Application of a logistic regression to assessing and mapping frequency of Springs in karst regions (Case study: Bojnourd Basin)

Assessment frequency of springs has become an important issue for land use planning, especially water resource identification and environmental protection.The purpose of this study is to produce a spring occurrence potential map in Bojnourd Basin, based on a logistic regression method using Geographic Information System (GIS) and remote sensing (RS). The locations of the springs (359 springs) were determined in the study area. In this study, 14 effective factors including spring were used in the analysis: lineament density, distance to lineament, distance to drainage, drainage density, normalized difference vegetation index (NDVI), profile curvature, tangential curvature, surface rate, vector dispersion, precipitation, elevation, geology, aspect and slope. Binary logistic regression coefficients of the variables by selecting 300 spring randomly. 59 another spring were used for validation that 80.6% of the springs were correctly predicted. The accuracy of the model was measured using ROC curves which showed that accuracy is 86.6 percent which indicates that the model shows high accuracy in the analysis of spring occurrence potential in the study area. The results showed that the distance of lineaments, distance of drainage, drainage density, vegetation index, profile curvature, tangential curvature, vector dispersion, precipitation and slope have the greatest impact on the occurrence of springs. Finally, spring occurrence potential map was divided into four probably classes of very low, low, medium and high. According to the survey results, this method can be used to identify sources of groundwater in karstic zones and has important role in better management of the karstic Basins.