Date of Award

Spring 1-1-2013

Document Type

Thesis

Degree Name

Master of Science (MS)

First Advisor

Balaji Rajagopalan

Second Advisor

Edith Zagona

Third Advisor

Subhrendu Gangopadhyay

Abstract

The Colorado River Basin (CRB) is the most important source of water in the southwestern United States. Since the Upper Colorado River Basin provides between 80-90% of the total CRB's annual flow, previous investigations of water supply reliability for the CRB have focused on the Upper Basin. Given the smaller contribution of Lower Colorado River Basin (LCRB) flows, and that the Gila River Basin is not explicitly incorporated in the legal and management structure governing the rest of the CRB, Lower Basin flows have been typically left out of CRB water supply modeling. However, as the demand on the CRB system approaches the supply, the 10-20% contribution from the Lower Colorado River Basin (LCRB) becomes more critical. In order for these flows to be incorporated into planning frameworks, it is important to understand their long-term variability and develop robust simulation methods to capture the variability. These research needs motivated the present study. This thesis has two main sections: (i) development of new techniques for tree-ring reconstructions that can capture streamflow characteristics typically not captured by traditional methods (such as intermittency) and (ii) incorporating the reconstructions in a CRB water supply model to assess the impact of LCRB flows, including the Gila River, on overall CRB water supply risk.

For the first section, we offer two new statistical methods for tree-ring reconstructions of streamflow. The first method employs a cluster analysis on a regional network of tree-ring chronologies to identify spatially coherent subregions that have a common climate signal, then Principal Component Analysis (PCA) is performed on the clusters to obtain the leading modes of variability. The leading modes are used as predictors in a local polynomial model to fit the observed natural streamflows. This improves upon a similar approach, where K-nearest neighbor resampling was used without the local polynomial, in that it can extend the reconstructed values beyond the range of the observed data and also capture nonlinearities. The second method introduces extreme value analysis (EVA) to reconstructing threshold exceedances of streamflow. Underlying this method is the premise that the Gila River only contributes to Colorado River in a meaningful way during high flow events. The EVA models the probability of threshold exceedance, and the magnitude of exceedances, and is especially suited for reconstructing intermittent streamflow. These two new reconstruction methods were tested on the LCR mainstem flows and Gila River flows at its confluence with the Colorado River. These methods provide the ability to capture additional aspects of streamflow that are difficult or not possible to reconstruct with traditional linear regressions.

For the second section, streamflow simulations resampled from the new LCR mainstem and Gila River flow reconstructions are combined with previous reconstructions of the Upper Colorado River Basin (at Lees Ferry, AZ) in a water supply model to investigate the risk of active system storage being depleted under different scenarios. The water supply model is driven by three climate change scenarios, two reservoir operation rule sets, and two projected demand levels. We find that including Gila River flows provides water supply risk mitigation capability. It reduces the Colorado River system risk by 4-17% during mid-2050 under a severe 20% reduction in mean flow due to climate change. Furthermore, including the Gila reduces the average shortage volume per year, increases the storage volume in the system and reduces the average number of shortages.

The new reconstruction methods coupled with the demonstration of their utility in assessing water supply risk provides an attractive framework to studying water supply management and planning in other river basins, especially in semi-arid regions.

Available for download on Thursday, July 18, 2019

Share

COinS