| Numerical models have become increasingly important tools for the study of marine ecosystems. While these models serve as valuable research and heuristic tools, they are not yet to the point of producing reliable predictions upon which to base management decisions. Further, most existing models of coastal systems have been constructed for relatively deep estuaries rather than shallow, nearshore systems which receive the bulk of anthropogenic nutrient loading and experience the most extreme degradation in water quality.; A numerical model was developed for shallow marine ecosystems which contain both pelagic and benthic compartments of the food web. A series of innovations were developed to overcome limitations of existing models, including (1) the calculation of phytoplankton production with the light x biomass or BZI model, (2) the calculation of total water column respiration as a temperature-dependent fraction of the moving average phytoplankton biomass, (3) the delivery of a constant fraction of daily phytoplankton production to the sediments, and (4) the inclusion of layering effects on production and respiration by macroalgae.; The model was applied to Greenwich Bay, R.I. and produced predictions which agreed well with observations for the water column state variables but less satisfactorily for benthic macroalgae. Use of the BZI regression resulted in improved estimates of phytoplankton biomass and production compared to the traditional approach. While the model was relatively robust, sensitivity analysis revealed the need for better constraint of the macroalgal kinetics formulations, C:Chl ratio, temperature dependence of the BZI relationship, and the rate of water column respiration.; The model developed herein represents a step towards making improved predictions for management applications. The model overcomes some of the limitations of existing models through the use of a series of simplifying innovations. Some of these innovations were based on robust, data-driven empirical functions which apply across several systems. Many also served to reduce model complexity by integrating a variety of processes and state variables into bulk functions. These simplifying, data-driven relationships provide an ideal substitute when more detailed mechanistic relationships are inadequate. |
