APPLICATION OF LOGISTIC REGRESSION TO REMOTE SENSING AND GEOPHYSICAL DATA IN GROUNDWATER VULNERABILITY MODELLING.

Abstract

In attempt to provide the most accurate and reliable vulnerability prediction in a given area, a situable model for the area needs to be developed. Report of environmental problems associated with mining communities had prompted the groundwater vulnerability study of Ilesha Gold mining area. Groundwater vulnerability of Ilesha Gold Mining Area Southwestern Nigeria had been evaluated using the integration of electrical resistivity method, remote sensing and geographic information systems. In order to evaluate the groundwater vulnerability of Ilesha Gold mining area, sixty eight (68) vertical electrical sounding points were occupied. Four subsurface geo-electric sequences of top soil, weathered layer, partially weathered/fractured basement and the fresh basement were delineated in the area.The effects of ten hydrogeological indices, namely drainage density, lineament density, slope, percentage of clay in soil (Soil media), and geophysical parameters such as unsaturated zone thickness, aquifer layer resistivity, aquifer layer thickness, total longitudinal conductance and total transverse resistance on groundwater vulnerability were established. Also, the physico-chemical parameters, major metals concentration and heavy metals present in the groundwater and streams were assessed using a maximum permissible level for safe drinking water by Nigerian Standard for Drinking Water Quality Threshold guideline.Zinc ion concentration was considered as indices variable based on satistifation of binary logistic regression criteria for dependent variable. The hydrogeological indices were assigned as independent variables. The independent variables was subjected to statistical anaylsis (Kurtosis) to determine non-normality and non-parametric nature of the groundwater system. Also, spearman rank correlation was employed to established the significant variable(s) and strength of the relationship of these variables. Logistic regression analysis technique was employed to develop a model for predicting probability of Zinc concentration above the consumption threshold using the significant independent variables. The model developed was statistically assessed using model significance test, Hosmer-Lemeshow goodness-of-fit test, and R-square values.The prediction accuracy values were also established. The model developed assisted in predicting the groundwater quality of Ilesha Gold mining area, and classifying the area into, vulnerable and non-vulnerable groundwater zones. Validation of the model from water samples and statistical tests suggest 85.7% accuracy