An NDVI analysis of vegetation trends in an Andean watershed

Abstract

We performed a Landsat 5-TM derived normalized difference vegetation index (NDVI) analysis in a semi-arid watershed (2700 km2) in the Andes of southern Peru from 1985 to 2010. There, pastoralists rely on wetlands (bofedales) particularly during dry season months and in drought. We calculated annual dry season NDVI for 20 of the 26 years from 1985 to 2010 and used the mean to delineate wetlands in the watershed. To investigate the trends in NDVI, a multiple regression model with the covariates precipitation, temperature, Julian day, and year of image acquisition was performed on each cell (three million individual regressions). Results indicate there is a modest increase in NDVI for the majority of cells (81 %) in the watershed. Approximately 30 % of wetland areas display a decrease in NDVI. Dry season NDVI is moderately correlated with wet season precipitation (R2 = 0.56, p < 0.05) but absent a trend in precipitation, NDVI trends are not explained by this variable. Changes in land management may result in more intensive use of wetlands, causing the decreasing vegetation trends in some locations.