Assigning dates to burned areas in Portugal based on NIR and the reflected component of MIR as derived from MODIS

Abstract

Accurate information about location and extent of burnt area is required and of particular interest for the scientific communities dealing with meteorological and climate models in what respects to reliable estimations of biomass burned. An automated procedure is here presented that allows identifying and assigning dates of occurrence to burned areas in Portugal using daily reflectance from NIR (near-infrared) and MIR (middle infrared) bands, obtained from the MODIS instrument on-board Aqua and Terra satellites. The algorithm detects persistent changes in the so-called (V, W) Burned Index time series, and the day of maximum change is then identified as the burning day. The procedure was applied to the extreme summer of 2005 and results were validated against a reference map derived from Landsat imagery. Comparison between the burned map as obtained using the developed procedure and the reference map resulted in a Proportion Correct of 96%. Probability of Detection was 63%, meaning that the algorithm was able to identify almost two thirds of the scars occurred in 2005. An assessment of the temporal accuracy of the dating procedure was also conducted. Results show that 75% of burnt pixels were correctly dated by the algorithm with differences to the hotspots dates less than five days.