Ground and satellite based detection of rangeland management in per humid grasslands of Argentina

Abstract

Ranging is the most important single agricultural activity in the department of Empedrado, in the province of Corrientes, Argentina. The present study focuses on three issues (i) the classification of land cover types, (ii) the evaluation of range intensity use and (iii) the estimation of aboveground biomass. First, a multi-temporal Landsat image analysis, including supervised classification and ancillary data, was used to enable an accurate land cover classification. Second, by plotting historical biomass information and Landsat calculated NDVI, a simple regression model was calculated. Third, with the previous land cover information and statistical, up to date stocking rates were calculated. By comparing radiometrically normalised and not normalised images it could be demonstrated that imagery normalization was not always necessary, when comparing NDVI values. Regarding biomass estimation, neither a simple nor an averaged point approach provided accurate results. The inverted distance weighted interpolation method showed better biomass estimates in per-humid grasslands. Two maps were achieved for Empedrado, one 26-class land cover map and one 8-class land grazing intensity map. The classification accuracy strongly relied on ancillary GIS based information. The resulting maps showed that intensive grassland use was the most common vegetation on the hills, whereas not intensively used and burned areas were prevailing on the plains. The calculation stocking rates of grasslands with Landsat images was possible as long as additional statistical data were available.