Assessing the damage of forests burnt in central Chile by relating index-derived differences to field data

To assess the damage produced by wildfires on forest ecosystems is a critical task for their subsequent management and ecological restoration. Satellite-based optical images provide reliable ex-ante and ex-post data about vegetation state, making them suitable for the aforementioned purpose. In this study we assessed the damage produced on two forested lands by the series of wildfires occurred in central Chile during summer 2017. Arithmetic differences from pre- and post-fire NDVI (normalized difference vegetation index), NDWI (normalized difference water index) and NBR (normalized burnt ratio) were retrieved from a Sentine1–2 image set embracing four near-anniversary summer dates: 2016 (ex-ante), 2017, 2018 and 2019 (ex-post). The nine index-derived differences resulting were correlated to CBI (composite burn index) data collected in the field during summer 2019, and a model constructed by a stepwise regression was formulated. Results show that planted forests exhibited a somewhat smaller biomass recovery than native ones, in pait due to their post-fire clearing and preparation, deriving in a smaller tree cover. CBI poorly performed because its calculation includes low vegetation strata largely recovered at the time of the field data collection. However, when overstory field data were used alone correlations noticeably increased (${r}$=0,66–0,74). This was because during the field campaign this stratum was still appreciably damaged, thus better matching with the data provided by the indices-derived differences, intrinsically more representative of uppermost vegetation layers. The bum damage was mapped on both study areas employing the best performing regression model, based on $\mathrm {N}\mathrm {D}\mathrm {W}\mathrm {I}_{2016-2019}, \mathrm {N}\mathrm {D}\mathrm {W}\mathrm {I}_{2016-2017}, \mathrm {N}\mathrm {B}\mathrm {R}_{2016-201\mathrm {S}}$ and $\mathrm {N}\mathrm {B}\mathrm {R}_{2016-2017}$ differences (adjusted $\mathrm {R}^{2}=0.72, p< 0.005,$ root mean square error =0.38). The use of approaches like this one in other areas of central Chile, where wildfires are increasing their frequency and intensity, might contribute to better lead post-fire management and restoration actions on their damaged forest ecosystems.