important, the apportionment can be weighted accordingly. The equations for calculating flows and setting rate-coefficients by this procedure are as follows: x j =2(Qj A i,j>/£ p j,i» *i,j =XjQiWi,j. P'i.j = AijPi.j c i,j ■ ( X j w i,j>/Qj- Flow balancing must start at the top of the food web with animal groups that experience no significant predation or where the predation level can be estimated independently. (In this particular case I started the flow-balancing calculation with the large scavenger compartment.) The essential inputs for determining flows and setting rate-coefficients by the flow balancing method are (1) standing stocks for all compartments, (2) respiration rate-coefficients for all animal compartments, (3) either respiration rate-coefficients or rates 187 of all inflows from outside the system for all food-chain-base compartments (i.e., phytoplankton and organic materials), (4) assimilation coefficients for each food of each predator, and (5) selectivity weighting factors for alternative prey of each predator. If assimilation rates are not known and cannot be even roughly estimated or if selectivity is not thought to be important, A^ j and W^ a can be eliminated from the above equations. Calculation of initial feedbacks of unassimilated food to organic material compartments and of mineralized nitrogen to the nitrogen compartment are included in this iterative procedure. A computerized routine calculates inflows and outflows of nitrogen at steady state in order to set the constants relating nitrogen excretion rates to respiration rates. In this routine, the inflow of nitrogen to each compartment is determined by multiplying the rate of feeding on each food source at steady state by the nitrogen concentration of that food source. The quantity of nitrogen used in growth and predation is determined by multiplying growth and predation rates by body nitrogen concentration. The remainder is apportioned between elimination in excrement (N) and elimination in fecal material (NF). (Excrement is the elimination of metabolized body wastes, usually in the form of urine. Fecal material consists of ingested material that was not assimilated.) Constants relating nitrogen mineralization rates to respiration rates are then calculated for use in the model. In ecosystem modeling, the usual method for calculating the flows of nitrogen in excrement and fecal material is to make the nitrogen concentrations in excrement (carbon-dioxide-equivalent sugar) and fecal material equal to the nitrogen concentration of the organism. Calculations based on values from a laboratory study by Darnell and Wissing (1975) indicate that nitrogen concentrations in excrement or fecal material are not equal to the nitrogen concentration of the organism or even to the weighted nitrogen concentration of its food sources. Furthermore, if such relationships are assumed, nitrogen inflows will not equal nitrogen outflows to each compartment when organic flows are in steady state, if nitrogen concentrations of the various compartments differ. The method used in quantification of this model should be more accurate than that ordinarily used and should assure more realistic results. The necessary input information for the nitrogen initialization routine is (1) body nitrogen concentrations for all compartments (except high-nitrogen organic, which had to be calculated in the routine) and (2) ratios of nitrogen released in excrement to nitrogen released in feces from the animal compartments. Nitrogen concentration of the high-nitrogen organic compartment was a variable of the model that depended upon the mix of zooplankton fecal pellets and fish discards in the compartment. Computer programs for the top-down iterative flow-balancing procedure and the model were written in Microsoft BASIC and were run on an Ithaca Intersystems Z80-based microcomputer. The Euler method of numerical integration was used in the simulations. The iteration interval for the simulation was 0.1 day. QUANTIFICATION The initial conditions, rate-coefficients, and other constants that were required to run the model are shown in table 1. The values used for each of these parameters and the calculations and references that were the bases for 188 <1> a c r-H CO jO Q> 0> s d CO CO ^jO 4-> rn *^o s tH E >>^ ^•S°E o 1 ^O •rH O 6 o <>-. C <»-■ o 2CT> i- 8 •rH (0 o r-H X OJ J-> 3 o -=r D o o •o c o CO o Q> r-H s to 3 c •H ■t-> 1 S 3 •n. o .5 (1) Eh •P •rH iH 8 2 3 2 O hH CO P 1 - t Eh O C -• •=T O •— «""» ^V ,— O •"" r-H Q) , X o be >> E to t>-v£> E"^0 "O invooo >-Ht- «— on • c\«- o o -riw o m ^ wv + «- x v. o o a C\J en 189 E 3 c o c I m i o 5 O (0 00 to • S r-H ^^ r ° S (.1(1)^ P O Loi 4h O Q >X> E on >> • <- ■H L O C > d) .h a) Bjj a) O 0) -h E-i <«-i 5 4-> v ~' •h ai c n Q. "H o BOX) O r- a) o (0 • n CM , (fl M In C c 5 3 O H 5 CM -H 3 -2 XI o 3 w vX> CM CD o 3 \o .— 1 C\J to > wO IT\X) O ■=r o • ^™N • OJ vo «- 3" OJ VO v V ^^s V_^^_^N^ ON m ON MD • • kO m LTl ON CO =T •* c 1 u o 2 ft •rH oj c x: m 2 § •h (0 t-i Mx: 8 11 CL, ^ o 8 oo & S & 8 190 CD E 3 § a c co 5 00 o> 50 ITl 00 •H CO CD 4-5 <2 o CO I- I m o\ d) •o to a a E O 4J O £s CTi (0 0) 4-> co Cm to S o • a. o 4-5 • SL (1) Dm • x; cr co 3 o /-c T3 • H C 4-5 4-> • OU ^N in O Os .a- *-* CO i-H O 4-> x: to •4-5 0U >>-u|o E 4J cd|lti H CO >> S* 95 cd I S jiti cu e -h •HjJCl 1) £ 3 QS S Rj £ (13 8 —I i C • /O §3| Q. ,SZ r-t 4-> M o cd M O O *U 4-> •r-l rrl f- CO Q. Cu 2 c 53 H O T3 O ■c 3 <2 0) hfi 8S 4J $ c e —4 (U 5 4-5 o > a) 3 L o x: (S o -p EC E •i-t 3 4-5 • CO to 40 o n o -o CO J3 o -o Q-. bO O X3 C o 3 3 cj r-H 5 CO •e.% • oo o 00 1- wo • o CO --n o «- ?o* ca x ^-N m i-h MOO «- o •3- O • • C\J vo o cn XC^ O (\J on • =r o .o LO i- "~ s o7 ro . a- • t— • Ov-^ r> w\ N»^ ^bc ^ V '-• to o-oo tf 00 «"■» (\i «- oo T— O •£ O «- o^ X^h X 0) in — «- & • c— - a-*o • ^■^ • «— oo o + ^ X o CM CO o c c •rH -r-l jz x: o. a rH 4-> o ca ft to •w *-* 4h CO iH >. g> o 5 4-5 X g I bo B >40 to S- o 4J co 0) D. « CM 00 CO L 0) 33 > co to m s, I n « II 8 S o oo o 191 s o •H 4-> CO cd JZ 4J 3 CD 3. 00 VO 0\ «~» o •<-> =r jj tiro Oco £§■-. 3 O O •p U^ c5 O •O^-n O VO C 4J VO Q. CD On fH i- UH — - C CD •h Q.-H C rH O CO CO -P XI i c "8 & a o o Q bO vO 3 CJ (3 CD > O C C CD •o CO hO VO C\J ON VO ITl c— On On CD r-l 4J 4-> CD SC bO x: •rH c bO o o a On co o 33 ^C 8 co a on t>- 0Q x: Q. o •o o § rH 4-> 4J o X) bO o co •H ■o >> c •H a CO bo o o On .2 oo On XI o CO rH CO •o ^^ ^•^ o jr*. ^^ CO CO oo OO oo c- c- ON t— f- On ON rH ON c^ 4J 4-> CD £ 03 CO CO co co •o CD 3 C •H 4-> c - E CD -h rH Q X) CO s& & H CC X 0) 00 J vO rH C— 01 > o c CO x: C\J CC 8 X C\J c— oo o o o c CD 4-> 5 X c— 8 b c CD s c 4-> •H c •& •H X © s & 8 X 8 o rH O 1 C\J rH cr X X X X X X •=r X 00 CO 0O 8 o § o oo o VO o o oo 8 o o o o o o ^ o CNT1 CC c 2 c CO rH a. 8 in cc s i CO bO (0 CD Ou VO CC CO o 5 s CQ tr- ee & o CJ 00 CC x: CO o 4-> o CQ CC co a •H bO r^ CD a 4-) CD |J x: a ■8 CO CO x: co J h o CC CC CC CO L •rH t. CO 8 e Q. CO &5 o CD CD CD 4-> bO bO CO 4-> CO CD 8 X ON 00 m o o x: CO •H *-*■ CD •h (0 Q. ■a S r^ 6 L a,^--co u w X CM 8 X 00 Ri o o O r- m o x: o 4J CO O x: CO o 4J 4-> o m Da 8 X O o 5 J3 "O x; CO 1 % o 4-> o m 3 o & X co o o o o CO o •H bO r^ 8. o 4-> e bO in DO 192 £ $- $ • O Cffi tcHVO p a> i-H m i — £3~ cd co in -o g ^ « o *** to -o ■P'H £ c -o (3 3h>, g ■H 0!£ 4-) 2 « CO 4J 3 O c 4J -M 0> S3 S • M i a S- a Cm CO > •M o X C H 1 cfl eg CQ d. a> •HO C O <<-i O 3 co en OCX c >> a) x q " i-H .C -P CO CD -U CO .X CD CD > CD CO U •* - O CO c o j-> c 33 i-h sp X O Q « Q • CD C\J O O C CD X • N O C O -O -H D ■a t. c 4-> 4-> ■8i£ a eg +J jJ CD > ECU •hoo • > j-j a. t. tox ■h l Q o O jj O O Cm C -D CO Q. N 4-> O ■O O C\J "H 4-> CD S- 4-> • C CO C o, ao i cd D. CD £ 60 O O CO bO SS t, £ X CO -H 3 O 4-> 0) 2 x M co X ■a CD X o Cm co ^ o o •fH v£> 4-> On e >— • "8 X o •r-t 4-> CO co 3 -P •M M -P CD •o M CD > M CM Q) 4-) 4-> c2 c 2 a 8 N M c2 as -p CD co c ■8 CQ «■" oo on .=r in £K OS CO (X X X X vO ITl ITl vO in C— r— O O o • O o o in "8 3 C •H 4-> c cS X 43 CO 2 -p c 8 i •a g 1 i § b o H > ix, •9 c • CO 8 1 * t- z 4-> •M 2 1 1 — ' c 2 I CO 0) re i 2 -p >> •M £ c a £c 40 O c 4-> a c o •M 4J o 3 4J co co o -P H C CD c P a> J-, m 3 o o M 4-> C •rH -M C E O CD •M 4J C w re >> bo co M O CD £5 X CD c S Q) .H bo a °£ 43 >. 2j X z a. 00 a, 8* 9 oo in 193 co CD g c o £ (0 oo $ •o •iH CO in CTv ■s ~ CO CO S (1) £ •O •r-t 0) o o s •r-l CO CO •iH c L £ eo co 5 3 CO £ I •H CO p e CO ^: o CO >> 3 rH S 5 CO 3 g.2 £^ v co o 40 CO Si •H c9<2 a) H Si x: a iH O TJ CO CO g CO r-t ■P 4-> O oo oo os a\ ■O "D • c cS X) Cl, 8 X ^ CO fc- *— * CSS o cfl CO a- 2 CO o 5 S CQ 2 CO 2 s x; CO •iH o 4-) o CD. 2 O cS x CO u o 4-> g bO o 2 OO w CO u oq co I I e > E o co D C •H 4-> c cS CO ^~» c o » •H *— ' ♦J M c o a 3 o ±: § iH a. CO CD 8 •rH to O C CD S ■•■ H o •-^>. •M gj. Cm o 4 c CO 8 2 & Cm O •H bO 0) D. 8* c c CO B 3 >> x: 0-, CU CO o X> m O 5 •H i CO 3 9 o co o bO m O 8 a e •H 2 CO M 9 x: co o 4-> O X) BBS s a a o •&■&■& 8 x: CO •«-4 Cm o •M bO a o 4-> c B c CO M CO - 5 ^ o ^ o 5 sz co M o •P 2 bO B co •H Cm O •i-4 bO I CD CU co M o 4-> CO •o a CD bO B x: to •M Cm O •rH V iH CD a. co u CD cf CD > CO CD co CD bO Si 8 8 x: to Cm O bO CD a, 8 & •M M x: to B § CO o 5 8 s x: CO o ■p ■>-> o X5 CO o s s x: CO % s o -P 4-> O XI o 5 x: CO M o 4J e bO o 5 ^ a x; to M o 4-) e bO B «- 5 CO £. O ■P CO to M CD bO C CD > CO o co CD CD bO bC co & B B B B 9 S S 8 8 I M x: x: x: x: CO to CO •M •M •M l fe fc o o o -p •P 4-) -p •P -P o O o CQ ca m a § LTi VO Sf sf a- < en 3" •a: VO vO vO < CO *- CO vO CO CO < o> en CO •a: 195 I C o .o "^ w co co 8\ 3 43 4 — - w i/j ot to n to i/i 243i3^^i3|| in c-- 5 CO a, in in t~- c~- C7> C* fO a) CD 4-> j-> CO CO a, a. 5 t— O O • • • O t- r- oo o o o «- O O .- r- ,- o O O • • O T- o o i-H o o o o cm m oo a) 3 C c cd X> s 4-> h Cu — « CM Ei O CO CO £- L CD O g> •s CD > e CO o a CO CD CD bO bO £ u CO 5 5 o i-H CO S- o 4-> e bO O o 5 CO 2 c\j 5 n c CD Jf CD > CO CJ co CD bO ^. cr CO £- O 4-> CO •D a. CD bO L «- CM o <- < 5 a - g 4-> 2 2 .3 # CO o § X) § CO o b 2 I 3 3 i •H £ to § CO o •i-H bO O 2 I 3 3 ® CO o 4-> 4-> o X) • — >. § I 5 CO J- CD CD > CO o CO CD o bO +5 CO ^N ^ 9 x: CO CD a. cj> j— CM cS CO £- CD bo § > CO o CO CD bO I s: CO 5 I bO O (0 4J o CM CO t, CD If CD > CO CJ co CD bO CM Or CO L, CD I? CD > CO o co CD bO k ° 5 a CO ^ o o •rH v_* bO CD CO a i fc 1 o jj> CD g -5 bo t. 2 £ £ § £ CM CM CM i- <- L *- « - CO «? o «- £ o t— «— jj S 3 3 O g 9 8 o x: 4J <2 o in o CD •H ■a CQ •■H co L bO C •S CD •H X 4J i— 1 4-> CO (0 o .-( *5 CO s CJ 5 CO c L CO CD •H 4-> a, cu CD CD " TO sp: CO c bid CD X - CO >> CcS X o CO T3 •r-l T-l ^^ O CD 4-) CX, 4-> 22 ^3 CO t.^^ 3 o >>< X is *J ,co LOO cl, a. tn 4-> m s o ■s •H o o 00 CM o 4-> co s o O ?3 2 CO X CO CD E CO i-H co O CM •in cr> « « go c co c 8 8 a. co c s (0 Ou co c u CO a. 5 in CVJ X ITi C\J C\J o X cu g 4-> c CD r-l CO c o co •-I *j -u •rH b 5 a o 8 a. •H \ — * N •rH s ■E o cd c •o 53 C •r-l (rt o 4JS •H CO Cw Cn L w O Cm C O § 4-> O CO •r-l i— I CO •rH L 9 > CO C 5 cS § JO x; b0 2 & CO T3 « f U >> § 2 <3 6 x; 60 J- CO c o -p s 8" CO 4-> a •rH U o CO L & CO s 197 these values are also shown in table 1. The model was quantified for "average" conditions, and seasonality was ignored; thus model forcing functions were constants. Values in the literature and fisheries statistical data were used for the quantification. Standing stock estimates are specific to the area offshore the Mississippi Delta. Fishery standing stocks were estimated from fishery landings data for 1975, using the relationship: N = C/F, Q = N(D/M), where N is population number, C is landings, F is instantaneous fishing mortality, D is average dry weight of individuals, and M is the area inside 93 m (50 fathoms) covered by the landings, or, in the case of migrating species, area covered in migrations. The F values used in these calculations were crude approximations and may be sources of error. Pelagic forage fish standing stock was based on menhaden catches for Louisiana and doubled to take into account other forage fish species in the area. The shrimp standing stock was also based on Louisiana landings. Bottomfish standing stock was based on landings and estimated bycatch of croaker and other bottomfish species in the area from Point Au Fer, La., to Perdido Bay, Fla. The model is quantified for the area from the eastern Louisiana border to Point Au Fer, with respect to the above three groups. The standing stock of migratory pelagics was based on king and Spanish mackerel catches for the entire U.S. Gulf coast. The mackerels are not fished commercially to any extent in the north-central Gulf, but those stocks fished commercially elsewhere probably spend about half the year in the northern Gulf. By dividing estimated total Gulf standing stocks by total area of the Gulf, I obtained an estimate for an "annual average" standing stock of mackerels in the model area. This value was doubled to provide an estimate of coastal pelagics standing stock, which includes other species in this same trophic group for which there are no commercial landings on which to base standing-stock estimates. Animal respiration rate-coefficients were specific to each compartment of the model and were based on values for a representative species in each compartment or a related or similar type species. Calculations of nitrogen flows are based on body nitrogen concentration of a representative species in each compartment or a related or similar type of species. For zooplankton, the N/NF ratio used (in the initialization routine) for setting nitrogen release rates relative to excretion rates was 0.3 to 0.7. The routine calculated a nitrogen concentration ratio in fecal pellets of 0.06455 at steady state (com- pared to a value of 0.0448 calculated from Johannes and Satomi (1966)). For bottomfish, the N/NF ratio employed was 0.99 to 0.01 (compared to a value of 0.9942 to 0.0058 from Darnell and Wissing (1975) for pinfish, Lagodon rhomboides ). Specific nitrogen information of this type was not available for the other animal compartments. The production of feces by the three higher trophic compartments (migratory pelagics, marine mammals, and large scavengers) was considered to be insignificant in the model, and all nitrogen loss was designated to excrement. For the other animal compartments, the N/NF ratio was 0.8 to 0.2. 198 Phytoplankton standing stock, (estimated from chlorophyll data) and net primary productivity values were from offshore Mississippi Delta studies. The Michaelis-Menten half-saturation constant relating phytoplankton production to nitrogen concentration in the water came from a laboratory study with a coastal diatom, using nitrate as the substrate. Phytoplankton respiration was estimated from generalized information and is not specific to the area or the species found there . Some assimilation coefficients were available from the literature, but most were approximated, based on qualitative information. Selectivity weighting factors of alternative food sources for benthic organisms, shrimp, bottomfish, and large scavengers were weighted differently to reflect qualitative information from the literature regarding feeding selectivity. High-nitrogen organic material was weighted higher than low-nitrogen organic material for both benthos and shrimp, assuming a preference by these groups for the high- nitrogen material. The weighting factor for feeding by benthos on low- nitrogen organic material was adjusted downward slightly to set the respiration rate-coefficient for high-nitrogen organic material (calculated in the flow balancing procedure) close to the value of the decomposition rate-coefficient for fecal pellets that was calculated from Johannes and Satomi (1966); the model value was 0.2071, as compared to the literature value of 0.1834. Benthos was weighted equal to high-nitrogen organic for feeding by shrimp. Shrimp received a low weighting relative to alternative prey of bottomfish to reflect the infrequent occurrence of commercial penaeid shrimp in the stomachs of bottomfish that has been noted by Sheridan et al. (1981). This would tend to minimize the effect that predation by bottomfish could have on shrimp standing stock in the model. Discards were weighted slightly more heavily than live fish in the food flow to large scavengers, reflecting the fact that sharks and others in this group have a scavenging tendency, although they also feed on live animals. Marine mammals were weighted lower than live fish in the feeding flow to scavengers because the size of adult marine mammals equals or exceeds that of large scavengers, and adults are capable of protecting both themselves and juveniles. Weighting factors were estimated on a log^g scale, as a first approximation. Simulation tests indicated that weighting factors of this magnitude were necessary for differences between simulations caused by differences in weighting factors to be detected on the scale of the simulation graphs. SIMULATION TESTS Using the model, tests were made of the effect of alternative strategies of handling bycatch on fishery yields and standing stocks. These strategies were as follows: 1. Present trawling and discarding practices. 2. Half of bycatch not discarded (utilized). 3. Shrimp trawl efficiency for catching fish cut in half. 4. Shrimp trawl efficiency for catching fish halved and directed fishing effort for bottomfish doubled. 199 In making these tests, I assumed that the basic structure of the system (i.e., the number of compartments and the links between compartments) did not change as an adjustment to new conditions. In each test, no rate-coefficient or other constant in the model program was changed other than the one (or in test #4, two) changed to simulate the test condition. A sensitivity test was run to determine how changes in the selectivity weighting factor of bottomfish for shrimp affected the behavior of the model under alternative strategies of handling bycatch and how this might affect the conclusions drawn from simulation results. A test was also made of the influence on model results of the relative standing stocks of bottomfish and the other groups. In a further test, marine mammals were made to feed on bottomfish discards as well as living bottomfish, a feeding flow that was not included in the original model. In the initialization routine, model coefficients were set at steady-state to represent the present practice of handling discards. Under present practices, all standing stocks were constant throughout the 5-year period of the simulation (fig. 2). Figures 3 through 5 are simulation graphs of the management tests. These graphs track the standing stock of each compartment for the 5 years following imposition of new conditions. Under test conditions, standing stocks were initially the same as under steady-state conditions, but changed in response to the particular conditions imposed, moving toward a new steady-state (or, in some cases, toward a stable oscillation). Shrimp stocks declined and leveled at a new steady-state approximately 28 percent lower than the former one when half of the bycatch was utilized rather than being returned to the system (fig. 3). When discards were reduced through the use of shrimp trawls with half the efficiency for catching fish, shrimp stocks declined initially but rebounded to present (steady-state) levels (fig. 4). When the combination of shrimp trawls half as efficient in catching fish and a doubling of directed effort at bottomfish was tested, shrimp standing stock declined by 10 percent (fig. 5). when shrimp standing stocks declined, shrimp harvests declined, as indicated in table 2. Other standing stocks also responded to the tested changes in harvesting practices. Marine mammals, zooplankton, and high-nitrogen organic material increased when shrimp trawls that reduced fish catchability relative to shrimp catchability were used, but decreased when the bycatch was utilized rather than discarded. Contrarily, stocks of pelagic fish and migratory pelagic fish decreased when shrimp trawls that reduced fish catchability relative to that of shrimp were used but increased when the bycatch was utilized. Standing stock of bottomfish and phytoplankton and the quantity of nitrogen in the water also changed under different harvesting conditions, but these changes were so small relative to the total quantity that they could not be seen on the graphs. An idea of the effect of test conditions on these larger, more stable compartments can be obtained by comparing total annual respiration of these compartments during the different test simulations (table 3). Respiration is directly proportional to standing stock and thus total respiration is an index of the level of standing stock over a period of time. In the first sensitivity test, weighting factors were changed to indicate less selectivity of bottomfish against shrimp relative to alternative prey. The change in weighting factors increased the predation rate of bottomfish on 200 ,ui/6uu ,W/biU o 9 9 o ~ ro - cu X X U 4J O 4J 4-> ig C 111 O O Q rf5 T3 -H ^ H CU 1 60 .o co r-i 3 nj to l-i co o i-i i-i 00 cu . o i— i cu ro C rfJ Vj co O. E •i-i g ro ■« B ■—1 9 o m H •» tj ■o g >^ fri o .-< e c • • CO i-J 4-> O O CU X •• cu c c y t3 ci) ■* «s i-l 4-i H O C3 G Vj o o E jj o o ro g )j 4-1 4J g co ro ro ^ ,ms c a c c ••> o • c B « ro --I •i-i ^-v •■-( o i-i i—i 4-> i — l O G. o- *- •H =s= co O O X T3 X X 4-1 O CO C 4J 01 O >> N -H o co ill ig £ 4-1 o a> ? a) cu •« 4-» -i-4 r-l S^ 4-1 » — rfl <4-l •" Vj C O <-< •» O 01 /-n v± O i-l 4-> CO CO N ^ • CO to CU C U O -i-l U U O 0) O Sj 00 a G Vj i-i ai ro CU ID 0) 00 60 6 •- T3 C VJ C O ■-! C O CU i-4 vj o o 3 O rf3 "O 4-1 tJJ . H C -^ C O CO e tu ro c ro cr < cu to 4J oo *> 4J CU • CO u X • > 4-1 CO cu to -i-l > CO "0 .C 3 C 4-J •s j-i o h o . ro to c tu 4-1 4-> ro cu ro i ■« > O to TJ >■> rf3 ""> ro 4-11 01 00 • o CO CO u"l CU Vj -i-l o to jd i r-i > to X O t»i v - x -i-" tu O T) 4-1 CO ■« D. 00 4J ro to Vj i— 1 6 Vj C0 CU -M VJ O O t4 ro 4-i CU O 4J O Vj 1-H oo to cu u o x ^-^ C 15 -> ro • to •H C "i-l CO 4-1 O CO T3 i-l rfJ ,M r« p CU 00 ^ 4"> Vj ro co 4-i cu C i-i o ro 4-1 -i-l C 3 t4 tO O rf: CO ~ O 4-i Vj O W 4J - in co D- cn ro to C >, HI H 6 o •\ O to Vj • X ^-\ •i-l CO CU O 4.) co 4-1 r-l CO C 4-1 -i-l C Vj ro oi t4 ro ro c cu o r-4 T3 X CL C « rfJ 4-1 3 o ro to i-i oo ro 6 B •o u •- T3 1-1 r-l P VJ O rH cu co oo ro o o o Vj C 4-1 -i-l C • a- -O -i-l C 4-1 cu cu o cu B o ro x oo cu 4-1 3 N H 4-1 O i 00 4J CO -i-l 3 Vj X Vj O to Vj g 4J 4J CO ro i— i ro o i-i i-l i-i -i-i !-H cm *— ' x: co 4-1 c 4-1 cu Vj 3 00 fa 201 ,w/6w ,w/6ai o m 2 2° - X a. a> o u 0) • CO X o 01 4J c CD X OI -v T3 U 4J 4-1 60 X 0> 0) ■H o to N X c <4-4 I-l 1-1 •--l H o 4J U-l i— i O •H 01 1-1 DO 4-) C O 60 4-1 h 1 i-l u 3 /-n 01 ■a 3 60 CO (0 X CU O CO r-l X TJ e ^H l— 1 l— 1 N— ' O O 5 CT! 0> ■U i-l 15 u •» D. to CO l-i CO i-l i— I ci a) O ••> co >, a • CO • r— 1 E X CO 1-1 o • 1 >, V-l O co 01 CO to 4-1 E X TJ 0) C c - 4-1 1 >. 0) O C Q) ITI e 4-> O C <4-l i-l 0) 4-1 i! 4JtI o ^- > M C rW r-l 1-1 CO to C CO u-t to 14-1 (X i— i co E i-l -* E a r-l CO 01 l-i o O D. ■•> £ 01 o u 4J O i-l 3 >i o 01 -i-l •s X X N 1 C XI 01 o D- -C o M CO »« CO 4-1 0) 0) •- i-H T-l X c 01 r-l <4-l 4-1 0) 3 U-l O ■« i-t to i-l o • i-H >1 3 oi X O co U u ^ •H o E ao o - i-l 4-1 0) 0> CM o c oi co 4-1 U i— l 4J 111 4J rl co CO 60 to 60 >i to CO o E i-i tO T^ C GO r-4 S X CO C ■u o <-t CO D- ■r-l •H i-l •*. 01 cn T3 C C r-l 13 i-l c co o • O CO C CO 60 • rH E u O 4-1 •• U O c •r-l CO to O •> tM O 4J 3 ---- O N CO 01 O C -C CO •i-l iH -C r-l 01 tO rJ to U 3 H i-H 00 i-l QJ J^ o e o o y-i 60 o JS •rJ 14-4 i-l s c o CO • 4J o OJ 4-1 0) a t0 -r-l 4-1 > tO X 0) u CO C 4J CO 4-1 XI o 1 o o 60 H 4-1 OI rC X CO C C CO r-l 60 ■H O co i-l ••> o) "O • oo X: "I 60 C C0 >-i c tO • rJ CO u en •r-l rJ •- O TO 4J 01 0) -a O i— 1 1—1 to XI >> C 4J O *» ^— ' a eg o o a. U-t 3 i—i 4-1 co o E to z 0) CO <4-l • -r-l ^; X) O rJ |J c 4-1 00 X to o •• e C C •» to X SB 0) •i-l r-l CO CO CO ■— 1 co •-. CO O 0> CO 'r-l ITI •" -« Z c r-l o ij o m 3 •H a. w tu o e OI 4J • »\ •H • CO rJ CO O r—S CO .— N Jd • E to CM QJ 0) OI - u ■a =s= 0) 4J 4J O to O aj 3 CO CO -r-l O 4J 4-1 4-1 •r-l c C C X co 4-1 tO X) •r4 i-l CO 4-1 TJ O OI ■o ■D 60 C OI r- 1 4-) ■sf rJ U rJ OJ rJ PL, w OJ O O X o. 0) V4 3 60 202 Ul/ftUj W/&UJ 8 O 8 2 J l_ o m 9 O O (VI - in - «■ > . — I I ro CM 0- 5 55 u Z CO Ul -1 i CO HE > o 5 < tr o z NITROG E MAMMA oc o z 2 _i < 2 r > / \ I I h \ 11 o "3rtt ro CM to O o X t- z Ul CD z < o 0u CO S-l ■- >-. 6 .* o l-l l-l •a 0) . r* O u CU «\ r-l ~ 4J X 3 to X O i—l CO co •■-I ^ a • <0 U X 0) CO o •r-l 60 IM CD 0) l-i -.-I O 4-> S •\ cu B C •■"! 14-1 c 4-1 >> 3 c B o a c «J o - u r-< O 03 1— 1 60 4-1 ^ «0 CO -r-l 4J 1—1 O 4-1 CO o 4J 4J CO o l-i O U C CO M-l 4-> X CU X. O .-< 0) •i-l 00 CO N 3 X CO C •» 3 •i-( i-l g H <0 1 Ul CU ch l-l ••-> x: • > O co CU 60 O CO CO X • l-l •.-I U X 0) ^ co X ^ co 4-1 0) X u ex X H o CO ■« E cu X 4-> 4-1 l-l l-l -rJ 60 u CO o O !-i l-i •H 3 • 4-1 O X co s O Vj 60 O co r-l CO R CO • N— ' a CO 0) •T-l CM o •» u l-l ,>> X) m co ■U a> c 60 ~o X CO co XI --I CO C ■-I O S-i tr >-. B CD 4-1 •H CO • CO < CO 3 XI CO r-l •H O X Ul s o CO l-l CO > 1—1 B 4-1 CJ (U *N CD 3 CO 4J CO ■« T> • • CO at (J OH O w CO B X 6 H C 0) • 4J ^ O -H IJ 0) O C y-x CM g a. 4J •i-l CU CO O CO ai CO C X M X. ij a CO O CO • cu •H 60 ••> 4J ^5 ^-n a> 0) ■a U i-l CO O co S 4J i-i O O T) o =S= tI CO o • CU 4J X c C O l-i CO 4J ■H CU cu ex co b •r-l X /-> 6 CO O r-l 4J CO 3 • CU CO X) XI ss CO ~0 3 r-l QJ O •M 3 -rA 3 O "H o ~ u 6 3 U • 4-1 J-i C co •i-l XI CU CO c ,M o B co CU ex l-i •> 0) 4J r-i 3 i-i CD CO CO >N l-l ex co 4J r-l XI CO O ■— i "> 4J S i 0) D. 4J exx o co m > E >>o » r-l 1-1 r-l •r( o X O -i-l PM 4-1 ^-^ 14-1 CJ ex n y-i CU 3 60 i-l 203 w/6w o o o If) * lO I I I I S 2 o \ / I a XI u z C/> 2 - oc o z < o cr o UJ o o cr < 2 < 5 z ^ 2 cr UJ o * z 2 o _J cr < 2 in to <• ft 3 i I < < \ > ( H~ > ( - z-' v «wu CO co C U o o n V in 1 l l o CM I _ ^H O CO •- 2 3 m CO 0) • CO 1 • 1 i_, !_■ u i_i rr <-> 1 1 -tO 4-> CO CO 60 O 60 1 ^ 4) C U ^ •» 1 a g >, i-i o x o. 1 : B 3 "O co e •HZ'— 1 C <4-i ••> i-l M CO CO .-h m co X 73 -U 00 O X 60 cO"OcoGOcom U S i-l O CO cm h 4J i— i . .- i— i ; ocjcoccooiti^-^ j Z co o o V 1 >> C co c/o •> O co 1 : u i-i « t-i • ^d 1 C • >-i co O (J : cu /-^ co ex • i-i co vl<4^ O S h »J3 : — ~ u=s=a)W4-»(uco(o i-l CU CO 4J O ; <4-i4J3(uCco£-» C4-I CO 1-1 4-> -r-4 g 4J i— 1 CU CO X (0 73 C : w c i-i o co - ,3 ^ <]- 1-4 O i-4 .0 -^ ; O CN 73 C CO 4J 73 M CO CO ••> M it 41 01 O £ UH O O i-l I-l 4J 1-1 O 4J • ,0 CO CO O • co x o x j-> o 73 CD O CO 4-> i-l C CU ft T> M "W (8 ■» u >4-i a) C 60 .C D. 1 _io 4J ,C o O o co 4-1 CO CO 4J U-t &0 ■H i-l l-l 73 i-l M 3 cm . CO CO C U CO y-v ,0 i-H | i-l i— 1 > 73 h co •» a co ^ CO C CO • i-l 1-4 1 i 1 3 a. co O ■« B 1 ,-1 O W 4J • i-H CO ) co l-i >* co C O • B { 73 60 CO CO CO O 1 Q 73 >% B *• -i o c i-t i CM4Ji-l>C04JOM s o o^-'-i-i a. ,y 4-j to co «-i B C ^ B i (0 l-i co O co C / : ££•<-< .M U O .-i CO •« o 73 co o a. i— i i-i > ! o co x: o a. 4J»4-l3 ^(J4JO •» coOi-icneo>-,Oj: < 60 4-1 CO D. t4 C M 4-1 CO CU •- 14-1 •H O C 3 O • - i-l ns i4-i cu i-l i-i >, I s : C i« m xi ^ O * U j| — — COCOCO U • i— 1 o E I ; 4-1 1-4 O O O co 4-i I < ! CO 73 D. CM 4J i-l CO h i C cu i-i co •> i-i u S : y-l CO U C CO 00 i / O CO 00 CO 4-1 i-l cm CO C C 60 co B u s : C iH ** co »H O B S : o co x a. 73 i-i \ i i-l,CC0CAG4JOi-l I i : 4-1 CO i-l i-l O • \ : C0C0i- | C4-lCC • i-H \ 1 i-l C co O CO CO O 3 O 4-1 i-l 60 •> ( i 1 : B C 4-i cu •• u ■«/-> i 1 A 1 i • H^onfaOiCn -— 1 CO _D N i — 1 [h ^ COM • BENTHOS ORGANIC SHARKS C Z2 o oc ° -H 3 O C i-l CO ' 7373wBi —| CO l 4-lO0 ' CUC0Oi-l»i-i00BC L fcfc 4->ox:co^:ooocu I X* ; 4J D O "4-1 M 4J > I 22 - << - _l_ ; O73C0C0O4-I4JC0 HU£jUHi400 pLiM4->HC0C0CX>C0 z c }ia ■ a a oc » i OH m o M> I CO i a M 00 1-1 204 Table 2. Simulated annual harvests over five-year period under tested harvesting strategies. Selective Gear 3 Present Selective Utilize Bycatch 13 Increase Strategy Gear a Fishing Year 1 Pelagic fish 1,483 1,42 1,499 1,442 Shrimp 79, .41 75, .68 71, .67 74.33 Bottomf ish 2 92, .4 308, .3 288, ,3 605.7 Migratory fish 7, .52 6, .37 7, .84 6.76 Bycatch 2,047 1,079 2,018 1,060 Discards 2,046 1,079 1,009 1,060 Total landings 1,862 1,810 2,875 2,128 Year 2 Pelagic fish 1,483 1,409 1,489 1,431 Shrimp 79, .41 81, .96 59. ,08 75.25 Bottomf ish 292, ,4 312, ,1 290. ,6 612.7 Migratory fish 7, .52 3, ,73 8. ,68 4.72 Bycatch 2,047 1,092 2,034 1,072 Discards 2,046 1,092 1,017 1,072 Total landings 1,862 1,807 2,864 2,124 Year 3 Pelagic fish 1,483 1,417 1,479 1,436 Shrimp 79. ,41 81. ,20 57. ,74 74.61 Bottomf ish 292. ,4 311. ,3 292. ,5 612.2 Migratory fish 7. ,52 2. ,251 8. ,74 3.30 Bycatch 2,047 1,090 2,048 1,071 Discards 2,046 1,089 1,023 1,071 Total landings 1,862 1,812 2,861 2,126 Year 4 Pelagic fish 1,483 1,415 1,481 1,435 Shrimp 79. ,41 79. ,92 58. ,27 73.34 Bottomf ish 292. ,4 312. ,2 2 92. ,2 613.8 Migratory fish 7. ,52 1. ,38 8. ,58 2.32 Bycatch 2,047 1,093 2,046 1,074 Discards 2,046 1,092 1,023 1,074 Total landings 1,862 1,808 2,863 2,124 Year 5 Pelagic fish 1,483 1,415 1,481 1,435 Shrimp 79. ,41 79. ,37 58. ,12 72.67 Bottomf ish 292. 4 312. ,5 292. ,2 614.6 Migratory fish 7. ,52 ■ ,84 8. ,49 1.6 Bycatch 2,047 1,094 2,0 46 1,076 Discards 2,046 1,093 1,023 1,075 Total landings 1,862 1,807 2,863 2,124 Reducing catchability of shrimp trawls for groundfish by 50 .percent. c Utilizing 50 percent of groundfish bycatch. Increasing directed fishing effort on groundfish by 50 percent, 205 Table 3. Total annual respiration for organic compartments under different harvesting strategies. Selective Gear a Present Selective Utilize Increase Compartment Strategy Gear 3 Bycatch Fishing c Year 1 a a a a Phytoplankton 3.11x10* 3.08x10; 3.09x10; 3.12x10; Low-N Organic 1.47x10° 1.47x10* 1.47x10; 1.47x107 High-N Organic 2.53x10; 3.00x10* 2.75x10* 2.18x10* Zooplankton 1.68x10; 2.24x10; 2.03x10; 1.56x10; Pelagic Fish 2.14x10* 2.05x10* 2.08x10* 2.14x10* Benthos 2.33x10:? 2.33x10:? 2.32x10:? 2.32x10:? Shrimp 1.18x10; 1.13x10; 1.11x10; 0.88x10; Bottomfish 1.30x10* 1.37x10* 1.34x10* 1.29x10* Migratory Fish 7.00x10^ 5.93x10* 6.29x10?: 8.07x10* Marine Mammals 3.28x10:* 5.21x10^ 4.54x10^ 2.99x10:? Large Scavengers 5.96X10 1 5.93X10 1 5.93X10 1 5.82X10 1 Year 2 Phytoplankton 3.11x10* 3.07x10* 3.08x10* 3.11x10* Low-N Organic 1.47x10° 1.47x10; 1.47x10; 1.47x10* High-N Organic 2.53x10^ 3.32x10* 3.02x10* 2.33x10* Zooplankton 1.68x10* 2.52x10* 2.26x10* 1.70x10* Pelagic Fish 2.14x10* 2.03x10* 2.06x10* 2.13x10* Benthos 2.33x10^ 2.34x10:? 2.33x10^ 2.34x10^ Shrimp 1.18x10; 1.22x10; 1.12x10; 0.86x10; Bottomfish 1.30x10* 1.38x10* 1.36x10* 1.30x10* Migratory Fish 7.00x10* 3.47x10* 4.39x10* 8.13x10* Marine Mammals 3.28x10^ 5.69x10^ 4.98x10^ 3.31x10^ Large Scavengers 5.96X10 1 5.85X10 1 5.84X10 1 5.73x10* Reducing catchability of shrimp trawls for groundfish by 50 .percent. Utilizing 50 percent of groundfish bycatch. Increasing directed fishing effort on groundfish by 50 percent. 206 shrimp by a factor of 100. Simulation results indicated that shrimp standing stock and the shrimp harvest were more sensitive to the reduction in bottomfish fishing mortality rate caused by trawls with a reduced efficiency for catching fish when predator selectivity against shrimp was eliminated. When there was no selectivity against shrimp by bottomfish (weighting factors set at 1 and predation rate increased by a factor of 100), the shrimp stock did not fully recover in the 5-year period of the simulation (fig. 6). In a second sensitivity test, the initial standing stock of bottomfish was reduced by one-half, which changed the relationship of standing stock of this compartment to that of the others. This change made little difference to the response of the model system to management tests, except that, under conditions that induced expansion of bottomfish standing stock, the expansion occurred more rapidly. For the third sensitivity test, a feeding link was established between discards and marine mammals. Discards were weighted equally with other mammal food sources. This change made almost no appreciable change in simulation results, except that the oscillation in marine mammals that typified the management-test simulations was eliminated from the simulation of the effect of utilizing half the bycatch. The simulation plots revealed oscillations in many of the standing stocks under non-steady-state conditions. These oscillations were intrinsic to the model system and not due to oscillating system inputs, because all inputs to the model system were constants. The frequency of oscillation of the marine mammal standing stock (fig. 3) is obviously an artifact of the model and cannot reflect the real world, because marine mammals have slow maturation rates and long gestation periods and the standing stock of marine mammals obviously would not fluctuate several times a year. Fluctuations such as those seen in other standing stocks of the model system are plausible, although, in the real world, seasonal variations in solar radiation and river discharge would superimpose seasonal cycles on any intrinsic cycles of the system. DISCUSSION To help understand the basis for the changes observed in the plots, calculations were made of total annual inflows and outflows to several of the compartments under the test conditions. These annual totals are shown in tables 4 through 12. Model simulation results indicate that reducing discards by either proposed strategy could have some detrimental effect on shrimp harvests initially, but readjustments in the system will allow the shrimp stock and shrimp harvest to recover within 2 years under the option of using trawls that reduced fish catchability relative to that of shrimp, if predation pressure of bottomfish on shrimp is low (selectivity weighting factor against shrimp of 0.01 or less). The simulations of the model suggest that bottomfish harvesting strategies could influence shrimp by more mechanisms than those considered in the question originally posed in the introduction. Some of the influences observed in the simulations were counter to what was expected. Although nitrogen remineralization was decreased when the discard rate was reduced through bycatch utilization, the rate of nitrogen remineralization was greatest when bottomfish fishing mortality was reduced by one-half through the 207 tu/buj ,Ui/blU 2 O - m -■«■ _L o. 1 1 <_> 2 cr Li- U. z < i >- cr o o cr o s z o: 1 o * - ro - <\J Z UJ -J o < O 2 Z 5 UJ z cr < O m 2 9 o o _ * to cj 2 O _J I I L I \ ( ( \\ 1L l - m - < - to V) cr < - CJ 8 1 1_ ^ o K o z u CD (OIZZ *co oo <2 ** crt i z o z << I ' -J O o-Q- m OO oi- X a. ■ H •r-l X o •H Ol 43 H C 4J C o 00 go rJ o I ■r-l u o X) 3 oc CO (4-1 4-1 • CO CD O CO rH O o ^-s M rH rH rH N.^ CO CO 0) CO 0) >>u-i •o rQ CJ • n CL, CO o E CO rH rH B 00 •r-l 3 O • M CO 0) e h Z CO • 1-4 P •r-l ■H 0) 1-4 o • p o 4-1 a. o CO •rl rC • 4-1 •\ E Un oo ►, CO c c •\ U-l •H CO r4 01 o c cu o> a; X> CO E 4-1 o c 3 1 CU 4-1 r^ 4-1 1-1 A >— ' en >. u c M rJ o rH CO CO a CO «j Q- V-*' 01 a. rH CO E TO e > E an o •r-l CO •rl o o a. • #* u r* CU CJ 4-1 1—1 x: ja 0J >> o CO 00 0) rJ rC rC N •\ •H 3 O CJ c X U-l 4-1 ■r-l CO ■ #• CO CO rO •\ CU >n rH i-l X C (0 rH U-l 4-1 ■H 4-1 M "4-1 o •s •H CO c 0) O > 1— 1 !* 3 Ml 0) 01 o CO l-i CO CO 3 -M •rl e CU 1-1 y •\ u 4-1 a) X )-i XI o C 01 CO 4-1 CO a 4-1 01 4-1 1-1 co •r-l Ol o CO 00 CO oo >s UH u CN o E i-i in -a i-l 00 u E E CO c 4-1 u i— i a •r-l CO •r4 •H •H • M 0) o X a ■a a c rH ■a i-i CO CO a CO o • o 00 Q. co oo • rH r-l CO 4-1 •• u o U-l CO CO CO •\ U-l o 4-1 C 3 •V ^— \ O E cu o a rC CO r^ o •rl rC rH 01 CO I-l oo 4-1 N 4-1 H rH 00 •r-l 01 ^ •rl •H CO o UH 00 o > rJ rH CM u E C o •rl o 3 • 4-1 o cu 4-1 4-1 -C E r« CO 1-1 4J > L0 CJ •rl CJ CO c 4J CO rH I-l en TJ u W c 4-1 o •N s — • a 0) CO CO CJ o C- 14-1 CO X> 0) 4-1 CO o E CO o £ >. CO U-l • ■H r* CiH 3 o rJ u c rH r5 r-l 4-1 00 rC CO o CO 0) C c -\ CO rC •I-l x X> 01 •H rH CO 4-1 •• O CO r-l CO • •• en 0) W E Ol CO •r-l en • n i— i e H u CJ u O rH 3 o O CO a. en cu o Z cu 4-1 • »\ •H >^ C H co o •^ 00 -Q •rl • E DO • 0) CU •\ M "O T> r"*S CO 4-1 4-J CJ CO O (1) CU O Jd CO CO ■rl 4J •u O O CU C c C X CO 4-> 3 r-l 01 •H •rl CO 4-1 X3 o ■u 3 X) X) oo c CD rH 01 >> -H u u u CU l-i fc r4 rQ rQ o o o rO Q- CU r4 D 60 208 Table 4. First-year inorganic nitrogen budget. Management Strategies Selective Utilize Present Gear Bycatch IHEIIXS Imported (J(D) 453,600 453,600 453,600 Recycled (N(3-12)) 85,491 85,662 85,399 OUTPUTS Exported (FLN) 85,919 85,920 85,878 Uptaken (P(l,2)) 39,895 40,065 39,842 Units are milligrams nitrogen per square meter per year, 2 no Table 5. First-year energy budget for phytoplankton. Management Strategies Selective Utilize Present Gear Bycatch IHPI1IS Gross primary production (J (2) ) 975,993 975,983 975,719 OUTPUTS Grazing by zooplankton (P(2,5)) Phytoplankton rain (P(2,3)) Respiration (R(2) ) 33,372 Grazing by pelagic fish (P(2,6)) 50,527 581,028 311,051 44,081 47,951 575,739 308,211 30,637 51,168 582,218 311,676 Units are milligrams dry organic matter per square meter per year. 210 Table 6. First-year energy budget for high-nitrogen organic compartment. 3 Management Strategies Selective Utilize Present Gear Bycatch Fecal pellet rain (F(5)) Discarded fish (D) 10,012 2,046 13,224 1,079 9,191 1,009 OUTPUTS Ingestion by benthos (P(4,7)) Ingestion by shrimp (P(4,8)) Ingestion by groundfish (P(4,9)) Respiration (R(4)) COMPETITION RATIOS P(4,8)/P(4,7) P(4,8)/P(4,9) 9,436 12.99 11,179 14.65 81.47 102.0 2,527 2,995 7,964 10.07 68.78 2,163 .001377 .1594 .001310 .1436 .001260 .1464 Units are milligrams dry weight per square meter per year, 211 Table 7. First-year energy budget for zooplankton compartment. Management Strategies Selective Utilize Present Gear Bycatch INPUTS Ingestion of phytoplankton (P(2,5)) 33,372 44,081 30,637 OUTPUTS Predation by pelagic fish (P(5,6)) Respiration (R(5)) Egestion (fecal pellet production) (F(5)) 10,012 13,224 9,191 RKLLQ£ P(2,5)/R(5) P(5,6)/R(5) P(2,5)/P(5,6) a Units are milligrams dry weight per square meter per year, 6,553 8,353 6,066 16,808 22,421 15,400 1.985 1.966 1.989 .3899 .3726 .3939 5.093 5.277 5.051 212 Table 8. First-year energy budget for pelagic forage fish compartment. 3 Management Strategies Selective Utilize Present Gear Bycatch iUPHT_£ Ingestion of phytoplankton (P(2,5)) 50,527 47,951 51,168 Ingestion of zooplankton (P(5,6)) 6,553 8,353 6,066 OUTPUTS Harvesting (H(6)) 1,483 1,420 1,499 Predation by migratory pelagics (P(6,10)) 68.35 55,54 71.95 Predation by marine mammals (P(6,1D) 2,208 3,351 1,900 Predation by large scavengers (P(6,12)) 39.64 37.76 39.71 Respiration (R(6)) 21,355 20,450 21,581 Egestion (F(6)) 31,929 31,153 32,104 a Units are milligrams dry weight per square meter per year, 213 Table 9. First-year energy budget for marine mammals compartment. 3 Management Strategies Selective Utilize Present Gear Bycatch IHEQIS Ingestion of pelagic fish (P(6,ll)> Ingestion of groundfish (P(9,1D) Ingestion of migratory pelagics (P(10,1D) OUTPUTS Predation by large scavengers (P(ll,12)) Respiration (R(1D) 2,208 1,489 24.12 3,351 2,499 31.96 1,900 1,251 21.35 .00082 .00139 .000745 3,279 5,209 2,792 Units are milligrams dry weight per square meter per year. 214 Table 10. First-year energy budget for benthic compartment. 3 Management Strategies Selective Utilize Present Gear Bycatch inputs Ingestion of low-N organic (P(3,7)) 494,859 492,884 488,868 Ingestion of high-N organic (P(4,7) 9,436 11,179 7,964 OUTPUTS Predation by shrimp (P(7,8)) 3,112 2,960 2,768 Predation by groundfish (P(7,9)) 19,523 20,557 18,974 Respiration (R(7)) 233,272 232,904 229,962 Egestion (F(7)) 248,356 247,560 245,231 COMPETITION RATI O P(7,8)/P(7,9) .1594 .1440 .1459 Units are milligrams dry weight per square meter per year. !15 Table 11. First-year energy budget for shrimp compartment. 3 Management Strategies Selective Utilize Present Gear Bycatch Ingestion of low-N organic (P(3,8)) 85.12 80.93 76.98 Ingestion of high-N organic (P(4,8)) 12.99 14.65 10.07 Ingestion of benthos (P(7,8)) 3,112 2,960 2,768 OUTPUTS Harvesting (H(8)) 79.41 75.68 71.67 Predation by groundfish (P(8,9)) 1.938 1.947 1.725 Predation by migratory pelagics (P(8,10)) 1.830 1.478 1.716 Respiration (R(8)) 1,184 1,128 1,068 Egestion (F(8)) 1,943 1,850 1,729 Units are milligrams dry weight per square meter per year. 216 Table 12. Second-year energy budget for shrimp compartment. 3 Management Strategies Selective Utilize Present Gear Bycatch INPUTS Ingestion of low-N organic (P(3,8)) Ingestion of high-N organic (P(4,8)) Ingestion of benthos (P(7,8)) 3,112 3,215 2,303 OUTPUTS 85.12 87.37 63.42 12.99 17.60 8.32 Harvesting (H(8)) 79.41 81.96 14.89 Predation by groundfish (P(8,9)) 1.938 2.135 1.433 Predation by migratory pelagics (P(8,10)) 1.830 .9343 1.570 Respiration (R(8)) 1 r 184 1 ,222 880.6 Egestion (F(8)) 1 ,943 2 ,010 1,437 a Units are milligrams dry weight per square meter per year. 217 use of shrimp trawls with a reduced catch efficiency for fish, despite a reduction in the rate that dead fish were returned to the system to close the nutrient cycle (table 4). Annual gross primary productivity was slightly decreased when half the bycatch was utilized, decreasing the quantity of dead fish returned to the model system. However, annual gross primary productivity did not increase when trawls that reduced fish catchability relative to that of shrimp were used to decrease discards, despite the fact that nutrient regeneration rates were highest under this condition (table 5). Perhaps this was because saturation concentrations for nitrogen with respect to phytoplankton photosynthesis were more frequently exceeded under the condition of half the shrimp trawl efficiency for catching fish than under the present strategy for handling discards. The direct effect of discard rate on the standing stock of high-nitrogen organic material was small, and the rate of zooplankton fecal pellet deposition was so great by comparison that it overshadowed discarding as a source of high- nitrogen organic material (table 6). Although decreasing shrimp trawl efficiency for catching fish resulted in a decrease in the rate of deposition of dead fish, this was greatly outweighed by an increase in zooplankton fecal pellet production that was an indirect result of decreasing shrimp trawl efficiency for catching fish. Both the rate of inflows and rate of outflows to the zooplankton compartment were influenced by the discard rate and the bottomfish fishing mortality rate (table 7). Predation of pelagic forage fish on zooplankton was decreased when the standing stock of pelagic forage fish was depressed by predation of marine mammals , which appears to have been due to the greater abundance of bottomfish, which are also their prey (table 9). The rate of fecal pellet production increased when zooplankton standing stock increased. Although predation by bottomfish did exert some detrimental influence on shrimp stocks and shrimp harvests, this influence was minor when compared to the influence of competition for common food between shrimp and bottomfish and between shrimp and benthos (tables 10 and 11). An increased availability of food for shrimp outweighed the increased competition for that food in the second year (table 12). How the shrimp stock in the model system reacted to a change in bottomfish standing stock was determined by the balance between production of fecal pellets, which provide food for shrimp, benthos, and bottomfish, and the pressure of competition for food with both benthos and bottomfish. Initially, when the fishing mortality of bottomfish was reduced with shrimp trawls with half the efficiency for catching fish, shrimp standing stock and the shrimp harvest reacted negatively. However, shrimp standing stock and the shrimp harvest recovered when the supply of fecal pellets increased to the point where an increased food supply overcame increased competition for food from the larger standing stock of bottomfish. Under any management strategy, dead fish formed a small component of the high-nitrogen organic compartment, relative to zooplankton fecal pellets, which represented 82.5 percent of compartment standing stock at steady-state. Furthermore, the nitrogen released from this compartment in decomposition was infinitesimal compared to that released in the decomposition of low-nitrogen organic material, and it was even small compared to that released in animal excrement. Nevertheless, utilizing the bycatch, which decreased the rate of discarding without increasing the standing stock of living bottomfish, did affect shrimp stocks and shrimp harvests in the model system. Conclusions from the simulations differ in this detail from conclusions of Browder (1981) and 218 Sheridan et al. (1981). The conclusion made in the 1981 paper was that shrimp production would not be affected by decreasing the discard rate. This conclusion was based only on relative rates of nitrogen remineralization from different sources, and food flows to shrimp from different sources. In the simulations, the response of phytoplankton production to changes in water nitrogen concentration was dependent upon the Michaelis-Menten half-saturation constant, relative to the concentration of nitrogen in the water, rather than on the relative rates of release of nitrogen from the various compartments. The simulation exercise demonstrated that it is possible for a change in a nitrogen inflow that is very small relative to other flows to affect model results. The model, however, does not include nitrogen fixation and denitrif ication. These flows are difficult to quantify. If they are large in this system relative to the nitrogen flows included in the model, they may influence the way that variations in other nitrogen flows affect the system. Model results were sensitive to the weighting factor determining the intensity of bottomfish predation on shrimp. Model results appeared somewhat sensitive to the apportionment of nitrogen between the excrement and fecal material of bottomfish and possibly also of zooplankton and other animals. Results are undoubtedly also sensitive to the Michaelis-Menten half -saturation constant that relates phytoplankton production to the concentration of nitrogen in the water. Although values for these parameters that were used in the model had some basis in the literature, none were specific to the north-central Gulf of Mexico. No specific sensitivity tests have yet been run on these parameters. The impact of variations in these values on model results and conclusions has not been fully explored. Do model results apply to the real-world system? In the development of an ecosystem model, there are at least four levels of procedure at which variations can affect model results: (1) assumptions of the conceptual model, (2) translation of the conceptual model into mathematical equations, (3) model quantification, and (4) model computerization. Ecosystem models are difficult to validate, but there are a number of things that can be done to improve confidence in an ecosystem model and its results. One is to test the validity of the assumptions independently of the model and to assess the dependence of the model on these assumptions. A second is to test the sensitivity of the model to variation in initial values, rate-coefficients, and other constants, such as those singled out above (bottomfish selectivity weighting factor for shrimp relative to alternative prey, apportionment of waste nitrogen between feces and excrement, and Michaelis-Menten half-saturation constant relating gross primary productivity to water nitrogen concentration). Quantification of a model requires many rough estimations. It is important to know whether any of these have major effects on model results, not only to evaluate the present reliability of the model but also to determine what further work is needed to improve the model's reliability. Sensitivity testing is done by varying values one at a time and comparing simulation results. Sensitivity test results can lead to a more intensive search of existing literature, which may yield improved estimates, or to field or laboratory studies designed to obtain the critical information that is needed to adequately answer the question. Sensitivity testing is needed not only for model input values but also for model structure. The model is, of necessity, a simplification. Does the addition of structural detail significantly alter system results? If so, this needs to be known. Finally, errors in program coding can cause responses that 219 cannot possibly reflect the real-world situation. Working with a model over a period of time increases the likelihood that errors such as these will be found. A model is a simplistic interpretation of the system, whether or not this model reflects the response of the real system to the conditions we are testing remains to be seen. It is certain, however, that even a simple system such as this 12-compartment model has complex and unanticipated reactions to harvesting pressure and its variation, due to multiple indirect effects operating along several pathways. It is therefore likely that the response of the real system to harvesting pressure and its variation is even more complicated and difficult to predict. This is the important message for management of the modeling effort at this point. REFERENCES Bennett, R.H. and D.N. Lambert, 1971. "Rapid and reliable technique for determining unit weight and porosity of deep-sea sediments," Mar . Geol . 11:201-207. Bishop, J.M., J.G. Gossellink, and J.H. Stone, 1980. "Oxygen consumption and hemolymph osmolality of brown shrimp, Penaeus aztecus , " Fish . Bull . , U.S. 78:741-757. Brett, J.R. and J.M. Blackburn, 1978. "Metabolic rate and energy expenditure in the spiny dogfish Squalus acanthias ," J_. Fish . Res . Board Can . 35:816-821. Brill, R.W. , 1979. "The effect of body size on the standard metabolic rate of skipjack tuna, Katsuwonus pelamis ," Fish . Bull . , U.S. 77:494-498. Browder, J. A., 1981. "Use of an energy flow model to evaluate alternative harvesting strategies in a multispecies fishery," p. 571-586. In W.J. Mitsch, R.W. Bossman, and J.M. Klopatek (eds.), Energy and Ecological Modeling , Elsevier, N.Y. Conover, R.J., 1959. "Regional and seasonal variation in the respiratory rate of marine copepods," Limnol . Oceanogr . 4:259-269. Darnell, R.M. and T.E. Wissing, 1975. "Nitrogen turn-over and food relationships of the pinfish Lagodon rhomboides in a North Carolina estuary, p. 81-110," In F.J. Vernberg (ed.), Physiological Ecology of Estuarine Organisms , University of South Carolina Press, Columbia, S.C. Eppley, R.W. and W.H. Thomas, 1969. "Comparison of half-saturation constants for growth and nitrate uptake of marine phytoplankton, " J_. Phycol . 5:375-379. Griffin, W.L. and J. P. Warren, 1978. "Costs and return data; groundfish trawlers of northern Gulf of Mexico," unpublished report fo Gulf of Mexico Fishery Management Council, Tampa, Fla. (Available from Gulf of Mexico Fishery Management Council, Lincoln Center, Suite 881, 5401 West Kennedy Blvd., GMFMC (Gulf of Mexico Fishery Management Council). Unpubl. Draft fishery management plan for groundfish in the Gulf of Mexico. Gulf of Mexico Fishery Management Council, Tampa, Fla. 233 p. (Available from Gulf of Mexico Fishery Management Council, Lincoln Center, Suite 881, 5401 West Kennedy Blvd., Tampa, Fla.) GMSAFMA (Gulf of Mexico and South Atlantic Fishery Management Councils), 1980. Fishery Management Plan, Final Environmental Impact Statement, and Regulatory Analysis for the Coastal Migratory Pelagic Resources. Gunter, G. , 1963. "The fertile fisheries crescent," J_. Miss . Acad . Sci. 9:286-290. Happ, G., J.G. Gosselink, and J.W. Day, Jr., 1977. "The seasonal distribution of organic carbon in a Louisiana estuary," Estuarine Coastal Mar . Sci . 5:695-705. 220 Hausknecht, K.A. , 1980. "Description of surficial sediments and suspended Tampa, Fla.) particulate matter," Vol. V. In W.B. Jackson and G.M. Faw (eds.), "Biological/ chemical survey of Texoma and Capline sector salt dome brine disposal sites off Louisiana, 1978-79," NOAA Tech . Memo . NMFS SEFC-20, 56 pp. (Available from NTIS, Springfield, Va.). Hettler, W.F., 1976. "Influence of temperature and salinity on routine metabolic rate and growth of young Atlantic menhaden," J_. Fish . Biol . 8:55-65. Hoss, D.E. , 1974. "Energy requirements of a population of pinfish Lagodon rhomboides (Linnaeus)," Ecology 55:848-855. Irving, L. , P.F. Scholander, and S.W. Grinnell, 1941. "The respiration of the porpoise Tursiops truncatus ," J_. Cell . Comp . Phys io 1 . 17:145-168. Johannes, R.E. and M. Satomi , 1966. "Composition and nutritive value of fecal pellets of a marine crustacean," Limnol. Oceanogr . 11:191-197. Jones, R.R. , 1973. "Utilization of Louisiana estuarine sediments as a source of nutrition for the brown shrimp Penaeus aztecus ," Ph.D. Dissertation, Louisiana State Univ. Baton Rouge, La. 130 pp. Leatherwood, S. , J.R. Gilbert, and D.G. Chapman, 1978. "An evaluation of some techniques for aerial census of bottlenose dolphins," J^. Wildl . Manage . 42:239-250. NMFS (National Marine Fisheries Service), 1978. "Fishery Statistics of the United States 1975," Statistical Digest No. 69. U.S. Dept. Commer. , NOAA, Washington, D.C. 418 pp. Odum, H.T. , 1982. "Systems," J. Wiley, N.Y. Pamatmat, M.M. , 1980. "Facultative anaerobiosis of benthos," pp. 69-92. In K.R. Tenore and B.C. Coull (eds.), Marine Benthic Dynamics . Univ. of South Carolina Press, Columbia, S.C. Parker, R.H. , A.L. Crowe, and L.S. Bohme, 1980. "Description of living and dead benthic (macro- and meio-) communities," Vol. I In W.B. Jackson and G.M. Faw (eds.) "Biological/chemical survey of Texoma and Capline sector salt dome brine disposal sites off Louisiana, 1978-79," NOAA Tech . Memo . NMFS-SEFC-25. 103 pp. (Available from NTIS, Springfield, Va.). Parsons, T.R. , M. Takahashi, and B. Hargrave, 1977. Biological Oceanographic Processes , Pergamon Press, N.Y. , 332 pp. Patella, F., 1975. "Water surface areas within statistical subareas used in reporting Gulf coast shrimp data," Marine Fisheries Review 37( 12) :22-24. Perret, W.S. , B.B. Barrett, W.R. Latapie, J.F. Pollard, W.R. Mock, G.B. Adkins , W.J. Gaidry, and C.J. White, 1971. Cooperative Gulf of Mexico Estuarine Inventory and Study , Louisiana , Phas e I , Area Description . Louisiana Dept. Wildl. Fish, 175 pp. Reitsema, L.A., 1980. "Determiination of seasonal abundance, distribution, and community composition of zooplankton," Vol. II. In W.B. Jackson and G.M. Faw (eds.). "Biological/chemical survey of Texoma and Capline sector salt dome disposal sites off Louisiana, 1978-79," NOAA Tech . Memo . NMFS-SEFC- 25. 103 pp. (Available from NTIS, Springfield, Va.). Ricker, W.E. , 1975. "Computation and interpretation of biological statistics of fish populations," Bull . Fish . Res . Board Can . 191, p. 382. Rivas , L.R. (Unpubl.) "Estimates of biomass for pelagic and coastal sharks in the Gulf of Mexico," Southeast Fisheries Center, National Marine Fisheries Service, Miami, Fla. Ryther, J.H. and R.R.L. Guillard, 1962. "Studies of marine planktonic diatoms, III. Some effects of temperature on respiration of five species," Can . J. Microbiol. 8:447-453. 221 Sackett, W.M. , 1972. "Chemistry," In S.Z. El-Sayed et al . (eds.). "Chemistry, primary productivity, and benthic algae of the Gulf of Mexico," Folio 22, Serial Atlas of the Marine Environments . American Geographical Society, N.Y. Sheridan, P., J. A. Browder, and J. Powers, 1981. In B. Rothschild and J. Gulland (eds.). Proc. Workshop on the Scientific Basis of the Management of Penaeid Shrimp , November 1981, Key West, Fla. Sidwell, V.D. , 1981. "Chemical and nutritional composition of finfishes, whales, crustaceans, mollusks, and their products," NOAA Tech Memo NMFS F/SEC-11. U.S. Department of Commerce. Southeast Fisheries Center, Miami, Fla. 432 pp. Sklar, F.H. , 1976. "Primary productivity in the Mississippi delta bight near a shallow bay estuarine system in Louisiana," Ph.D. Dissertation, Louisiana State Univ. Baton Rouge, La., 96 pp. Strickland, H.D.H. , 1960. "Measuring the production of marine phytoplankton," Fish . Res . Bd . Canada Bull . 122:172. Tietjen, J.H. , 1969. "The ecology of shallow-water meiof auna in two New England estuaries," Oceologia 2:251-291. 222 MATHEMATICAL MODEL OF OXYGEN DEPLETION IN THE NEW YORK BIGHT: AN ANALYSIS OF PHYSICAL, BIOLOGICAL, AND CHEMICAL FACTORS IN 1975 AND 1976 Andrew Stoddard and John J. Walsh Brookhaven National Laboratory Division of Oceanographic Sciences Upton, NY 11973 223 INTRODUCTION The increasing use of the sea for urban waste disposal requires an understanding of the subtle biological impacts of nutrients, organic matter, and other pollutants discharged into coastal waters. Significant water quality problems frequently result from the transient response of natural waters to some perturbation such as changes in circulation patterns, climatic conditions, nutrient enrichment, organic carbon loading, or phytoplankton species composition. In particular, an extensive bloom of the dinof lagellate Ceratium tripos developed throughout the Middle Atlantic Bight continental shelf from January through July 1976. From July through September 1976 bottom water oxygen was also progressively depleted to less than 2 ml 02^"* over a large area (8600 km ) of the New Jersey coastal region from the 20 m to 60 m isobath (fig. 1). Presence of this anoxic region resulted in the formation of hydrogen sulfide and mass mortalities of demersal fishes and shellfish (Swanson and Sindermann 1979). The occurrence of the C. tripos bloom, and the onset of anoxic conditions, suggested that the abundance of C_. tripos had imposed an unusually large oxygen demand on the subpycnocline layer off the New Jersey coast and was causally related to the anoxic episode (Malone et al. 1979; Falkowski and Howe 1976). A marine ecosystem model, designed to evaluate the relative significance of natural physical and biological processes and anthropogenic waste inputs on oxygen depletion in the New York Bight, described the continental shelf of the Bight as a two-layer system separated by the pycnocline (Stoddard 1983). The model equations specified the interactions of carbon, oxygen, and nitrogen in an analysis of oxygen depletion, nutrient dynamics, and phytoplankton distribution during the stratified summer season. A comparison of a year of high bottom oxygen content (1975) and the large- scale anoxic episode during 1976 is the problem setting for the analysis. Previous studies of the 1976 anoxic episode (Falkowski et al . 1980; Swanson and Sindermann 1979; Falkowski and Howe 1976) examined the influence of urban waste inputs, and climato logical , physical, chemical, and biological forcing on oxygen depletion in the New York Bight. The model quantitatively incorporates the hypotheses presented in these studies coupled with a quasi-time-dependent circulation sub-model (Han et al. 1980) to assess the relative significance of the various forcing terms on the development of anoxia. RESULTS Verification of the model included a synthesis of physical, chemical, and biological data collected by Brookhaven National Laboratory and other institutions during the MESA New York Bight Project. In general, the model is capable of simulating most of the ecological behavior of the Bight where the calculations are in reasonable agreement with observations during a year of high bottom oxygen (1975) (fig. 2), and during the anoxic episode in 1976 (fig. 3). The significance of transport processes on water quality distributions in the Bight are clearly demonstrated in an analysis of: the component sources and sinks of nitrogen, Ceratium , and dissolved oxygen for the 1976 verification case; and onshore subpycnocline penetration of high nitrate across the shelf during June - October for the 1975 verification case. The model also clearly demonstrates the influence of the flow field reversal on the accumulation of Ceratium and other particulate substances during June - July 1976 (fig. 4). 225 TEMPORAL PROGRESSION OF ANOXIA NEW YORK BIGHT Figure 1. Temporal progression of anoxia in the New York Bight: July- October 1976 226 10.0 8.0 6.0 z UJ o >- X o Q UJ > _l o CO CO COMPARISON OF MODEL RESULTS TO OBSERVED DATA DISSOLVED OXYGEN Z 4.0 CM O 4 2.0h 0.0 1 NJ MIDSHELF 30-60m ABOVE PYCNOCLINE MAY 1975 # OBSERVED MEAN MODEL MEAN u.u 1 i i i i BELOW PYCNOCLINE 8.0 - 6.0 'j x ^m ^^^^*- ll - 1 %Cr)r^rvv^~^">'^^ 4.0 — L ^^^-MXOCXXC)OOCL^-^a-xa^3CXJ^^- 2.0 - - 00 i B 1 1 1 1 JUN JUL AUG SEP OCT NOV Figure 2. Temporal comparison of calculated and observed dissolved oxygen for the New Jersey midshelf : 1975. 227 10.0 COMPARISON OF MODEL RESULTS TO OBSERVED DATA DISSOLVED OXYGEN Ld o 0.0 x 10.0 o UJ > 8.0 _i o t/> 5 " 4.0 T ~l 1 1 ABOVE PYCNOCLINE NJ MIDSHELF 30-60m _L 125 100 75 -50 25 o < 2.0 0.0 BELOW PYCNOCLINE ■N 4 'c^^^ # OBSERVED MEAN MODEL MEAN * — MODEL SATURATION J_ ± J_ B 25 < CO UJ l00 o > x O 75 50 25 MAY JUN JUL AUG SEP OCT NOV 1976 Figure 3. Temporal comparison of calculated and observed dissolved oxygen for the New Jersey midshelf : 1976. 228 I o 4. >- X a. o cr o _i x o COMPARISON OF MODEL RESULTS TO OBSERVED DATA CHLOROPHYLL 5 4 - NJ MIDSHELF 30-60 m ? • + ABOVE PYCNOCLINE OBSERVED MEAN CHL OBSERVED NET CHL MODEL NET + NANO CHL MODEL NET CHL b 1 1 1 i i BELOW PYCNOCLINE 1 ~^^\ 4 /! { 3 - /' \ - *»\ s> Vy 2 - V" k vv _ \\ \^o > I _ \ + - N ■*• -^_ X ^ + ^ B i i i - .- - MAY 1976 JUN JUL AUG SEP OCT NOV Figure 4. Temporal comparison of calculated and observed chlorophyll for the New Jersey midshelf : 1976. 229 The occurrence of anoxia in 1976 resulted from temporal sequence of climatological , physical, and biological processes that were significantly different in 1976 in comparison to 1975, or other years of record. Analysis of the observed and computed components of the midshelf oxygen budget (fig. 5) demonstrates the interaction of physical transport mechanisms and biological/ chemical processes on the cross-shelf and seasonal distribution of dissolved oxygen in 1976. The dominant physical and biological factors influencing the observed spatial and temporal distributions of oxygen included: (1) Circulation regime over the New Jersey shelf with upwelling predominant from May-July and downwelling occurring in August. (2) Cross-shelf vertical distribution of viable Ceratium populations and the balance between photosynthetic oxygen production and respiratory oxygen demands, i.e., compensation depth, in relation to the anomalously deep pycno- cline during May-August. (3) Decline of the Ceratium bloom and decomposition of the additional organic carbon load within the water column and on the seabed resulted in the onset, and progression, of anoxia over the New Jersey shelf during July-September, (4) Increased wind mixing, surface cooling, and erosion of the pycnocline resulted in replenishment of oxygen during late September-October above, and below, the pycnocline with the isothermal vertical dispersion coefficient ( 10- 15 cm^sec - *) estimated from the observed data and a one-dimensional, vertical calculation for the New Jersey Midshelf. Oxidation of particulate organic carbon derived from sewage effluents represented a negligible component (1%) of water column oxygen consumption over the New Jersey midshelf in 1976. Even within the Apex region, where the ocean dump sites are located and sewage carbon concentrations are higher, the oxygen demand from sewage-derived materials represented a minor component (4%) of oxygen consumption below the pycnocline. This suggests that the major impact of carbonaceous waste discharges is limited to the vicinity (30 km) of the discharge location, i.e., a relatively localized process. The results of the oxygen budget analysis demonstrate the necessity of including a kinetic process in the model that accounts for the conversion of phytoplankton biomass to nonliving organic detritus in the water column (O'Connor et al. 1981). Incorporation of such a mechanism would then account for the distribution of particulate organic carbon observed in August-September 1976 with mineralization of the observed biomass accounting for the oxygen depletion rate observed during July-September 1976 over the midshelf area off New Jersey. MODEL PROJECTION OF APEX WATER QUALITY The validated model was used to evaluate the impact of a ten-fold increase in present urban carbon and nitrogen loading (Mueller et al . 1976) on eutrophi- cation and oxygen depletion in the Apex (Segar and Berberian 1976; Garside et al . 1976). Steady-state August distributions are used to examine the water quality response in the Apex to present and increased anthropogenic loading. Substantial increases in sewage sludge carbon diminished light penetration near the ocean dumpsites and reduced primary productivity. In the outer Apex, increased nitrogen abundance increased primary productivity slightly over an 230 10.0 COMPONENT SOURCE/SINKS OF DISSOLVED OXYGEN BELOW PYCNOCLINE T NJ MIDSHELF 30-60 m SOURCES I -lateral advection 2-downwelling 3-lateral diffusion 4-vertical diffusion 6-nanopl photosynthesis 12- C.tripos photosynthesis NOV Figure 5. Component sources and sinks of dissolved oxygen below the pycnocline for the New Jersey midshelf : 1976. 231 40 o 45 ' •DREDGE SPOILS A OXYGEN (ML0 2 /L) BASE RUN 40° 30' 40° 15' 40° 00' M iBELOWPYCNOCLINE 40°45' 40° 30' r ' /OXYGEN-(MLO?/L) - /wc^i-iox z _ SEWAGE SLUDGE B ' /OXYGEN SAT. (%) BASE RUN ■■ ■•■/ 80A BELOW PYCNOCLINE 40° 1 5 ' V 40 o 00 ' 1/ iBELOW PYCNOCLINE 74°00 73°30' 74°00 iBELOW PYCNOCLINE 73°30' Figure 6. Spatial below a ;hr mPariS ? n ° f / issolved oxygen and oxygen saturation below the pycnocline for the base run and 10X projection- August. 232 area proportional to the ten-fold increase in waste discharges (Garside et al . 1976). Within the inner Apex, the increased oxygen demand from the additional urban carbon loading resulted in anoxia below the pycnocline (fig. 6) and a reduction in surface layer oxygen to 60-70 percent of saturation. The similarity between simulated detrital carbon content in the inner Apex (50-100 gm C m ~2) and the estimated POC content of the Ceratium bloom within the Apex (80 gm C m~2) and the New Jersey midshelf (25-125 gm C m~2) in June 1976 suggests that the model response of anoxia to the additional urban carbon loading is a realistic calculation of carbon-oxygen dynamics within the Apex. These results, consistent with preliminary estimates of the assimilative capacity (37 gm C m~ 2 ) of the New York Bight (Goldberg 1979), suggest that a critical detrital carbon content of the water column is 50-100 gm C~2 for the New York Bight. ACKNOWLEDGMENTS We thank Dr. G. Han for contributing results from the AOML diagnostic circulation model and we thank D.A. Dieterle for computing assistance in transformation of the transport fields and compilation of the data base. We thank our colleagues, Drs . T. Whitledge, T.C. Malone, T.S. Hopkins, G.T. Rowe, J. Vidal, F.G. Falkowski, and C.D. Wirick, for their data, insight, and numerous helpful discussions. We should like to thank Dr. S.O. Howe for contributing the numerical integration scheme used in the analysis. Financial support was provided by the Marine Ecosystems Analysis (MESA) New York Bight Project of the National Oceanic and Atmospheric Administration under Contract No. NA-80RAG-02206. Additional support was furnished under Contract No. DE-AC02-76CH00016 with the Department of Energy to Brookhaven National Laboratory with Dr. John J. Walsh as the principal investigator for both contracts. 233 REFERENCES Falkowski, P.G. and S.O. Howe, 1976. "Preliminary report to IDOE on the possible effects of the Ceratium tripos bloom in the New York Bight, March-July 1976," p. 57-62. I_n: Proc . Special Symp . Conf. by International Decade of Ocean Exploration (IDOE), Washington, D.C., 15 October. , T.S. Hopkins, and J.J. Walsh, 1980. "An analysis of factors affecting oxygen depletion in the New York Bight." J. Mar . Res . 38 (3): 479-506. Garside, C. , T.C. Malone, O.A. Roels , and B.F. Sharfstein, 1976. "An evaluation of sewage-derived nutrients and their influence on the Hudson estuary and the New York Bight." Est. Coast . Mar . Sci. 4: 281-289. Goldberg, E.D. (ed.), 1979. "Assimulative capacity of U.S. coastal waters for pollutants, U.S. Department of Commerce, National Oceanic and Atmospheric Administration, Boulder, Colo. Han, G., D.V. Hansen, and J. A. Gait, 1980. "Steady-state diagnostic model of the New York Bight." J. Phys . Oceanogr . 10 ( 12) : 1998-2020. Malone, T.C, W.E. Esaias, and P. G. Falkowski, 1979. Chapter 9. "Plankton dynamics and nutrient cycling." Part I. "Water column processes." In : R.L. Swanson and C.J. Sindermann (eds.), Oxygen Depletion and Associated Benthic Mortalities in New York Bight, 1976. NOAA Prof . Pap . No. 11, U.S. Dept. of Commerce, Rockville, Md. Mueller, J.A. , D.S. Jeris , A.R. Anderson, and C.F. Hughes, 1976. "Contaminant inputs to the New York Bight." NOAA Tech Mem . ERL-MESA- 6, Boulder, Colo. O'Connor, D.J., J.L. Mancini, and J.R. Guerriero, 1981. "Evaluation of factors influencing the temporal variation of dissolved oxygen in the New York Bight Project, Stony Brook, N.Y. Segar, D.A. and G.A. Berberian, 1976. "Oxygen depletion in the New York Bight Apex: Causes and consequences." ASLO Spec . Symp . 2:220-239. Stoddard, A., 1983. "Mathematical model of oxygen depletion in the New York Bight; An analysis of physical, biological, and chemical factors in 1975 and 1976." Ph.D. thesis, University of Washington, Seattle, Wash. Swanson, R.L. and C.J. Sindermann (eds.), 1979. "Oxygen depletion and associated benthic mortalities in the New York Bight, 1976." NOAA Prof . Pap . No . 11 , U.S. Department of Commerce, Rockville, Md. 234 Panel A - Ecosystem Modeling as a Fisheries Management Tool Panelists: Marvin Grosslein, Chairman Sharon LeDuc, Chairman Kenneth Dixon John Finn Brenda Nor cross Robert Pedrick Mark Reed Stephen Re illy Peter Saunders William Schaaf Michael Sissenwine Nancy Pola Swan I. INTRODUCTION The focus of the Panel was on major fishery management concerns and the role of numerical models in providing the information necessary to meet these concerns or needs. First, the Panel outlined fishery management needs and the general kinds of models currently used to meet those needs. The utility of these models was discussed, and an attempt was made to assign priorities to future research in terms of potential for improvement in predictive capability. No attempt was made to prepare a complete list of all the variations of models — just the general types of models. Next, the Panel discussed the role of ecosystem models in fishery management, There was general recognition by the Panel that full-scale mechanistic ecosystem models (i.e., involving quantitative and dynamic linkages among all the principal physical and biological processes which control organic production in the aquatic environment) are not being widely used for fishery management. This is because quantitative knowledge of the natural ecological mechanisms which control fish production is still very limited; in particular, the factors which control survival in the first year of life of fishes are not well understood, and this is the life stage when mortality is both high and variable. The Panel considered the critical role of the recruitment process in driving fishery fluctuations and discussed the importance of modeling in helping to clarify this process. II. FISHERY MANAGEMENT NEEDS VERSUS MODELS IN CURRENT USE The products of ecological models in current use which meet basic needs of fishery managers can be put into three general categories: 1) short-term projections of fishable stock, 2) long-term harvest strategies, and 3) environ- mental quality. A brief outline was constructed of the various types of models relative to these categories. Short-Term Stock Projections The basic tool of the fishery manager is still the traditional single species population dynamics model which can yield good forecasts of fishable 235 stock size providing that reasonably accurate data on current stock abundance and age structure are available, and assuming that the recruiting age class is a small proportion of the average harvest. The theory underlying these models is well developed, and the models are being utilized at or near their full poten- tial wherever adequate fishery data bases are available. The limitations of these models are that they do not take account of species or environmental interactions and they are of limited value where annual recruitment represents a large fraction of potential harvest. Pre-recruit (i.e., juvenile fish) estimates of abundance often are correlated with year-class strength and thus can provide useful forecasts of recruitment. Recruitment estimates together with the traditional population models add a further improvement to short-term stock projections. These approaches are also being utilized at or near full potential given sufficient time series of data for establishing a regression. Improvements in sampling procedures can sometimes help, but the general priority for research in this area is low. The lead time achieved with pre-recruit abundance estimates is usually only a few years, and precision of estimates is seldom high. Single species and raultispecies Virtual Population Assessment (VPA) methods are very useful for estimating parameters needed for traditional population dynamics models. The VPA is a calculation procedure more than a model, but it provides good "hindcasting" ability in terms of mortality parameters and stock sizes in previous years, and it is still under development. VPA calculations do not provide forecasts, and estimates of the most recent year or two are subject to the greatest errors due to uncertainties of fishing mortality rates applicable in the most recent years. On the other hand, multispecies VPA methods do provide a method for estimating natural mortality of pre-recruit stages caused by predation. Further advances along these lines are clearly feasible given the data on abundance of pre-recruits and predation rates, and these results would have useful applications to both short-term forecasts and long-term harvest strategies. Since the major natural fluctuations in fish populations result chiefly from variation in recruitment, a top priority for improving predictions of future population abundances is the development of models incorporating definitive insight into the physical and biological mechanisms controlling the recruitment process. It is conceivable that empirical relationships might be derived from a long time series of recruitment and general environmental data (e.g., temperature, timing of spawning), but it is unlikely these will ever be adequate for anything but gross predictions (e.g. , recruitment always poor after abnormally cold spawning seasons). More likely, extensive field and laboratory studies will be needed to clarify the relative importance of mortality processes including food supply, predation, physical transport, and disease. At present, models of these processes do not have much predictive ability because our understanding of the mortality processes is still largely qualitative (even the timing of first year mortality is seldom known let alone the causes). Recruit- ment process models offer the largest potential improvement in predictive capability, but the logistic problems are enormous for unraveling such complex large-scale phenomena as the growth, dispersal, and survival of larvae even for one species. A number of conceptual models of the process are available, but empirical measurements for testing the validity of the models are extremely rare. 236 Long-Term Harvest Strategies Short-term predictions of abundance are the bases of short-term decisions, e.g., what should the catch be for next year? The specific management decisions are based on a combination of predicted abundances and long-term exploitation strategies which, in turn, are based on balancing tradeoffs between harvesting now or later and between target population sizes and target species. Recruitment is also the critical issue in the development of models in support of long-term harvesting strategies. Here we are more concerned with long-term trends in abundance (long-term average abundance) rather than annual fluctuations, and the average relationship between spawning stock size and recruitment is the feature of major concern. Traditional stock/recruitment models (e.g., Ricker, Beverton, and Holt) are not adequate because the parameters essential to long-term management strategies are not estimated with adequate precision. New approaches are needed which include stochastic elements, i.e., assessment of the risk of transition from the condition where recruitment is sufficient to support fishing to a condition where probability of collapse of the fishery is high. The new paradigm for stock/recruitment models probably should be based on a world-wide analysis of the empirical behavior of various classes of exploited fish populations. At the same time, experimental studies on mortality mechanisms are needed on these same classes of fish populations because changes in the environment can alter the relative importance of certain mechanisms, thereby confounding the relationship between stock size and recruitment. Another type of model in current, if limited, use for evaluating long-term strategies is the surplus production model. This approach has been applied to both single species and multispecies situations. The predictive capability of this model is low because it does not explicitly take account of the recruitment process, nor does it incorporate any definitive environmental linkages. In essence, all the complex physical and biological processes controlling fish production are assumed to follow a certain pattern, but the model provides no basis for testing the validity of the assumptions. Once the stock/ recruitment problem has been clarified, multispecies fisheries models incorporating recruitment effects will have significantly improved predictive value for charting long-term management strategy. For the time being, multispecies extensions of traditional single species models can be used to help evaluate alternative strategies. For example, multispecies yield- per-recruit analysis (post-recruit stages only) is feasible given the same data required for a single species VPA. With this approach, tradeoffs between exploitation of different species can be evaluated to a significant degree in advance of solutions to the multispecies stock/recruitment problem. The most recent multispecies VPA models are also designed to incorporate pre-recruit stages as well. However, predation rates and absolute abundances of pre-recruit stages are necessary to model this aspect, and these data are seldom available in any fishery. Biomass balance models and holistic ecosystem models have also been used to help develop guidance for management of multispecies fisheries. However, these approaches make assumptions about energy transfers and linkages which are not firmly grounded in our quantitative understanding of the actual control mechanisms. Hence, their predictive power is unknown and unverif iable. 237 Two other approaches are also used to help formulate long-term strategy: 1) statistical models of yield (or population abundance) and long-term environmental trends and 2) gross energy budgets. Use of environmental trend- fishery yield correlations requires a long consistent time series of data, and is fraught with risk as experience has shown time after time. Energy budgets are another way of helping analyze long-term strategies, particularly through estimation of limits to total fish yield in an ecosystem. However, they are either heavily dependent on theoretical transfer efficiencies between trophic levels, or they require a very comprehensive data base on actual production (not yield) and food consumption rates of all major biological components of an ecosystem. Such data bases, particularly the predator-prey interactions, are not available for full scale ecosystems. Environmental Quality Quality of the environment is, in the long run, the most important aspect of ecosystem-fishery management since it controls carrying capacity. Effects of overfishing generally are reversible and usually can be mitigated within a relatively short time simply by reducing fishing pressure. In contrast, effects of a degraded environment may not be reversible, or only after a long recovery period, assuming cause of degradation is removed. Numerous environmental impact models have been constructed to deal with short-term effects of oil spills, power plant discharges, and other pollution such as acid wastes, sewage, etc. In all cases, these models have utility assuming worst case estimates for target organisms. However, their predictive power is generally poor because the cumulative synergistic effects on the ecosystem are not known, and even for target species (e.g., fish), the effects are only local and often do not apply to the entire population. Thus, it is usually impossible to estimate the real significance of a pollution impact, particularly for continental shelf populations. Detailed knowledge of hydro- dynamics as well as biological effects are required for quantitative modeling in these cases, and input data are seldom adequate for ocean environments. The same general models are applicable in the case of chronic pollution, but the problem of estimating biological effects is even more difficult than in the short-term case because of the enormous expansion of the relevant scales of time, if not also space. Nevertheless, since we are dealing with chemical effects on organisms, biological effects of pollutants can be measured in laboratory experiments for a limited but controlled set of environmental conditions and then extrapolated to the natural environment using the model and ambient concentrations of the pollutant in the animals and environment. Direct verification of pollution effects on organisms at the population level in the natural environment is, however, extremely difficult. Summary of Models A brief outline of the various types of models was prepared by the Panel and is summarized in table 1. General research priorities were assigned according to expected payoff in terms of improved predictive capability taking into account cost-benefit considerations for various kinds of research. In this case, research implies a new initiative, i.e., developing new methodology (model) or obtaining new ecological information. 238 o a 01 a 01 a* c co X u 0) J3 09 oo co 2 01 T3 g «J CO >» 09 O a w 01 c co (9- CO 09 U CO 09 4J .H 3 09 01 u M 3 co X9 CO H o H M Z H W < cad u gs U Cl, 5! J CO w w a a- 2£ &-s I 4J 01 X) 01 to a c x -H oi oi x> X! CO CO o fei, 0) a. o i-i > 0) G 01 01 M S CO 01 V w 09 a co S j3 CO 01 0> 4J ij h id CO U T3 4J go -H 01 0) 01 CO «J • 01 -H Q. 01 01 U c« X 3 oi cr §3 ■H 01 4J O 3 i 3 a co j3 CO 5 £ P I iH 4J x: n oc u •H IM .H a) O 3 *j >> 8 00 4J M •H >vrH >, i-H J3 (0 CO U -J. 01 -H C CO 0) > bo co 09 G a. o •H -H x: *j 3 X) O -H •H 1-1 3 C 00 o 3 -H 00 M iH CO CJ 01 01 ISO w a. § XI -H £ CO 0) T3 M 4J C 3 Xi •H 09 T3 09 M 01 CO U 3 bo m 01 O as «-i a S o co ~«. c 4J CO 09 4-1 01 00 H O •H CJ c oi x: > 60 3 -H 8 T3 4-1 "8 l-l l-l *J c <4-l 01 e 3 -H cm 3 01 M 3 3 M -H BO .H C CO -H 01 09 M CO 01 *J CO 5 s S y oi m oi i-i co o 01 a 09 H < 3 CVi a > ro a B C 01 ■H O H 4-1 01 •H X> CJ H 4J CO M 01 O £ cj n *J •H 09 o m "O -H *-> u 01 u-i 09 CL j= l-i CO PLl oi c oo CO 3 00 01 00 O C- 09 01 5 -H 09 CJ j2 XS 01 gg o m >> B o a H O M l-l -H Pi i-t o *■> CO Q <0 01 09 O. rH CO H 01 0) 01 *J I-i -C a. c u 00 01 o •Q iJ «-> 01 • •H 00 01 CO •a 3 -H 3 , Oj -o 4-1 T3 S 01 IJ CO ■H C C > H ICO H CO C S ij bO co I-i H co U C 01 tJ n tH H u B C co oi 5 oo C 4) § « CO -H c C O. U 01 CO -H XI l-l •h a •H *J c B o > CJ 01 0) 4-1 c fl S£tj 0) O 4J 4J 01 T3 I-i 0) O 01 a 0- XI CD C a 33 Q W H 4-1 C 01 B 3 I-i cj oi 01 01 Ij ^H CO XI cm ■H O CO 09 Ct go go •H 01 -c a oc 00 c q cm co O O M ■H U II U • CO 09 C rH 3 0) s 01 CO l-l o MUM •M Oi 3 4J cj Xi cj a •H B 4J t»T3 3 09 4J C CO ■H -H O C M >-. 3 CO o 01 x: "H xi CJ M IS oi a co •H 2 0) CM x; M •H bO c • M • "H CO CO 01 oi x: xi 4J > > cj CO c O iH T> 3 Xi CO B «T| MO C CJ 01 01 o ^ « u > -• 00 CO B 0) •H 01 4J OO H 09 0) 01 4J 00 M <0 C Ij ■g CJ 01 M 0) M a co 3 00 4J s^ CO M >-, «J 4J 09 "H iH C -H O X> ■H CO 4J O. CO co N CJ •H H 01 •H > 4-> TJ 3 4J o M -H bO O TJ "O Uj 0) 01 •o a S ° q 8 8 H M -1 * «-i CO •H 01 •H 2 ^ CJ CJ > 01 CD H a H c CO -H J *J CO 4J 3 8. c 0) £ ^_ / Q 239 •a v I c o ■8 E-> Z o H H Z H W < o a. 51 i-l C/3 a Qu > co •a C o> 0! C/3 o z He I en CO •H CJ c u s 01 0 Ij •H H ■8 5 > c ■a co -g 0) CO X co 4-1 tJ CO s e ■a 4-1 O CD ■H CD •H -O c U •H 8 <-> B T3 4J CD s 4-1 CD 4-1 ZM 4-1 1 4-1 c DO m"S CO C 4-1 O CD h H > ^ a u cd -i; 3 CO s o O -H c CO X cj CD a Cd £ H £ M ►J C o co M TJ C • O CD •H iH «J A u ctj 3 ^H 73 -H O <0 U > o. co o m 4-1 CD I a CD o ^-1 CO 4-1 CO 1 B •H X O H u a) a. u a -h CO l-i •H 4-1 a CD B U CD •H «-i o c u O "41 C -H CD H j= a * §• c o u •H O 4-1 CO CD BOrH ■H XI 4-1 «J CD H CD -H S» CO C > •H CO >> co M i-l a) a h •H CD e n S 8 CD tJ <4H a. o l-i CD O D- n-i X Hut • 3 4-i HI U CD 1WHTI «H cd S 5 jr J3 CO iH CD 4J CO 3 ^H Z O CJ Q 1 DO Ij CD C 01 u tJ •H 0) 4-1 St CO 4-1 1 CO x> h a i o c 4J — I O 8*J CJ >, CJ ^H AJ CO CO CD C9 U iH > CD CD -C -H U 4-1 4-1 4-1 bO C H m 01 td -C B 4J ■H o 01 0) 1-1 CO a o CD l-i >J CO C <« CD CO >> 4J H 3 ■a a a 0) 00 o C CD CD 01 X 4-1 X 4-> .H O o i-i oj e Q. CO U CD M-l 3 C U-l -H 3 CD T) 01 u e ■g CO 4-1 CD 3 4-1 CO T3 •H 01 X M 01 co a bo B c o •H CJ CD CO tJ 4-1 -p CO CD OJ oi a. t-i 4-i BO CO C « U U O CD OJ •H CJ C CO 4-1 C >iiH 3 eg a rH 4-1 C i-l tJ -V U £o C 01 O. CO cj O -H CO • ^1 > T3 CD CO OJ > 4-1 CD CO CD 1-" c CD O CO J= -H 1 bO 4-i CD •H CO 4-1 J= B -H c £ 01 T3 4-1 CO >» CO o o CD o J= ■8 CO 01 -a 01 o c CO rH CO J2 CO CO o •H a o •H % BO I-l CD c OJ CO 01 •o 3 r-l o a o 01 a CO 3 240 III. ECOSYSTEM MODELS AS A FISHERY MANAGEMENT TOOL There have been few attempts so far to apply full-scale ecosystem models/ to fishery management problems. Full-scale means a quantitative understanding of the dynamic processes controlling production rates of all the major biological components including the linkages to the physical driving forces. Perhaps the Andersen-Ursin model for the North Sea comes closest in this regard. However, this model falls short of real predictive or even diagnostic power because we do not have a quantitative understanding of the mechanisms controlling the recruitment process in fish. We cannot link survival of young fish quantitatively and mechanistically to either primary or secondary (zooplankton) production nor to the circulation dynamics which drive the whole process of organic production in the sea. In fact, we are just beginning to measure predation mortality on pre-recruit fish by recruited sizes. Determining the relative importance of this predation, which is applied largely to post-larval stages, requires that we also measure larval mortality rates, a prodigious logistic problem. There was a clear consensus in Panel A that more fundamental understanding of this recruitment process is required before we can expect ecosystem models to achieve predictive capability really useful for short-term fishery management needs. The possibilities are perhaps somewhat better for helping evaluate long-term strategies with ecosystem models, but even here the major priority is for better quantitative understanding of ecological processes rather than development of more and better ecosystem models. There are already plenty of models available. Definitive input data is the critical need. In spite of these problems the Panel felt that there should be a continuing effort toward the application of ecosystem models as a mechanistic framework within which to help clarify the nature of processes controlling fish production. For example, given qualitative but size-specific data on predator-prey interactions in the natural environment and experimental estimates of energy requirements for normal growth, it may be feasible to explore the robustness or reasonableness of alternative hypotheses on pre-recruit mortality processes taking into account time dependent aspects and physical dynamics affecting larval fish food pro- duction. Furthermore, multispecies models, involving predator-prey interactions properly scaled by at least an empirically based energy budget, may provide some insight into the potential for affecting long-term average biomass and species composition of finfish populations through alternative harvest strategies. Ecosystem models may also be helpful in resolving resource allocation problems (e.g., recreational vs. commercial fishery disputes over the relative impacts of the respective fisheries on resources). Since an appropriate model depends critically on the available data as well as the specific questions posed, the development of generic ecosystem models would appear to be of limited value (except in very broad terms). On the other hand, comparisons between different models using the same data base might be helpful in some cases for clarifying properties of the models. It would be perhaps more useful to apply a given model to two generally comparable ecosystems which occur in different locations and which have independent measures of the critical quantities (biomass, production, consumption) and mechanisms (growth, mortality). On the critical problem of the recruitment process, empirical models have clearly been useful for developing hypotheses, but understanding the mortality mechanisms and evaluating them with dynamic models will be necessary for testing 241 those hypotheses. The known mortality mechanisms of pre-recruit fish and their stochastic nature implies that recruitment predictions of high accuracy are not likely for any reasonable cost. Therefore, fishery managers will need to utilize strategies which, on the average, will optimize benefits as opposed to looking for precise predictions on an annual basis. One possible area of "high payoff" research for modest cost might be an in-depth review of patterns of stock recruitment data for major species groups on a global scale as a basis for a first approximation of relative potential risk of fishery collapse. The panel discussed, in general, the relative merits of dynamic versus static models in fishery applications but came to no firm conclusions other than each approach had advantages and disadvantages depending on the available data base. For a given data base, the question was posed as to the relative effectiveness of dynamic versus static models for uncovering counterintiutive characteristics of complex systems. Here again, comparison of results among different ecosystems of generally comparable structure seems to be a worthwhile and prudent approach. 242 Panel B - Ecosystem Modeling as an Environmental Management Tool Panelists: Bernard C. Patten, Chairman Kenneth Webb, Chairman Walter Boynton Joan Browder William Forster Peter Grose Wiley Kitchens C. Bruce Koons James Kremer Donald Scavia Eric D. Schneider Peter Schroeder Jay Taft I. INTRODUCTION The Panel identified six subtopics under the general heading "Ecosystem Modeling as an Environmental Management Tool" for discussion. These subtopics are: 1) state-of-the-art with respect to management application; 2) technical basis of modeling; 3) interactions among managers, modelers, and research scientists; 4) transferability of models; 5) translation of modeling results to environmental quality criteria; and, 6) use of existing data bases. The Panel formed subpanels, each of which was tasked with addressing one of the specific topic areas. A Panel consensus was formed for each specific topic, and a priority list of recommended high return research activities which would enhance the usefulness and effectiveness of ecosystem modeling as a resource management tool was generated. II. STATE-OF-THE-ART WITH RESPECT TO MANAGEMENT APPLICATIONS Overview Ecosystem modeling is an attempt to describe mathematically the structure and function of ecosystems. This mathematical representation of natural systems is perhaps the most difficult task in ecology since it requires an understanding of the physical, chemical, and biological variables, processes, and interactions that give rise to ecosystem states. The science of ecology is hampered by several inherent problems that make mathematical modeling of ecosystems or ecosystem components difficult. Ecology is an immature science and as such it lacks the theoretical richness of the physical and chemical sciences. It has few, if any, theories that allow precise predictions of organismic or ecosystem state as a function of abiotic and biotic forcings. For example, physics has its force and energy laws that allow rather precise estimates of at least average physical states, and chemists can calculate reaction kinetics and byproducts. In contrast, ecologlsts have not yet been able to tie their science to proven mechanistic or empirical relationships of cause and effect. 243 Another inherent difficulty that ecological modeling must contend with is the plethora and interaction of biotic and abiotic variables that control ecosystem state. Physical variables such as temperature, radiation, and turbulence are driving variables that control ecosystem structure and function. Chemical variables ranging from nutrients to toxics can interact to alter ecosystem state and response. Biotic components such as species present, population densities, competition, and behavioral patterns combine to provide the ecosystem modeler with a complex matrix of interacting variables which is almost impossible to untangle. This problem becomes more complex when one tries to integrate the synergies between all three classes of variables. It is true that the wise scientist tries to minimize extraneous variables emphasizing only those factors subjectively believed to be "important." However, he or she must enter into that exercise with the knowledge that the underlying assumptions and conceptual formulations may be incompletely defined. Finally, ecological data which are crucial to ecosystem modeling needs are often lacking. Perhaps most important is the lack of basic physiological data for many species. For example, verifiable toxico logical data are available for only a handful of chemicals and species. The data that exist are for single compounds tested under limited environmental or experimental conditions; knowledge of synergistic effects of toxics and toxic effects over a wide range of environmental conditions is virtually nonexistent. Important ecosystem rate kinetics are unknown as are fluxes, routes, and reservoirs of many elements and compounds in ecosystems. These constraints in theory, complexity, and data are facts that ecosystem modelers must face before they promote their potential products to resource managers and decisionmakers. Engineers can predict safe loading levels for bridges, and the public expects ecologists to be able to predict safe loading levels for pollutants in ecosystems. Ecologists and resource managers must be eminently cognizant of the constraints to ecosystem modeling and the limitations to model applications and predictions. State-of-the-Art The term "ecosystem model" encompasses a very broad range of approaches, each with strengths and weaknesses, each with different data requirements for development and evaluation, and thus each relevant to different aspects of management and decisionmaking. None of these approaches, however, can now or in the foreseeable future provide results that are conclusive and reliable enough to be the sole basis for any significant management decision. Further, modelers do not believe that model output should ever be the sole or primary basis for a management decision. The state-of-the-art of ecosystem modeling is such that it can produce information that is germane and useful to management, but such information must always be closely integrated with many other forms of information input: expert opinions, direct experiments, and socio-economic, health, and political deliberations. It is important to recognize quantitative models of all types as extensions of the conceptual perceptions of how ecosystems function. As such, models are never more than tests of necessarily incomplete hypotheses. As these hypothesis evolve with our experience, modeling inherently becomes an iterative intellectual process. In this context, it is important to recognize the necessity of including the modeler and the modeling process throughout the decisionmaking 244 process or, conversely, including the decisionmaker throughout the modeling process (see Section III). While this may seem to be a serious limitation to modeling, it is an essential perspective to take when attempting to specify the state-of-the-art of modeling and how models can provide useful input to management. The extent and quality of scientific information about different topics varies greatly. Similarly, there is a spectrum of approaches to ecosystem modeling, each with its specific data requirements and each capable of contributing information at different degrees of resolution to management decisions. Modelers vary widely in their interests and convictions about what model types are best. Perhaps a more appropriate view is to see the various modeling approaches as members of a spectrum of strategies. Relatively simple empirical models are at one end of the spectrum while highly detailed mechanistic models incorporating many functional forms of complex cause and effect interactions are at the other end. Throughout this broad spectrum there is the consideration of deterministic versus stochastic modeling techniques. The deterministic approach is the simpler of the two since a unique set of outputs, free of variability, results from each unique set of input data. Stochastic approaches, on the other hand, incorporate variability (uncertainty) in the model structure often resulting in nonunique outputs which are viewed as probabilities of occurrence within the natural variability of the system being modeled. The remainder of this section is a brief summary of a general framework within which the various approaches may be viewed. Empirical Models A number of purely statistical techniques exist for describing apparent relationships among or between empirical observations. In the simplest case, linear regression can relate a single dependent variable (Y) to an independent variable (X). Multiple linear regression relates the range of observed Y-values to more than one X or independent variable. Polynomial regressions are more complex, and assume that the variation in Y is related as some nonlinear function of one or more X variables. The statistical methods are well known and readily available on most computers. Thus, these methods are relatively quick and inexpensive to implement if the data are available. The most serious limitation is that such regression methods do not necessarily reveal anything about causal relationships, though it is often tempting to suspect they do. Mechanistic Models At the other extreme of modeling strategies is the family of models that attempt to represent the functional mechanisms that regulate the processes taking place in the natural ecosystem. These models are based upon massive amounts of data, often at the level of physiological responses of organisms to important features of their environment. Such data are almost never available for all or most of the species in any given system, and assumptions must be made about specific features unique to each locale. The predictions of these mechanistic models are usually compared to data of a fundamentally different kind: synoptic observations of the standing stocks of the state variables through time. Thus, this modeling approach requires the greatest amount of 245 data at both the physiological and ecosystem level and incorporates a large number of assumptions. The strength, however, is that these models potentially can "predict" the responses of many components of the ecosystem to new or extreme conditions, even when such conditions historically have never existed. These predictions, though they may be important, are based on an explicit series of identifiable statements about the underlying causal mechanisms. When used correctly, with continuous appreciation of the assumptions, these models may be the only quantitative method for evaluating certain environmental scenarios. Intermediate Strategies In between the extremes of empirical and mechanistic models falls a broad range of approaches, many combining, to various degrees, elements of both. Mechanisms can be specified, but with various theoretical perspectives. For example, many approaches attempt to relate changes in certain state variables to forcing factors in the environment using simple or complex statistical expressions. Thus, linear or nonlinear terms may be used to define the dynamic interactions among the compartments at the ecosystem level. In contrast to the mechanistic strategy, the data required for these formulations of processes are not physiological, but are the synoptic observations of the state variables and forcing functions. Thus, the data requirements are more limited and generally site specific, but again causal mechanisms are less fundamentally represented. Deterministic and Stochastic Modeling Approaches Output of deterministic models can be seriously misleading if the modeled environment is characterized by variable or uncertain conditions. This is particularly true when both controllable and uncontrollable variables have stro