Term of Award

Fall 2024

Degree Name

Master of Science, Applied Geography

Document Type and Release Option

Thesis (open access)

Copyright Statement / License for Reuse

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.

Department

Department of Geology and Geography

Committee Chair

Munshi Rahman

Committee Member 1

Wei Tu

Committee Member 2

Meimei Lin

Abstract

Globally, as extreme weather patterns intensify, flash floods have emerged as one of the most destructive and immediate environmental threats. In Maryland, flash floods are particularly concerning due to its diverse topography and increasing urban development, which exacerbates runoff and overwhelms drainage systems. The state has experienced significant flash flood events, highlighting the need for effective models to manage risks and inform mitigation strategies. While regression models such as the Negative Binomial (NB) and Zero-Inflated Negative Binomial (ZINB) are commonly used for count data analysis, their application to flash flood modeling in the USA, including regions like Maryland, remains limited and warrants further exploration. This study addresses that gap in the literature by using these models to model flash flood occurrences across Maryland. The objectives are to evaluate the association between key factors (i.e., precipitation, elevation, slope, topographic wetness index, drainage density, population density, and land use land cover) and flash flood risks and identify and map high-risk areas. Flash flood data obtained from the NOAA Storm Events Database (2013–2023) includes 373 flash flood points and 2,651 non-flood points. The data are structured on a 4 x 4 km grid with the flash flood influencing factors in raster and vector data formats. Model performance was assessed using the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and McFadden's Pseudo R². The ZINB model demonstrated superior performance over the NB model, achieving lower AIC (2822.343 vs. 2856.105) and BIC (2924.587 vs. 2946.32), higher log-likelihood (-1394.172 vs. -1413.053) and McFadden's Pseudo R² (0.1588 vs. 0.1474), as well as a reduced MAE (0.37528 vs. 0.4246156), highlighting its better fit. Analysis of the ZINB model reveals that the significant factors of flash flood occurrence include precipitation (β = 0.1121, p < 0.001), which suggests that higher precipitation increases the likelihood of flash floods, drainage density (β = 3294, p < 0.001), indicating higher flood risk in areas with poor drainage, population density (β = 0.06808, p < 0.001), where urban areas are more vulnerable. Building density (β = 832.8, p < 0.001) suggests dense urban development increases flash flood risk. Conversely, land use/land cover - water (β = -1.48, p < 0.001) shows a negative association, implying that areas with water bodies are less prone to flash floods. Flood risk modeling based on the ZINB model indicates that highly urbanized areas, including Baltimore, Montgomery, and parts of Prince George's County, face the highest risks. In contrast, rural regions like Garrett and Allegany counties display minimal risks. The study recommends that vulnerable urban counties, including Baltimore, Montgomery, and Prince George's, invest more in advanced stormwater management systems like green roofs, permeable pavements, and rain gardens to mitigate future flash flood risks.

Research Data and Supplementary Material

Yes

Download

Share

COinS