Developing nitrogen removal models for stormwater bioretention systems

Bioretention systems have the potential of simultaneous runoff volume reduction and nitrogen removal. Internal water storage (IWS) layers and real-time control (RTC) strategies may further improve performance of bioretention systems. However, optimizing the design of these systems is limited by the lack of effective models to simulate nitrogen transformations under the influences of IWS design and environment conditions including soil moisture and temperature. In this study, nitrogen removal models (NRMs) are developed with two complexity levels of nitrogen cycling: the Single Nitrogen Pool (SP) models and the more complex 3 Nitrogen Pool (3P) models. The 0-order kinetics, 1^st order kinetics, and the Michaelis-Menten equations are applied to both SP and 3P models, creating six different NRMs. The Storm Water Management Model (SWMM), in combination with each NRM, is calibrated and validated with a lab dataset. Results show that 0-order kinetics are not suitable in simulating nitrogen removal or transformations in bioretention systems, while 1^st order kinetics and Michaelis-Menten equation models have similar performances. The best performing NRM (referred to as 3P-m) can accurately predict nitrogen event mean concentrations in bioretention effluent for 20% more events when compared to SWMM. When only calibrated with soil moisture conditions in bioretention systems without internal storage layers, 3P-m was sufficiently adaptable to predict cumulative nitrogen mass removal rates from systems with IWS or RTC rules with less than ±7% absolute error, while the absolute error from SWMM prediction can reach -23%. In general, 3P models provide higher prediction accuracy and improved time series of biochemical reaction rates, while SP models improve prediction accuracy with less required user input for initial conditions.