Coal fires are a persistent threat to major coal-producing countries in the world. Thus, it is very important to delineate the potential risks of coal fires. This study presents a method for the delineation of coal fire risk areas from Landsat-8 TIRS data. Land surface temperatures (LSTs) were first retrieved from the Landsat-8 TIRS images. The degree of spatial autocorrelation among these LSTs was then identified using local Moran's I statistic. Thermal-related anomalies for the delineation of coal fire risk areas were identified by setting the MEAN+2*SDEV (SDEV is the standard deviation), MEAN+3*SDEV, and MEAN+4*SDEV formulas as thresholds on the local Moran's I statistic. These coal fire risk areas were finally validated using known coal fire sites and cross-validated by comparing them with those obtained from hot spot analysis. A case study of the Na Duong coal field (northern Vietnam) has shown that coal fire risks at moderate and high levels were mainly detected in the center of the coal field. The higher values of local Moran's I statistic, the higher levels of coal fire risks. These coal fire risks were mainly concentrated around known coal fire sites. These results reveal that Landsat-8 TIRS data can effectively delineate coal fire risk areas.