Four decades of phosphorus pollution history in the Yangtze River revealed through Landsat monitoring

Total phosphorus (TP) is a key indicator of eutrophication in aquatic ecosystems. However, long-term and large-scale monitoring of TP remains challenging, particularly in rivers due to their complex hydrodynamic conditions and optical water heterogeneity, as a consequence of river catchment conditions. In this study, we propose a novel framework based on Landsat images and field data that integrates optical water type (OWT) with an extreme gradient boosting (XGBoost) model to estimate TP concentrations in the Yangtze River from 1984 to 2024. The results demonstrate that the OWT-XGBoost model, calibrated according to four OWT related to turquoise, greenish and brownish waters, effectively retrieves TP concentrations. Specifically, the models achieved validation correlation coefficients (r) of 0.71, 0.86, 0.81, and 0.81 for OWT3-OWT6, respectively, with corresponding mean absolute percentage errors (MAPE) of 10.48 %, 9.99 %, 9.21 %, and 8.83 %. Landsat-derived TP concentrations in the Yangtze River over the past four decades have shown a decreasing trend at a rate of 2 × 10−4 mg/L/yr, accompanied by significant spatial heterogeneity and seasonal fluctuations. A distinct spatial pattern of “high upstream, low downstream” was observed, with average TP concentrations of 0.063 mg/L, 0.051 mg/L, and 0.051 mg/L in the upper, middle, and lower reaches, respectively. Surface runoff and phosphorus fertilizer application were identified as the primary drivers of TP variability. Furthermore, the construction of the Three Gorges Dam has profoundly altered downstream phosphorus dynamics by modifying sediment retention and flow regimes. In the context of global change, this study offers important insights into the long-term, basin-scale monitoring of river TP concentrations using Landsat data.