Year of Award

2022

Document Type

Dissertation

Degree Type

Doctor of Philosophy (PhD)

Degree Name

Geosciences

Other Degree Name/Area of Focus

Hydogeology

Department or School/College

Department of Geosciences

Committee Chair

W. Payton Gardner

Commitee Members

Douglas J. Brinkerhoff, Benjamin P. Colman, Kristopher L. Kuhlman, Marco P. Maneta

Keywords

environmental tracers, hydrogeology, integrated hydrologic models, model calibration, uncertainty analysis

Abstract

Groundwater flow and transport processes strongly influence and are inextricably linked to the integrated hydrologic and biogeochemical dynamics within catchments. Yet, groundwater system understanding and model predictions remain uncertain owing to the unknown subsurface property distributions, errors in atmospheric forcing conditions, and limited observations to constrain groundwater fluxes. In this dissertation we investigate the use of environmental tracer observations that inform hydrological processes over broad timescales to reduce uncertainties in groundwater transport prediction uncertainties. We further develop environmental tracer data assimilation and uncertainty quantification techniques to enhance integrated hydrological and groundwater process understanding at two distinct field sites: a semi-arid region in central Wyoming with minimal topography, and a snow-dominated mountain catchment in Colorado.

Environmental tracer observations are typically used to derive “apparent” groundwater ages, which require assumptions regarding the residence time distribution of a sample. We demonstrate reductions in permeability and infiltration rate parameter uncertainties when using environmental tracer concentrations, rather than apparent age, to calibrate a numerical model of a field site located near Riverton, Wyoming. We then extend the model uncertainty analysis technique to robustly quantify the full parameter joint posterior distributions with Markov-chain Monte Carlo (MCMC) sampling and Bayes’ theorem. To circumvent the intractable computational expense required by the MCMC method, we train a computationally frugal Artificial Neural Network to emulate the process-based groundwater transport model. We show that the parameter inference that assimilates 3H observations reduce the uncertainty in the permeability field and infiltration rates, relative to assimilating hydraulic head observations alone. However, CFC-12 transport predictive uncertainties do not reproduce the validation dataset, highlighting the influence of model and observation data structural errors on the parameter inference. Uncertainties in environmental

tracer interpretations are further investigated using an observation dataset (3H, SF6, CFC’s, and 4He) sampled from bedrock groundwater wells in the East River Watershed near Crested Butte, Colorado. We develop MCMC techniques to quantify uncertainties in the noble gas recharge thermometry parameters and the resulting groundwater residence time distributions. The inferred residence time distributions suggest that the shallow bedrock groundwater contains a mixture of waters characterized by residence times that are modern (<70 years) and pre-modern (>70 years). The findings that shallow fractured bedrock hosts groundwater with residence times ranging from decades to centuries informs the integrated conceptual model of how mountain systems store and transmit essential water resources, and how these resources will respond to perturbations in the hydrologic cycle.

Download

Share

COinS
 

© Copyright 2022 Nicholas E. Thiros