Obtaining reliable estimates of hydrogeologic properties from nuclear magnetic resonance (NMR) measurements requires the ability to measure NMR relaxation parameters that are most sensitive to pore-scale geometry. Conventional surface NMR measurements of the free induction decay yield accurate estimates of the relaxation time parameter , but it has been shown that this parameter can exhibit limited sensitivity to pore size and permeability. We evaluated an improved surface-NMR scheme that uses spin-echo signals to estimate the more robust and readily usable relaxation parameter . The acquisition methodology builds upon previous spin-echo schemes and incorporates robust phase-cycling procedures, which remove responses that can potentially interfere with the echo signals. A new two-stage linear inversion was used to derive quantitative estimates of with depth. The method was evaluated in two field experiments at sites in the central and western United States. At one site, NMR logging measurements in a nearby borehole provided the first opportunity to compare -values estimated by surface NMR to -values determined from the logging data. The surface and logging results showed very close agreement at depths where is long, but echoes cannot be detected from depths where is shorter than the minimum echo time. As anticipated, we found that derived from spin echoes was generally much longer than , derived from the free induction decay. We explain the observed differences by considering the magnitude of inhomogeneity in the background magnetic field. We note that exhibited greater variation and sensitivity to pore size than in coarse-grained materials, while provided greater sensitivity in fine-grained materials where no echo signal was detected. Given these complementary advantages of and measurement, we advocate adoption of a framework combining spin-echo and free induction decay data to improve characterization of groundwater aquifers.