A crucial component in sustainable freshwater management is the reliable and cost-effective characterization of groundwater aquifers. A technique that allows noninvasive characterization of shallow () aquifers is surface nuclear magnetic resonance (surface NMR). The measured parameter longitudinal relaxation time provides a link to pore-scale properties and can be used to estimate the hydraulic conductivity of the sampled region. The recent development of an optimized acquisition scheme, phase-cycled pseudosaturation recovery (pcPSR), has significantly advanced our ability to acquire surface-NMR data. Building on these findings, we developed an inversion scheme that can reconstruct the depth-distribution of from pcPSR data. To stabilize the inversion, we took a staggered approach: We first determined the distribution of water content and effective transverse relaxation time , and then we resolved the structure. We tested the capability of the inversion in a synthetic study using models containing an unconfined and confined aquifer under conditions of low and high levels of ambient electromagnetic noise. The results allowed us to optimize the design of pcPSR experiments, finding that acquiring pcPSR data at three interpulse delay times is the best choice when acquiring surface-NMR data in an area with no prior information. We have verified our inversion approach in a field study in Nebraska, USA, where we had access to borehole lithologic information and logging-NMR measurements. Although the structural details () detected in logging NMR were not resolved in surface NMR — highlighting the general resolution limitation of surface NMR — the two methods consistently revealed a zone of elevated corresponding to a sand and gravel unit, overlying a zone of lower corresponding to a sand/sandstone unit. Our developed inversion scheme is an important step toward the reliable interpretation of surface-NMR data for aquifer characterization.