Cave entrance location model using binary logistic regression: the case study of south Gombong karst region, Indonesia

Cave entrance data are crucial as the primary indicators in the underground water inventory of a karst area. The data collection was traditionally conducted by field survey, but it is very costly and not efficient. Remote sensing and Geographic Information System (GIS) can help estimate cave entrance locations more efficiently. In this study, variables for cave entrance identification were determined using remote sensing and GIS. In addition, the accuracy of the Cave Entrance Location Model (CELM) derived from binary logistic regression was examined. Several remote sensing and geological data were used including ALOS PALSAR Digital Elevation Model (DEM), Digital Elevation Model Nasional (DEMNAS), topographic and geological map. Topographic elements were extracted by using Toposhape and Topographic Position Index (TPI). Contours derived from the topographic map showed the highest accuracy for extraction of topographic elements compared to ALOS PALSAR DEM and DEMNAS, hence it was used for further analysis. Binary logistic regression was applied to estimate the probability of cave entrance locations based on the variables used. The result shows that three topographic variables: ravine, stream, and midslope drainage had a significant value for estimating cave entrance location. Using these variables, logit equation was formulated to generate a probability map. The result shows that cave entrances are likely to be located in a dry valley. The accuracy assessment using the field data showed that 52.77% of cave entrances are located in medium to high potential areas. This suggests that the moderatehigh potential area can indicate potential water resources in karst area.