Prediction of long-term changes of weak diurnal stratification in shallow lakes using artificial neural networks

Abstract

ABSTRACT Thermal stratification plays a key role in lakes' ecosystems. In contrast to deep lakes, the thermal structure of shallow polymictic lakes is characterized by a weak stratification with an apparent diurnal cycle. Long-term changes in stratification are governed by climate change and anthropogenic effects such as water level regulation. We developed a simple and robust model system consisting of an energy balance model to estimate depth-averaged water temperatures and an artificial neural network (ANN) model to predict stratification with high temporal resolution. One novelty of our approach is that instead of directly estimating water temperatures at different depths, we simulated the potential energy anomaly index, the indicator of stratification's strength. The ANN-based model's performance was assessed against a physical-based one-dimensional model (General Ocean Turbulence Model) by modeling a 40-year-long period from 1981 to 2020. The new model accurately predicts a shallow lake's weak stratification and its diurnal cycle. Besides, the model proved reliable on longer time scales, capturing the effect of climate change, anthropogenic water level regulation, and their synergistic interaction on the change of stratification's intensity and duration.