Study on complex inversion of magnetic resonance sounding signals

Magnetic resonance sounding (MRS), also known as surface nuclear magnetic resonance (SNMR), is used for non-invasive direct groundwater determination and aquifer characterization. Among other parameters, the electrical conductivity of the subsurface causes a complex-valued MRS signal. We show in our study that the real and imaginary parts of the signal evolve from different depth volumes, and so they contain complementary information. Generally, the imaginary part is more sensitive to deep structures than the real part of the signal, i.e. in conductive media, signals arising from deep layers have a significantly greater imaginary part than an equivalent signal from shallow depths. Statistical analyses of the inversion result of synthetic data show the advantages of a complex inversion scheme for determining the water content in the subsurface: model ambiguities are significantly reduced and the depth resolution is increased. Also the investigations on artificially noise-perturbed data show a clear improvement in the stability and resolution of model boundaries. For field data, amplitude and complex inversion schemes yield significantly different results. However, the various influences on the phase are still undergoing investigation, and at present the complex inversion is limited to selected (explainable) field data sets.