Abstract: To enhance the adaptability of eutrophication assessment and prediction in Poyang Lake, monthly monitoring data from representative locations within the lake area from 2012 to 2020 were selected as model training samples. Key physicochemical parameters of the lake were selected as independent variables for the model. The water quality integrated index (WQII) was calculated to assess the water quality changes in recent years. A multiple linear stepwise regression equation (MLSR) with Chl-a as response variable, and a seasonal autoregressive summation moving average model (SARIMA) were established respectively. The concentration values of Chl-a from June to August 2020 were predicted and compared with the measured values to assess the applicability of the two models. Results indicated an improvement in the overall water quality of the representative monitoring sites in Poyang Lake in 2018, with better conditions during flood season. The average WQII ranks in an order of Kangshan (2.91), Duchang (3.01), Banghu (3.11), and Shehan (3.31). Moreover, the SARIMA model demonstrated higher accuracy in predicting Chl-a concentrations compared to the MLSR equation, thereby offering an optimized theoretical framework for early warning of algal outbreaks in large river-connected lakes.

Key words: river-connected lake, WQII method, chlorophyll-a, SARIMA, stepwise regression