Extração automática de nascentes a partir de modelos altimétricos: avaliação da influência da resolução espacial e do tipo de modelo

Abstract

Springs are essential hydrogeomorphological environments for hydrological dynamics, biodiversity conservation, and human water supply. Despite their environmental and social relevance, they remain largely underrepresented in spatial mapping, compromising management and conservation efforts. Discussions on spring mapping are often accompanied by the challenge of defining a reliable methodology for detecting these features in the landscape. In this context, this doctoral thesis aimed to evaluate the application of an automatic extraction method as a supporting tool for spring identification. Specifically, it analyzed the influence of spatial resolution and elevation model type on the automatic extraction of springs through the generation of Hydrologically Conditioned Digital Elevation Models (HCDEMs). Four Digital Elevation Models (NASADEM, COPDEM, SRTM/ASF, and TanDEM-X) and two Digital Terrain Models (ANADEM and FABDEM), all derived from radar interferometry, with spatial resolutions of 12 m, 12.5 m, and 30 m, were employed. The study was conducted in three experimental areas located in the Baturité Massif, Ceará State, Brazil. Automatic spring extraction was performed in the QGIS environment using the SAGA GIS tool. The results were evaluated based on an adapted completeness analysis, following the Technical Specification for Geospatial Data Quality Control (ET-CQDG), using georeferenced field survey data as reference. This approach enabled the assessment of both the spatial correspondence and the operational usefulness of the extracted data. The results indicated better performance for models with higher spatial resolution, although no consistent superiority was observed between Digital Elevation Models and Digital Terrain Models. The best performance was found in the Bananal area, associated with gentler relief and higher moisture conditions, whereas the Ouro and Rolador areas, characterized by steeper terrain, presented greater limitations

for spring identification. These findings demonstrate the potential of high- resolution orbital elevation models for supporting automatic spring extraction.

Therefore, this study highlights the promising application of altimetric models in identifying priority areas for the detection and investigation of spring environmental systems.