Masters Theses

Date of Award

8-2025

Degree Type

Thesis

Degree Name

Master of Science

Major

Geography

Major Professor

Yingkui, Li

Committee Members

Sally Horn, Dimitris Herrera

Abstract

Wildfires significantly alter landscapes, increasing susceptibility to debris flows during subsequent rainfall events. In Washington State’s Methow Valley, the 2021 Cedar Creek, Cub Creek 2, and Muckamuck fires burned over 56,000 hectares, but produced markedly different geomorphic responses. This study examines the factors driving these differences using machine learning and high-resolution environmental data. I modeled post-fire debris flow susceptibility across 1,378 basins using Random Forest (RF) and Extreme Gradient Boosting (XGBoost) algorithms, incorporating basin morphology, soil properties, burn severity, vegetation indices, and short-duration precipitation intensity. These models were compared to the widely used USGS M1 logistic regression model to assess regional accuracy. Results show that 15-minute rainfall intensity was the dominant trigger of debris flows. Terrain features like Melton’s Ruggedness Number and slope, as well as soil characteristics such as sand content, were also influential predictors. Among the RF models, the one using USGS M1 model-aligned variables showed the strongest performance, offering improved regional hazard mapping over the M1 model alone. Satellite precipitation data proved useful for climate context but insufficient for capturing localized storm peaks, reinforcing the need for dense rain gauge networks in mountainous terrain. This research advances post-fire hazard assessment in the Pacific Northwest by emphasizing regional model calibration and the integration of high-resolution datasets.

Download
Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Included in

Geomorphology Commons

Share

COinS