Estimating aquifer transmissivities — on the value of auxiliary data

Abstract

Groundwater quality modelling relies heavily on the knowledge of preferential flowpaths such as buried stream channels and their distribution within the aquifer. This paper examines the extent to which these patterns may be identified by including auxiliary data, such as transverse electric resistances or specific capacities, when estimating the transmissivity field. The analyses are based on two hypothetical aquifers. The first involves a high transmissivity flowpath. The second is a realization of a correlated random field with the same spatial moments as the organised case. Monte Carlo simulations and cokriging estimates are used to analyse the effect of the number of samples and their correlation with transmissivity on the width of the capture zone of a well. Results indicate that, in the organised case with no auxiliary information, the estimated widths are substantially biased. This bias can be reduced significantly by including auxiliary data, even when poorly correlated to transmissivity. Auxiliary data also reduce the scatter (i.e. standard deviation) of the estimated widths significantly, which is a measure of the accuracy of the estimates. In the example used here, 70 samples of auxiliary data, which are correlated to transmissivity by r = 0.6, outweigh the information from 12 additional pumping tests. For the case of the correlated random field, the benefit of using auxiliary data is much less pronounced both in terms of removing the bias and in terms of accuracy (i.e. standard deviation). It is concluded that auxiliary data are useful for estimating transmissivity fields in the context of groundwater quality modelling, particularly when channelised flow is to be expected.