The Impacts of Climate Change and Land Use on Urban Water Quality in Indianapolis

Abstract

Urban water systems are increasingly vulnerable to microbial contamination driven by climate change, aging infrastructure, and intensified land use. This dissertation investigates long-term trends in E. coli concentrations in Indianapolis, Indiana, and explores the influence of climatic and environmental drivers on urban water quality using a combination of statistical analyses and machine learning models. Chapters 1 and 2 investigated the temporal and spatial trends of urban water quality in Indianapolis, with a focus on climate and land use as key drivers. In Chapter 1, long-term analysis of E. coli concentrations in the Pleasant Run watershed revealed a significant upward trend, with values exceeding the Indiana standard of 235 MPN/100 mL since 1998. Concentrations increased from 111 MPN/100 mL in 1999 to 911 MPN/100 mL in 2019. Precipitation and stream discharge explained 60% of the observed variability. Under the RCP 8.5 climate scenario, E. coli levels are projected to increase by up to 58% by the 2080s. Chapter 2 expanded the analysis to include nitrate, sulfate, and chloride concentrations across 12 sites in Pleasant Run and Fall Creek. Seasonal and spatial differences were observed, with E. coli peaking in summer and chloride in March due to road salt runoff. Key drivers included 7-day antecedent precipitation and snow, urban built-up area, tree cover, and NDVI. Together, the results underscore the combined influence of climate variability and land use patterns on urban water quality, highlighting the need for integrated, climate-informed management strategies. Chapter 3 applies machine learning approaches Random Forest, XGBoost, and LightGBM-based quantile regression to model non-linear patterns of E. coli concentrations in a weekly scale. These models capture complex interactions among environmental variables and outperform linear models in predictive accuracy. Quantile regression further provides probabilistic estimates, enabling risk-based assessments of high E. coli levels. Overall, this research demonstrates that urban water quality in Indianapolis is significantly impacted by climate variability and land use dynamics. The integration of long-term monitoring data with interpretable machine learning models offers valuable tools for predicting microbial risks and informing climate-resilient water quality management in urban environments.