Publication Date

5-2014

Date of Final Oral Examination (Defense)

11-21-2013

Type of Culminating Activity

Dissertation

Degree Title

Doctor of Philosophy in Geosciences

Department

Geosciences

Major Advisor

Warren Barrash, Ph.D.

Advisor

James P. McNamara, Ph.D.

Abstract

In this dissertation, I address the lack of knowledge of unsaturated flow in coarse, conglomeratic sediment by determining if functional θ-ψ-K relationships, specifically van Genuchten-Mualem (VGM) relationships, developed to predict unsaturated flow in relatively fine-grained sediment can be directly applied to coarse, conglomeratic sediment. In the summer of 2011, a field-scale infiltration test was conducted at the Boise Hydrogeophysical Research Site to determine if functional ψ-θ-K relationships could be applied to infiltration in coarse, conglomeratic sediment, and to estimate parameter values for the VGM relationships. Vertically and laterally distributed ψ(t) and θ(t) measurements were made within the infiltration volume during the test, and geophysical data and core samples were used to determine material structure and distribution for model development. A four-material, 1D layered model was first used with a Metropolis-Hastings search to fit partial ψ(t) and θ(t) data and determine if VGM relationships are appropriate for unsaturated flow in coarse, conglomeratic sediment. The 1D model accurately fit a subset of the observed data, implying that VGM relationships were applicable, and predicted low uncertainty in θ(ψ) and K(ψ) curves for three of the four different materials but high uncertainty was observed in individual parameter values (σ/μ > 50 %). A four-material, 2D model was then constructed to incorporate variations in material thickness (lateral heterogeneity) and to fit all ψ(t) and θ(t) data. A Direct-search optimization with this model showed that fitting θ(t) and ψ(t) data simultaneously was not possible due to additional, lateral heterogeneity within one of the material layers so a five-material, 2D model was constructed. Direct-search optimization using this model successfully fit the full θ(t) and ψ(t) data sets and Latin-hypercube sampling was used to estimate final parameter uncertainty. These results showed further reduction of uncertainty in parameter values compared to the 1D model (σ/μ < 15 % for all parameters and up to 36 % reduction of σ/μ for some individual parameters).

Results from both the 1D and 2D models show that unsaturated flow relationships developed for agricultural soils (e.g., the VGM models) may be used to predict flow and moisture distribution in coarse, conglomeratic sediment. This implies limited obstruction by cobbles at low-saturation and a very high capacity for infiltration in these types of materials under natural conditions. A method was also developed and presented in this dissertation which uses reflection travel-time from time-lapse ground-penetrating radar (GPR) profiles to estimate changes in θ in the vadose zone. The method was applied to the infiltration test data but failed to accurately reproduce the observed GPR travel-time data which was attributed primarily to uncertainty in picking GPR reflection times.

Share

COinS