DYNAMICS OF LAND USE AND COVERAGE IN LAKE POOPÓ (BOLIVIA) BETWEEN 1985 AND 2017 USING OBJECT-ORIENTED CLASSIFICATION IN LANDSAT DATA

Abstract

The Poopó Lake is the second largest lake in Bolivia and is currently undergoing a severe water crisis, that some authors directly associate the change of land occupation. In this work a classification of land use and coverage was carried out in the P6 sub-basin of the lake between the years of 1985 and 2017. Initially, the performance of the algorithms was analyzed: SVM (Support Vector Machines), KNN (K-Nearest Neighbor) e MaxVer (Maximum-Likelihood). The classification with better acuracy was performed by the SVM algorithm, where the Kappa index value was 82.28% and 83.7% for Landat-5 and Landsat-8 images respectively, and Global Accuracy was 92% for both images. With the classification it was verified that the most significant changes occurred in the native vegetation, agriculture and wetland areas. The loss of wetland area in the sub-basin has been occurring since 1995, 15 years before the increase in agricultural activity, which started of 2010. Thus, several factors may be contributing to this accelerated reduction in the lake, for example, the local climatic variations and the anthropic activities that interfere in the hydrological cycle regionally.