Targeted calibration of one-dimensional process-based lake models

Process-based one-dimensional hydrodynamic models are standard tools in physical limnology due to being robust, computationally inexpensive and often sufficiently accurate to derive management decisions. These models are usually calibrated using objective functions that compare temperature differences between model and observations. Such an example is the root mean square (RMS) which is currently used to maintain up-to-date 1D models for 54 Swiss lakes. However, it is not clear (and rarely investigated) whether such a RMS-based calibration is optimally suited to the various subsequent uses of model outputs: primarily, quantifying vertical transport processes and water column stratification. To investigate this issue, we compose a master error function that includes a weighted sum of use-case specific error functions, related to thermocline depth, stratification stability and heat content. We then investigate the effect of these error function weights on the accuracy, utility and robustness of the Simstrat 1D hydrodynamic model calibration. We place particular emphasis on the effect of low-resolution historical data on this new calibration. We will present first results from this ongoing investigation on a range of selected Swiss lakes.