Short-term prediction and attribution analysis of algal bloom in lower Hanjiang River based on LSTM

In order to timely warn of algal blooms in the lower reaches of Hanjiang River and explore the preventive measures for algal blooms in the Hanjiang River, a long short-term memory neural network model was constructed to simulate the change process of Chl-a concentration in the water body near the Xiantao Hydrological Station in the lower reaches of Hanjiang River. The contribution of ammonia nitrogen, total phosphorus, water temperature, flow and temperature changes to Chl-a concentration in the water body during the algal bloom were evaluated through scenario analysis. The results showed that the LSTM neural network model could simulate the change of Chl-a concentration well. The concentrations of ammonia nitrogen and total phosphorus in water bodies show a significantly positive correlation with Chl-a concentration, while flow rate exhibits a significantly negative correlation with Chl-a concentration. Chl-a concentration demonstrates a unimodal variation pattern with water temperature, which isinitially increasing then decreasing as temperature rises. In contrast, it exhibits a nonlinear response to temperature changes, initially enhancing and promoting the effect, then weakening it, and finally promoting it again as temperatures rise. The change of water temperature and flow has a relatively large impact on the change of Chl-a concentration in the water body. The research results can provide a reference for the early warning and prevention of algal blooms in relevant departments.