Automatic Burned Land Mapping From MODIS Time Series Images: Assessment in Mediterranean Ecosystems

Abstract

A novel automatic burned area mapping algorithm for Mediterranean ecosystems based on Moderate-Resolution Imaging Spectroradiometer (MODIS) time series data is presented in this paper. This algorithm is based on a two-phase approach. The first phase detects the most severely burned areas, using spectral/temporal rules computed from dynamic temporal windows. The second phase improves the discrimination of burned areas around those “seed” burned pixels using contextual algorithms based on edge detectors. The use of those filters improved the performance of the contextual algorithm, by refining the discrimination of fire perimeters while restricting the segmentation process. The algorithm was validated over six Mediterranean regions during the fire season of 2003, where reference data was generated using Landsat TM/ETM+ images. Omission and commission errors were below 20%, with an overall Kappa value of 0.846. The validation based on regression scattergraphs of 5 × 5 km grids showed good agreement as well (R2 = 0.972). The standard MODIS burned area product (MCD45A1) showed lower accuracy than the proposed algorithm, with higher omission errors (38.6%) and lower Kappa (0.704) and R2 (0.838) values.