Thematic accuracy of hotspots and wildfires in Michoacán, Mexico

Introduction: Wildfire monitoring programs were first implemented in Mexico at the end of the last century; however, there are no studies that report their accuracy in detecting fires and hotspots. Objective: To evaluate the thematic accuracy of hotspots and wildfire records in a region of Michoacan de Ocampo, Mexico. Materials and methods: Rapid Eye high spatial resolution images were classified to detect fire scars. The images were crossed with hotspot maps from the MODIS sensor, processed by the National Commission for the Knowledge and Use of Biodiversity (CONABIO), and from the MODIS, AVHRR and GOES sensors, processed by the National Meteorological Service (SMN), as well as with wildfire records from the National Forestry Commission (CONAFOR). Detection rates and confusion matrices were generated to assess accuracy. Results and discussion: The detection rate was higher with the CONABIO MODIS sensor (>75 %) in categories greater than 50 ha. The overall accuracy of hotspots was 62 %; however, the errors of omission (underestimation) and commission (overestimation) were 85 and 34 %, respectively. The fires reported by CONAFOR lack spatial coincidence with respect to scars, so the accuracy analysis was discarded. Conclusion: The use of hotspots, according to detection rates, is ideal for fires exceeding 50 ha. The thematic accuracy of hotspots and fire records should be evaluated at the beginning of a study.