Estimation of Biomass Burned Areas Using Multiple-Satellite-Observed Active Fires

Abstract

Biomass burning releases a significant amount of trace gases and aerosols into the atmosphere and affects climate change, carbon cycle, and air quality. Accurate estimates of emissions depend strongly on the calculations of burned areas. Here, we present an algorithm that is used to derive burned areas by blending active fire observations from multiple satellites which are provided in the Hazard Mapping System (HMS). The HMS consolidates automated fire detections from Geostationary Operational Environmental Satellite (GOES) Imager, Advanced Very High Resolution Radiometer (AVHRR), and MODerate resolution Imaging Spectroradiometer (MODIS). Our goals are to derive burned areas in each GOES fire pixel across contiguous United States (CONUS) from 2004 to 2007 and to validate the estimates using Landsat Thematic Mapper/Enhanced Thematic Mapper plus (TM/ETM+) burn scars and National Fire Inventory data. The results show that annual fire events burn 0.4% <formula formulatype="inline" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex Notation="TeX">$(3.4\times \break 10^{4}\ \hbox{km}^{2})$</tex></formula> of total land across CONUS, which consists of 0.49% of total forests, 0.64% of savannas, 0.68% of shrublands, 0.40% of grasslands, and 0.30% of croplands. The large burned areas are dominantly distributed in the western CONUS, followed by the states in the southeast region and along the Mississippi Valley. Extensive validation shows that MODIS+AVHRR+GOES instruments greatly improve the determination of fire duration and fire detection rate compared to single instrument detections. The detection rate of small fire events <formula formulatype="inline" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex Notation="TeX">$(&lt; 10 \ \hbox{km}^{2})$</tex></formula> from multiple instruments is 24% and 36% higher than that from MODIS and GOES, respectively. The error in the burned-area estimate is less than 30% in individual ecosystems, and it decreases exponentially with the increase of burn scar size. Overall, the accuracy of total burned area across CONUS is 98.9% when compared to TM/ETM+-based burn scars and 83% when compared to national inventory data.