Application of the Dozier retrieval to wildfire characterization: a sensitivity analysis

• Published in: Remote sensing of environment 2001-07, Vol.77 (1), p.34-49
• Main Authors: Giglio, Louis, Kendall, Jacqueline D.
• Format: Article
• Language: English
• Subjects: Applied geophysics; Dozier retrieval; Earth sciences; Earth, ocean, space; Engineering and environment geology. Geothermics; Exact sciences and technology; Fire; Fire area; Fire detection; Fire temperature; Internal geophysics; Pollution, environment geology; Sensitivity analysis
• ISSN: 0034-4257; 1879-0704
• DOI: 10.1016/S0034-4257(01)00192-4

The demand for improved information on regional and global fire activity in the context of land use/land cover change, ecosystem disturbance, climate modeling, and natural hazards has increased efforts in recent years to improve earth-observing satellite sensors and associated methods for fire information retrieval. While sensor development has made considerable headway for fire monitoring with the recent launch of Terra in late 1999 and upcoming missions such as bispectral infrared detection (BIRD), FOCUS, and the National Polar-Orbiting Operational Environmental Satellite System (NPOESS), the retrieval of fire properties from satellite observations remains problematic. Current methods for determining properties of actively burning fires, for example, continue to be based on a simple two-component model developed by Dozier in 1981 [Remote Sens. Environ. 11 (1981) 221.]. Its limitations in terms of reliable operational fire property retrieval have been known for some time, although not examined in detail to date. With new measurements from recently deployed and evolving satellite systems, which offer key advantages for fire remote sensing, it is appropriate now to more comprehensively examine the heritage methodology for fire property retrieval. The results of a sensitivity analysis indicate that under realistic conditions the random errors in fire temperature and area retrieved using Dozier's method are ±100 K and ±50% at one standard deviation, respectively, for fires occupying a pixel fraction greater than 0.005 (this corresponds to a 5000-m 2 fire within a 1-km pixel). For smaller active fires, larger random and systematic errors are likely to occur.