Use of remote sensing-derived variables in developing a forest fire danger forecasting system

Our aim was to develop a remote sensing-based forest fire danger forecasting system (FFDFS) and its implementation in forecasting 2011 fire season in the Canadian province of Alberta. The FFDFS used Moderate Resolution Imaging Spectroradiometer (MODIS)-derived 8-day composites of surface temperature, normalized multiband drought index, and normalized difference vegetation index as input variables. In order to eliminate the data gaps in the input variables, we propose a gap-filling technique by considering both of the spatial and temporal dimensions. These input variables were calculated during the i period and then integrated to forecast the fire danger conditions into four categories (i.e., very high, high, moderate, and low) during the i + 1 period. It was observed that 98.19 % of the fire fell under very high to moderate danger classes. The performance of this system was also demonstrated its ability to forecast the worst fires occurred in Slave Lake and Fort McMurray region during mid-May 2011. For example, 100 and 94.0 % of the fire spots fell under very high to high danger categories for Slave Lake and Fort McMurray regions, respectively.