Post-fire burned areas and impervious features mapping with two new spectral models: conception and complementarity

Sentinel2A/B-Multispectral Instrument, S2-MSI, spectral domain is rich in potential for land surface cover changes, including fire disasters-related dynamics. So far, a dual and complementary pair of spectral indices has not yet been proposed to map burned area scar and extract included impervious features. As a tentative solution to this, we develop a 10m-resolution Sentinel2-Normalized Difference Charred Land Cover Index, NDCLCIS2 plus an Imperviousness Discrimination Burned Area Index, IDBAI. Both show a strong correlation to existing indices recording coefficients of determination, R2 up to 0.92 and 0.56 respectively, with algorithmically closest indices. NDCLCIS2 yields satisfactory spectral discrimination index values between burned classes, reaching, ɱ= 2.752, versus ɱ= 2.134 for the best challenging index. IDBAI records an overall accuracy, up to, OA=0.979, while crossing the training seed with S2-MSI sampling. In addition, their ability to perform simultaneously is measured by a crossed-ɱ up to 11.8, and satisfactory collinearity parameters, such as, variance inflation factor, VIF: [1.0003-3.18], and tolerance, Tol: [0.301-0.999]. While, the OTSU-segmentation is applied to reinforce imperviousness modelling and predicts burned impervious surface. Therefore, the novelty of algorithms and the complementarity of both spectral models offer additional insights, especially their dual usage for the accurate mapping in a post-fire’s context.