**2. Remote sensing of fire**

Information derived from satellites has many advantages to traditional in situ data collection. A global perspective can be achieved to observe various Earth system processes allowing monitoring of spatially distinct, inaccessible, or remote locations. Regular monitoring is possible from polar orbiting satellite platforms such as National Aeronautics and Space Administration's (NASA) Aqua and Terra satellites and National Oceanic and Atmospheric Administration's (NOAA) Advanced Very High Resolution Radiometer (AVHRR). In the case of the former, the combination of Aqua and Terra provide nominally 4 daily "looks" of most locations on Earth. In addition, geostationary satellites such as NOAA's Geostationary Operational Environmental Satellites (GOES) and the European Organization for the Exploitation of Meteorological Satellites' (EUMETSAT) Meteosat provide high temporal resolution (15-30 minute), continental wide observations.

Many of the channels available from a particular satellite sensor are useful for fire monitoring, for example aerosols can be monitored using the the visible and near-infrared bands or burn scars can be monitored with the visible, near, and middle infrared bands. Burned area mapping, a commonly used metric, is important for estimating total biomass consumed and thus emission estimates. Advanced new algorithms for accurate estimation of burned area now address the effect of bi-directional reflectance (Roy et al., 2005); an effect which is a function of the sun-target-view geometry influencing the directional dependence in reflectance and a potential source of error when using time-series data.

The retrieval of fire hot-spots provides additional monitoring and measurement capabilities. The foundation for fire detection is the enhanced middle infrared radiance emitted during flaming or smoldering combustion as described by the Planck function. In the case of an observed fire, radiance values generally peak around 3.7 – 4.0μm, whereas the peak for background terrestrial surface temperatures is near 10.0 – 11.0μm; thus temperature anomalies can be flagged as potential fire "hot-spots". As a result of this rather simple relationship, remotely sensed data has had a significant contribution in fire science and monitoring. Heritage systems such as AVHRR and GOES, though not necessarily intended to include fire detection or monitoring missions, have proven valuable for this task nonetheless (Boles & Verbyla, 2000). Fire characterization from satellites, such as subpixel temperatures and flaming area, was obtained from a method developed by Dozier (1981), who introduced a theoretical procedure that exploits the different responses of two channels (3 and 4) aboard AVHRR (3.75μm and 10.8μm, respectively) for sub-pixel hot spot detection; an approach that set the framework for future sensor fire detection and characteristic methodologies (Giglio & Kendall, 2001; Justice et al., 2002; Giglio et al., 2003; Wooster et al., 2003, 2005).

The application of hot-spot detections has been employed in numerous studies and for a variety of uses. Legg & Laumonier (1999) demonstrated the effectiveness of hot spot detections using AVHRR and ATSR (Along Track Scanning Radiometer), as well as burned

In order to understand the spatial and temporal global distribution of biomass burning, and ultimately the potential impacts to the biosphere and atmosphere, regular, broad scale monitoring is necessary. Satellite sensors provide daily, synoptic observations to detect and analyze fires (Justice et al., 2002) and therefore a great deal of research to characterize fires from remote sensing systems has been performed over the past several decades (e.g. Dozier, 1981; Kaufman et al., 1998; Giglio et al., 2003; Wooster et al., 2005; Ichoku & Kaufman, 2005).

Information derived from satellites has many advantages to traditional in situ data collection. A global perspective can be achieved to observe various Earth system processes allowing monitoring of spatially distinct, inaccessible, or remote locations. Regular monitoring is possible from polar orbiting satellite platforms such as National Aeronautics and Space Administration's (NASA) Aqua and Terra satellites and National Oceanic and Atmospheric Administration's (NOAA) Advanced Very High Resolution Radiometer (AVHRR). In the case of the former, the combination of Aqua and Terra provide nominally 4 daily "looks" of most locations on Earth. In addition, geostationary satellites such as NOAA's Geostationary Operational Environmental Satellites (GOES) and the European Organization for the Exploitation of Meteorological Satellites' (EUMETSAT) Meteosat

Many of the channels available from a particular satellite sensor are useful for fire monitoring, for example aerosols can be monitored using the the visible and near-infrared bands or burn scars can be monitored with the visible, near, and middle infrared bands. Burned area mapping, a commonly used metric, is important for estimating total biomass consumed and thus emission estimates. Advanced new algorithms for accurate estimation of burned area now address the effect of bi-directional reflectance (Roy et al., 2005); an effect which is a function of the sun-target-view geometry influencing the directional dependence

The retrieval of fire hot-spots provides additional monitoring and measurement capabilities. The foundation for fire detection is the enhanced middle infrared radiance emitted during flaming or smoldering combustion as described by the Planck function. In the case of an observed fire, radiance values generally peak around 3.7 – 4.0μm, whereas the peak for background terrestrial surface temperatures is near 10.0 – 11.0μm; thus temperature anomalies can be flagged as potential fire "hot-spots". As a result of this rather simple relationship, remotely sensed data has had a significant contribution in fire science and monitoring. Heritage systems such as AVHRR and GOES, though not necessarily intended to include fire detection or monitoring missions, have proven valuable for this task nonetheless (Boles & Verbyla, 2000). Fire characterization from satellites, such as subpixel temperatures and flaming area, was obtained from a method developed by Dozier (1981), who introduced a theoretical procedure that exploits the different responses of two channels (3 and 4) aboard AVHRR (3.75μm and 10.8μm, respectively) for sub-pixel hot spot detection; an approach that set the framework for future sensor fire detection and characteristic methodologies (Giglio & Kendall, 2001; Justice et al., 2002; Giglio et al., 2003; Wooster et al.,

The application of hot-spot detections has been employed in numerous studies and for a variety of uses. Legg & Laumonier (1999) demonstrated the effectiveness of hot spot detections using AVHRR and ATSR (Along Track Scanning Radiometer), as well as burned

provide high temporal resolution (15-30 minute), continental wide observations.

in reflectance and a potential source of error when using time-series data.

**2. Remote sensing of fire** 

2003, 2005).

area estimates using SPOT (Satellite Pour Observation de la Terra) imagery, during the 1997 Indonesian fire season. Legg & Laumonier (1999) also employed the United States Defense Meteorological Satellite Program (US DMSP) to help eliminate spurious daytime hotspots by detecting highly reflective pixels at night, presumably from fire, while excluding known human related bright spots. Kaufman et al. (1990) used AVHRR fire counts to assess trace gas and aerosol emissions in the tropics. Although some of the assumptions about burned area and fire detections would later be shown to be erroneous, their research set in motion the development of new approaches for using remotely sensed fire information for emission estimates (Kaufman, 1998). Aragão et al. (2007) examined drought and fire spatial distribution in Amazonia using NOAA-12's AVHRR and MODIS hot spot detections and later Aragão et al. (2008) examined specific interactions between precipitation, deforestation, and fires related to the 2005 Brazilian Amazon drought. More recently, Aragão & Shimabukuro (2010), again using MODIS and NOAA-12 hot spots, showed the co-varying nature of deforestation and fire activity trends over the past several years in the Brazilian Amazon. Their research has implications for fire and emission policies such as the United Nation's REDD+ (Reducing Emissions from Deforestation and Degradation). Giglio et al. (2010) used active fire detections to expand their burned area product to pre-MODIS data using the ATSR and Visible and Infrared Scanner (VIRS) aboard the Tropical Rainfall Measuring Mission (TRMM). The active fire burned area was developed using relationships based on regression between MODIS-Terra 500 meter burned area reference maps and Terra active fire counts. Morton et al. (2008) showed a clear correlation between fire hot spot frequency and land use patterns in Amazonia. Their research concluded that trends in land use intensity and fire frequency were linked. Such work offers promise for developing monitoring schemes to characterize land use transitions to inform policy makers.

In addition to the above research, a variety near-real time applications of active fire detections exist. The U.S. Forest Service Active Fire Mapping Program (http://activefiremaps.fs.fed.us/) is an operational system providing invaluable, near-real time information about location and timing of fire activity in the United States and Canada allowing fire managers to efficiently monitor fires and allocate resources. The Fire Information for Resource Management System (FIRMS, http://maps.geog.umd.edu/firms/) delivers timely fire detections, made by MODIS and processed through the Rapid Response System (http://rapidfire.sci.gsfc.nasa.gov/), to fire managers around the world.

Development of "new tools" such as fire radiative energy (FRE) can aid in estimating the biomass combusted and rates of atmospheric loading trace gases and aerosols. Calculated by determining the amount of energy emitted during fire, FRE may offer an accurate measurement of the fire intensity and vegetation consumed per unit time, as will be discussed in more detail below.
