Date of Completion


Document Type

Honors College Thesis


Department of Mathematics and Statistics

Thesis Type

Honors College, College of Arts and Science Honors

First Advisor

Donna Rizzo


streamflow drought, Colorado River Basin, clustering, time series data, watershed attributes, artificial neural network


Streamflow drought, a subset of hydrological drought and important measure of a watershed’s health, is a complex environmental concern that impacts ecosystems and society in many ways. It is, therefore, important that we understand the development of streamflow drought, especially with the increasing effects of climate change. In this thesis, I explored the 20% level of streamflow drought using time series and watershed data collected and calculated by the USGS over the past 40 years across the Intermountain West, Rocky Mountains, Southwest, and High Plains regions of the United States. I used a new, unsupervised artificial neural network known as SOMTimeS (Self-Organizing Map for Time Series) alongside K-means clustering for the time series analysis. I used a Random Forest feature selection to examine a set of watershed attributes, with a particular focus on human-modified ones, in an attempt to back out important attributes associated with clustered watersheds. Results suggested that elevation and variables relating to timing and magnitude of water runoff impact the 20% streamflow drought threshold. Examining a low-elevation cluster of watersheds showed that attributes related to agricultural land use likely impact this type of streamflow drought. Furthermore, shrub and Evergreen Forest land cover appeared to impact this region. These findings support the hypothesis that the vegetation and land use within a watershed might significantly impact the 20% level of streamflow drought. Future research includes examining each watershed attribute as a component plane and comparing USGS streamflow drought model performance across the different clusters identified in this thesis.

Creative Commons License

Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 License.