SALMO-OO : a process-based simulation library for lake ecosystems.

Abstract

Over the past three decades numerous lake ecosystem models incorporating algal population dynamics have been developed and published. However, most of these models have been constructed, calibrated and validated ad hoc to suit one specific lake application. Even though many models, including SALMO (Benndorf and Recknagel, 1982; Recknagel and Benndorf, 1982), were designed and validated as being generic for a range of lake properties they were always rigid in their process equations and parameter values. This study discusses the concept, implementation and testing of SALMO-OO towards a more generic simulation library for lakes by taking advantage of object oriented design and Java programming. A library of three phytoplankton growth and three grazing process models have been implemented in SALMO-OO, with the aim to increase the generality and flexibility of SALMO-OO for simulations of lakes with different trophic states and mixing conditions. The initial focus was on phytoplankton models that were of the form of ordinary differential equations (ODEs) that displayed a similar model rationale to the original SALMO model. Three phytoplankton growth and three grazing models implemented in the full object-oriented version of the model (SALMO-OO) as a simulation library of alternative process models. Combinations of different growth and grazing functions were tested within the simulation library to find generic model structures for lakes with different trophic state and mixing conditions. The validation of the SALMO-OO simulation library was based on comparison between the simulation library experiments and the results produced by the original SALMO-OO growth and grazing functions for phytoplankton biomass, zooplankton biomass, phosphate concentration and algal functional groups abundances. Root-mean square error (RMSE) and r2 values are given as a quantitative measure of fit between the measured data and the model outputs for each state variable. The results demonstrate the ability of the SALMO-OO model to simulate a variety of trophic and mixing conditions for freshwater lakes using a generic approach, and the ability of the simulation library to improve the validation results for each lake simulated. Generic model structures were found for different categories of lakes based on trophic state (eu-/hyper-, meso- and oligotrophic) and mixing conditions (dimictic and warm monomictic). A key factor that has determined a particular generic model structure has been the realistic simulation of phytoplankton functional groups dynamics. By providing the simulation library as an additional validation toolbox this has improved the overall model performance to give more accurate and realistic results for phytoplankton dynamics. As a result, the SALMO-OO model is a more comprehensive decision support tool for lake and reservoir management, which can be used to support the ranking of management scenarios and to base decisions on understanding and expert knowledge. Future research for the SALMO-OO simulation library includes the integration of a multiple parameter optimisation option based on evolutionary algorithms. This will calibrate parameter values within their range of variance to improve the accuracy of simulation results. It is concluded that the object-oriented implementation of ODE based ecosystem models significantly improves its knowledge base, functionality and accuracy.