EVALUATION OF APPLICABILITY OF SPARSE IDENTIFICATION TO AQUATIC ECOSYSTEM MODEL

Abstract

 A model structure, as well as model equations and parameters tailored to the characteristics of the water body, is essential due to the uncertainties in the output values of the ecosystem model. In this study, sparse identification was applied to time series data of water quality state variables generated by the ecosystem model to evaluate its applicability. The results indicated that the correct equation forms and optimal parameters were identified for phytoplankton, particulate organic matter, dissolved organic nitrogen, and dissolved organic carbon. The application of sparse identification to actual observed data suggests the potential for identifying model equations and parameters that reflect the characteristics of the water body and reduce uncertainties in the ecosystem model. On the other hand, for dissolved organic phosphorus, phosphate phosphorus, ammonia nitrogen, and nitrate nitrogen, sparse identification proved challenging. The applicability of sparse identification appears to be influenced by the candidate functions for constructing the library and the amount of time variation of state variables.