Abstract: Due to the different spatial and temporal resolution of different sensors, the research period of NDVI dynamic change is relatively short. Based on EOT algorithm, taking Danjiangkou Water Source Area as an example, with the help of GIMMS NDVI and MODIS NDVI data in overlapping period, the 8 km resolution GIMMS NDVI data is resampled to 1 km, and the long-term NDVI data set of 1982-2018 is constructed. On this basis, the temporal and spatial dynamic changes of NDVI in Danjiangkou Water source area are analyzed. The results show that EOT algorithm has good applicability in spatial resampling. The fit between EOT NDVI and MODIS NDVI was higher (r=0.929, P<0.01, n=24), and there was a significant correlation between EOT NDVI and Landsat NDVI (r=0.702, P<0.01, n=200). In space, the EOT algorithm is affected by the mixed pixel in the area with complex land use types, the prediction accuracy is reduced, and higher in the area with continuous vegetation distribution. In space, the EOT algorithm is affected by the mixed pixel in the area with complex land use types, the prediction accuracy is reduced, and higher in the area with continuous vegetation distribution. Based on the analysis of time series NDVI data, the NDVI showed an upward trend of 0.002 9 year-1. Spatially, the NDVI in Danjiangkou Water source area increased significantly (P<0.05), and the annual average NDVI increased significantly in Hanzhong City, Ankang City, Shangzhou city in the north, Zhuxi County in the South and other regions. The areas with decreased annual NDVI are mainly distributed in Danjiangkou reservoir, Shiyan City, Nanyang City, Pingli County and other areas with rapid economic development. The Mann Kendall test showed that there was no mutation in the annual NDVI time series, but the NDVI in some regions was different at different times.