Rethinking Green Infrastructure Performance: PMT removal in GAC-Amended Column Experiments under Extreme Operating Conditions

Rapid urbanization has expanded impervious surfaces, enhancing pollutant buildup and the transport of Persistent Mobile (PMT) substances through stormwater runoff. Green infrastructure, which is designed for flood mitigation and aquifer recharge, can inadvertently transfer polar contaminants into the soil–groundwater systems. As climate change drives more intense storms, measurements of higher-throughput stormwater are necessary.  Understanding their limits in removing dissolved PMTs and metals is essential to improve current mitigation strategies.

To investigate these hydraulic and geochemical performance constraints, we conducted a series of controlled fixed-bed column experiments simulating diverse green infrastructure operating conditions. Different PMT loadings, adsorbent dosage, competitive interactions with co-solutes (dissolved metals and dissolved organic matter, DOM) under three infiltration-rate regimes (4.5 – 25.5 cm·h^-1) were tested. Columns were packed with a mixture of sandy-loam soil and granular activated carbon (0.5, 2, and 5 %wt.; (GAC) to evaluate the breakthrough behaviour and adsorption capacity towards 8 representative PMTs with different physicochemical molecular properties. Our preliminary results show that at high flow rates, associated with low residence times, substantially decrease adsorption performance, particularly under high inorganic contaminant loads with DOM. Although the 5% GAC amendment achieved the highest overall removal capacity towards the studied PMTs regardless experimental conditions, it also introduces hydraulic limitations, producing pronounced tailing effect driven by micropore diffusion and extended intra-particle residence times.

To interpret the observations from experiments, HYDRUS will be applied to simulate reactive solute transport, enabling inverse calibration of hydraulic properties and solute transport parameters from column breakthrough data. Subsequently, HYDRUS-derived parameters, together with experimental variables, will be integrated into a machine learning framework to identify key removal predictors, and forecast the removal of challenging PMTs under different stormwater conditions.