Methodology

Active Fire Detection Algorithm

Fire detection is performed using a contextual algorithm (Giglio et al., 2003) that exploits the strong emission of mid-infrared radiation from fires (Dozier 1981, Matson and Dozier 1981). The algorithm examines each pixel of the MODIS swath, and ultimately assigns to each one of the following classes: missing data, cloud, water, non-fire, fire, or unknown.

Pixels lacking valid data are immediately classified as missing data and excluded from further consideration. Cloud and water pixels are identified using cloud and water masks, and are assigned the classes cloud and water, respectively. Processing continues on the remaining clear land pixels. A preliminary classification is used to eliminate obvious non-fire pixels. For those potential fire pixels that remain, an attempt is made to use the neighboring pixels to estimate the radiometric signal of the potential fire pixel in the absence of fire. Valid neighboring pixels in a window centered on the potential fire pixel are identified and are used to estimate a background value. If the background characterization was successful, a series of contextual threshold tests are used to perform a relative fire detection. These look for the characteristic signature of an active fire in which both the 4 micron brightness temperature and the 4 and 11 micron brightness temperature difference depart substantially from that of the non-fire background. Relative thresholds are adjusted based on the natural variability of the background. Additional specialized tests are used to eliminate false detections caused by sun glint, desert boundaries, and errors in the water mask. Candidate fire pixels that are not rejected in the course of appyling these tests are assigned a class of fire. Pixels for which the background characterization could not be performed, i.e. those having an insufficient number of valid pixels, are assigned a class of unknown.

Burned Area Mapping Algorithm

The MODIS burned area algorithm maps the approximate day of burning using multitemporal land surface reflectance data based on a method described by Roy et al. (2002). The algorithm is applied independently to geolocated pixels over a long time-series of reflectance observations. A bi-directional reflectance model is inverted against multitemporal reflectance observations to provide predicted reflectances and uncertainties for subsequent observations. A statistical measure of the difference between the observed bi-directional surface reflectance (BRF) and the predicted BRF at the viewing and illuminating angles of the observation is used to quantify change from a previously observed state. Large discrepancies between predicted and measured values are attributed to change. A temporal consistency constraint is used to differentiate between temporary changes considered as noise and persistent changes of interest. The Southern Africa Fires and Atmospheric Research Initiative (SAFARI 2000) was selected as the first regional test for a prototype regional 500m MODIS burned area product.

References

Dozier, J., 1981, A method for satellite identification of surface temperature fields of subpixel resolution. Remote Sensing of Environment, 11:221-229.

Giglio, L., Descloitres, J., Justice, C. O., and Kaufman, Y., 2003, An enhanced contextual fire detection algorithm for MODIS. Remote Sensing of Environment, 87:273-282.

Matson, M., and Dozier, J., 1981, Identification of subresolution high temperature sources using a thermal IR s ensor. Photogrammetric Engineering and Remote Sensing, 47:1311-1318.

Roy D.P., Lewis P.E., Justice C.O., 2002, Burned area mapping using multi-temporal moderate spatial resolution data - a bi-directional reflectance model-based expectation approach. Remote Sensing of Environment, 83:263-286.

Send comments to MODIS FIRE User Support at University of Maryland Authorized by Christopher Justice, Fire and Thermal Anomalies Principal Investigator