How much carbon and aerosols are released during Santa Ana fires?

Mapping combustion completeness and mortality from remote sensing data (JPL)

We will make use of high spatial and spectral resolution airborne data in Southern California collected by Ikhana/Wilfire sensor and MASTER (Fig. 3) with spatial resolutions of 5-15 meters to relate the field measured combustion completeness (CC) and mortality estimates with spectral data and severity map. The wildfire sensor has 12 spectral channels from the visible through thermal infrared (0.4-14 !m) and MASTER has 50 channels of data over the same spectral region (Hook et al. 2001). NASA Ikhana has acquired airborne data sets over severe wildland fires since 2007 in response to emergency requests. We have compiled all these airborne data as part of previous projects. The data will be preprocessed for georegistration and removing atmospheric effect using a radiative transfer model (MODTRAN and ATCOR). Various indices will then be produced from the surface reflectance, emissivity and temperature data. Satellite data from ASTER, Landsat (5, ETM+) and MODIS will also be acquired and processed in a similar manner. We will then use a range of indices such as Normalized Burn Ratio (NBR) to map burned area and severity (Epting et al. 2005; Smith et al. 2007a,b). The higher resolution airborne data will be used to validate the satellite results. We will also apply the spectral mixture analysis (SMA) method to delineate burned areas, to calculate the fractional cover of vegetation and char within burned pixels, and to infer CC and mortality (Smith et al. 2007a,b; Cochrane and Souza 1998; Hudak et al. 2007). We will focus on reflectance endmembers for four cover types: green vegetation, non-photosynthetic vegetation, bare soil, and burned vegetation (areas with char and ash). We will also identify a shade based cover in areas of rugged terrain. We will use reflectance endmembers that sample a range of species and floor coverings and that are drawn from our field measurements, two reference libraries for Southern California fires from Fire Research and Management Exchange System, and a vegetation library from Dar Roberts at UCSB. Reference library spectra will be manipulated to derive candidate reference endmembers for each cover type using optimization techniques (Rogan and Franklin 2001). We will explore logistic regression methods to calculate combustion completeness and mortality from spectral reflectance and the various indices, based on the field measurements of CC and mortality. We will evaluate the performance of these algorithms across a range of vegetation types and fires and select the best algorithm for mapping. We then collect Landsat TM/ETM data from 1983 covering southern California, which are freely available from the USGS Global Visualization Viewer (http://glovis.usgs.gov/). At least one cloud-free imagery in wet growing season and one after typical fire season for each Landsat path/row will be selected for each year. These data will be georectified and corrected for atmospheric effects and sensor variability. The derived burn extent, CC and mortality from recent years will be compared with those aggregated from Wildfire and MASTER high resolution maps. The Landsat burned area maps will also be compared with FRAP fire perimeters on an annual basis. This will give us knowledge on the accuracy and the consistency of the FRAP data through time. This 30 years of fire history will allow us to better understand the various aspects of fire characteristics, how differently vegetation recovers under various fire patterns, and to better quantify the spatial and temporal pattern of emissions.