Analysis of the eutrophication in a wetland using a data-driven model

Eutrophication is a major problem in the international Anzali wetland (northern Iran). The present research initially aimed to determine the trophic state index (TSI) in ten sampling sites in the main parts of the Anzali wetland (western, eastern, central, and Siahkeshim parts). After determining the TSI in the wetland, a data-driven method (classification tree model with a J48 algorithm) was implemented to predict the trophic condition in the wetland based on a set of water quality and physical-structural variables. One hundred twenty samples related to chlorophyll-a (the model's output) and environmental variables (the model's inputs) were measured monthly during 1-year study period (2017-2018). Based on the TSI calculation, the western, Siahkeshim, eastern, and central parts of the wetland are classified as eutrophic, super-eutrophic, hyper-eutrophic, and hyper-eutrophic, respectively. When all environmental variables were introduced to the model (with five-time randomization effort, pruning confidence factor = 0.01, and seven-fold cross-validation), eight variables (bicarbonate, pH, water temperature, electric conductivity, dissolved oxygen, total phosphorus, water depth, and water turbidity) were predicted by the model. The model predicted that an increase in total phosphate, water turbidity, and electric conductivity concentration may contribute to the hyper-eutrophic state of the wetland. In contrast, the hyper-eutrophic of the wetland is associated with a decrease in water depth, dissolved oxygen, and pH concentration. According to ANOVA test, the trophic condition in the wetland can be affected by spatial and temporal patterns. Anthropogenic pressures such as the influx of chemicals particularly the nutrients (phosphorus and nitrogen) are the main cause of water enrichment (eutrophication problem) in main parts of the Anzali wetland ecosystem.