Abstract
Availability of digital elevation models (DEM) of increased spatial resolution has triggered interest in texture-based methods for automated geomorphometry. This prospect is all the more appealing concerning tropical countries for which mapping of geomorphic entities has remained limited despite its relevance for natural resource assessment and land planning. In this paper we applied to the Shuttle Radar Topography Mission (SRTM) DEM a multiscale texture analysis based on Fourier 2D periodogram (FOTO method) to reach a spatially coherent mapping that would agree with current knowledge over the conterminous territories of French Guiana and Amapa (Brazil). Through a principal component analysis, FOTO-extracted textural features were found complementary and combined with six common physiographic criteria (convexity, wetness index, dissection, elevation, and percent of steep and low slopes) calculated from the DEM. This provided a primary semiautomated mapping of 16 classes of landform types. To increase legibility, we aggregated into eight landscape unit classes that were the most similar in terms of texture and relief criteria while being frequently contiguous. The resulting map was compared with existing maps that were, however, more local or of lesser resolution and informative content over the region of interest. Multiscale Fourier spectra proved useful to depict land surface texture and, in synergy with relief variables, allowed identifying a large variety of landforms for broad-scale geomorphic mapping in the humid tropics.