Research on the impact of cloud cover and artificial structures on the effectiveness of monitoring the forestry fire situation by MODIS spectroradiometer satellite images was carried out. It is shown that the regular fire detection algorithm MOD14 often assumes reflections at the dense clouds edges as thermal anomalies. The false detection of the hot pixels by the adaptive algorithm proposed by us occurs only when these are the edges of a very dense cloud cover. It is shown that the probability of false alarm from the adaptive algorithm is 15 times less than that from the MOD14 algorithm. Improving the efficiency of fire detection zeroing of cloud areas of the image is proposed. To this end, the MOD35 algorithm to build cloud masks is used. Thus, the use of the adaptive algorithm allows one to exclude the false hot pixels. As for the interference of artificial structures, it is proposed to exclude them by means of county masks based on Landsat images. It is shown that the use of the proposed masks significantly reduces the false thermal anomalies.