Coupled hydrogeophysical inversion using time‐lapse magnetic resonance sounding and time‐lapse gravity data for hydraulic aquifer testing: Will it work in practice?

Abstract

Temporal changes in water content can be directly related to the time‐lapse signals retrieved using magnetic resonance sounding (TL‐MRS) and relative gravimetry (TL‐RG). Previous studies suggest that TL‐RG measurements can potentially provide accurate estimates of aquifer characteristics in an aquifer pumping test experiment when used in a coupled hydrogeophysical inversion approach. However, these studies considered highly idealized conditions. The aim of this paper is twofold: first, we investigate three major issues which likely limit the practical utility of TL‐RG for pumping test monitoring: partially penetrating pumping wells in anisotropic aquifers, delayed drainage effects, and typical data errors for TL‐RG. Second, we introduce TL‐MRS in a similar coupled hydrogeophysical inversion framework and compare the performance of TL‐MRS and TL‐RG for pumping test monitoring. For this purpose we consider a virtual pumping test, for which we generate synthetic drawdown, TL‐MRS and TL‐RG observations, and subsequently determine the aquifer parameters in an inverse parameter estimation approach. The inclusion of TL‐RG and TL‐MRS data did slightly improve parameter estimates for the specific yield and hydraulic conductivity when considering a fully penetrating well and minimal data error. Using more conservative TL‐RG and TL‐MRS data error estimates according our own field experience strongly limited the informative value of the TL‐RG data; TL‐MRS data was less affected by this. For a partially penetrating well under anisotropic conditions, parameter uncertainty could be reduced more effectively compared to a fully penetrating well. Delayed drainage effects did not limit the ability of the TL‐MRS and TL‐RG data to reduce parameter uncertainty significantly. The incorporation of representative measurement error correlation in the TL‐RG data did not affect its informative value. A local sensitivity analysis showed that observations were most sensitive to the pumping rate and the thickness, specific yield, and hydraulic conductivity of the aquifer. The inclusion of TL‐MRS data proved to be more effective to constrain the aquifer parameters compared with TL‐RG. The inclusion of both TL‐RG and TL‐MRS had limited added value compared to TL‐MRS only. We conclude that this particular application of coupled hydrogeophysical inversion has limited potential for TL‐RG, while TL‐MRS appears to be a more promising method.