INTRODUCTION 



During the past 10 years the Great Lakes Basin has been an exciting proving 

 ground for developing, testing, and applying numerical, aquatic ecosystem models. 

 Several research efforts developed during the early 1970's have been brought to 

 bear on issues of overenrichment and contamination of these magnificant inland 

 seas. A certain amount of maturity has evolved in the modeling field from the 

 experience. Where models were originally touted as mystical algorithms with 

 the ability to make decisions for managers, they are now in better perspective 

 and have become useful aids for the decisionmakers (e.g., PTSTF, 1980). The 

 applications and explored limits of models of the Great Lakes have important 

 implications, not only for analysis of limnological systems, but for marine 

 ecosystems as well. 



Probably the most significant insight gained from our experience has been 

 to not rely on a single model or even a single class of models for analysis of 

 the Lakes. The collective efforts on the Great Lakes have included models from 

 the simplest, empirical correlations between nutrient levels and in-lake water 

 quality to more theoretical (and less verified) models simulating dynamic, time- 

 dependent mechanisms of biological, chemical, and physical process interactions. 

 Heidke's (1979) summary of the modeling efforts on the Great Lakes lists as 

 many as 50 different water quality models that have been calibrated and/or 

 verified for one or more of the Great Lakes. 



Rather than repeat or augment that report, I will present herein results 

 from several studies carried out at NOAA's Great Lakes Environmental Research 

 Laboratory that demonstrate uses and limits of ecosystem models. I will describe 

 briefly an ecological model developed and calibrated for data amassed during 

 the International Field Year for the Great Lakes (IFYGL) on Lake Ontario and 

 illustrate how its use has provided a better understanding of the structure and 

 function of that lake. I will then describe results from studies that explored 

 the limits to this and similar models by assessing, first, a consequence of 

 nonunique coefficient estimates in light of field measurements and, second, 

 error propagation from model inputs to predictions. 



THE IFYGL MODEL 



The ecological model (Scavia 1980a) simulates phytoplankton, zooplankton, 

 cycles of phosphorus, nitrogen, silicon, and carbon, and oxygen balance; and 

 calculations of carbonate equilibrium (fig. 1). Each model compartment is 

 described by a differential equation representing biological and chemical 

 processes. For example, the phytoplankton equation includes terms for gross 

 primary production, respiration, excretion, grazing, and sinking; and the 

 zooplankton equation includes terms for grazing assimilation, respiration, 

 excretion, defecation, and predation. 



Gross phytoplankton production is modeled as a single-step process that 

 assumes, for the time scale of the model, that rates of uptake and growth are 

 in equilibrium. The process is modeled with a temperature-dependent maximum 

 growth rate times a reduction factor for light and nutrient limitation. 

 Potential light limitation is modeled after Steele (1965) and potential nutrient 

 limitation is expressed as a saturating function for each nutrient. The 



59 



