Abstract/Summary

Ongoing increase in human and climate pressures, in addition to the lack of monitoring initiatives, makes the Caatinga one of the most vulnerable forests in the world. The Caatinga is located in the semi-arid region of Brazil and its vegetation phenology is highly dependent on precipitation, which has a high spatial and temporal variability. Under these circumstances, satellite image-based methods are valued due to their ability to uncover human-induced changes from climate effects on land cover. In this study, a time series stack of 670 Landsat images over a period of 31 years (1985–2015) was used to investigate spatial and temporal patterns of land-cover clearing (LCC) due to vegetation removal in an area of the Caatinga. We compared the LCC detection accuracy of three spectral indices, i.e., the surface albedo (SA), the Enhanced Vegetation Index (EVI) and the Normalized Difference Vegetation Index (NDVI). We applied a residual trend analysis (TSS-RESTREND) to attenuate seasonal climate effects on the vegetation time series signal and to detect only significant structural changes (breakpoints) from monthly Landsat time series. Our results show that SA was able to identify the general occurrence of LCC and the year that it occurred with a higher accuracy (89 and 62%, respectively) compared to EVI (44 and 22%) and NDVI (46 and 22%). The overall outcome of the study shows the benefits of using Landsat time series and a spectral index that incorporates the short-wave infrared range, such as the SA, compared to visible and near-infrared vegetation indices for monitoring LCC in seasonally dry forests such as the Caatinga.