The impact of spatial and spectral resolution on burned area classification accuracy: A multi-sensor approach

Accurate forest fire mapping allows land managers to assess environmental damage and support reforestation efforts. Satellite remote sensing offers significant advantages for environmental monitoring and represents an effective tool for mapping and monitoring fire-affected areas across a wide range of geographic scales. Information on the expected mapping accuracy is essential for selecting the most appropriate satellite imagery characteristics for accurate burned-area delineation. In this study, data from nine remote sensing satellites (ALI, ALOS, ASTER, DMC, IKONOS, Landsat, MODIS, RapidEye, and SPOT-4) were used to evaluate the influence of their spatial and spectral characteristics on burned-area mapping accuracy. The study focuses on the large wildfire that occurred in East Attica, Greece, in August 2009. All satellite images were radiometrically and geometrically corrected and subsequently classified using the Maximum Likelihood Classifier (MLC). The accuracy of the resulting classifications was then assessed and reported. To investigate the effects of sensor characteristics on classification performance, correlation analyses were conducted between image pixel size, number of spectral bands, band separability, and various classification accuracy metrics. The results revealed a clear relationship between the spatial and spectral resolution of satellite data and the accuracy of burned-area mapping, highlighting the importance of sensor characteristics in wildfire assessment applications.