Improving Lake Winnipeg Integrated Environment Modelling with OpenMI

Abstract

We have applied OpenMI standards for two lake models applied to Lake Winnipeg to improve interactions among models and model calibration. These two models are OneLay and PolTra, which combine to form a 2-D horizontal, vertically mixed lake model. The source codes for the two models were modified for OpenMI to enable data exchange at each time step. Our approach is to calibrate drag coefficient and bottom coefficients in the OneLay model and settling coefficient, resuspension critical shear velocity and resuspension coefficients in the PolTra model for total suspended sediment. Model calibration statistic RMSE is used to measure the effectiveness of the calibration and is updated on an on-going basis as the simulation runs. This allows for the model simulations to be stopped partway through when the statistics are not improving, which greatly reduces the calibration time. A modelling database is developed to store key results of the multiple simulation runs from the calibration process. A decision support system with an implementation of a genetic algorithm is being implemented to demonstrate the system can be further automated.