EVALUATING SINKHOLE HAZARD SUSCEPTIBILITY USING LOGISTIC REGRESSION MODEL IN KHLONG I PAN SUB-WATERSHED, SURAT THANI AND KRABI PROVINCE, THAILAND

Abstract

Sinkholes have frequently occurred over the past 20 years in the Khlong I Pan sub-watershed (KIPs) in Surat Thani and Krabi Province, Thailand. It was found that the earth collapsed more than 34 times. The objective of this research is to evaluate the sinkhole susceptibility using Logistic Regression (LR) analysis at the sub-watershed scale. This methodology used 14 variables affecting sinkhole occurrence to analyze the area, and create a sinkhole susceptibility map using LR. The results found that the variables that affect sinkhole formation include Well Density (WD), geology, Land Use (LU), Total Hardness (TH), Total Dissolved Solids (TDS), slope, Chlorine (Cl), distance to stream, elevation, Topographic Wetness Index (TWI), distance to village, soil, distance to active fault, and distance to well, respectively. All such variables are expressed by the exp β value coefficient. When prepared as a Karst sinkholes (KS) susceptibility map, it was found that a very high sinkhole susceptibility level covers an area of up to 399.86 km2 (19.16% of the total area). They appear mainly in the eastern region of the KIPs, especially at the confluence of the Khlong I Pan stream and the Khlong Trom stream. The other area is the central mountain range and the western mountain range, where geological structures with a casque topography are found. The results of this research suggest using the KS Susceptibility Map as a guideline for planning and monitoring potential future sinkholes.