Predicting the short-time-scale variability of chlorophyll a in the Elbe River using a Lagrangian-based multi-criterion analog model

Early prediction of excessive phytoplankton growth is especially important in rivers that are sources of potable water. It is difficult to predict algal growth accurately due to high variability of chlorophyll a over a brief period in rivers. To address this issue, we proposed an analog model based on a Lagrangian method. Despite a lack of mechanistic details, the model also accurately predicted hourly chlorophyll a concentrations at the Geesthacht Weir station on the lower Elbe River. In the analog model, water temperature (Wt), silica (Si), light intensity (Li), and chlorophyll a (Chl-a) were evaluated as impact factors both individually and in combination. A synthetic index was developed to serve as a multi-factor criterion and was evaluated based on weighted similitude indexes (SIs) of the four impact factors (Wt, Si, Li, and Chl-a). Factor weights were selected based on comparative analysis and were manually set to 0.5, 0.25, 0.5, and 1. Data preprocessing resulted in substantially improved prediction accuracy. The use of appropriate weights and SI values in our model allowed the model's three-day predictions to accurately describe hourly variability in chlorophyll a concentrations. Through analysis of parameter sensitivity, we verified that model predictions were robust to parameter variation.