NETLAKE STSM 2015 to Budapest University of Technology and Economics, Hungary
Mechanistic models are established tools to learn about the processes controlling oxygen and carbon dynamics in lakes. Our aim is to calibrated existing models against high-frequency data, with the long-term goal of using such models to provide robust short-term forecast of water quality. Providing model-based early warning of anoxic evens would, for instance, allow implementing measures to avoid triggering fish kills. However, calibrating against high-frequency data reveals systematic flaws in the models, as they cannot capture stochastic processes that are now visible in the datasets. We thus propose to complement the mechanistic model with a statistical module (for instance based on autoregressive methods), which will improve the model performance and forecasting ability.