Investigating the thresholds of groundwater table depth and salinity in soil salinization zones is crucial for effective salinization management and ecosystem stability. The Salinization Detection Index (SDI) model, developed using remote sensing image data, enables rapid interpretation of the spatial distribution of soil salinization and can be applied to identify groundwater ecological thresholds, thereby significantly improving the efficiency of salinization control. In semi-arid and semi-humid regions, soil salinization is often associated with lakes and marshes, while the traditional SDI model exhibits limited applicability for the study area with surface water bodies. This study takes the Songnen Plain as a typical research area. Based on field ecological, hydrogeological, groundwater investigation and sampling analysis, the traditional SDI model was improved by incorporating the Tasseled Cap Transformation (K-T transform) fusion of multi-source data. The obtained WNSAI (WI + NDVI + SI + Albedo index) model demonstrates a significantly improved correlation with measured soil salinity, and the interpretation accuracy has been improved by 32% compared with the traditional SDI model. The WNSAI model was used to interpret the soil salinization spatial distribution, and groundwater ecological thresholds were efficiently determined. The ecological threshold of groundwater table depth was >= 4.0 m, and salinity was <= 860 mg/L. By synergistically controlling these two indexes, a scientific and systematic soil salinization control system can be established. The findings of this study contribute to enhanced identification and precise prevention and control of soil salinization, provide deeper insight into the mechanism by which groundwater influences salinization evolution, and offer scientific support for the precise prevention and control of soil salinization in semi-arid and semi-humid regions.