AutoCalibration of Integrated Lake Water Quality Models for the Lake Winnipeg Basin Initiative

Machine learning (ML) and genetic algorithm (GA) are well known techniques which can be used to calibrate environmental models. This paper investigates the calibration of the 2-D horizontal, vertically mixed lake models OneLay and PolTra using ML and GA routines. A GA was used jointly with the Open Modelling Interface (OpenMI) wrapper approach on a single powerful server. Explicit and implicit gridded approaches with the Probably Approximately Correct (PAC) learning were used as ML techniques. Monte Carlo simulations were used to generate the model input parameters for explicit and implicit gridding. Parallel computing using Shared Hierarchical Academic Computing Network (SHARCNET) was used for explicit and implicit gridded approach calibration.