1-D test-bed calibration of a 3-D Lake Superior biogeochemical model

Complex circulation models are commonly coupled with ecosystem models to characterize large-scale biogeochemical processes. While the reliability of these models is highly dependent upon accurate parameterization, the large computational expense associated with general circulation models generally prohibits the application of formal parameter estimation techniques to ecological model components in situ. Here, a 1-D model, driven by 3-D model output, is developed to provide an efficient test-bed environment in which model parameters are estimated using a Markov Chain Monte Carlo (MCMC) approach. The spatial and temporal uncertainty of model predictions due to parameter estimation error is quantified. A simple ecosystem model is calibrated for Lake Superior that is capable of reproducing most of the major features in observed concentration profiles of nutrients, dissolved organic carbon, and chlorophyll at the calibration location in the western basin of the lake. However, the optimized model is unable to reconcile observations of these variables with measured primary productivity during the stratified period. The test-bed calibrated parameters perform well in the 3-D framework at off-shore locations throughout the lake, and result in a 43% improvement in fit to validation data over manually adjusted parameters. The test-bed approach presented here represents a practical approach to the calibration of 3-D coupled models and has the potential to significantly improve model performance.

► A simple 1-D test-bed is developed for 3-D model calibration.
► The 1-D model is forced using 3-D hydrodynamic output.
► Parameters are estimated using a Markov Chain Monte Carlo (MCMC) approach.
► Model fit is improved with respect to literature-derived and hand-tuned parameters.
► This is a viable strategy for efficiently calibrating complex ecological models.