Enhancing spatial prediction of sinkhole susceptibility by mixed waters geochemistry evaluation: application of ROC and GIS

摘要

Many buried karst areas in Iran, and in other parts of the world have not yet been mapped using detailed geological or geophysical studies to delineate susceptibility to sinkhole development. The purpose of this paper is to investigate the possibility of using the results of hydrogeochemical analysis with routine measurements of physicochemical parameters to evaluate and detect areas prone to sinkhole develop. Sixteen spatial maps were prepared using analyzed data from 77 water samples from monitoring water wells in the Kabudar Ahang, Razan, and Qahavand (KRQ) sub-catchments of the Hamadan province, western Iran. By use of geographic information system tools 16 thematic maps for physicochemical parameters (EC, pH., TDS, and groundwater temperature), major cations (Ca2+, Mg2+, Na+, and K+), anions (HCO3-, SO42-, NO3-, and Cl-), and calcite, dolomite, gypsum, and partial pressure of CO2 saturation indices (SIC, SID, SIG, and SIpCO(2)) were prepared. It was hypothesized that the anomalies of each parameter concentration could be consistent with sinkhole prone areas. To evaluate this assumption, the area under the Receiver Operating Characteristic (ROC) curve was calculated by 100 points as a true sinkhole pixel (50 positive true) and non-sinkhole point (50 true negative). The areas under curve of ROC for these thematic maps were calculated for the 16 variables. Results show that the dolomite and calcite saturation indices (0.49 and 0.43, respectively) are poor indicators, whereas HCO3- and pCO(2) saturation indexes (0.83 and 0.78, respectively) are good indicators of sinkhole susceptibility in the study area. The result confirmed application of hydrogeochemical anomaly analyses and the ROC validation method in covered karst can be a useful tool for prediction of sinkhole prone zones forming in region, where sparse data are available.