Maximum Entropy Model-Based Spatial Sinkhole Occurrence Prediction in Karap?nar, Turkey

Abstract

Sinkholes in Karapinar and their rapidly increasing occurrence rate are considered one of the main hazards that threaten arable lands and human life. The sudden occurrence and unavoidable characteristics of sinkholes make them more dangerous and challenging to avoid. More than 300 sinkholes have been recorded in the Karapinar region of Konya province in Turkey. There are intensive agricultural activities in the region, and therefore over 60,000 water wells are used to meet the demand. Thus, drought, the effects of climate change and decreasing precipitation rate reveal stress on sinkhole occurrence due to the geological structure of the region and its high tendency to sinkholes since ancient times due to its volcanic history.The primary purpose of this study is to predict possible sinkhole occurrence probabilities in Konya, Karapinar region based on historical occurrences and to report to the authorities to raise awareness about this problem. The Maximum Entropy (MaxEnt) model is applied for sinkhole susceptibility mapping by evaluating 17 variables affecting sinkhole occurrence in meteorological, topographic, environmental, and geological aspects. The results indicated that 458.52 km2 (2.48%) of the study area is highly susceptible to sinkholes. 100 sinkholes were assigned as sample data, and 45 sinkholes were set as test data for the MaxEnt model. The AUC values of training data with 0.978 and test data with 0.963 were calculated where a good correlation was provided. The variables Annual Mean Temperature, Precipitation Seasonality (Coefficient of Variation) Geology, and precipitation, which are mostly responsible for sinkhole formations, have been calculated.