Abstract:Efficient detection and analysis of natural sinkholes is essential for human sustainability in karst area. In this work, two methodologies were applied to detect and quantify natural sinkholes in southwest China, using of DEMs derived from ASTER and SRTM remote sensing images and topographic maps. The methodologies were a semi-automatic approach and a field-based approach. The first identifies the sinkholes based on DEM. The latter integrates the field survey, visual interpretation of high-resolution aerial photographs and topographic maps. The comparison between the semi-automatic approach and field-based approach generated the threshold values of the sinkhole shape, area, and TPI (topographic position index), which were used to distinguish true sinkholes and improve the identification accuracy. The result indicated that the semi-automatic approach produced the best accuracy by using the thresholds area > 60 m 2 , ellipticity > 0.2, and TPI ≤ 0 with the DEM from the topographic maps. With these realistic thresholds, the semi-automatic model shows well performance that model accuracy ranges from 0.78 to 0.95 for DEM resolutions from 3 to 75 m. The morphometric analysis was then obtained by using parameters derived from the semi-automatic approach and field-based approach. The sinkhole morphometric parameters are area, perimeter, diameter, shape, orientation, and volume. The results demonstrate (1) morphometric characteristics of sinkholes derived from the semi-automatic approach are coincident with those from the field-based approach. (2) Sinkholes in the region are relatively old and skewed to irregular or elliptical shape. (3) Sinkhole diameter and spatial alignments are highly controlled by faults. (4) Sinkholes argue about solution-only origin in the region.

Morphometric Analysis of Sinkholes Using a Semi-Automatic Approach in Zhijin County, China