The Southeast Bering Sea Carrying Capacity (SEBSCC) Planning Workshop plenary summary has given an overview of current (1995) understanding of the Eastern Bering Sea ecosystem and of the goals of the SEBSCC project. Six discussion or "breakout" groups, organized by research approach, met at the Workshop following the plenary talks (see agenda), with discussions structured around a series of scientific questions (Table 3). Groups were instructed to discuss the merits of the existing questions, to suggest additional questions, and to suggest research that would address issues critical to the southeastern Bering Sea ecosystem and the role of walleye pollock within it, with special attention given to factors that affect pollock recruitment.
From the wide range of issues discussed, the following narrative summarizes common insights of the groups. Material is organized into four categories that correspond to the research components of the PICES/GLOBEC program: Physical Forcing Processes, Lower Trophic Level Responses, Upper Trophic Level Responses, and Ecosystem Interactions. Detailed data needs identified for monitoring, retrospective, and modeling studies of these components are summarized in Table 4 a-c. (Additional hypotheses that were not discussed are listed in Table 5).
Mechanisms that might supply nutrients to the shelf emerged as a major issue of interest, since nutrient supply appears to be an important constraint on primary production over the shelf during summer. Mechanisms suggested included nutrient-rich eddies propagating from the slope onto the outer shelf and dissipating, nutrient storage in the "cold pool," and tidal mixing at structural fronts observed in the vicinity of the Pribilof Islands. Other possible mechanisms (such as upwelling induced by flow in submarine canyons on the slope, breaking internal waves, and storms)(4) received less discussion.
Eddies are observed seaward of the shelf break, predominantly south of 58°N and in summer.(2) Offshore of the 100 m isobath, circulation is more vigorous than inshore, and well-developed eddies propagating into the vicinity of the 200 m isobath may provide a source of nutrients.(7) Once crossing onto the shelf, they dissipate, delivering their contents into the shelf ecosystem.(2),(3) The inshore movement of larger eddies (>100 km), which current moorings often fail to detect, can be monitored using satellite altimetry, but smaller ones are more difficult to find (Table 4a). (2),(7)
Evidence from PROBES suggests that nitrogen may be enhanced in the Pribilof Islands area by eddy diffusion from deeper water.(7) There are other shelf/slope exchange processes, including filaments or jets, meanders, and less energetic exchange of waters. Some way of calculating the relative importance of these processes to nutrient and production dynamics relative to nutrient regeneration over the shelf would be important for understanding primary production. (2)
Considerable discussion also focused on the "cold pool," a body of water with a temperature of 2°C or less that resides just above the bottom on the shelf throughout the summer over the approximate area occupied by sea ice in the winter. Routine monitoring of sea ice extent should continue because of ice's ecological effects.(7) Because of the large temperature difference between the upper wind-mixed layer and the lower tidally-mixed cold pool, which enhances vertical stability of the water column, these layers mix only very slightly with each other during summer. Nevertheless, the cold pool may act as a reservoir for nutrients that may contribute to upper-layer primary production later in the year - it may represent about 30% of the nitrate for new production. Cold temperatures may also impact the microbial loop and remineralization of nutrients.(2)
The cold pool of bottom water that forms over the shelf in winter may have dramatic effects on the Bering Sea ecosystem. How the cold pool influences ecological processes appears to be a key factor in understanding variability in the Bering Sea ecosystem. The vertical stability of the water column related to the cold pool is hypothesized to have direct effects on higher trophic levels organisms, including pollock. The cold pool especially may impact coupling between primary and secondary production and pelagic/benthic energy flows. (2)
Advective Transport
Advective transport emerged as a second major issue of interest, in part because ocean currents may supply nutrients to the shelf. Waters over the eastern Bering Sea shelf are affected by water entering from outside the Bering Sea. Relatively warm, nutrient-rich water may enter through Unimak Pass. Water entering the Bering Sea through passes further to the west determines many properties of the Bering Slope Current, which may supply nutrients to the shelf.(2) It is important to maintain a few moorings and/or conduct CTD transects in representative subareas, especially Amchitka, Amukta, and Unimak Passes, to document flow and water properties (Table 4a).(7)
More importantly, however, currents carry planktonic eggs and larvae of fish, including pollock, and the resulting trajectories are hypothesized to play an important role in determining survival rates of these critical life history stages. The Pribilof Island region is hypothesized to possess current features that may be critical for production of higher trophic levels. Between the 50 and 100 m isobaths (the middle shelf domain), the net circulation is very weak or insignificant (<1 cm/s, but organized flow is generally present just north of the Alaska Peninsula and to the west of the Pribilof Islands. Tidal flow is rotary and ~20 cm/s.(7) Both juvenile pollock and their prey are hypothesized to be retained in the residual tidal circulation associated with a frontal system in water between 40- and 70-m depths around the Pribilofs.(3) For purposes of predicting flows to aid in larval surveys, it was suggested that drifters be deployed and drifter data be assimilated into a physical model to help supply local mesoscale detail.(6)
Modeling of Physical Forcing Processes
For modeling purposes, major physical processes of interest were classified by their spatial and temporal scales, as summarized below.(6)
Feature Spatial Scale Temporal Scale microstructure and turbulence less than 1 m seconds-minutes nutrient flux, mixed layers and blooms 1-10 m vertical days-weeks fronts 1-10 km days-perpetual eddies 10-100 km weeks-months ice and the cold pool 100-1000 km seasonal-annual storms 1000 km days regime shift 1000 km decadal
Unimak Pass, Amukta Pass, Amchitka Pass, the N. Aleutian Slope, the Bering slope current, the Bering Strait, and western outflow points were identified as boundary conditions and "pulse points" important to monitor for driving physical models and/or verifying their output. Output from an existing 6-layer Naval Research Laboratory model, for example, overestimates flow through some passes by an order of magnitude, which demonstrates the importance of verifying model results with real data.(6)
The magnitude of primary production may set the upper limit on biomass production in the ecosystem.(7) Annual primary production on the Bering Sea shelf varies by as much as 50% between years. Monitoring of primary productivity is needed to assess this variability and how it affects higher trophic levels. Phytoplankton species vary in their food value for herbivores. so phytoplankton species composition should be monitored along with productivity (Table 4a).(7)
The existence and importance of the spring bloom on the shelf are well documented. The shelf bloom arises usually during mid- to late April or early May, in response to stabilization of the water column through a period of calm weather, or as a result of salinity stratification through sea ice melting. At this time, nutrients are abundant in the water column from in situ regeneration, winter mixing, and convective overturn, and there is adequate light.
In late spring and summer, phytoplankton production is nutrient-limited over the middle and outer shelf.(2) Considerable uncertainty remains about the physical processes that regulate primary production over the shelf during this period. It appears that the sequence of storms and wind mixing during and following the spring bloom is critical. The accumulated effect of winds on nutrient entrainment and development of the spring bloom appears to determine the magnitude of primary production for the entire year.(2) If a prolonged period of calm weather allows the water column to become very stable early in the season, nutrient supply to the euphotic zone and hence primary production may be greatly reduced through the summer. If a series of moderate storms keeps the water column mixed, the nutrient supply and primary production may be enhanced. This pattern is complicated by variations in sea ice extent across the shelf and resulting changes in the timing and magnitude of water column stabilization, the spring bloom, and surface nutrient depletion.
In addition, episodic propagation of eddies from the Bering Slope Current onto the outer shelf could enhance nutrient conditions for phytoplankton and/or zooplankton and may entrain significant biomasses of organisms.(2),(3) The Pribilof Islands structural front is hypothesized to enhance primary production relative to other parts of the shelf, particularly in late summer.(3)
There are fewer studies of primary production and other aspects of the ecosystem later in the season. Since many important processes influencing juvenile pollock are occurring in summer, ship time should be available for process-oriented studies in summer. In order that linkages may be made with fisheries questions, the ship involved needs to be capable of doing both oceanographic and fisheries research.(2)
Secondary Production
There is also uncertainty about the magnitude, distribution, and species composition of zooplankton production on the shelf. Two related major questions are: a) what continues to supply food to fuel secondary production in summer? and b) is zooplankton production during summer sufficient to support the observed biomass of higher trophic levels? The theoretical additional sources of nutrient supply to the shelf, perhaps in localized eddies or fronts, might sustain primary and secondary production on the middle shelf during summer.
Species composition, biomass, and physiological condition of the zooplankton community should be monitored (Table 4a).(7) The ice edge bloom, normal spring shelf bloom, and temperature effects on zooplankton production may be important factors regulating food supplies for larval pollock in spring. Summer phytoplankton and zooplankton production and their variability may be of crucial importance to production of predators, and, in turn, the impact of predators on zooplankton may be critical in summer or late summer (when juvenile pollock biomass seems to peak).(2) Physical features such as shelf-slope eddies and the Pribilof Islands structural front may concentrate plankton and at least temporarily enhance foraging opportunities for zooplankton and vertebrate predators relative to other parts of the shelf, particularly in late summer.(3)
Annual production processes of many larger zooplankton species such as euphausiids, which are important prey for higher trophic levels, are not well understood. Production of large copepods and euphausiids, in particular, may be important to monitor during summer because:(2)
Retrospective and Modeling Studies of Physical and Lower Trophic Level Processes
Retrospective Studies Physical and biological variability in ecosystems respond to forcing functions that have long periods (e.g., lunar nodal tidal cycles) or may be aperiodic (e.g., regime shifts). Thus, the effective design and interpretation of short-term (five-year) field studies requires knowledge of the ecosystem's characteristic temporal modes of variability. This constraint highlights the importance of retrospective studies.(4)
The goal of these studies should be to document and quantify the properties of event-scale, seasonal, interannual, decadal, and centennial biophysical variability in the southeast Bering Sea ecosystem. Some of this variability appears to co-vary and perhaps be linked to variability in the northern north Pacific Ocean. Therefore, relevant retrospective studies may include processes that occur over a larger geographic region than the Bering Sea shelf.(4)
Table 7 lists questions to address the scientific issues relevant to retrospective analysis. Our ability to address ecosystem questions is limited, however, by the quality of available historical data. In general, the duration of physical time series exceeds the duration of biological time series. This discrepancy presents a serious problem to examining biophysical processes with decadal and centennial periodicities. Table 4c summarizes data that are needed for retrospective analyses, and identifies data sets thought to span at least 30 years.(4)
Appropriate analytical tools for performing retrospective analyses are not well established. Typical parametric procedures like linear regression and correlation are obvious first steps, but they require some restrictive assumptions (e.g., assumptions of linearity, no autocorrelation, and normal distribution of errors) that we know are violated by marine biotic and abiotic time series. Some newer statistical methods (e.g., tree regression modeling, General Additive Models [GAM], time series/intervention analysis, and neural networks) have less restrictive assumptions. These newer methods may prove useful, and their application to retrospective analyses should be encouraged.
Modeling Heat fluxes, winds, salinity, and insolation(6) data were identified as important components of models designed to reproduce features of physical and lower trophic level components of the ecosystem (Table4b). Insolation was considered important in setting vertical stratification later in the year; salinity is far more important in the spring. There was general agreement that the timing of the spring bloom on the middle shelf (though not its intensity or spatial details) could be predicted in warm years using meteorological data(6) , specifically the arrival of a multi-day period of calm weather any time after mid-April, when there is sufficient insolation to support a bloom.
The shelf current is considered well-behaved and predictable, but the slope current is not.(2) Canyons could be resolved by a 4 km model grid. The mean slope current might be predicted a week in advance using an appropriate data-assimilating model.(6)
Modeling of the shelf/slope exchange of nutrients was considered important, but there was disagreement about its tractability.(2),(5) Cross-shelf fluxes are difficult to model and would likely require assimilation of data. This exchange has not yet been modeled successfully, and the uncertainties about the processes involved present a modeling challenge. A model of currents and hydrography is a first order necessity. To succeed with such a model, heat flux and insolation data are necessary (Table 4b). A model also would need to incorporate mixed layer dynamics and the effects of tidal mixing, since these mechanisms determine the two-layered structure of the middle shelf domain.(2) Predictive skill would generally be low for mesoscale detail, given only biophysical platform data. In general, without data assimilation (2), the best we can hope for is models that capture the statistics of the flow adequately (e.g., the correct number of eddies). Model studies with "pseudo-data" can indicate where time series would yield the greatest improvement in forecasts/hindcasts.(6)
Models of Differing Scale Modeling lower trophic levels with a nutrient-phytoplankton-zooplankton (NPZ) model linked to the shelf/slope nutrient exchange model was considered important, because current understanding of nutrient dynamics is inadequate to explain energy flow to upper trophic levels.(5) An NPZ modeling effort would help address scientific issues 3 and 4 (Table 3). Three potentially nested approaches were suggested for coupling a physical model to an NPZ model:(6)
An issue of major interest was the dependence of pollock recruitment on the processes by which the adult pollock stock (or stocks) select spawning areas, and pollock larvae are subsequently transported by currents. Improvements in our limited understanding of pollock spawners that may return to natal spawning locations would be critical to elucidating spawning stock structure. Recent advances in otolith microchemistry may be fruitful in this area. Mortality processes and transport conditions that give rise to larval supply are also poorly understood.(2)
Transport of pollock larvae to favorable "nursery" waters is hypothesized to enhance year-class success. Juvenile pollock are hypothesized to be advected from spawning areas to the south toward the Pribilof Islands region, where they may concentrate.(3) The Pribilof Islands area may have unique features that enhance primary and secondary production, particularly in late summer, and that concentrate and sustain abundant prey resources for forage fish. Hence, the Pribilof Islands region may be a superior foraging environment for young pollock.(3) The movement capabilities and patterns of juveniles in relation to oceanic features and prey conditions are not well known, but are feasible to study using advances in technology.(2)
Vertical stratification of the water column also was hypothesized to have direct effects on pollock larvae, for example, by affecting horizontal transport. Vertical movements of larvae are not well understood and may be critical to models of larval transport. This is particularly true when there is vertical shear within the vertical distribution range of the larvae.(2)
Most discussion, however, surrounded the effects of vertical stratification on trophic interactions among pollock larvae and their prey and predators. Seasonal changes in larval feeding and condition should be linked to changes in temperature, physical structure of the water column, spring bloom development, and production of the larval prey field. The cumulative influence of winds on the seasonal development of prey and resulting effects on larval pollock condition should be investigated.(2)
Modeling Higher Trophic Levels
It may be possible to develop statistical correlative models to predict fish stocks in the Bering, based on appropriate environmental indices. Hindcasts and process-oriented simulations may help suggest appropriate indices for this type of model.(6) There have been successful statistical models that explain much of the observed variance, but as more data are collected, they change, often radically. Most biological time series so far have been too short to support reliable statistical models.(2)
A model of mortality, biomass, and energetics of juvenile pollock coupled to process studies over fine spatial and short time scales (2-4 mos.) was discussed.(5) Individual-based modeling (IBM) was suggested for spatially explicit modeling of pollock through the juvenile stage. An IBM-type model for pollock would help address central scientific issue 3 (Table 3). An IBM model could be combined with an eulerian NPZ model (including benthic compartments) and an age-structured, spatially aggregated adult pollock model (with appropriate spatial weighting derived from field survey data for representative years).(6)
Monitoring of diets of adult pollock and other fishes was considered important, because it might provide insights into important trophic dynamics in the ecosystem. Likewise, the physiological condition of the fishes could provide additional clues to variability in production processes at lower trophic levels.(7) However, it is difficult to discriminate between production variability and spatial variability between predators and prey.(4)
Vertical water column structure also was theorized to affect the trophic interactions of juvenile pollock. The vertical distribution of forage fish may be influenced by the vertical distribution of their prey, and the vertical distribution of forage fish may determine their availability to predators.(3) Vertical differences in age-1+ pollock foraging may, in turn, affect forage fish distribution. Birds and juvenile marine mammals require concentrations of fish in near-surface waters; adult pollock are believed to cannibalize young pollock mainly at depth. Factors that affect the vertical distribution of these forage fish are thus important in establishing the carrying capacity for high-trophic level predators.(3)
Pollock Cannibalism
Cannibalism and the processes that control its magnitude were considered to be an important aspect of population control of pollock. Vertical water column structure was theorized to have particular relevance to spatial and temporal variability in rates of cannibalism. Seasonal changes from a two-layered to a vertically mixed water column may influence vulnerability of juveniles to cannibalism, and spatial variability in the rate of cannibalism between domains may be determined by water column structure.(3)
The extent of the cold pool on the middle shelf, in particular, was identified as a major feature that might affect pollock distribution and abundance and the resulting incidence of cannibalism. Pollock, especially juveniles, appear to avoid the cold pool. The size and extent of the cold pool may influence the horizontal distribution of juvenile pollock and hence their potential for interaction with cannibalistic adults. Furthermore, the timing and location of pollock spawning and hence the subsequent drift of larval fish may be determined in part by the extent of the cold pool on the eastern Bering Sea shelf.(3)
Fluctuations in populations of juvenile pollock and their prey also were hypothesized to influence vulnerability of juvenile pollock to cannibalism. For example, an apparent seasonal decrease in the abundance of copepods and other food in the upper mixed layer might result in movement of juvenile pollock deeper in the water column, where they could come into contact with adult pollock, thus fostering cannibalistic predation.(3)
Two alternative hypotheses were offered to account for heavy predation on age-0 juveniles. Predation may increase proportionally to prey abundance, so that the increase in late summer reflects the abundance of juveniles available at this time. Alternatively, the rate of prey consumption may change nonlinearly with prey abundance. Under this condition, the summer peak in juvenile pollock biomass, and the resulting competition with adults, may stimulate adults to switch prey from euphausiids to juvenile pollock. Each of these hypotheses is suggested by data from particular years. A study to resolve these competing hypotheses could be important to the project.(2)
Recent observations about the apparently increasing abundance of large gelatinous zooplankton over the southeastern Bering Sea shelf suggested hypotheses concerning possible interactions with juvenile pollock and other forage fish. Forage fish prey abundance in the upper mixed layer may be affected by grazing of gelatinous zooplankton. Juvenile pollock may gain protection from cannibalism by schooling or by using medusae as cover.(3)
Marine Birds and Mammals
Additionally, the Pribilof Islands region is important for the large aggregations of marine birds and mammals which breed on the islands and forage in surrounding waters.(3) Seabirds and fur seals on the Pribilof Islands and on St. Matthew Island prey on pollock and compete with pollock for common prey.(7) It was considered important to monitor birds and mammals along with fish, both as a way of maintaining a broad ecosystem perspective, and as a way of obtaining more useful information about distribution, abundance, and trophic interactions of target fish species. Bird and mammal populations reflect lower trophic level production integrated over multi-year variability in the ecosystem.(4)
On-going population monitoring programs exist for pollock and other groundfish, seabirds, and marine mammals (Table 4a). Improvements are needed in the program to better assess the abundance of juvenile pollock and other species of forage fishes of importance to marine birds and mammals. Diets and reproductive parameters of seabirds are sensitive to fluctuations in the abundance of prey and can be used to monitor local abundance of zooplankton and fish. Improvements in the monitoring of diets and foraging of birds and mammals could provide additional information on the role of pollock and other food web organisms of importance to the ecosystem of the Bering Sea.(7)
Benthic Community Relationships
Some discussion concerned the processes that control the proportion of primary production that reaches the benthic food web rather than being retained in the pelagic food web. It is generally believed, on the basis of PROBES data and ice-edge studies, that early spring blooms on the middle shelf are not heavily grazed by zooplankton (which are inhibited by low temperature), so that most of the resulting primary production sinks to the bottom. Processes that affect coupling between primary and secondary production are believed to regulate pelagic/benthic energy flow(2), and thereby may have critical effects on production of benthic invertebrates and flatfishes. These organisms have significant commercial values and are an important food source for marine mammals.(4)
Additional discussion concerned effects of fishing on pelagic and benthic communities. Fisheries harvests are major sources of biomass removal, and have a strong influence on the size class distribution and species composition of the fish communities of the eastern Bering Sea. These effects may also impact non-target species by changing prey availability or concentrations. These effects may be important for pelagic fish and their predators, but the impacts are likely hard to resolve. In contrast, benthic communities are more tractable to study.(3)
Ecosystem Models
Neither NPZ nor pollock IBM models address issues concerning adult fish and marine mammals, in which population responses tend to occur on a decadal scale.(6) Models including upper trophic level processes relating to pollock were considered important. In addition, some discussion focused on the importance of the whole suite of upper trophic level models, including those relating to other forage fish, not just directly to pollock. A spatial model of pollock cannibalism, including other upper trophic level predators, linked to a spatial model of pollock spawning and larval transport and survival was considered both highly important and highly tractable. These models should in turn be linked to the NPZ model.(5)
Models which addressed broader ecosystem concerns, particularly those relating to the benthic system, were also given high importance. Topics suggested in this category included:
a) Causative factors explaining increase in flatfish populations ;
b) Ecosystem effects of raising the 2 million metric ton (mmt) cap on groundfish removals;
c) Causative factors explaining past changes in species composition and abundance (regime shift);
d) Incorporation of salmon outmigrants and their effects on shelf forage ;
e) Effects of trawling on benthic populations.(5)
One recommendation was a highly aggregated but nonlinear whole ecosystem modeling effort, to proceed in parallel with the spatially explicit models, and where feasible interacting with them. The level of aggregation envisioned was similar to that displayed in the PROBES carbon budget: ~10 compartments representing various trophic levels and guilds integrated over large areas (the entire southeastern Bering Sea shelf, or subsections of it) with a primary production source term and fishing mortality as a specified drain on pollock. Pollock could be broken out into year classes as appropriate (e.g., juveniles and adults). The fluxes between boxes could be formulated using integrated mass fluxes and temperatures from physical models, standard ecological theory (e.g., predation = f[predator biomass, prey biomass]), and hypothesized fish migrations (e.g., avoidance of the cold pool by juvenile pollock).
As in physical modeling, spatially explicit biological models can help suggest proper parameter values and functional forms for more spatially aggregated models. The spatially explicit model results can be integrated spatially for use in the aggregated model, to study ecosystem sensitivity to juvenile pollock. Aggregated models, in turn, can help set the forcing functions for spatially explicit models which do not include the whole ecosystem, although proper spatial weighting of the aggregated results in the spatially explicit model is difficult. Such an aggregated model could serve as an investigative tool for managers. Present technology would allow such a simple model to be used by all interested parties through the World Wide Web.(6)