To optimize the management of non-point source (NPS) pollution in a changing environment, the cellular automata (CA) and Hydrological Simulation Program-Fortran (HSPF) models were used to study the response mechanism of NPS pollution loads to land use change and different precipitation scenarios. Taking the Dongjiang River Basin as a case study, the land use situation and its spatial distribution patterns in 2020, 2030 and 2050 were predicted by the logistic regression-based CA model. The trends of the NPS pollution loads under different land use and precipitation scenarios were quantitatively evaluated. The results show that the total accuracy of the land use change simulated by the CA model was 81%. Both the HSPF model and the CA model were highly applicable to this basin. Precipitation is proven to be the main driving force of NPS pollution. From 2020 to 2050, the annual load, average monthly load, maximum and minimum monthly load of the BOD and TP show an upward trend. TN shows a slight downward trend, which is related to the reduction in cultivated land area and the use of nitrogen fertilizer. In view of the future trend of NPS pollution, the basin should continue to control TN pollution and focus on strengthening BOD and TP control to achieve high-quality management of the water environment.

为了优化在不断变化的环境中非点源（NPS）污染的管理，使用了元胞自动机（CA）模型和Fortran水文模拟程序（HSPF）模型来研究NPS污染负荷对土地利用变化和响应的响应机制。不同的降水情景。以东江流域为例，运用基于logistic回归的CA模型预测了2020年，2030年和2050年的土地利用状况及其空间分布格局。定量评估了不同土地利用和降水情景下的NPS污染负荷趋势。结果表明，CA模型模拟的土地利用变化的总准确性为81％。HSPF模型和CA模型都非常适用于该盆地。降水被证明是NPS污染的主要驱动力。从2020年到2050年，BOD和TP的年负荷，平均月负荷，最大和最小月负荷呈上升趋势。TN略有下降趋势，这与耕地面积减少和氮肥的使用有关。鉴于NPS污染的未来趋势，流域应继续控制TN污染，并重点加强BOD和TP的控制，以实现水环境的高质量管理。