Examining the Relationship Between Water Quality and Land Use Land Cover in the Huron River Watershed

Abstract

The nature of the landscape that composes a watershed plays a role in determining the chemistry of runoff that makes its way into a body of water. The goal of this thesis was to assess the impacts of different patterns of land use land cover (LULC) on water quality outcomes in the Huron River Watershed. Focusing on two different subwatersheds, Mill and Allen Creek, I used RStudio to create a series of statistical models for the prediction of five different water quality indicators using percent cover by urban area, percent cover by agriculture, and percent cover by impervious surfaces as explanatory variables. The water quality indicators in question were total dissolved solids, total suspended solids, total phosphorus, nitrite, and nitrate. I also performed additional analyses in the form of two sample t-tests, hierarchical clustering, principal components analysis (PCA), and logistic regression.

I experimented with several different types of models, including generalized additive models, support vector machine regression, random forest regression, neural network regression, and a linear model in OpenBugs. These methods were met with varying degrees of success as indicated by R-squared values and mean squared error (MSE). As the ways in which humans interact with land continue to evolve, it may be useful to devise viable methods of predicting how these changes will affect the quality of the water that we rely on for both drinking and recreation. I concluded that percent cover alone does not provide sufficient information to accurately predict the parameters in question and that it might be helpful to include additional predictors such as temperature, precipitation, and flow in the analysis.