Abstract: The land surface temperature and cropland information of Fujian were retrieved based on MODIS data. The land surface temperature retrieval model and cropland classification system were derived respectively by using the split windows algorithm and expert decision tree method. Furthermore, the low temperature risk assessment model of cropland was established by Surfer and ArcGIS software. The results showed that land surface temperature monitoring had high accuracy (83.56%) with the split windows algorithm. There was same in distribution of high and low temperature to observed data from meteorological stations, which could provide fine temperature difference and make up insufficient numbers of meteorological stations. The cropland information was extracted exactly from normalized difference vegetation index (NDWI) according to expert decision tree method. Three grades of low temperature risk assessment for major crops in Fujian province were divided, which included light (0.45-1.00), medium (0.24-0.45) and heavy (0-0.24) by using normalization method. The results revealed the surface low temperature layout and risk pattern of crops in detail, which could provide reference for crops regionalization and low temperature risk assessment.

Key words: MODIS, Low temperature, Remote sense, Risk assessment