Evaluating fire growth simulations using satellite active fire data
Ana C. L. Sá, Akli Benali, Paulo M. Fernandes, Renata M. S. Pinto, Ricardo M. Trigo , Michele Salis, Ana Russo, Sonia Jerez, Pedro M. M. Soares ,Wilfrid Schroeder, José M.C. Pereira
Remote Sensing of Environment, DOI: http://dx.doi.org/10.1016/j.rse.2016.12.023
The recognized shortcomings of fire spread modeling systems and the widespread use of their outcomes in fire management decisions render the evaluation of fire simulation results crucial for model calibration and improvement. Additionally, new methods are essential to avoid model misapplication in fuel and fire management decision processes. This study proposes an exploratory evaluation framework of fire growth simulations using satellite active fire data. It uses nine very large fires that occurred in Portugal between 2003 and 2012. Their fire growth was simulated using the Fire Area Simulator (FARSITE) and compared with active fire data from theMODerate-resolution Imaging Spectroradiometer (MODIS). The evaluation is based on two spatial measures that quantify the absolute (SpD) and relative (NRSpD) spatial discrepancies between fire growth simulations and satellite active fires. Both measures account for the uncertain location of the fire front(s) inside the active fire footprint. Results highlight the contribution of the spatial discrepancy measures to locate areas of low/high spatio-temporal agreement between simulated fire growth and MODIS active fires, thereby aiding the assessment of potential sources of simulation error. Results also show that despite the coarse spatial resolution of MODIS active fires, these data are able to capture the spatial dynamics of fire growth. Limitations on the spatial measures were identified, particularly the lack of independence of evaluations using the NRSpD. In spite of being exploratory, this study represents a novel contribution to the evaluation of fire growth simulations because: 1) it uses satellite active fire data and is independent of the collection of reference burnt area perimeters; 2) it proposes two simple quantitative spatial discrepancy measures; and 3) it can be applied to a large number of wildfires, regardless of their geographical location. This innovative approach represents a potential cost effective evaluation scheme, that can be used systematically whenever evaluation of a large number of large wildfires is required and reference burned area perimeters are not available.