Predictability and trends of annual pollutant loads in Midwestern watersheds

Nutrient loads;Midwestern watersheds;Trends;Load estimation;Climate Change;Data mining

Verma, Siddhartha

关键词: Nutrient loads;    Midwestern watersheds;    Trends;    Load estimation;    Climate Change;    Data mining;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/46851/Siddhartha_Verma.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

The effect of multiple stressors on global water resources have been increasing rapidly over the past few decades. Anthropogenic activities such as rapid industrialization, urbanization, deforestation and increased application of agricultural nutrients have led to a decline in overall quality of our aquatic environment. Additionally, these activities have increased greenhouse gas concentrations globally, warming the earth’s atmosphere and eventually having a detrimental effect on global water and energy balances. The global water cycle has been altered, leading to its overall intensification and an increase in frequency of extreme events and such as floods and droughts. Also, the demands of higher quality water have been rising globally attributed to a burgeoning world population, further stressing the water resources. To address the increased water demands worldwide coupled with declining water quality and depletion of water resources requires new approaches in water management along with improvement in water use efficiencies. To facilitate development of newer approaches of water management and solutions to alleviate global water problems requires an overall comprehensive assessment of our water resources. A key step in these assessments is water quality monitoring which will help improve our ability to predict water quantity, quality and distribution on a global scale. In this research, I aim to improve our knowledge of anthropogenic and natural impacts on global water resources, largely focusing on water quality monitoring by evaluating and refining the science of predicting pollutant (nutrient and sediment) export from large scale watersheds. To enable these goals, this research is centered on large watersheds in Midwestern United States, which have been some of the primary sources of nutrient and sediment export to downstream water bodies such as the Gulf of Mexico and Lake Erie leading to massive eutrophication. In total, fourteen watersheds with extensive water quality datasets are analyzed in different stages of this research. Typically, these large watersheds are predominantly agricultural with intensive row-cropped farmlands having a network of sub-surface tile drain systems. The science of pollutant export and various hydrological processes associated with it have been simulated using three major modeling approaches namely statistical and empirical modeling, physically-based modeling and data mining methods. In this research I improve, apply or evaluate all three approaches to meet specific objectives related to annual pollutant load predictions and trend assessments. In the first part of this research, I use regression techniques to assess the role of large load events in predicting annual pollutant (Suspended Solids (SS), Total Phosphorus (TP) and Nitrate-Nitrogen (NO3-N)) loads. In doing so, a novel baseflow separation technique based on mechanistic differences in nutrient and sediment export is proposed and applied. Then, I assess the spatio-temporal patterns of pollutant export from large Midwestern watersheds using circular statistics. This enables identification of critical periods of high load export and also gaging impacts of landuse, management practices, and sources of pollution on overall annual loads. These analyses constitute the first such application of these approaches on a large spatio-temporal scale especially for nutrient export dynamics. I next calibrate a physically-based SWAT model for hydrology and water quality predictions in the largest watershed in the Lake Erie basin. I use this calibrated model to gage the impacts of future projected climate changes from the mid-century and late-century time periods on the hydrology and water quality in the watershed. Further, I evaluate two data mining techniques namely the nearest-neighbor method and decision trees which have scarcely been used in hydrology to predict missing NO3-N concentrations for two extensively monitored watersheds in the Lake Erie basin. Lastly, I evaluate the impacts of available water quality data for concentration and load predictions and trend calculations based on traditional statistical methods and some new improved and modified approaches which have not yet been applied extensively.