Sprules, W.G.1, Johannsson, O.E.2, Millard, E.S.2, Munawar, M.2, Stewart, D.S.3, Tyler, J.4, Dermott, R.2, Whipple, S.J.5, Legner, M.1,2 Morris, T.J.1, Ghan, D.2, and Jech, J.M.4
1Department of Zoology, University of Toronto, Mississauga, ON; 2Department of Fisheries and Oceans, Burlington, ON; 3State University Of New York, Syracuse, NY; 4Great Lakes Environmental Research Laboratory, Ann Arbor, MI; 5National Marine Fisheries Service, Woods Hole, MA
Introduction
The Lake Erie ecosystem has experienced major perturbations such as reductions in phosphorus loading, variations in commercial fish harvests, and the invasion of exotic species such as dreissenid mussels and white perch (Morone americana). These perturbations have precipitated food web changes that include reductions in the abundance of many fish species, shifts in composition and productivity of the algal community, emergence of a more diverse littoral invertebrate community, and the virtual loss of the deepwater amphipod, Diporeia hoyi, from the east basin. Formulating management plans for an ecosystem undergoing such fluctuations is almost impossible without models that provide a detailed quantification of the complex of ecological processes. In this paper we present one such model that quantifies the transfer of material from prey to predator at each of the major trophic links in the Lake Erie pelagic food web.
Objectives
Our goal is to
Our model is not a dynamic simulation, nor does it explicitly represent the many detailed processes involved in the transfer of energy through a food web. Compartments in the model represent the mean biomass of component trophic groups over moderate spatial and temporal scales. The input to each trophic group is consumption of prey, and output from the group is consumption by its predators. Production of each trophic group is also retained in the model. Since the model is not dynamic the biomass of each trophic group does not change in response to gains and losses during a time period. The model simply provides a summary of these seasonal gains and losses, and the biomass, for each trophic group.
The model is designed to identify linkages in the food web that may be approaching an unstable state. In reality the biomass of any trophic group would fluctuate through time according to variations in consumption and losses. Only a persistent imbalance between consumption and losses would lead to a change in biomass. Trophic groups with high biomass would clearly have a higher capacity to sustain persistent net losses than groups with low biomass. Since our model represents only the mean state of the Lake Erie food web, we consider a trophic group to be unstable if there is an excess of losses to predators over gains from prey, and if this excess is large relative to the biomass of the group. Such instability would be evidenced by excessive energy demands by predators on zooplankton (planktivorous fish and carnivorous invertebrates) compared to total algal energy consumption and resultant production by zooplankton, particularly if zooplankton biomass were low. In this instance an appropriate management response could be to increase stocking rates of piscivorous fish or to increase harvesting rates of planktivorous fish.
The strength of our model lies in an extensive and highly coordinated database of field measurements. This is made possible by ensuring uniformity of personnel and techniques on all research cruises and by employing, where possible, automated sensors such as hydroacoustics, an Optical Plankton Counter (Sprules et al. 1992), and flow cytometry (Legner et al. 1999).
The model comprises 12 compartments or state variables (Fig. 1). The example shown is for the west basin of Lake Erie - a similar model is constructed for the central and east basins. The compartments or state variables represent mean seasonal biomass (May – October) of the trophic groups in fresh grams m-2 and the arrows indicate consumption of prey by predators in fresh grams m-2 season-1. Trophic groups comprise varying numbers of species that are considered to have a common trophic position.
Fig. 1. The food web of the west basin of Lake Erie as represented in the trophic transfer model.
Consumption by each trophic group is determined by applying a measure of growth efficiency to estimates of production. Depending on the trophic group, seasonal production is based either on direct measurements in the field or laboratory, or on various algorithms that convert biomass to production. Phytoplankton photosynthesis (>net production) is measured by exposing integrated epilimnetic water samples labelled with C14 to a light gradient in a shipboard incubator. Photosynthetic parameters are derived from the photosynthesis vs. light relationship determined in the incubator experiments. These parameters, along with data on transparency, chlorophyll, and mixing depth, are used to calculate daily rates of integrated water column photosynthesis using computer programs (Fee 1990). Production of bacteria, ciliates + rotifers, Dreissena, and benthic invertebrates is computed by multiplying mean seasonal biomass by a turnover or growth rate taken from the literature. Zooplankton production is estimated either from direct egg-ratio calculations or from biomass measurements multiplied by temperature- or mass-dependent production:biomass ratios taken from the literature. Fish production and consumption is based on bioenergetic models (Hewett & Johnson 1992) which use species-specific physiological rates in a mass balance of feeding gains against activity, respiration, digestion, excretion, growth and reproduction for a cohort of known numbers and size at the beginning and end of the season.
Production is scaled to consumption through division by growth efficiencies taken from the literature for the various trophic groups. The only exception is fish for which consumption is estimated by the bioenergetics model.
The biomass estimates underlying these production and consumption values are based on extensive field programs run on Lake Erie since 1992. Within the constraints of ship availability we endeavoured to sample the whole lake three times each year. In practice this was rarely achieved, but between the years 1992-1996 we managed to obtain good data on all trophic groups during spring before thermal stratification, during the summer period of full stratification, and during the fall when stratification was weakening. A network of about 50 sampling stations covering all basins as well as offshore and near shore areas was established. Many of these stations were oriented along a series of six transects running across the breadth of the lake or between islands in the west basin. During a typical research trip of roughly two weeks, data taken during the day at each station would include thermal and light profiles, nutrients, chlorophyll, phytoplankton, ciliates, rotifers, and zooplankton. At night, hydroacoustics and the OPC would be towed along the transects to estimate fish and zooplankton biomass, abundance, and size distributions. Collections of benthic organisms were made with a Ponar grab or box core at most of the sampling stations in the summers of 1992 and 1993.
In addition we sampled a smaller series of reference stations once every two weeks to obtain density, biomass and production data on the smaller organisms with higher growth rates.
To estimate production and consumption for the model, data from all stations/transects for a particular basin were averaged in a seasonally-weighted manner over years to give a mean growing season value (May - October). Hence the final model is a representation of the mean state of the ecosystem for the period 1992 to 1996.
Targeted State Variables
Nutrient concentrations
Nutrients are not explicitly incorporated in our model. Primary production is a seasonally and spatially averaged carbon uptake rated based on water column light intensities, chlorophyll levels and physiological properties of the phytoplankton. However seasonal phytoplankton photosynthesis rates have been related to seasonal mean total phosphorous concentrations for relatively unimpacted lakes or basins (Millard et al. 1996). Hence it would be possible to use this relationship to predict changes in photosynthesis caused by changes in phosphorus, or to determine whether seasonal photosynthesis falls above or below the expectation for unimpacted lakes. Increased or decreased seasonal photosynthesis would then be available as additional input to the grazers in our model.Total Algal Biomass
Both appear as state variables in our model. Algal biomass was estimated from microscopic examination of samples, but production was determined from independent experiments.
Blue-Green Algal Biomass
Walleye Biomass
Blue-green algal biomass can be estimated from the microscope data. However the current representation of our model includes only total algal biomass. Furthermore we do not have adequate data to compute zooplankton consumption of blue-green algae specifically. PISCIVOROUS FISH seasonal production is a summation of production for walleye, lake trout, rainbow trout, and coho and chinook salmon, each of which is derived from species-specific bioenergetic models. Thus it would be possible to specify production for each species separately, and to recover biomass data from inputs to the bioenergetics models.
Fish Body Burdens of Bioaccumulative Chemicals
Our model incorporates no measures of chemical contaminants at all. The original goals were to estimate only seasonal production and consumption of trophic groups, and there was no attempt to model the effects of contaminants.
Richness and Evenness of Intermediate Fish Trophic Levels
Our state variables at this level of the food web include only PLANKTIVOROUS FISH (smelt) and OMNIVOROUS FISH (yellow and white perch). Since our goal was to relate prey production to predator consumption at major trophic linkages only, we felt it was adequate to model two intermediate fish groups. The abundance and biomass data for bioenergetic models of these groups came principally from bioacoustic monitoring. It is difficult to identify acoustic targets to species. In the east basin we considered all acoustic targets to be smelt; in the central basin we considered those in less than 20 m of water to be yellow and white perch, and those offshore to be white perch or smelt depending on target size and thermal stratum; in the west basin we considered all targets to be yellow and white perch, the proportions taken from trawl catches.
State Variables in Relation to Stressors
No stressors are explicitly modeled. The model is a static representation of seasonal supply/demand at various trophic linkages for each basin of the lake. Including stressors was not one our original objectives, nor would this make much sense because there are no dynamic processes and feedbacks in the model. Addressing Lake Erie Millenium Management Issues
Primary productivity is included in our field measurements and is retained in our model. The empirical relationship between seasonal algal production and mean phosphorous concentration could be used to simulate changes resulting from eutrophication. However because of the nature of our model, this would simply generate higher primary productivity without any dynamic effect on any other state variable. It would simply increase algal supply in relation to the demands for it in the model.
The only exotic species explicitly included in our model are the dreissenid mussels. Other exotics such as white perch appear in the OMNIVOROUS FISH compartment, and Bythotrephes in the CARNIVOROUS ZOOPLANKTON. Any new exotics predicted to enter the foodweb could be added to our model if knowledge of potential biomass, annual production, predators, and prey were available. The model would specify whether there is adequate prey production to support such an invader, or whether consumption of the invader, when added to that by existing species, would lead to overexploitation of the particular prey groups.
Additional harvesting of predatory fish in our model could be simulated by reducing the biomass, and hence production and consumption, of PISCIVOROUS FISH. No changes to the state variables in the model would result because our model is not dynamic. Such a simulation would simply indicate what additional harvesting of piscivores would be required to relieve energy demands on the planktivorous fish.
One of the major goals of our trophic transfer model is to identify potentially unstable links in the food web. Such instabilities would exist if there is evidence of a sustained excess of predator demand over prey supply, particularly if prey biomass is low. In this sense, then, our model addresses issues related to ecosystem stability.
There are no features of the habitat that explicitly appear in the model. Some habitat information is included indirectly through, for instance, allowing zooplankton production to be partially determined by vertical temperature stratification. Similarly the near shore habitat (less than 20 m deep) was used to segregate "yellow and white perch acoustic targets" from "smelt acoustic targets" in the central basin. However none of this habitat information is used explicitly in the model to determine trophic interactions.
No contaminant information is included in the trophic transfer model, nor was our intention ever to simulate such effects.
Management Problem Being Addressed
The principal objective of our modeling approach is to identify linkages in the food web of Lake Erie at which there is an imbalance in energy supplied by prey and that demanded by the predator. For example preliminary analyses indicate that the energy requirements of dreissenid mussels far outstrip what is available from algae, rotifers and ciliates, and pelagic bacteria - their principal prey in our model. This suggests a) severe resource competition between mussels and other algal grazers (zooplankton, benthic detritivores), b) that mussels are accessing alternate energy sources, such as benthic detritus, not explicitly included in our model, or c) that mussel biomass will decrease through time, although the current biomass is very large. In any instance the simulation confirms that mussels are having a large impact on the Lake Erie ecosystem.
The management problem is that exotic invaders such as the mussel can seriously modify the pattern of energy flow through the Lake Erie food web. Algal production that would normally flow to zooplankton and then to fish is now concentrated in mussels which are largely unutilized by any predators in the lake (the flow from mussels to omnivorous fish diagrammed in our model is minor). Mussels thus represent an energy sink, at least on a short to moderate time scale. It does not appear that mussels can easily be removed from the lake, so our analysis confirms that a new state of the ecosystem exists, and that a return to past configurations that produced high yields of smelt and yellow perch is unlikely.
Modeling Assumptions
Our primary assumptions are embedded in the patterns of energy flow indicated by the arrows in Fig. 1. Secondary assumptions relate to the proportion of the total energy consumed by a predator that comes from its various prey groups. Feeding pathways and allocation of consumption are based either on our collective "expert knowledge" of these matters for Lake Erie, or on direct analyses of diets such as those required to determine fish consumption and production from bioenergetic models. We also assume that reasonable estimates of production can be derived by multiplying mean seasonal biomass of a trophic group by a seasonal growth rate, and that consumption can be determined by dividing production by growth efficiency. Finally, we assume that averages taken over the whole lake or across seasons or years constitute a reasonable "snapshot" of the state of the Lake Erie food web – effectively an assumption that the system does not change much from one year to the next.
Relevant Space and Time Scales
The time and space scales over which averages are taken to generate the "snapshot" of the lake Erie food web referenced above depend on the state variable being measured. In theory it should be possible to standardize the scales, but our experience has been that this is not always easy. Thus algal production data might be based on one particular year in which rather complete sampling of the lake wide stations was achieved for two or three seasons. On the other hand, estimates of zooplankton production from the towed Optical Plankton Counter might be based on several years of data which in total cover most parts of the lake in most seasons. Furthermore, expensive lake wide cruises that last up to two weeks cannot be performed very often, so it is important to collect companion data more frequently at fewer reference stations. This is particularly necessary for smaller organisms with more rapid turnover rates. Data from these reference stations can be combined with lake wide station data to benefit from the high temporal resolution of the former and the good spatial resolution of the latter.
Data/Monitoring/Research Needs
The data requirements for our Trophic Transfer Model are considerable. It requires good field sampling estimates of the biomass of all component trophic groups at an extensive series of lake wide stations or transects, and during each of the major stratification periods within a year. Since these data can rarely be obtained at high enough frequency to capture the principal dynamics of the smaller organisms (algae, microbes, zooplankton), an additional series of reference stations that can be visited more frequently (say every two weeks) is necessary. Since the model is not dynamic, it cannot generate the complete response of the food web to some new condition such as higher nutrient loadings or increased piscivore stocking. It can only compare supply and demand at the various trophic linkages of the existing static model under a scenario of increased algal production or increased piscivore production/consumption. To modify the model to reflect a new state of the ecosystem would necessitate extensive new field data on all trophic groups.
Overall Model Utility
We feel our trophic transfer model can help to address management issues by pinpointing major pathways of energy flow and their susceptibility to perturbations. It indicates linkages which need further investigation, and whether there is enough energy flowing through a linkage to warrant the time and expense of investigation. Finally, it provides a holistic snapshot of conditions and production in the lake against which process-oriented models can be calibrated.
References
Hewett SW & BL Johnson. 1992. Fish bioenergetics model 2. University of Wisconsin Sea Grant Inst., Madison, WI, Tech. Rpt. WIS-SG-92-250.
Fee EJ. 1990. Computer programs for calculating in situ phytoplankton photosynthesis. Can. Tech. Rpt. Fish. Aquat. Sci. 21: 465-475.
Legner M, WG Sprules, RJ Daley & ED Fillery. 1999. Flow cytometry for the unicellular plankton of the Laurentian Great Lakes. In: M Munawar & R Hecky [eds.]. Great Lakes of theWorld, World Ecovision Series (accepted).
Millard ES, DD Myles, OE Johannsson & KM Ralph. 1996. Phytoplankton photosynthesis at two index stations in Lake Ontario 1987 – 1992: assessment of the long-term response to phosphorous control. Can. J. Fish. Aquat. Sci. 53: 1092-1111.
Sprules WG, B Bergström, H Cyr, BR Hargreaves, SS Kilham, HJ MacIsaac, K Matsushita, RS Stemberger, & R Williams. 1992. Non-video optical instruments for studying zooplankton distribution and abundance. Arch. Hydrobiol. Beih. Ergebn. Limnol. 36: 45-58.