| dc.description.abstract | Deforestation arises deep concerns about several issues related to the planet‘s future like biodiversity loss, erosion increase, land degradation, carbon emissions, temperature fluctuations, and changes in hydrological cycle, global warming, climate change, social impacts, among others. Remote sensing methods and techniques play a key rol on monitoring activities and assessing decision makers. Several methods have been developed over the last decades, principally based on vegetation indices analysis, which only uses a portion of the spectrum, while other methods less used have tried to gain information by dimension reduction techniques. The access to new Satellite Constellations like Sentinel-2, with a higher spatial resolution and richer spectral information than other Satellite programs as Landsat brings important opportunities to enhance the quality and accuracy to identify forest disturbances around the world. This study evaluates the performance of Sentinel-2 imagery detecting structural changes in forests based on a Empirical Fluctuation Process (e.g. OLS-MOSUM) over Normalized Difference Vegetation Index (NDVI) and Principal Component Analysis, to indicate which of both data sources is more feasibly. On the other hand, the present research evaluates if the method can perform with short time-series datasets in cloudy areas.
The results of the study depicts higher figure of merit‘s accuracy, user‘s accuracy, producer‘s accuracy, and overall accuracy using NDVI time series; however, though marginally lower accuracy was obtained using the PCA analysis, consider the PCA scores have 20 meter spatial resolution while the NDVI time series has 10 meters spatial resolution, the use of PCA still shows its potential. | |
