Implications of a quantitative fire radiative power distribution for multi-sensor fire activity estimation

Abstract:

Satellite-based estimation of fire activity is hindered by limitations of detection of smaller fires. Satellite-based inventories can be reasonably accurate at some scales, and this is because fire activity has strong spatial and temporal correlation permitting smaller fires to be extrapolated based on observed larger fires in many fire regimes. In order to combine observations with different sensitivity, a quantitative model must describe the relative prevalence of fires of different sizes. One such model developed at NRL for application to satellite retrievals of fire radiative power (FRP) shows good results for correction of angular variation in sensitivity of MODIS and VIIRS FRP retrievals. These polar orbiting wide-swath satellite sensors detect fires at view angles from nadir to more than 60 degrees, which corresponds to a factor of 8 change in the land surface footprint per detector. The NRL model is designed to correct retrieved FRP to account for fires below the detection threshold based only on the pixel footprint size and the effective atmospheric transmittance in the 4 micron and 11 micron bands of the sensor.

This paper describes the construction of the NRL FRP distribution model, its application to correcting MODIS and VIIRS fire detection data, and the results of applied corrections for single-sensor and multi-sensor consistency of estimated fire activity.

This paper also includes a theoretical discussion of the implications of this FRP distribution model for additional sensors. The capability of this model to harmonize observations from different sensors is examined in detail to identify relevant sensor properties not included in the current model. Further, the implications for describing the distribution of FRP as it would be observed by sensors with spatial resolution better than 375m are discussed.