Identification of model structure for aquatic ecosystems using regionalized sensitivity analysis

The Regionalized Sensitivity Analysis (RSA) was developed in 1978, for iden tifying critical unknown processes in poorly defined systems, thus directin g the focus of further scientific investigations. Here, we demonstrate its application to model structure identification, by ranking the constituent h ypotheses and identifying the critical elements for progressive revision of the model. Our case study is Lake Oglethorpe - a small monomictic impoundm ent in South-eastern Georgia, USA. Recent studies indicate that the warm te mperate regional climate affords an extended growing season -typically from March to October - which promotes bacterial productivity in the lake. The result is a summer food web dominated by microbial processes. in contrast t o the conventional phytoplankton-dominated food chains typically observed i n the cold temperate lakes of Europe and North America. Starting with a sim ple phytoplankton-based food web model and a qualitative definition of syst em behaviour, we use the RSA procedure to establish the critical role of ba cteria-mediated decomposition in Lake Oglethorpe, thus justifying the inclu sion of microbial processes. Further analysis reveals the importance of siz e-dependent selective consumption of phytoplankton and bacteria. Finally, w e discuss important practical implications of this novel application of the RSA regarding sampling efficiency and statistical robustness.