The objective of this study is to develop statistical models to predict the depth limit stop of water drilling in crystalline and crystallophyllian rocks using physical parameters of drilling. The study was conducted using data from physical parameters (depth of drilling, depth of the first water supply significant, alteration thickness and thickness of base drilled) from geological units of Archean and Paleoproterozoïcal domains of Ivory Coast. The methodology consisted of first analyze the physical parameters from the simple statistical features and frequency class distribution of these parameters. Then, a normalized principal components analysis (NPCA) was applied for the identification of explanatory variables relevant and expressive of drilling depth limit. Finally, a calibration was performed with a sample of 1,605 wells representing two thirds of the total sample to determine the coefficients of the linear regressions and the associated standard errors. The analysis of the distribution of physical parameters of drilling shows that only the depth of drilling is homogeneous (24.50%) with an average of 68.17 m. The most relevant and expressive variables for predicting the depth limit stop of water drilling in crystalline and crystallophyllian rocks are the depth of the first significant water arrival and the thickness of alteration. The errors associated with the regression coefficients generally low (below 1%) reflect the close relationship between the dependent variables and the explanatory variables. Two models for predicting the depth limit stop of water drilling (PROLIFE models) in crystalline and crystallophyllian rocks were developed.

The objective of this study is to develop statistical models to predict the drilling flow on rock aquifers. The study area is the former region of N'zi-Comoé (Central-East of Ivory Coast) that is located between longitude 3° 40' and 4°55' West and latitude 6° 20' and 8°10' North. The geological formations are composed of igneous rocks and metamorphic rocks. The methodology has been first applied to the approach of principal component analysis normalized, for identifying relevant and expressive variables of the drilling flow. Then, a calibration was performed with a sample of 100 drillings representing two thirds of the total sample to determine the coefficients of linear regressions and associated standard errors. Finally, an evaluation of the developed models was performed using the criteria of performance and robustness. The most relevant and expressive variables for modeling the drilling flow are the transmissivity, the thickness of healthy base drilled, the depth of the first significant water inlet and the total number of water inlet. Thus, two models were developed. Models assessment shows that they are performing with correlation coefficients ranging around 0.7, biases varying between 10-7 and 10-2 and quadratic errors in the order of 2.5 to 2.9. These models have also demonstrated their robustness with rates fluctuating between 3.2 and 5.3%. However, the model 2 is more efficient and more robust than model 1.