ABSTRACT
ABSTRACT: On the basis of case analysis, this study selected the daily highest temperature, daily minimum temperature, daily minimum relative humidity and no-precipitation days (day precipitation invalid < 0.3 mm), the daily mean wind speed and sunshine time, the shortest distance from roads and settlement and the NDVI value the eight indexes to build grassland fire early warning model and identify of grassland fire. By testing the model using typical grassland fire cases from 2000 to 2013, the results show that it has the high feasibility of this grassland fire early warning model by using Support Vector Machine (SVM) method. Combining spatial data, this study can realize large scale grassland fire early warning, and this will have great significance on grassland fire management.
