Abstract:Soil salinization in the Yellow River Delta (YRD) is a key factor restricting local agricultural productivity and ecological stability. In order to accurately grasp the spatial distribution of saline soils and improve the spatial prediction accuracy of soil salinity, based on measured soil salinity at 193 sampling points and two soil layers in the YRD in May, 2022, and data such as digital elevation model and Landsat 9 remote sensing imagery were combined, the Geographically Weighted Regression (GWR) model was used to construct interval-type soft data, and then the Bayesian Maximum Entropy (BME) model was established to predict the distribution of soil salinity in the study area, and finally the prediction results were compared with the traditional geographic statistical model Ordinary Kriging (OK) and the GWR model. The results showed that the prediction accuracy of the BME model was higher than those of the other two models. Compared with OK, the prediction errors (RMSE) of BME in the soil surface (0–30 cm) and bottom (90–100 cm) layers were decreased by 25% and 21%, respectively, and the R2 were improved by 0.543 2 and 0.352 7, respectively. As the best spatial prediction model in this study, BME showed the advantages of multi-source data integration and nonlinear estimation. The salinization rate (88%) of the surface layer in the YRD was higher than that of the bottom layer (68%). Generally, soil salinity was increased from southwest to northeast, the coastal areas were more serious than the inland areas, and soil salinization was most prominent in the northern part of the YRD.