Abstract
With climate change scenarios and rapid urbanization projected for the coming decades, urban environments are facing more frequent extreme precipitation events and higher pollutant loads. While static bioretention systems have the potential to retain stormwater runoff volume and remove pollutant loads, their effectiveness in preventing floods is challenged by increasingly frequent extreme events. Although bioretention systems are effective in removing sediment, heavy metals, and phosphorous, nitrogen removal rates vary significantly with location, age, maintenance, and design of bioretention systems, and net nitrogen export has also been observed.
To enhance nitrogen removal in bioretention systems, submerged zones can be designed to provide long periods of high saturation in soil layer that facilitate denitrification. However, a deep submerged layer and higher saturation rate in the soil layer can hinder the system's ability to handle stormwater during subsequent events, leading to flooding. Balancing the conflicting requirements of hydraulic retention time from both volume reduction and nitrogen removal goals can be achieved by implementing real-time control (RTC), or valve control rules. To select the best valve control rules based on specific stormwater management goals, weather conditions, and bioretention designs, it is helpful to have modeling results that predict nitrogen loads or concentrations in bioretention underdrain effluents under different operational and environmental conditions in the field.
In this dissertation, a literature and a case study were first conducted to find a readily available modeling tool that accurately simulate nitrogen removal rates or transformations in bioretention systems under the impacts of environmental and operational conditions that varies over time. Statistical models lack accuracy for specific bioretention systems and are not able to predict discharged nitrogen loads during events. Current stormwater and agricultural models complement each other with their strengths on simulating hydraulic processes in bioretention systems and nitrogen transformations under impacts of environmental factors, respectively. A previous attempt has proved that adding a process-based nitrogen model to the Storm Water Management Model (SWMM) improved prediction accuracy on nitrogen removal for wet pounds, but similar models have not been reported for bioretention systems. Therefore, we conclude that a process-based nitrogen model (NRM) needs to be developed for bioretention systems, and the most efficient way is by modifying the nitrogen module in agricultural models as extensions to the hydraulic modules in SWMM.
Six NRMs (SP-0, SP-1, SP-m, 3P-0, 3P-1. 3P-m) were developed with two model structures and three reaction kinetics. These NRMs were calibrated and validated using one set of laboratory data. The validation results show that 0-order kinetics is not suitable for NRMs. The SP-1, SP-m, 3P-1, and 3P-m models improved the prediction accuracy of percent removal of total load and event mean concentrations of total inorganic nitrogen in underdrain effluent by up to 20%. 3P-1, 3P-m outperformed SP-1 and SP-m in describing the impacts of environmental and operational conditions accurately.
The 3P-1 and 3P-m models were then updated and applied to simulate a field bioretention system. When simulating total dissolved nitrogen, the 3P-m model improved predictions of percent removal of total load by 5.5% to 10.6% and reduced scaled root mean square error by 16.2% to 53.0% when compared to SWMM. Statistical analysis confirmed that the 3P-m model accurately captures the impacts of environmental and operational conditions, and its simulated denitrification aligns with field isotope tests, providing strong evidence that the 3P-m model correctly describes the biochemical processes of nitrogen cycling in field-scale bioretention systems. Time-series generated by the models revealed that the 3P-1 model's calibration results are less reliable, but the 3P-m model can provide valuable insights to assist in the design of real-time control rules.
Despite some acknowledged limitations, the 3P-m model has demonstrated potential to significantly improve prediction accuracy on removal of nitrogen species, and to provide insights on nitrogen transformations under multiple environmental and operational conditions. Further updates and applications of the 3P-m model are encouraged, and its application in bioretention and valve control design is recommended as a potential research opportunity.
