With the development of optoelectronic technology, low-light-level (LLL) imaging technology and its application have recently become a research focus. In LLL remote sensing images, it is difficult to completely separate forest fires from industrial combustion and urban heat sources based on radiance or temperature alone. Meanwhile, because of frequent data saturation in the low-light band, the existing fire detection products only provide detection information. To identify forest fires among various heterogeneous heat sources with high brightness and further improve the pixel-level characterization of fire detection products, a new spectral index, the enhanced noctilucent fire disturbance index (ENFDI),is proposed based on the principle that surface temperature decreases with the increasing vegetation density by Latent heat transfer. According to the results, ENFDI enhances the differences in spectral characteristics between forest fires and city lights and improves the ability of forest fire identification under low-light conditions. The forest fires’ ENFDI are significantly higher than those of urban heat sources.Moreover, ENFDI can also effectively relieve the impact caused by LLL band’s proneness to saturation. Not only can ENFDI clearly distinguish flame glow differences within potential saturation zones and enhance the distinguishability of forest fire’s pixels, but the correlation (R) between ENFDI and the mid-and far-infrared brightness temperature difference is as high as 0.94–0.97, which is considerably higher than that of NTL (0.82-0.83).Furthermore, ENFDI is relatively stable —it is not affected by lunar phases in that forest fires at night are identified with or without moonlight. ENFDI recorded an 87.66% forest fire identification accuracy in this study, which is higher than the 83.91% accuracy of the conventional TMIR method. The forest fires identified using the ENFDI show a good overall correspondence with the NPP/VIIRS active fire product (VNP14IMG), with a positional tolerance within 628 m. Therefore, ENFDI is sensitive, stable and accurate for identifying forest fires. Further, it may serve as a feasible reference for achieving further pixel-level characterization of fires.