Abstract

Accurate information is needed about burn severity caused by fires to design strategies for ecosystem recovery. The Mediterranean zone of Chile holds a high level of plant endemism. In summer 2017, nearly half a million hectares were severely burnt by fire in this zone. In this paper, we assessed the performance of three spectral indices used in mapping burn severity (difference Normalized Burn Ratio [dNBR], Relative difference Normalized Burn Ratio [RdNBR], and Relativized Burn Ratio [RBR]) on fragmented populations of the endangered ruil tree Nothofagus alessandrii, an endemic species of central Chile. The indices were obtained from Sentinel 2 images and their results contrasted with field measurements of severity recorded using the Composite Burn Index (CBI). The indices estimated the severity categories with a high overall accuracy obtained from error matrices and reaching values of 77.3, 76.0 and 69.3% for RdNBR, RBR, and dNBR, respectively. The surface estimated for the categories without damage (ND), low (L), and medium (M) were greater for the RdNBR and RBR indices than for the dNBR, while an opposite tendency was observed for the high severity category (H). Our results properly described the spatial variability of burn severity. This information may be useful for creating and managing plans for the recovery of highly fragmented and threatened deciduous ecosystems such as those dominated by ruil. We suggest that strategies for recovery or restoration should be oriented to the high fire severity zones, where reproductive individuals were lost, and the successful seedling recruitment is expected to diminish.