The response of spatial and temporal vegetation changes to runoff is a complex process involving the interaction of several factors and mechanisms. Timely and accurate vegetation and runoff change information is an important reference for the water cycle and water resource security. The Ebinur Lake Basin is representative of arid areas worldwide. This basin has been affected by climate change and human activities for a long time, resulting in the destruction of the basin’s ecological environment, and especially its vegetation. However, there have been few studies that have focused on watershed vegetation and runoff changes. Therefore, we combined Generalized Information System and remote sensing technology, used SWAT and InVEST models based on the Google Earth Engine platform, and used the vegetation normalization index method to calculate the spatial distribution of vegetation and water production from 2000 to 2020 in Ebinur Lake. Sen’s trend analysis and the M–K test were used to calculate vegetation and runoff trends. The relationship between vegetation and runoff variation was studied using bivariate spatial autocorrelation based on sub-basins and plant types. The results showed that the Z parameter in the InVEST model spanned from 1–2. The spatial distribution of the water yield in a watershed is similar to the elevation of the watershed, showing a trend of higher altitude leading to a higher water yield. Its water yield capacity tends to saturate at elevations greater than 3500 m. The local spatial distribution of the Normalized Difference Vegetation Index(NDVI) values and water yield clustering in the watershed were consistent and reproducible. Interannual runoff based on sub-basins correlated positively with the overall NDVI, whereas interannual runoff based on plant type correlated negatively with the overall NDVI.