Date of Award

Spring 1-1-2016

Document Type


Degree Name

Doctor of Philosophy (PhD)

First Advisor

Balaji Rajagopalan

Second Advisor

William Kleiber

Third Advisor

Subhrendu Gangopadhyay

Fourth Advisor

Ben Livneh

Fifth Advisor

Joseph Kasprzyk


In the semi-arid western US, water resources infrastructure such as dams, levies, and canals provide a reliable water supply, flood mitigation, and hydropower generation. The ability to manage these aging infrastructure efficiently for hydroclimate extremes such as floods, in a nonstationary climate, is crucial for the sustainability of the region. Traditional methods for analyzing hydroclimate extremes assume stationarity of climate in space and time and do not capture uncertainty in a robust manner, leading to inaccurate estimation and misrepresentation of risk.

This dissertation is motivated by the need for a mechanistic understanding of hydroclimate extremes and a general framework to model them across the entire western US, while maintaining a robust quantification of uncertainty. To this end, this dissertation makes four contributions: (1) Identification of eight seasonally dependent and spatially coherent regions for precipitation extremes in the western US and the dominant moisture sources and delivery pathways for each region. Extreme value clustering and storm trajectory analysis were used in this effort. (2) Development of a Bayesian hierarchical spatial model for precipitation extremes in large regions, such as the western US, incorporating data from thousands of observation locations. This model can produce maps of gridded return levels and to simulate gridded precipitation extremes at any resolution, with associated uncertainties. (3) Development of a Bayesian hierarchical model for joint nonstationary precipitation and streamflow frequency analysis in rainfall-runoff dominated basins such as those in the southwest US. Peak streamflow is modeled as a consequence of spatial precipitation extremes, allowing for temporal nonstationarity from climate covariates. (4) Development of a general hierarchical Bayesian multivariate nonstationary frequency analysis framework with an application to dam safety analysis. In this, the peak snow depth and peak streamflow are linked to reservoir water level and, consequently, dam safety. These contributions provide new insights into the mechanisms influencing hydroclimate extremes in the western US and offer new frameworks for spatial, nonstationary, and multivariate modeling of extremes with an emphasis on the robust quantification of uncertainties. These contributions have significant applicability to engineering design, planning, risk assessment, and mitigation strategies for managing aging infrastructure under changing climate hazards.