Near Real time Multisensor Algorithm for Deforestation Alert Over the Dry Chaco Forest

The objective of this work is to evaluate the performance of a change detection algorithm for deforestation monitoring based on MODIS Enhanced Vegetation Index (EVI) and Land Surface Temperature (LST) datasets. The algorithm is based on the statistical characterisation of the EVI and LST historical values for each pixel. The goal is to monitor abrupt decreases in those variables dynamics, which could be associated with deforestation events. This model was tested on real satellite time series available over a pilot area in Dry Chaco Forest, Argentina, where deforestation was common in the 2004-2016 period. As validation, a comparison with a benchmark product based on Landsat images is also presented, showing a 51.6% precision, 74.4% recall and an F-score of 0.60 for deforestation in 2013. These results show the advantages of using an automatic algorithm to monitor break points in the time series and the potential of combining LST and EVI to identify deforestation events.