In coal-bearing areas of the circumpolar North, a region rich in carbonaceous deposits, coal outcrops on south-facing slopes are particularly vulnerable to catching fire as they receive substantial amounts of solar radiation during the long summer days. In this study, we use remote sensing to map thermal anomalies associated with coal fires in a coal field in interior Alaska, following a two-step process: (1) thermal anomaly detection on individual thermal infrared (TIR) images from the Landsat satellites; and (2) persistent anomaly detection from a time series of these TIR images. In step 1, a Gaussian mixture model combining two independent normal distributions is fitted to every TIR scene. Subsequently, pixels are identified as thermally anomalous (hot spots) if their surface temperature is more than 4 standard deviations above the mode of the dominant normal distribution. After this individual analysis, the full series of TIR images is stacked and an anomaly occurrence index is calculated for each anomalous pixel. Only pixels that have valid data and are not masked out due to clouds or active fire are taken into account. Finally, the distribution of anomaly occurrence indices of the image stack is plotted over an interval between 0 and 1, and by observing the falloff of the distribution a pixel is counted as persistently anomalous if it appears in more than one-third of the scenes in which it is present. This automated processing flow yields coal fire hazard maps that are useful for fire and forest managers and commercial operators.