Surface NMR, also called magnetic resonance sounding (MRS) for 1D applications, has emerged as a promising geophysical method that is used to assess aquifer properties. A typical coincident-loop MRS measurement consists of a large wire loop deployed on the surface, acting as both a transmitter and a receiver. In this study, we present central-loop MRS data and discuss how this loop configuration helps better estimate the subsurface model as compared with the typical coincident loop configuration. We demonstrate sensitivity kernels for coincident and central loop configurations and describe advantages of central-loop MRS data in terms of superior behavior of the sensitivity function, reduced noise level, higher signal level, and improved resolution matrix. With no extra time and effort, the results of our field example represent an improved model of the area when central loop configuration is employed. Our results are in a very good agreement with complementary geophysical and borehole data and with hydrologic model of the area.