Estimating Unsaturated Hydraulic Functions for Coarse Sediment from a Field‐Scale Infiltration Experiment

Abstract

A field-scale infiltration experiment was conducted in coarse conglomeratic soil with high gravel fraction. Unsaturated flow properties were estimated from modeling of infiltration using the van Genuchten–Mualem model and a Metropolis–Hasting optimization scheme. Results provide optimal unsaturated flow parameters for a soil type that is underrepresented for vadose zone flow. Conglomeratic alluvial sediments (sand–gravel–cobbles) are common in fluvial, periglacial, and tectonically active regions but have received little attention with respect to unsaturated flow, specifically moisture–tension–conductivity relationships, due to difficulty in making measurements in the field or laboratory and lack of agricultural value. We used a field-scale infiltration experiment, a one-dimensional layered forward model, and parameter estimation modeling to examine in situ flow behavior between residual and partial saturation in a four-layer system under steady infiltration (0.84 cm h−1) for 19 h. Prior information from ground-penetrating radar, grain-size distributions from core samples, and long-term tension (ψ) and moisture (θ) monitoring were used to define geologic structure, simulate test behavior, and provide initial parameter estimates. Vertically distributed measurements of ψ(t) and θ(t) from the experiment were matched using four parameters (θs, α, n, and Ks) of the van Genuchten–Mualem (VGM) relationships for each material layer and a Metropolis–Hastings (MH) search with multiple, independent-chain runs to 106 samples each. Scale reduction factors indicated convergence of independent chains for 11 of 16 parameters. Final distributions of individual parameters varied from normal to nearly uniform distributions, and some parameter pairs showed high cross-correlation (R2 \u3e 0.9). Results showed that (i) VGM relationships can be applied to these coarse, conglomeratic soils to characterize unsaturated flow behavior across the natural range of partial saturation, (ii) even under high sustained infiltration rates, these coarse conglomeratic soils remain well drained, despite relatively low porosity and significant cobble fraction, and (iii) high uncertainty and nonconvergence of MH chains does not lead to significant misfit of the observed data. These findings imply that a significant cobble fraction does not markedly reduce infiltration at low saturation levels that develop under natural recharge conditions

Estimating hydrologic parameters from water table dynamics using coupled hydrologic and ground-penetrating radar inversion

“…For the hydrologic model, we utilize Hydrus 10 which is freely available and solves Richard's equation for 10 vertical flow. We use the van Genuchten Mualem model [3,4] to simulate the soil moisture retention curve and partially saturated hydraulic conductivity as it has been shown to produce reasonable results for the vadose zone materials at the BHRS [5].…”

Soil hydraulic material properties and layered architecture from time-lapse GPR

“…Infiltration experiments (e.g., Léger et al, 2014;Thoma et al, 2014; are fast (hours) and provide indirect information about the near-surface material properties. Through its dependence on the form of the infiltration front or plume, the resulting GPR signal can get rather complicated to reproduce when used for quantitative evaluation.…”

“…The NMS was further developed to the shuffled complex evolution (SCE-UA, Duan et al, 1992) which has become a standard tool in hydrology and was also applied on GPR measurements (e.g., Jadoon et al, 2012;Léger et al, 2014Léger et al, , 2015. Additionally, Markov chain Monte Carlo (MCMC) methods (e.g., Scholer et al, 2011;Thoma et al, 2014;Jonard et al, 2015) and data assimilation approaches (e.g., Tran et al, 2014;Manoli et al, 2015; have been successfully applied so far.…”

Soil hydraulic material properties and subsurface architecture from time-lapse GPR

“…Carlo (MCMC) methods (e.g., Scholer et al, 2011;Thoma et al, 2014;Jonard et al, 2015) and data assimilation approaches (e.g., Tran et al, 2014;Manoli et al, 2015;Rossi et al, 2015) have been successfully applied so far.…”