Graduation Year

2023

Document Type

Thesis

Degree

M.A.

Degree Name

Master of Arts (M.A.)

Degree Granting Department

Geography

Major Professor

Joni Downs Firat, Ph.D.

Committee Member

Mark Rains, Ph.D.

Committee Member

Kai Rains, Ph.D.

Keywords

Florida, Spring, Kentucky, Maximum Entropy (Maxent), Oklahoma

Abstract

The World Resources Institute reveals that 17 countries face extremely high levels of water stress. Moreover, with increasing population and industrialization, the gap between water supply and demand increases day by day around the world. Groundwater is a key freshwater source, and springs are important resources as they enable to access groundwater. Therefore, it is crucial to monitor, protect, and manage groundwater springs. The first step in spring management is to recognize and define freshwater resources and to determine the locations of groundwater springs that serve as natural discharge points. Traditionally, field studies have been employed to determine the locations of springs. However, in some cases, fieldwork can be difficult, expensive, or even impossible. In this study, one of the general-purpose machine learning approaches, Maxent modeling, was used to predict the locations of springs in Florida, Oklahoma, and Kentucky, U.S.A., using topographic and bioclimatic data. Elevation, slope, flow weighted slope, distance to flowlines, planform curvature, profile curvature, CaCO3, sand, silt, clay, topographic wetness index, rainfall, and terrain ruggedness index data were used to predict spring locations. The goal of the study was to both determine if Maxent modeling can accurately predict spring locations in diverse geographic areas and to identify which environmental factors are the most predictive. For Florida, Kentucky, and Oklahoma, respectively, the model yielded AUC values of 0.89, 0.72, and 0.92. Distance to the river was the most important model contributor in Florida, with clay soil in Kentucky and profile curvature in Oklahoma.

Download

Share

COinS