Item

Abstract

We screened the algorithm 2 (308 K threshold) European Space Agency (ESA) World Fire Atlas (WFA), for the period 1997–2002, using ancillary land cover, night-lights and volcanic activity datasets, combined with statistical techniques to detect the occurrence of space-time clusters of anomalous observations. The WFA is built using night time data from the Along Track Scanning Radiometer (ATSR) onboard the Second European Remote-Sensing Satellite (ERS-2). The spatial resolution of the data is 1 km and the satellite revisiting period is 3 days at the equator. The WFA is the first and longest archive of global fire observations and has been used in numerous biomass burning studies. Known limitations of the WFA are the inclusion of warm surfaces, gas flares, and city lights, and an underestimation of actual global fire activity, due to the time of satellite overpass. Nevertheless, it has been considered that the WFA contains a relatively small proportion of observations that do not correspond to vegetation fires, which is not corroborated by our findings. During the study period, the annual percentage of false alarms and non-vegetation fires varied from a minimum value of 20.6% in 1997 to a maximum of 27.9% in 1998. Gas flares and hot bare soils are the major sources of false alarms and non-vegetation fires.