Caracterización SIG del uso de la tierra y geomorfología del Macizo del Cajas - Ecuador.

Abstract

With the intention of generating through remote sensing useful information for the proper management of the valuable natural resources of southern Ecuador, the preliminary results obtained from the combination of various remote sensing products are presented, as well as the treatment of Lansat 8 satellite images. Three main analyses were carried out in this study: firstly, the vegetation cover and land use in the Cajas Massif Biosphere (MzC) - Ecuador was defined through the application of three supervised classification methods, namely, Maximum Likelihood (MV), Artificial Neural Networks (ANN) and Mahalanobis Distance (MH). Secondly, the depths of 135 lakes and lagoons of the Cajas Antional Park (PNC) were estimated, through the application of the Single Band algorithm (BU). Finally, the analysis of the correlation of limnimetric variables of 145 lakes and lagoons of the PNC and physical variables characterising the surrounding of these lakes and lagoons was performed through a Principal Component Analysis (ACP). The performance of each of the algorithms applied in this study was evaluated, particularly through absolute relative error, as well as through the use of frequency histograms. The evaluation suggested that for the study of land use/coverage, the best image classification product was obtained using the ANN algorithm. On the other hand, of the 135 lakes studied, 89 were classified as shallow lakes and 46 as deep lakes. In the third analysis, the ACP did not suggest an acceptable correlation between the limnimetric and physical variables of the contours of the lakes and lagoons. Based on these results, future work should include the review of the training polygons (ROIs) used in this investigation, as well as the possible reduction of the number of classes considered, in order to minimize the problems of automatic classification due to the superposition of the spectral response for land use/coverage.