Discrimination of tower-, cockpit-, and non-karst landforms in Guilin, Southern China, based on morphometric characteristics

Depressions are well developed in certain types of karst landform assemblages. These natural depressions should not be simply filled and removed during the preparation of DEMs. This study presents a research to distinguish karst landform assemblages which tend to have natural depressions from other karst landform assemblages and non-karst landforms in the Guilin area of China, by examining their morphometric characteristics derived from a 30-m resolution DEM. The variations in the morphometric characteristics are examined at neighborhood level instead of pixel level. The DEM is divided into square tiles with a specific spatial scale. Statistical indicators of typical morphometric characteristics such as the area ratio, elevation, slope, and curvature are calculated for each tile. Discriminant analysis (DA) is then performed to discriminate tower karst, cockpit karst, and non-karst landforms. These procedures are repeated at the scales of 0.45, 0.9, 1.8, 2.7, 3.6, and 4.6 km. Comparison of the mapping results with a reference geomorphic map shows that the DA works best for the 2.7 km tiles with an overall accuracy of 80.06%. The resulting map can be used to guide whether depressions should be retained or removed during DEM preparation. This method, with appropriate modifications and improvements, can also be used to map the karst landforms of the whole of southern China. With such a comprehensive map, geomorphologists would be able to examine the development of karst landform assemblages at a broader view to unveil their genesis and evolution processes.