It is difficult to detect small and cool fires using current remote sensing algorithms because these fires do not emit sufficient radiation to penetrate dense canopies and cannot be easily distinguished from non-fire background radiation. To date, most algorithms are designed for global fire detection, and rely on identifying hot spots using thermal infrared (TIR) channels. The limitation of that technology is that false alarms are occasionally generated over certain surface types during the day time, and small, cool fires are oftentimes missed using relatively high thresholds optimized for global fire detection.