Inferring the heterogeneity, transmissivity and hydraulic conductivity of crystalline aquifers from a detailed water-table map

•Innovative approach for upscaling transmissivity (and hydrau. cond.) is proposed.•Inferring the heterogeneity, T and K of crystalline aquifers.•Method tested on a crystalline aquifer, in New Caledonia.•LogT map uses statistics from detailed water-table maps.•71% of simulated LogT varies less than 20% from local measurements. Estimating the transmissivity or hydraulic conductivity field to characterize the heterogeneity of a crystalline aquifer is particularly difficult because of the wide variations of the parameters. We developed a new approach based on the analysis of a dense network of water-table data. It is based on the concept that large-scale variations in hydraulic head may give information on large-scale aquifer parameters. The method assumes that flux into the aquifer is mainly sub-horizontal and that the water table is mostly controlled by topography, rather than recharge. It is based on an empirical statistical relationship between field data on transmissivity and the inverse slope values of a topography-reduced water-table map. This relationship is used to compute a transmissivity map that must be validated with field measurements. The proposed approach can provide a general pattern of transmissivity, or hydraulic conductivity, but cannot correctly reproduce strong variations at very local scale (less than10m), and will face of some uncertainties where vertical flows cannot be neglected. The method was tested on a peridotite (ultramafic rock) aquifer of 3.5 km2 in area located in New Caledonia. The resulting map shows transmissivity variations over about 5 orders of magnitude (average LogT: −5.2±0.7). Comparison with a map based on measured water-level data (n=475) shows that the comparison between LogT-computed values and LogT data deduced from 28 hydraulic tests is estimated with an error less than 20% in 71% of cases (LogT±0.4), and with an error less than 10% (LogT±0.2 on average) in 39% of cases. From this map a hydraulic-conductivity map has been computed showing values ranging over 8 orders of magnitude. The repeatability of the approach was tested on a second data set of hydraulic-head measurements (n=543); the mean deviation between both LogT maps is about 11%. These encouraging results show that the method can give valuable parameter estimates, and can characterize aquifer heterogeneity. The computed LogT and LogK maps highlight the spatial distribution of parameters that show a pattern clearly controlled by the fault network of thi