| Abstract |
Estimating hydraulic properties of highly heterogeneous reservoirs is still a problem difficult to solve, especially when available data are few and poorly distributed. Such an estimate is still necessary to develop numerical models for resource management of these reservoirs. In the Republic of Djibouti, the basaltic reservoirs are the main source of water supply. Djibouti, the capital city, is especially supplied with water from the Gulf basaltic aquifer. The sustainable management of this reservoir is currently a priority, since its resources are overexploited. The objective of this work is to analyze several geostatistical methods (ordinary kriging, cokriging and kriging with external drift) and compare their performances to estimate the properties of highly heterogeneous reservoirs. The Gulf basaltic aquifer is strongly heterogeneous. The transmissivity (T, m²/h) data are few and clustered mainly along the coastline. However, several geophysical surveys were undertaken over this basaltic reservoir. At present, a great deal of geoelectric data (transverse resistance, R ohm.m²), almost uniformly distributed, are available. Performances of the geostatistical methods are assessed using the cross-validation procedure. The results show that ordinary kriging, which is a univariate method, is the least suitable method to assess the transmissivity of the basaltic aquifer. The use of the transverse resistance (R), as an auxiliary variable, much more densely sampled than the principal variable (T), can significantly improve the estimation of T. Comparing the cokriging and the kriging with external drift, shows that KED is the best estimation procedure of the transmissivity of this highly heterogeneous reservoir. |