Comparative analysis of automated extraction of dolines from digital terrain models

Abstract

Topographic maps at semi-detailed and detailed scales (1:50,000 and 1:10,000) do not adequately represent the complexity of the drainage network and the morphological features of karstic areas. This paper evaluates the automated extraction of dolines, typical morphological features of polygonal karst, from digital terrain models generated from stereopairs of aerial photographs at the 1:25,000 scale and from SRTM (Shuttle Radar Topography Mission/NASA) sensor data with resolution of 30 and 90 m. The study area corresponds to the karst plateaus of the Ribeira do Iguape River Valley, southern São Paulo state, which contains important remnants of the Atlantic Forest and a rich speleological heritage. The digital terrain model obtained from aerial photographs was generated by the software Geomatica (Module OrthoEngine), using parameters of the internal calibration of the camera and control points extracted from topographic maps at scales of 1:10,000 and 1:50,000, and by GPS (Global Positioning System). Automated extraction of dolines from the models generated from aerial photographs and SRTM data was made in a GIS (Geographic Information System) environment. Based on morphometric analysis, the results were compared with topographical maps at 1:10,000 and 1:50,000 and with conventional photointerpretation. The models generated by Geomatica and SRTM (30 m resolution) detected several dolines not represented in the topographic maps nor in the SRTM (90 m resolution) model, similar to the results of conventional photointerpretation. Thus, the higher the resolution, the more sensitive the method is for extracting the dolines. Comparing the area and perimeter of the dolines, the values obtained by the Geomatica software were very close to those for photointerpretation. The use of different models thus represents an important tool that can help in conventional photointerpretation, especially in regions where there is not detailed cartographic information.