Measurement of the size distribution of water-filled pores at different matric potentials by stray field nuclear magnetic resonance

The water retention characteristic provides the traditional data set for the derivation of a soil's pore-size distribution. However, the technique employed to achieve this requires that assumptions be made about the way pores interconnect. We explore an alternative approach based on stray field nuclear magnetic resonance (STRAFI-NMR) to probe the water-filled pores of both saturated and unsaturated soils, which does not require information relating to pore connectivity. We report the relative size distributions of water-occupied pores in saturated and unsaturated samples of two sets of glass beads of known particle size, two sands, and three soils (a silty loam, a sandy loam and a loamy sand), using measurements of the NMR T-1 proton relaxation time of water. The T-1 values are linearly related to pore size and consequently measured T-1 distributions provide a measure of the pore-size distribution. For both the sands and the glass beads at saturation the T-1 distributions are unimodal, and the samples with small particle sizes show a shift to small T-1 values indicating smaller voids relative to the samples with larger particles. Different matric potentials were used to reveal how the water-occupied pore-size distribution changes during drainage. These changes are inconsistent with, and demonstrate the inadequacies of, the commonly employed parallel-capillary tube model of a soil pore space. We find that not all pores of the same size drain at the same matric potential. Further, we observe that the T-1 distribution is shifted to smaller values beyond the distribution at saturation. This shift is explained by a change in the weighted average of the relaxation rates as the proportion of water in the centre of water-filled pores decreases. This is evidence for the presence of pendular structures resulting from incomplete drainage of pores. For the soils the results are similar except that at saturation the T-1 distributions are bimodal or asymmetrical, indicative of inter-aggregate and intra-aggregate pore spaces. We conclude that the NMR method provides a characterization of the water-filled pore space which complements that derived from the water retention characteristic and which can provide insight into the way pore connectivity impacts on drainage.