Application of multiple linear regression models for predicting drilling water flow in the hard rock aquifers: Example of N'zi-Comoé area (Central-East of Ivory Coast)

Abstract

The aim of this study is to develop parsimonious statistical models that predict the drilling water flow in the hard rock aquifers using hydrodynamic and physical parameters estimable by the geophysical methods for optimizing the hydrogeological prospecting. The study area is the region of the N'zi-Comoé (Central-East of Ivory Coast). The calibration was performed with a sample of 100 drillings (2/3). The evaluation of the quality of the models in validation phase reveals strong correlation coefficients varying from 0.647 (model 2) to 0.660 (model 1). The biases fluctuate between 0.298 (model 3) and 0.377 (model 2) and tend towards 0. The robustness criterion values range from 6.4% (model 1) to 8.0% (model 2) and remain below 10%. It appears that the four models designed are powerful, robust and parsimonious. However, the models 1 and 3 stand out as the best tools for the exploration of groundwater in the crystalline and crystallophyllian aquifers.