Abstract: It is universally acknowledged that coalfield fire has direct impacts on the efficient utilization of coal resources and economic development,so does the coalfield of east Junggar Basin in Xinjiang Uyghur Autonomous region. In this paper,Land surface temperature(Ts)and normalized difference vegetation index(NDVI) were obtained from Landsat-8 satellite data,and were used to create the feature space of Ts-NDVI,then a dryness index-Temperature Vegetation Dryness Index(TVDI)was calculated through the dry and wet line equation. The accuracy was verified and the reference upper boundary was determined by adopting MODIS LST product data. Through quantitative analysis of the information on surface heat anomalies using remotely sensed data,the dynamic variation rules and tendencies of sub coal fires were summarized. Because TVDI is much more sensitive to Ts than NDVI,extracting the coalfield fire boundary with TVDI method in the coalfield with fewer vegetation coverage is better. The results show as follows:(1)the correlation coefficient of the inversion value verified by using field measured soil moisture and measured value is R²=0.66,this demonstrates that the accuracy of TVDI is acceptable;( 2)TVDI presents a high sensitivity to temperature,while a lower sensitivity to NDVI,thus can be used to identify the fire area of high temperature;(3)the biggest area extracted by TVDI-MW method is 5.54 km², and the longest boundary is up to the 93.7 km extracted by TVDI-SW2. TVDI method performs well in extracting the information in uncomplicated underlying surface,however,the coal dust could be deposited at the land surface and the root of vegetation around the mining area,which may result in a higher value of Ts and also disturb the accuracy of Ts. Therefore,there is a wide range of high-temperature coal fire area,and the potential deterioration may be more serious.

Key words: Landsat-8, Temperature Vegetation Dryness Index, boundary extraction, coalfield fire