Five years evolution of hydraulic properties of engineered soil of experimental bioretention cell planted with perennials

This study presents a five-year experimental investigation conducted within the Horizon Europe project NBSINFRA on a bioretention cell located at the University Centre for Energy Efficient Buildings of the Czech Technica University in Prague. The system was constructed as a multilayer system comprising a biofilter layer, a sand layer, and a drainage layer, and planted with perennial vegetation (Aster novae-angliae, Hemerocallis, Molinia caerulea and Eupatorium 'Phantom'). The biofilter consisted of 50% sand, 30% compost, and 20% topsoil. The bioretention cell was connected to roof of neighboring building and hydraulically isolated from the surrounding soil using a waterproof membrane to allow for water balance monitoring and equipped with a set of monitoring sensors. Soil water content within the biofilter was measured using four Time Domain Reflectometry (TDR) probes, while five tensiometers were installed to record soil water potential. Outflow from the bioretention cell was measured by a tipping-bucket flowmeter, and inflow was estimated from precipitation using a rain gauge.

Over the five-year monitoring period, the study analysed water balance, biofilter water regime, and vegetation development to investigate their influence on water retention and detention in the bioretention cell. Rainfall–runoff episodes were analysed individually to quantify changes in episodic runoff coefficients, peak flow reduction, and runoff delay, and a semi-quantitative approach was applied to assess the effect of inter-annual vegetation development on evapotranspiration. Hydrological modelling was performed using two-dimensional simulations in HYDRUS 2D/3D, solving the Richards equation for variably saturated flow. The model represented vertical water flow through the multilayer bioretention profile, including infiltration from roof inflow and direct precipitation, drainage outflow, evapotranspiration, and root water uptake with a time-evolving root zone. Simulations focused on three representative 14-day study periods in August 2019, 2020, and 2023, selected to ensure comparable initial conditions and vegetation states. Model calibration and evaluation were based on measured outflow and pressure head dynamics. Parameter sensitivity was assessed using an informal Bayesian framework (GLUE) combined with Latin Hypercube Sampling, focusing on soil hydraulic parameters of the biofilter and sand layers.

The results showed a gradual decrease in the episodic and annual runoff coefficient over time, mainly driven by increasing inter-annual evapotranspiration and lower initial biofilter saturation at the beginning of rainfall events. Peak flow reduction ranged from 30% to 100%, with a median value of 88%, while runoff and peak runoff delays exhibited median values of approximately 30 and 56 minutes, respectively, with increasing variability in later years. Hydrological modelling and sensitivity analysis identified the saturated hydraulic conductivity of the sand layer and the van Genuchten parameter of the soil water retention curve n of the biofilter as the most influential parameters. Simulation results further indicated a decline in saturated hydraulic conductivity in both the biofilter and sand layers with the aging of the bioretention system. However, these changes did not impair the overall hydrological performance of the bioretention cell.