Based on Landsat remote sensing imagery and in-situ measurements, machine learning methods were used to construct inversion models for total nitrogen (TN) and total phosphorus (TP) mass concentrations across the entire Poyang Lake. Combined with water quality data from the five rivers (the Ganjiang, Fuhe, Xinjiang, Raohe and Xiushui rivers) flowing into the lake, the spatiotemporal distribution characteristics of TN and TP mass concentrations in Poyang Lake from August 2021 to January 2023 were systematically analyzed. The results show that the TN mass concentration inversion model constructed using a back propagation (BP) neural network achieves a good fit ( R 2=0.91), and the TP mass concentration inversion model based on the GA-optimized BP neural network (GA-BP) performs well ( R 2=0.82). The TN and TP mass concentrations in Poyang Lake exhibit strong seasonal variation, and the water quality in the wet season is generally better than that in the dry season. The spatial distribution of TN and TP mass concentrations in the lake area during the wet season shows strong heterogeneity. Areas with high TN and TP mass concentrations in Poyang Lake are mainly concentrated in the central and northeastern parts of the lake area and the entrances of the five rivers flowing into the lake. The extreme drought exacerbated water quality deterioration in Poyang Lake, leading to an increase in the TN and TP mass concentrations across the entire lake.