Caracterização geomorfométrica multiescala de modelos digitais de terreno.

Abstract

The study of landforms through geomorphometry science that quantifies the topographic surfaces demands that multiple scales are considered simultaneously. This fact sums itself up with the inherent spatial continuity of the topography, which rarely exhibits crisp boundaries, to provide landforms a vague character. This double vagueness , however, is rarely approached explicitly in scientific papers which, normally, make use of arbitrary and subjective thresholds and definitions in order to describe the relief. In this context, the work presents an innovative method for the multi-scale geomorphometric characterization of digital terrain models (DTMs). The proposed approach is based on Wood's (1996) method for morphometric feature extraction, but modifies it in a number of ways: (i) increasing the number of classes identified; (ii) turning it into a fuzzy system, with automatically parameterized fuzzy sets and (iii) locally limiting the maximum scale of analysis in an unsupervised way. As a result, a fundamental feature map and a fundamental scale map are produced which together synthesize the multi-scale structure of surfaces. Also, fuzzy membership and confusion index maps are produced for every analyzed scale, as well as multi-scale versions of the corresponding maps. In order to evaluate the transferability and the unsupervised character of the method, five DTMs from different databases, with diverse resolutions and spatial extents were analyzed, comprising continental regions, the ocean floor and the surface of the planet Mars. The results achieved were evaluated with respect to the use of fixed scales and parameterization, attesting for the method's ability to undertake more complete geomorphometric characterizations than standard approaches. Two applications were further proposed: the multi-scale geomorphometric parameterization and the development of a multi-scale morphometric signature, depicting clear paths for the continuity of the research. As a general conclusion, in spite of several limitations detected, it was considered that the presented work reached its objective of providing the multi-scale characterization of digital terrain models.