Assessment of the U.S. Outer Continental Shelf Environmental Studies Program: II. Ecology (1992)

As discussed in Chapter 1, the overall goal of the Environmental Studies Program is to develop information needed to assess and manage environmental impacts associated with oil and gas development on the outer continental shelf and to ensure that the information is available for the decision-making process. The assessment generally consists of three parts: inventory of the resources at risk, estimation of the likelihood of adverse impacts (estimation of risk), and retrospective estimation of damage caused by OCS activities (impact assessment). This report deals with the application of that kind of assessment to marine ecosystems. Marine ecosystems are complex, and many of the interactions within their biotic communities and between biotic communities and the physical environment are poorly understood, so estimation of risk and impact assessment cannot be very precise. Although the ESP is not primarily an ecological research program, many of its projects have provided useful data on the functioning of marine ecosystems, and all its results must be interpreted in an ecosystem context. Evaluation of the success of the ESP in meeting its overall goals requires evaluation of the extent to which it has addressed the specific ecological problems that arise in estimation of risk and impact assessment.

This chapter summarizes the ecological context of the ESP. Specifically, it formulates the principal questions that arise when one assesses impacts of OCS activities on marine ecosystems and the extent to which the questions have to be answered before risks and impacts can be assessed successfully. Chapter 3 summarizes the program itself and evaluates the extent to which it addressed the principal questions.

The mandate of the ESP includes the entire continental shelf of the United States subject to federal leasing—i.e., the area between 3 and 200 miles from the coast—and the adjacent inshore waters and coastal areas. The shelf adjoining the continental United States has been divided into 13 regions (Rabalais and Boesch, 1987). Each region can be divided into domains characterized by oceanographic and biological features, e.g., outer, middle, and inner domains of the Bering Sea shelf, as defined by Iverson et al. (1979) and Coachman (1986) according to hydrographic structure and nutrient flux. Although many species occur in more than one region, generally each region or domain is characterized by a different community of species and often by different relationships among the species and between the species and the food webs on which they depend (e.g., Schneider et al., 1986).

Some knowledge is transferable from one region to another (e.g., effects of oiling on birds does not differ among regions), but the risks posed by OCS activities to different components of marine ecosystems can vary markedly among the regions and domains. Thus, although generic studies can be of great value, data on numbers, population characteristics, ecological relationships, and other factors that affect vulnerability in one region cannot necessarily be applied to another. Similarly, the results of studies conducted elsewhere (e.g., in northern Europe) cannot necessarily be applied to populations in the United States, even in cases where the same species occur in apparently comparable ecological circumstances. Thus, population studies must be conducted in all regions, at least until a basis for extrapolation between regions can be established. It is not necessarily desirable to allocate efforts equally to all regions; allocation of effort should be based on careful evaluation of the data needs in each region.

Continental shelf habitats of North America contain diverse communities and constitute important economic assets; they are extremely productive and support some of the world's most important commercial harvests of fish and shellfish. Given the proximity of dense population centers in coastal areas, continental shelf habitats are subjected to the impacts of many human activities, including waste disposal, commercial transportation, commercial fishing, and mineral-resource exploitation. Those activities sometimes conflict with one another and compromise recreational and aesthetic concerns and the use of resources in continental shelf habitats. Multiple-use impacts on continental shelf habitats are highly variable in space and time, and such impacts, their ecological ramifications, and the prospects for community recovery are difficult to predict.

Added to the variability of multiple-use impacts is the inherent variability of continental shelf habitats in abundance and distribution of dominant species in time and space. The structure of any ecological community is a product of the processes that determine the distribution and abundance of component populations. These patterns of distribution and abundance result from the integration of several things, including:

• Relative availability of important resources.

• Recruitment and differential mortality resulting from

  • species interactions, such as predation and disease,

  • physical disturbance, and

  • physiological stress.

• Direct and indirect effects on communities, including those of contaminants in the food chain.

Natural variability on both spatial and temporal scales determines the detectability of an ecological effect. The problem of detection is complicated by the spatial and temporal covariation of many variables. What appears to be a statistically significant decline in a biological variable during a period of several years of anthropogenic activity (such as OCS oil and gas exploration) might no longer be significant when autocorrelation is taken into account (see Sissenwine and Saila, 1974, for an example). The characteristic scales of natural variability differ in different organisms and processes. Whereas benthic communities can appear to be less

variable in time and space than pelagic communities over short periods, benthic communities might not be seen as less variable when annual production scales are considered. For comparison, variability can be scaled to specific generation times of individual components of the ecosystem.

The optimal design of an environmental studies program to discern impacts of OCS operational activities must:

• Define what impacts are likely to occur and the temporal and spatial scales of their occurrence.

• Differentiate between such impacts and natural variability.

An optimal program should also require collection and archiving of environmental data and animal and plant specimens as well as information on the nature, scope, and timing of industrial activities to permit retrospective analysis necessary to determine the likely causes of observed changes in the variables being monitored.

It is normal in environmental monitoring to test for the adverse effects of perturbations. The perturbations of interest here are hydrocarbon-related discharges and perturbations of the marine and coastal environments. The natural variability of ecological systems makes the effects of human-caused perturbations difficult to detect. If the null hypothesis is that there is no effect, and that null hypothesis cannot be rejected, then the power of a test (i.e., the probability of detecting an effect of a specified size that is present) must be known. If no effect is detected and one is to interpret the test result, one must know how small an effect could have been detected. For example, if a test for increased mortality of some benthic species shows no effect, one needs to know what increase could have been detected; if a mortality increase of 50% could not have been detected, then the test has low power. Because of natural variability and other factors, tests for the effects of pollutants in nature often have low power.

Power can be increased by increasing sample size or increasing accuracy of detection; both are usually expensive. It is important to decide, in monitoring, how big an effect is “serious,” so that the test can be designed with appropriate power. A test with insufficient power to detect a “serious” effect might be relatively inexpensive, but it could be worthless.

An alternative approach is to use a different null hypothesis. This will not affect sampling design, but might affect the outcome of statistical tests and decisions based on those tests. If the null hypothesis is “there is an effect of size s” (e.g., a 22% increase in mortality), then rejection of that null hypothesis is likely to be much more informative than failure to reject a null hypothesis of no effect. The failure to reject the null hypothesis of “a 22% increase in mortality” also provides greater protection of the resource at risk than the failure to reject the null hypothesis of no effect. The approach suggested here is to assume that something causes damage unless it is shown not to, instead of assuming, as is common, that there is no damage unless damage is shown. Because tests for effects often have low power, real effects often fail to be detected until they are quite serious. The current approach, generally used, has led to serious losses, e.g., of fish stocks, of air quality, of water quality, of wetlands, and of diverse other habitats.

The roles of the allocation of important resources and of survivorship have received most of the attention of ecologists who study pelagic and benthic communities; in contrast, the causes and ramifications of recruitment phenomena have received most of the attention of fisheries biologists. Studies of birds and mammals have emphasized descriptions of individual behavior, distribution, and population parameters. The difference in focus is probably because benthic systems and the various vertebrate populations are amenable to different types of approaches to investigation. Important progress has been made in the last decade as ecologists have started to integrate their own research programs with programs directed at an understanding of physical processes and biogeographic phenomena. But much work remains to be done to address the processes affecting natural variability of marine populations, to identify the effects of disturbance (natural and anthropogenic) on such populations, and to define patterns of recovery after disturbance. Thus, to understand the impact of anthropogenic activities on components of continental shelf habitats, we need to define the mechanisms of resistance, elasticity, and recoverability of such communities, as well as the various successional pathways in operation within the many assemblages of pelagic and benthic species.

The past decade has seen considerable advances in both our understanding of the potential impacts of OCS development (Boesch et al., 1987) and the design of sampling programs that are suitable for testing hypotheses and distinguishing between anthropogenic effects and natural variability (Green, 1979; Carney, 1987; Clarke and Green, 1988). Even with those advances, however, prediction of broad-scale ecological consequences of anthropogenic activities, such as OCS development, is a highly uncertain process.

Because of legal requirements in the ESA and MMPA, and the high valuation placed on them by the public and the perception that they are at especially high risk, marine birds and mammals and endangered species (including marine turtles, as well as endangered marine birds and mammals) have received special focus in the ESP. Specific concerns are that those groups have been adversely affected in the past directly by sedimented and floating oil and other activities associated with hydrocarbon development in the marine environment and indirectly through effects on the food chain. Not all those species are equally valued or at equally high risk, however, and it is not possible to study them all. A rational study plan should allocate scarce resources for study so as to maximize the value of the information obtained.

Species must be selected whose status can also serve as an indicator of environmental change. Among the marine birds and mammals and endangered species, selection of species for study should take account of several criteria:

• Representativeness. To the extent feasible, species selected for study should include representatives of various geographic regions, of various oceanographic regimes, of various foraging types (e.g., among birds, aerial feeders, surface feeders, plunge divers,

underwater pursuers, and bottom feeders), and of various morphological types (e.g., bare-skinned versus furred mammals).

• Vulnerability. The species selected for study should include those judged relatively vulnerable to the effects of OCS activities (e.g., birds that swim at the surface of the water while foraging and hence are vulnerable to oiling), as well as some that are less likely to be affected and might be able to serve as controls for other environmental variables when impacts of OCS activities are judged.

• Rarity. The species selected for study should include species that are endangered, rare, or localized and hence that are at risk of severe population effects even if impacts are localized.

Within a species selected for study, geographical populations or breeding colonies should be selected for study according to the following criteria:

• Representativeness. To the greatest extent feasible, the populations selected for study should include populations that represent different ecological circumstances, e.g., large and small breeding colonies, dense and sparse feeding aggregations, inshore and pelagic locations, and central and peripheral parts of a range.

• Ease of Study. Subject to all the other considerations listed above, it is obviously desirable to select populations or colonies that are easy to study (e.g., species that occur in accessible places, are easy to see and identify, are easy to mark and trace, are tolerant of disturbance, etc.). Locations where representative populations of several species can be studied together have many advantages.

The principal goal of inventories done by the ESP is to map the seasonal distribution and abundance of the species at risk. That information is necessary for three purposes: to delineate the biological resources at risk, to identify places and times when populations are especially vulnerable to local impacts, and to provide a historical record for measurements of change. Although inventories should focus on the species selected for detailed study, useful information can sometimes be gathered simultaneously on many other species. However, at other times it is inappropriate to combine surveys. For instance, the altitudes appropriate for aerial surveys of birds are not appropriate for turtles and even less appropriate for cetaceans. Ideally, these early studies would lead into monitoring of selected sites so that natural rates of change and processes could be documented before oil is produced.

Information on distribution and abundance is necessary as the basis for estimation of risk and impact assessment, but it is not sufficient for either. Some information on ecological processes themselves is necessary for predicting the consequences of OCS activities on populations and identifying the causes of observed changes in populations. This section summarizes four types of ecological information that are required for those purposes.

Estimation of risks to vulnerable species requires calculation of the probability that individual members of the species will encounter and be affected by oil spills (or other adverse consequences of OCS activity). That requires prediction of the distribution and numbers of members of vulnerable species in relation to hypothetical oil spills and knowledge of possible direct and indirect effects. However, the distribution of marine birds, mammals, and turtles is typically very patchy and information is not complete, so there is a large variance in these predictions. Improved understanding of the factors that control the patchiness in distribution of marine species would result in improved predictions of the damage from specific activities. It requires study of the relationships between the distribution of marine species and oceanic features on various scales (Hunt and Schneider, 1987; Winn et al., 1987). On the largest scales, the distributions of individual species are related to broad physical and chemical oceanic characteristics—such as temperature, salinity, and nutrient concentrations—and to the abundance of prey species, which themselves depend on these oceanic characteristics; seasonal distributions and migrations of marine species are related to seasonal changes in temperature, insolation, ice cover, and food availability. On smaller scales, distributions are related to such oceanic features as eddies, fronts, upwellings, and (in the breeding season) the proximity of islands or other breeding sites. Many marine species are social feeders and breeders and aggregate in response to the sight of other organisms (of their own or other species) actively feeding (Hoffman et al., 1981). The timing of feeding aggregations depends on the behavior of the prey; temporal variability in distribution on larger scales is difficult to study, and factors controlling it are poorly understood.

Four types of study are required to understand the relationships between the distributions of marine species and oceanic features: statistical characterization of the variability of distributions in space and time; correlation of the distributions with oceanographic features; study of the behavioral and trophic characteristics that underlie the correlations (see below); and modeling of the relationships (discussed near the end of this chapter). The four types of study themselves have conflicting requirements: the first requires extensive, periodic, systematic surveys; the second requires integration and coordination with studies conducted by scientists in other disciplines; the third requires focused, detailed studies of selected individuals and groups; and the fourth requires systematic, comparable data from the other three. Successful study of ecological relationships will therefore require especially careful planning and integration of studies conducted in different disciplines, at different times and places, and by different methods.

Information regarding trophic relationships—including diets, food webs, foraging methods, predator-prey interactions, and energy transfer—is important for estimation of risk for several reasons. First, foraging methods often have an important influence on vulnerability; for example, birds that swim at the surface between dives for prey are disproportionately vulnerable to floating oil (Bourne, 1976). Second, the distribution and behavior of prey species are important in forming the patterns of distribution of predators that were discussed earlier (Schneider and Hunt, 1984). Third, the impact of OCS activities on prey species, competitors, or predators might affect some species indirectly; understanding of indirect effects will require knowledge of trophic relationships. Fourth, energy flow is important in population dynamics, including recovery times (Ford et al., 1982). Finally, trophic relationships are often important in

natural variations in reproductive success—variations that might erroneously be ascribed to OCS activities if their causes are not understood (Ainley and Boekelheide, 1990).

OCS activities can kill or injure mature birds or mammals or impair their reproductive success. Some local effects might be concentrated in a single breeding population, but OCS activities can affect breeding stocks over a large geographic area. Knowledge of migration and dispersal is needed to assess the potential geographic scope of impacts, especially those occurring outside the known breeding season.

Identification of the population consequences of mortality or reproductive impairment requires knowledge of the dynamics of the populations affected. Most marine mammals, birds, and turtles are long-lived and have low reproductive rates and overlapping generations; some have mature individuals that do not breed every year. Development of models that could be used to predict the effects of OCS activities on populations of those species requires knowledge of several population characteristics, including age-specific mortality and fecundity, emigration and migration rates, and numbers of breeding and nonbreeding adults (e.g., Ford et. al., 1982). The effect of frequency of environmental perturbations on populations with different life histories should also be considered. Because of the large time scales of population processes in marine birds and mammals, long-term studies (often exceeding 20 years) are needed to measure such characteristics (Nisbet, 1989). These models could also identify the extent to which low-frequency natural perturbations influence population processes.

Marine birds, mammals, and turtles interact with each other and with other species and other components of the environment. Interactions occur on a wide variety of spatial and temporal scales and lead to fluctuations in the distribution, size, structure and vital rates of populations. The interactions are complex and difficult to understand, so population fluctuations are difficult to interpret and predict. Understanding the fluctuations is one of ecology's central problems. However, within the context of the ESP, it is important to determine the most important community and ecosystem processes that are vulnerable to perturbation by OCS activities. That task requires both field studies of interactions at the community and ecosystem levels and the development of models to describe the interactions. The models serve several functions; e.g., they help in clarifying questions and directing field studies to the most critical measurements of population processes and they help in predicting population responses to specific perturbations. Such models are necessary if effects of OCS activities (other than large, localized kills) are to be identified or predicted against the background of natural fluctuations or the effects of other human activities such as fishing.

Marine birds, mammals, and turtles can encounter oil while swimming at the surface of the water, while diving or when they come ashore on oil-coated beaches, rocks, or other

substrates. External oiling may also affect terrestrial species that encounter spilled oil in the intertidal zone or forage on oiled carcasses. Even if the distribution, movements, and characteristics of spilled oil can be predicted reliably, prediction of the numbers of animals likely to make external contact with the oil requires several other types of information: knowledge of the seasonal distribution and abundance of each species in the area and knowledge of the behavioral characteristics of each species (e.g., swimming at the surface or hauling out on beaches) that would affect the likelihood of their encountering oil. The likelihood that an animal that encounters oil will die as a consequence depends on many factors—the degree of contact, characteristics of the species, characteristics of the oil, environmental temperature, etc. When an oil spill actually occurs, mortality of birds or mammals is often assessed according to number of oiled animals that come ashore; such assessments require knowledge of the behavior of oiled animals (e.g., the tendency of birds to swim ashore when stressed or the probability that oiled carcasses will remain afloat long enough to drift ashore), as well as of drift trajectories.

Although lightly oiled birds and mammals might not succumb to the direct effects of external oiling, they could be affected by oil ingested during feeding, preening, or grooming, or absorbed through the skin or by inhaling toxic vapors (Geraci and St. Aubin, 1987); reproduction of birds and turtles can be affected by oil that reaches their eggs (King and Lefever, 1979; Ainley et al., 1981; Lewis and Malecki, 1984; Parnell et al., 1985; and Fry et al., 1986). Several studies have shown that birds that ingest crude oil can develop persistent hemolytic anemia, which would compromise their survival in the wild (Fry, 1987). Other studies have shown that ingestion of off impairs growth, osmoregulation, and other physiological functions in young birds (Peakall et al., 1980, 1982, 1983; Fry et al., 1986). The significance of those and other toxic effects is likely to depend on the degree of oiling, characteristics of the species, and characteristics of the oil. All those factors need to be taken into account in estimating risk.

In addition to direct effects on animals that come into contact with oil, OCS activities might affect birds, mammals, and endangered species indirectly, by changing or contaminating populations of their prey or of predators. Hypothetical examples of such effects include effects on populations of zooplankton, fish, or benthic organisms that are used as food by birds or mammals and effects on populations of predators that may otherwise control populations of marine birds and mammals. Although trophic effects of those types are of great theoretical interest, the panel is unaware of any published studies that have demonstrated the occurrence of such effects in response to oil spills or other OCS activities. Their potential occurrence and importance could be demonstrated by modeling of trophic interactions and measuring or modeling effects on the trophic level of interest.

Many OCS activities—including construction, seismic exploration, and boat and helicopter traffic—result in noise and other stimuli that are potentially disturbing to sensitive

wildlife. In addition, the growth in human populations and development of roads, ports, and airfields that follows OCS developments might result in increases in hunting and other potentially disturbing activities. Known effects of such disturbance include abandonment of breeding colonies of birds and mammals and changes in migration routes and feeding areas. In such cases, displacement may result in higher densities in adjacent areas, and have effects beyond the area abandoned. In other cases, there may be suitable alternative habitats so that the only practical effect is a shift in distribution. However, some species can adapt rapidly to predictable human activities, so the effects of disturbance are not necessarily uniformly adverse or proportional to intensity (Schreiber and Schreiber, 1980). Assessment of the potential effects of noise and disturbance must be carried out case by case, and results of short-term studies are not necessarily predictive of long-term effects.

Other potential effects of OCS activities on birds, mammals, and endangered species include degradation of salt marshes by construction of pipelines and disposal areas, changes in tidal currents resulting from construction of islands and causeways, and secondary effects of residential development. Polar bears may be attracted to industrial activities and killed to defend human life or as a result of increased hunting pressure (Lentfer, 1990). Industrial activities can also disturb polar bear denning habitat, and there is the potential for cumulative effects on migratory species, as well as unforeseen effects that could be discovered through monitoring. Any such effects must be assessed on a site-specific basis, and the types of activity that are likely to occur, characteristics of the local environment, and sensitivity of local species must be taken into account.

Long-term impacts of OCS development are most likely to occur in the benthic environment from chronic discharges during development and production and from the accumulation of contaminants in sediment reservoirs (NRC, 1985; Boesch et al., 1987). Detecting changes in contaminant concentrations in benthic environments resulting from OCS oil and gas development will require a better understanding of the sources of contaminants (including other anthropogenic sources) and the processes that affect their fate in benthic environments. Such a task requires a combination of generic experimental approaches, observational studies, process-oriented studies, and regionally focused field assessments to establish realistic exposure scenarios (in both space and time) and to distinguish the effects of OCS activities from natural variation (Boesch et al., 1987; Capuzzo, 1987; Carney, 1987).

Ease of sampling is an obvious feature of benthic organisms that makes them useful in answering recruitment questions. Many are relatively sedentary (and consequently can be exposed to pollutants for long periods) and can be resampled over time. Coastal benthic habitats are heavily exploited for biological resources, including finfish, shellfish, echinoderms, and kelps. Because a large proportion of the U.S. population lives near the coast, human impacts are intense and well documented. Thus, the ability to know is coupled with the need to know.

Benthic systems are thus useful in understanding how populations and communities change in the sea. One way to assess present capability is to ask how well, using benthos, we

can predict and observe the full recruitment process from the production of new individuals (birth) through individual survival to successful reproduction and renewal of the cycle. Besides assessing the roles of dispersal and initial settlement, one must ask how well the crucial intervening growth and survival to the next reproductive event can be predicted. Understanding the acquisition of nutrients by the suspension and deposit feeders that dominate and characterize the seafloor has assumed a major role in such prediction, because these organisms both spend and gain much of their resources for growth and reproduction in feeding. Effort in understanding benthic feeding processes appears well spent also because the same physical processes that deliver food to the benthos deliver larvae as well.

Carney (1987), in his review of OCS benthic monitoring programs, concluded that none of the studies conducted before 1985 could define a priori what they were looking for; therefore, they could not use optimal design techniques. Although some intense point-source impacts of anthropogenic activities might be unambiguously recognized without a need to understand the sources of natural variability in benthic processes, detection of most anthropogenic activities requires an understanding of both the magnitude and the natural causes of variation in the absence of human intervention. Whenever an experimental effect is to be detected, it is necessary to know the magnitude of control or reference (“baseline”) variability in what is being tested. Without an understanding of the processes that cause natural variation, the mechanism of human influence is not likely to be established. Knowledge of mechanisms helps to confirm causation and is necessary for prediction under new conditions. It is difficult to establish causation in the presence of confounding variables, and so it is probably necessary to understand the processes that control natural variability to detect the more subtle impacts of human activities that are pervasive over larger areas or longer periods.

As Carney (1987) documented in great detail, it is insufficient for a monitoring program to document an effect without also providing an understanding of the mechanistic processes that produced it. First, an appreciation of the mechanism makes more credible the conclusion that an observed pattern was in fact caused by human intervention in the natural ecosystem. In the absence of a mechanistic explanation, causation remains in doubt and the observed pattern may be the result of spurious correlation. Second, a phenomenological understanding of mechanisms is necessary for generalization, extrapolation, and prediction based on the observed results. The transfer of results from one system or geographic region to another will be made more reliable if underlying mechanisms are understood.

Establishing mechanisms in benthic processes requires that well-designed field or laboratory experiments complement the monitoring process. Many variables are inextricably confounded in nature, and their effects can be separated only through experimentation (Carney, 1987). Mesocosms, flumes, or small experimental setups are potentially important in making possible the observations of organism behavior and various biological interactions that are necessary for a complete understanding of benthic processes. Mesocosms and other laboratory systems must be used appropriately to prevent laboratory artifacts from confusing the results (Capuzzo, 1987; Underwood and Peterson, 1988). Although laboratory and mesocosm studies are especially constrained to processes on small spatial scales, they provide a valuable link between field investigations by defining the biological and geochemical factors responsible for contaminant transport and effects. Furthermore, only through careful experimentation can complex interactions, such as nonadditive effects and higher-order interactions among variables, be detected and appreciated.

Understanding how anthropogenic activities, such as oil and gas exploration and development, affect benthic organisms on the OCS requires collaboration of biologists with physical oceanographers and biogeochemists. Without knowledge of physics, the scales and

boundary conditions of important transport processes cannot be determined. Without knowledge of geochemistry, the fates of materials and rates of transformation cannot be established. Those components of a complete monitoring study cannot be designed and conducted in isolation from biology. An integrated study of processes that affect benthic organisms is necessary for unequivocal demonstration of the anthropogenic impacts on benthic populations. With regard to specific, subtle, pervasive impacts of anthropogenic activities, study of physical and geochemical processes is necessary for deciphering the mechanisms of natural biological variation in benthic populations.

The benthic zone of the inner shelf from the beach or surf zone to a depth of about 30 m is poorly studied but crucially important because it is here where oil spills will encounter land. An understanding of the subsequent physical dynamics of transport, geochemical transformations of the hydrocarbons and other petroleum compounds, and biological impacts is required for a responsible and rigorous set of predictions of vulnerability and effects. In addition, understanding physical effects of perturbations, such as the effects of channelization on Louisiana's wetlands, also requires an integrated study of physics, chemistry, and biology. This zone of the ocean is too shallow for access by traditional oceanographic vessels and too far away and dynamic to be accessed easily from land. It has been insufficiently studied despite the complexity of its dynamics and its importance. As an indication of the need for process understanding of this zone, the NSF-nurtured research initiative on coastal oceanography (CoOP) has prepared and NSF has released an RFP (Joint Oceanographic Institutions, Inc., 1991) for interdisciplinary oceanographic studies, identifying this system as one of the two top priorities for immediate attention.

Assessment of the effects of OCS activities on fishery resources and ecosystem processes is constrained by the same difficulties as discussed above—the difficulties in predicting and detecting effects within the natural background of high spatial and temporal variability. MMS has a responsibility to assess the effects, regardless of the difficulties involved, but cannot be expected to make predictions of fisheries recruitment and productivity when these topics still confound fishery scientists. Assessment of OCS impacts requires a strategy that provides information to support reasonable short-term decisions based on current knowledge about how fishery resources and ecosystems respond to perturbations. The strategy must also provide a better long-term understanding of variability, ecosystem processes, and robustness of natural systems.

The most basic information that should be considered for assessing the potential effects of OCS activities is the spatial and temporal distribution of harvested species and key ecosystem components. Life-history patterns should be understood in sufficient detail to identify critical habitats, whose degradation can jeopardize populations. Spawning grounds, areas of concentration of eggs and larvae, nursery grounds, and migratory routes are particularly important.

One problem peculiar to assessment of fishery resources (stock assessment) is the great variation in recruitment of the species. Natural variability in reproductive success of the species makes it difficult to determine the relationship between population size and recruitment, and thus predict the size of the population a few years in advance and to assess the long-term capability of a population to compensate for stress (caused by overfishing, OCS activities or other human activities, or the interaction of multiple natural non-human and human forces). Numerous models are used to estimate the effects of fishing on fishery resources (Ricker, 1954).

OCS activity—which could be analogous in some respects to fishing—potentially has a chronic effect on a population as a result of effects on growth, natural mortality, and reproductive variables or as a result of effects on prerecruit survival. It might be useful to express the effect of OCS activity on a population in the same manner as fishing mortality, so that the vast amount of information on the effects of fishing can be used to help in assessing the potential effects of OCS activity. Traditional fish population-dynamics models can be expanded to account for perturbations other than fishing activity (e.g., Schaaf et al., 1987).

Ecological phenomena (e.g., succession, reproduction, feeding, survival, evolution) are characterized by processes that occur at different rates; those rates vary from one place to another. Mathematical models are a component of ecosystem studies that facilitate both understanding of underlying mechanisms and prediction of consequences and can generate new insights into basic systems. No model can achieve the goals of generality, realism, and precision simultaneously (Levins, 1968); for understanding and managing the environment, a variety of models are needed. In addition, problems of scale are increasingly recognized as important; development of useful models on one spatiotemporal scale probably cannot provide insights into phenomena operating on other scales. The concept of equilibrium is inseparable from that of scale. The insights from any investigation are therefore contingent on the choice of scale, and there is no single correct scale of observation.

Two major modeling approaches have evolved (NRC, 1990a), both of which are sensitive to the above limitations. Highly descriptive, parameter-rich models incorporate much system detail, but suffer from specificity (lack of generality) and statistical difficulties in estimation of parameters. Simplified models can attain generality, but only at the expense of development of the underlying mechanisms.

MMS has used two broad classes of models to assess the effects of OCS activities: descriptive models, in which observational data are fitted to statistics, and statistical parameters are derived for use in prediction or hypothesis testing; and process models of complex ecological systems, which are constructed from submodels with data derived from process studies. Empirical parameters derived from descriptive models of ecological processes are often used to set parameters for process models, which can then be used to predict the response of the system to external perturbations (e.g., Ford et al., 1982). Models of both types require validation before they can be linked and used this way to predict system responses. Validation of mathematical models is a perplexing problem, because of the natural variability and complexity of marine ecosystems. The need to resolve that problem highlights the importance of conducting long-term, process-oriented studies.

Because complex systems will always have components that are not well understood, conclusions will always depend on the validity of model assumptions, which should be stated explicitly. Similarly, models cannot take account of all components of a system, so conclusions can be valid only if system components not accounted for are constant. Finally, it is important that data used in models be representative of the variables that are being modeled. However insightful a model might be, if it is based on inadequate data, conclusions based on the model's output will be faulty. Indeed, one important use of models is to guide the collection and interpretation of data.

In many efforts to model particular ecological situations, irrelevant details are introduced on the mistaken premise that somehow more detail assures greater truth. In fact, there can be

no one “correct” level of aggregation for a given study. If the taxonomic species, for example, is used as the unit of classification, the differences among the individuals within it are automatically ignored. In ecology, functional systems of classification are often preferable to taxonomic ones, and a failure to recognize this relationship can lead to difficulties.

The results of the surveys and process studies discussed in the previous sections will ultimately be used for either of two purposes: to predict the consequences of hypothetical oil spills or other OCS activities, or to interpret and assign causes to observed changes in patterns of distribution and abundance. In either case, observations made at one place and time will be applied to circumstances that prevail at other places and times and that often involve different individuals or even species. Such applications are possible only through the use of models; thus, modeling is an essential element in OCS assessment.

Identification of cause-effect relationships of human activities is difficult, because of the large number of variables in biological systems (Rothschild, 1986). The variability and biological interactions present an impediment to both empirical prediction models and experimental approaches, in that it is impractical to account for all variables. Study designs must incorporate methods to assess both ecosystem perturbations and recovery rates. Statistical methods based on Markov-process models (e.g., Rothschild and Mullen, 1985; Saila and Erzini, 1987) are available to estimate how seriously a system has been perturbed and its characteristic recovery time. Those models assume, however, that a system is stationary, and that is often an invalid assumption.

Assuming that models can be constructed to describe or predict the responses of marine ecosystems to perturbations, the perturbations must be characterized. Potentially, OCS activities can perturb marine systems in several ways, such as a direct impact of oil on organisms that encounter it, effects mediated through trophic relationships, effects of noise or other disturbance on sensitive species, and effects on coastal wetlands.

In summary, ecosystem modeling holds the promise of increasing insight and the ability to predict effects. But there are many difficulties along the road. It seems likely that ecosystem models are currently capable of guiding the collection and interpretation of data. However, it appears that more data and validation of models against field observations are needed before ecological and physical oceanographic models can be linked in complex ways to predict the effects of OCS activities on ecosystems.