Parameter and Uncertainty Estimation of Physical, Chemical, and Biological processes in Environmental Models using Stochastic Inverse Modeling
Surface water is a vital resource for drinking water, food/fishing, recreation, transportation, industry, and agriculture. Human population growth and urbanization threaten this already precious resource via agricultural intensification and waste production. When existing vegetation is cleared for farmland, overland flow from rain is unimpeded and can more easily suspend and transport sediments. As topsoil is removed, farmers must increasingly rely upon fertilizers. Both agricultural runoff and municipal wastewater can contain contaminant like heavy metals and pathogens, as well as excess nutrients that pollute the water bodies they flow into. In small, balanced amounts, nutrients like nitrogen are necessary for healthy vegetation and wildlife. The nutrient loading from runoff and waste water, however, can cause potentially toxic algal blooms followed by heterotrophic bacteria that consume oxygen while feeding on the decaying algae. This results in hypoxia, a complete depletion of dissolved oxygen, and thus wildlife suffocation. Land management and wastewater treatment are used to prevent algal blooms, but the former has been difficult to evaluate quantitatively and the latter is very energy intensive.This dissertation focuses on modeling these two systems: surface runoff and advanced nitrogen removal from wastewater. Environmental models are often complex combinations of physical, chemical, and biological process with many parameters, some of which cannot be directly measured or have a wide range of reported values in literature. As these models ofteninform environmental policy and decision-making, it is also important that predictions incorporate uncertainty and confidence. I have used Bayesian inference with Markov-Chain Monte Carlo simulations to estimate parameters and uncertainty for these systems.To quantify agricultural runoff, elemental fingerprints were collected from sediment samples of several land use types (agriculture, forest, unpaved roads, and streambanks) in the watershed of Laurel Hill Creek, PA. Fluvial samples to collect suspended sediments were taken following 6 rain events between 2010 and 2011. I used a Chemical Mass Balance (CMB) approach to assess the contribution of each source into the fluvial samples. The results confirm the influence of agricultural runoff as up to 80% of the sediments could be attributed to this land use. We also found that agriculture contribution increases with more intense rain events. This is important as climate change is predicted to increase the intensity of weather events.Most modern treatment facilities perform nitrogen removal via activated sludge which convert ammonia to nitrogen gas via nitrite and nitrate. Some of the bacteria require oxygen (aerobic autotrophs) and others require an additional carbon substrate (heterotrophic) to thrive and accomplish these goals, but aeration and chemical addition are costly and contribute to the plant’s carbon footprint. An experimental pilot reactor incorporating a new anaerobic ammonia oxidizing autotrophic bacteria process called anammox was operated from 2014-2015. Observations from this reactor and batch tests of sludge were used for model discrimination and parameter estimation using inverse modeling. When anammox bacteria are limited by both electron donors and acceptors, only the minimum of the two – rather than the product – was found to effect the growth rate. The probability density functions of estimated parameters can be used to optimize treatment operation by promoting anammox growth and reducing the need for aeration energy or chemical additions.
StatsViewed 36 times
Downloaded 7 times