Fall 2022

Document Type

Master's Thesis (Open Access)

Degree Name

Master of Science (M.S.)


Applied Environmental Science


Water chemistry affects organisms at all levels of the food web in aquatic habitats. Water quality alteration from natural conditions can seriously degrade habitat quality, human health, and the survival rate of native species. Estimating background (i.e., historic or baseline) water quality of impaired streams can be difficult due to the effects of anthropogenic involvement in aquatic systems over decades to centuries. The ability to model natural background water quality levels would aid in overall stream management. The natural background predictions provided by a model would allow for increased understanding of stream condition and would allow us to determine the amount of divergence between current stream conditions and natural background.

To better predict baseline water chemistry levels, we created random forest models for ionic concentrations and integrated water quality measures of ionic balance, including chloride, calcium, magnesium, sodium, sulfate, alkalinity, hardness, total dissolved solids, and specific conductivity. Water quality measurements from minimally disturbed reference sites across the United States were used as response variables for model training. We developed these models using both static (e.g., geology, soils, etc.) and dynamic (i.e., monthly evapotranspiration, precipitation, and temperature) EPA StreamCat and PRISM predictor variables. The models explained 66% to 98% of the variation in samples from California streams and 55% to 85% of the variation across the US. The top predictors across models include yearly temperature averages, yearly precipitation averages, percent lithological sulfur, and base flow index. The baseline water chemistry estimates produced by these models will help California establish site-specific water quality standards and manage habitat in various situations, including urban development projects, habitat restoration, and endangered species monitoring.