Calibration versus computation: Comparison between 1D and 3D phytoplankton simulations in western Lake Erie

Abstract

Numerical models are commonly used tools to simulate hydrodynamics and water quality of lakes. Model dimensionality (0D, 1D, 2D, or 3D) requires different simplification levels of physical-biogeochemical processes, computational power and calibration strategies and metrics against observations. To investigate these modelling considerations, the 1D (vertical) Aquatic Ecosystem Dynamics – General Lake Model and the 3D Aquatic Ecosystem Model were applied to western Lake Erie in 2008 and 2011-14. The performance of the models was evaluated by comparing the simulations against observations of water temperature, total phosphorus, orthophosphate, nitrate, total chlorophyll-a and cyanobacteria at three stations located along a transect from the Maumee River mouth to mid-basin, as well as to the basin-averaged cyanobacteria index. The 3D model showed better skill in qualitatively reproducing seasonal and spatial variations of nutrients and phytoplankton and had lower average root-mean-square error, especially through the algal plume near the Maumee River mouth. However, the horizontally averaged 1D model performed better in qualitatively capturing the cyanobacteria bloom years, as this model was extensively calibrated to basin-average values. We conclude that models should be selected and calibrated to provide the required decision support information, rather than the highest resolution or lowest error metrics at discrete sites.