Dynamic Simulation of Urban Bioinfiltration Systems

Current designs of stormwater green infrastructure (GI) mostly utilize static principles and do not include the dynamic aspects of infiltration during the storm event and evapotranspiration (ET) to restore soil storage capacity for the next storm event. Long term monitoring of GI systems indicate that GI systems do not behave in a static manner. For example, several small rainfall events in a row may diminish the soil infiltration capacity for the next storm event. Also, ponding times for the same size event may be different throughout the year. A continuous simulation using long term hydrologic input data is needed to consider dynamic aspects of infiltration and ET and further benefit from what GI systems have to offer.

Bioinfiltration systems can be defined as vegetated GIs with no underdrain that allow the stormwater runoff to be exfiltrated into the surrounding in-situ soil. Two types of bioinfiltration systems, rain gardens and tree trenches (a small footprint system of tree pits that are connected by an underground continuous stone trench) are the focus of this research. Both GI systems are popular in US cities. A bioinfiltration rain garden in Villanova, PA and a tree trench in Philadelphia, PA, were simulated using both continuous and event-based models, yielding four discrete models. Rainfall measurements and continuous monitoring of the rain garden and the tree trench using water level sensors have been conducted by Villanova University since 2003 and by Philadelphia Water Department since 2012, respectively. Additional monitoring of the tree trench by Villanova University has been executed since April 2018. All four models were calibrated and validated against observed data (R² = 0.66-0.99, NSE = 0.725-0.92, and PBIAS = -30.6%-54%). Overflow duration, time to zero water depth, and overall curve fit were selected as calibration metrics in this research. The main recommendation was to use an appropriate and seasonally variable saturated hydraulic conductivity value to simulate exfiltration and incorporate dynamic design principles. The tree trench also required the addition of a virtual outlet to adequately mimic the effect of head-dependent infiltration. The results and discussions provide insights for dynamic modeling and design of increasingly popular bioinfiltration GI systems in urban areas.