Confronting the natural variability and modeling uncertainty of nonpoint source pollution in water quality management
Date
2017
Authors
Tasdighi, Ali, author
Arabi, Mazdak, advisor
Bledsoe, Brian, committee member
Bailey, Ryan, committee member
Hoag, Dana, committee member
Journal Title
Journal ISSN
Volume Title
Abstract
Nonpoint source pollution is the primary cause of impaired water bodies in the United States and around the world. Hence, managing the water quality is hinged mainly on controlling this type of pollution. However, characterization of nonpoint source pollution is extremely difficult due to high inherent natural variability and uncertainty. Nonpoint source pollution loads depend on climate, land use, and other environmental conditions that are highly variable by nature. On the other hand, since it is often infeasible to measure pollutant loads from nonpoint sources within a watershed using monitoring campaigns, models are increasingly used to estimate these loads. Models are simplified representations of reality. Consequently, they are subject to various sources of uncertainty including: model parameters, input data (climate, land use, etc.), model structure (conceptualization), and measurement data (streamflow, nutrient concentrations or loads, etc.).
The overarching goal of this dissertation is to characterize the natural variability and modeling uncertainty of nonpoint source pollution and probabilistically quantify the water quality benefits of conservation practices. To achieve this goal, first the relationship between land use and stream water quality was explored under various climatic conditions using multiple linear regression models. This analysis showed that strong and significant relationships exist between land use and ambient water quality. The strength and significance of these relationships changed with climatic conditions. Higher contribution of nonpoint sources in degrading water quality during the wet climate conditions was notable. Second, various sources of modeling uncertainties were characterized in simulating the hydrologic budgets specifically streamflow regimes, for various spatial scales and upstream land use conditions. The results of this analysis highlights important implications for the selection and application of appropriate rainfall-runoff methods within complex distributed hydrologic models, particularly when simulating hydrologic responses in mixed-land use watersheds. Third, a total uncertainty estimation framework was developed to assess the effectiveness of conservation practices in reducing nonpoint source pollution. The Bayesian-based framework entails a two-stage procedure. First, various sources of modeling uncertainties are characterized during the period before implementing Best Management Practices (BMPs). Second, the effectiveness of the BMPs are probabilistically quantified during the post-BMP period. Results indicate that the modeling uncertainties in quantifying the effectiveness of BMPs vary based on hydrologic conditions. Higher errors were observed in simulating nonpoint source pollution loads during high flow events. The results of this study have important implications for decision making when models are used for water quality simulation and management.
"What man really needs is not just more knowledge, but more certainty." Bertrand Russell, 1964.
The overarching goal of this dissertation is to characterize the natural variability and modeling uncertainty of nonpoint source pollution and probabilistically quantify the water quality benefits of conservation practices. To achieve this goal, first the relationship between land use and stream water quality was explored under various climatic conditions using multiple linear regression models. This analysis showed that strong and significant relationships exist between land use and ambient water quality. The strength and significance of these relationships changed with climatic conditions. Higher contribution of nonpoint sources in degrading water quality during the wet climate conditions was notable. Second, various sources of modeling uncertainties were characterized in simulating the hydrologic budgets specifically streamflow regimes, for various spatial scales and upstream land use conditions. The results of this analysis highlights important implications for the selection and application of appropriate rainfall-runoff methods within complex distributed hydrologic models, particularly when simulating hydrologic responses in mixed-land use watersheds. Third, a total uncertainty estimation framework was developed to assess the effectiveness of conservation practices in reducing nonpoint source pollution. The Bayesian-based framework entails a two-stage procedure. First, various sources of modeling uncertainties are characterized during the period before implementing Best Management Practices (BMPs). Second, the effectiveness of the BMPs are probabilistically quantified during the post-BMP period. Results indicate that the modeling uncertainties in quantifying the effectiveness of BMPs vary based on hydrologic conditions. Higher errors were observed in simulating nonpoint source pollution loads during high flow events. The results of this study have important implications for decision making when models are used for water quality simulation and management.
"What man really needs is not just more knowledge, but more certainty." Bertrand Russell, 1964.
Description
Rights Access
Subject
hydrology
modeling uncertainty
water quality
land use
BMP
statistical analysis