Data from the International Stormwater Best Management Practices (BMP) Database were used to compare the phosphorus (P) control performance of six categories of stormwater BMPs representing traditional systems (stormwater pond, wetland basin, and detention basin) and low-impact development (LID) systems (bioretention cell, grass swale, and grass strip). Machine learning (ML) models were trained to predict the reduction or enrichment factors of surface runoff concentrations and loadings of total P (TP) and soluble reactive P (SRP) for the different categories of BMP systems. Relative to traditional BMPs, LIDs generally enriched TP and SRP concentrations in stormwater surface outflow and yielded poorer P runoff load control. The SRP concentration reduction and enrichment factors of LIDs also tended to be more sensitive to variations in climate and watershed characteristics. That is, LIDs were more likely to enrich surface runoff SRP concentrations in drier climates, when inflow SRP concentrations were low, and for watersheds exhibiting high impervious land cover. Overall, our results imply that stormwater BMPs do not universally attenuate urban P export and that preferentially implementing LIDs over traditional BMPs may increase TP and SRP export to receiving freshwater bodies, hence magnifying eutrophication risks.

Phosphorus (P) export from urban areas via stormwater runoff contributes to eutrophication of downstream aquatic ecosystems. Bioretention cells are a Low Impact Development (LID) technology promoted as a green solution to attenuate urban peak flow discharge, as well as the export of excess nutrients and other contaminants. Despite their rapidly growing implementation worldwide, a predictive understanding of the efficiency of bioretention cells in reducing urban P loadings remains limited. Here, we present a reaction-transport model to simulate the fate and transport of P in a bioretention cell facility in the greater Toronto metropolitan area. The model incorporates a representation of the biogeochemical reaction network that controls P cycling within the cell. We used the model as a diagnostic tool to determine the relative importance of processes immobilizing P in the bioretention cell. The model predictions were compared to multi-year observational data on 1) the outflow loads of total P (TP) and soluble reactive P (SRP) during the 2012-2017 period, 2) TP depth profiles collected at 4 time points during the 2012-2019 period, and 3) sequential chemical P extractions performed on core samples from the filter media layer obtained in 2019. Results indicate that exfiltration to underlying native soil was principally responsible for decreasing the surface water discharge from the bioretention cell (63 % runoff reduction). From 2012 to 2017, the cumulative outflow export loads of TP and SRP only accounted for 1 % and 2 % of the corresponding inflow loads, respectively, hence demonstrating the extremely high P reduction efficiency of this bioretention cell. Accumulation in the filter media layer was the predominant mechanism responsible for the reduction in P outflow loading (57 % retention of TP inflow load) followed by plant uptake (21 % TP retention). Of the P retained within the filter media layer, 48 % occurred in stable, 41 % in potentially mobilizable, and 11 % in easily mobilizable forms. There were no signs that the P retention capacity of the bioretention cell was approaching saturation after 7 years of operation. The reactive transport modeling approach developed here can in principle be transferred and adapted to fit other bioretention cell designs and hydrological regimes to estimate P surface loading reductions at a range of temporal scales, from a single precipitation event to long-term (i.e., multi-year) operation.

Phosphorus (P) export from urban areas via stormwater runoff contributes to eutrophication of downstream aquatic ecosystems. Bioretention cells are a Low Impact Development (LID) technology promoted as a green solution to attenuate urban peak flow discharge, as well as the export of excess nutrients and other contaminants. Despite their rapidly growing implementation worldwide, a predictive understanding of the efficiency of bioretention cells in reducing urban P loadings remains limited. Here, we present a reaction-transport model to simulate the fate and transport of P in a bioretention cell facility in the greater Toronto metropolitan area. The model incorporates a representation of the biogeochemical reaction network that controls P cycling within the cell. We used the model as a diagnostic tool to determine the relative importance of processes immobilizing P in the bioretention cell. The model predictions were compared to multi-year observational data on 1) the outflow loads of total P (TP) and soluble reactive P (SRP) during the 2012-2017 period, 2) TP depth profiles collected at 4 time points during the 2012-2019 period, and 3) sequential chemical P extractions performed on core samples from the filter media layer obtained in 2019. Results indicate that exfiltration to underlying native soil was principally responsible for decreasing the surface water discharge from the bioretention cell (63 % runoff reduction). From 2012 to 2017, the cumulative outflow export loads of TP and SRP only accounted for 1 % and 2 % of the corresponding inflow loads, respectively, hence demonstrating the extremely high P reduction efficiency of this bioretention cell. Accumulation in the filter media layer was the predominant mechanism responsible for the reduction in P outflow loading (57 % retention of TP inflow load) followed by plant uptake (21 % TP retention). Of the P retained within the filter media layer, 48 % occurred in stable, 41 % in potentially mobilizable, and 11 % in easily mobilizable forms. There were no signs that the P retention capacity of the bioretention cell was approaching saturation after 7 years of operation. The reactive transport modeling approach developed here can in principle be transferred and adapted to fit other bioretention cell designs and hydrological regimes to estimate P surface loading reductions at a range of temporal scales, from a single precipitation event to long-term (i.e., multi-year) operation.