Semi-Automatic Algorithm to Map Forest Fires and Estimate Vegetation Recovery Using Landsat-8 Images

Abstract

This study proposes a reproducible methodology for the semi-automatic identification of large forest fires and the estimation of plant recovery using Landsat 8 (OLI) satellite images, and the analysis of information obtained from the calculation of spectral indices. For this purpose, an algorithm was developed using information from geo-information servers and spatial analysis packages from the R software. The processed satellite images were obtained from the EarthExplorer platform of the United States Geological Survey (USGS) before, during and after the forest fire in Quilanga, Loja-Ecuador in 2019. For this purpose, the calculation of spectral indices was carried out. Those spectral indices were employed to identify the indices of best applicability by means of a visual analysis and a Pearson correlation test with a 95% confidence interval, corroborating the visual analysis. The best correlation with reference to the NBRI was obtained. According to the results, it was possible to delimit the affected area and estimate the early recovery of the vegetation, however, it was not possible to identify the recovered species.