Towards a model for road runoff infiltration management

Road runoff, generated by rainwater leaching from roads, transports a complex mixture of micropollutants (fuels, oils, brake fluids, tyre wear, de-icing agents, metals, PAHs, salts, microplastics, pesticides, and other toxicants) that pose risks to human health and the environment. Management practices such as underground infiltration, porous pavements, rain gardens, bioswales, wetlands, and reuse are employed worldwide, but the chemical and molecular content of road runoff is increasingly recognized as harmful and must be managed at source. Infiltration systems recharge groundwater and mitigate peak flows, yet concerns remain about soil and groundwater contamination, necessitating risk-benefit evaluation. Infiltration ponds can filter and adsorb pollutants, but most prior studies addressed only shortlists of contaminants (e.g., metals, chlorides, PAHs), often in laboratory columns with artificial runoff; few real-world studies exist. Critically, how soil physico-chemical properties govern micropollutant capture during infiltration is poorly understood. This study measures micropollutant abundance and soil physico-chemical properties in a full-scale road runoff facility (sedimentation pond followed by infiltration pond), characterizes pollutant distribution across matrices and depths, and correlates pollutant abundances with soil properties to propose a holistic model for optimizing runoff infiltration management.

Previous work on road runoff treatment has largely focused on selected pollutant families (notably metals, chlorides, and PAHs) and on laboratory or column studies assessing infiltration performance and fate processes. Some real-world studies indicate that infiltration reduces environmental impacts and enhances aquifer recharge. Broader screening and non-targeted HRMS approaches have revealed unexpected compounds such as pharmaceuticals and pesticides in runoff and groundwater, underscoring the need for comprehensive monitoring. Studies have shown that pollutant behavior is influenced by chemical properties (e.g., hydrophobicity/logP) and soil factors (pH, organic matter, texture) affecting sorption and partitioning; PFAS soil-water partitioning depends on both chemical and soil properties and solution pH. However, integrated assessments linking a wide spectrum of micropollutants to in situ soil physico-chemical properties in operational facilities remain scarce, motivating the present work.

Study site: Wolfisheim (Alsace, France) along department road RD45; facility commissioned in 2013. A lined concrete sedimentation pond collects runoff from 3.66 ha of road (≈5400 vehicles/day, ~400 lorries), followed by an infiltration pond. A weather station recorded 94 rain events totaling 410.8 mm between July 2018 and July 2019. Sampling: Water and settled solids (“sediment”) were sampled at the sedimentation pond inlet. In the infiltration pond, soils were cored at 0–10 cm, -80 cm, and -165 cm; bank soil served as control. Three replicates per matrix were collected, stored cold, and processed without drying. Four surface soil subsamples around the coring point were collected to quantify root versus soil proportions. Soil physico-chemical analyses: Standard NF/ISO procedures measured water content (ISO 11465), particle size distribution by laser diffraction (ISO 13320) and texture classes (clay <2 µm; silt 2–50 µm; sand 50–2000 µm), removal of organic matter and flocculants for dispersion, bulk density and porosity (NF P94-410-3), organic matter by loss on ignition at 550°C (24 h), and pH (NF EN 15933). Granulometric percentiles d10 (1.26–1.38 µm), d50 (20.7–22.73 µm), d99 (101.1–161.2 µm) and curvature coefficient were determined. Metals (Al, Cd, Cu, Ni, Pb, Hg) and chlorides were quantified by accredited methods; concentrations were below French regulatory thresholds and not pursued further. Micropollutant extraction and analysis: Soil (10 g) was extracted twice (acetonitrile:water (90:10) +1% acetic acid, overnight; then isopropanol:acetonitrile (90:10), 15 min), centrifuged, pooled, freeze-dried, reconstituted for LC-HRMS or GC-MS/MS. Water (50 mL) was freeze-dried and reconstituted for LC or GC analysis. LC-HRMS (Dionex Ultimate 3000 + Bruker Impact II QTOF) applied a TargetScreener method targeting 2072 pollutants (848 pesticides, 1224 toxicants) with retention time, exact mass, and fragments; C18 column; positive ion bbCID; 30–1000 Da. GC-MS/MS (SCION 436 triple quad) targeted 328 compounds (pesticides, PAHs, toxicants) using MRM on an Rxi-5Sil MS column; PAHs were quantified using EPA 610 B standards. Identification reached Schymanski level 1 based on RT, exact mass, isotopic pattern, and fragment ions (LC) or ≥2 daughter ions (GC). Internal standards assessed repeatability; S/N≥3; LC mass accuracy 3 ppm, GC mass tolerance 0.5 Da. Statistical analyses: Peak areas were used. For differential analysis, Wilcoxon rank-sum (nonparametric due to small n) with thresholds P≤0.05 and fold change ≥2 or ≤-2. Chemical similarity enrichment used ChemRICH with PubChem identifiers; significant clusters at P≤0.05 and fold change thresholds. To relate soil properties to pollutant abundance, Spearman correlations were computed for three chemical clusters spanning hydrophobicity: aniline compounds (median XlogP<2), pyrethrins (~4), and PAHs (benzopyrenes+fluorenes, >6). Correlation coefficients were reported when P<0.05.

• Scope: Of 2406 micropollutants assayed, 700 were identified across matrices (sedimentation pond water and sediment; infiltration pond soils at 0–10, -80, -165 cm). Categories most represented: pesticides, drugs, PAHs, toxic industrial chemicals.
• Sedimentation pond performance: Chemical enrichment showed all compound families were increased in sediment versus water, indicating broad retention irrespective of hydrophobicity. Comparing infiltration pond (all depths) to sedimentation pond (water+sediment), nearly all clusters were decreased, evidencing limited transfer to infiltration pond. Mean trapping in sedimentation pond was 79% of total micropollutant abundance, reaching 91–98% for PAHs.
• Exclusive identifications and counts: Sediment contained many exclusive compounds (pesticides 48, drugs 142, PAHs 4). Few compounds were specific to water (33 drugs, 11 pesticides, 1 toxic industrial chemical). In infiltration pond surface soil (0–10 cm): 184 pesticides, 48 drugs, 16 PAHs, 16 toxic industrial chemicals were identified; deeper layers had fewer.
• Vertical distribution in infiltration pond: Micropollutants detected across all soil layers were generally most abundant at 0–10 cm. Overall, 39–99% of micropollutant abundance was trapped in the surface layer; for PAHs specifically, 90–99% of abundance was in the surface layer. Average trapping at the surface was 86% of total micropollutant abundance.
• Chemical enrichment clusters consistently impacted across comparisons included aniline compounds, organothiophosphorus compounds, benzene derivatives, pyrethrins, benzopyrenes, and fluorenes; retention did not align simply with logP expectations.
• Correlations with soil properties (Spearman): Trends were consistent across anilines, pyrethrins, and PAHs, with higher absolute coefficients for lower median XlogP. • Aniline compounds: Positive correlations with depth, coarse elements, sand, density, and organic matter (all r≈0.88, P<0.01). Negative with water content and pH (r≈-0.88, P<0.01). • PAHs: Sand proportion r=0.64 (P<0.001), clay r=0.46 (P<0.05); silt r=-0.73 (P<0.001). • Particle size: Strong positive correlations with large particles d99 (>100 µm): r≈0.87, 0.85, 0.79 (P<0.05 to <0.001 across families). Smaller particles showed negative correlations for pyrethrins and PAHs (d10 ~ -0.44, -0.46; P<0.05). Coarse elements also positively correlated (up to r=0.88). • Water content (19–21%) and pH (6.7–7.1) varied little; both negatively correlated with abundance across families (e.g., anilines r≈-0.8 to -0.88). Porosity showed no significant correlation. • Strong inter-parameter correlations: Sand proportion was highly negatively correlated with pH (-0.99, P<0.001) and positively with organic matter; increasing sand would lower pH and raise organic matter.
• Biotic context: In the top 10 cm, roots constituted 10.9–35.5% of dry mass, indicating substantial plant presence potentially interacting with micropollutants; root and soil water contents were similar, facilitating exchange.
• Soil grading: Curvature coefficient 4.05–4.44 indicated not well-graded soil; largest particles (>100 µm) best retained micropollutants.

The study demonstrates that a two-stage facility—sedimentation followed by infiltration—effectively manages a complex mixture of road runoff micropollutants. The sedimentation pond is pivotal, retaining most identified compounds and the majority of their total abundance, thereby sharply limiting transfer to the infiltration pond. Within the infiltration pond, the surface layer (0–10 cm) is the main barrier, accumulating the bulk of micropollutant abundance and preventing deeper percolation. Chemical enrichment showed broad sediment retention across hydrophilicity ranges, challenging simplistic logP-based expectations for environmental distribution. Correlation analyses linked higher micropollutant abundance to greater sand content, coarse elements, larger particle sizes (>100 µm), higher density, and higher organic matter, while silt, finer particles, higher pH, higher water content, and curvature coefficient were negatively associated. These consistent trends across families spanning median XlogP from <2 to >6 suggest generalizable soil-property controls. Mechanistically, sand (silica) likely enhances adsorption via large, accessible surfaces and supports biofilm development; biofilms are known to improve micropollutant removal. Organic matter further enhances PAH retention. The strong negative correlation between sand and pH, and positive correlation between sand and organic matter, position sand as a central lever: increasing sand proportion simultaneously shifts multiple soil parameters in directions favorable to retention. Biodiversity (roots, microbial communities) likely contributes additional trapping and transformation capacity in the surface layer. Together, these insights inform a practical management model prioritizing sedimentation and optimizing infiltration media to maximize retention and minimize groundwater contamination.

This real-world study screened 2406 micropollutants and identified 700 across a road runoff treatment facility, confirming: (1) the essential role of sedimentation ponds in retaining the majority of micropollutant abundance (79% on average; 91–98% for PAHs), and (2) the critical function of the infiltration pond’s surface layer in trapping most residual pollutants (on average 86% of abundance at 0–10 cm). A holistic model links micropollutant abundance to soil physico-chemical properties across diverse chemical families (median XlogP <2 to >6). Properties positively associated with retention include higher sand content, coarse elements, larger particle size (>100 µm), density, organic matter, and clay (for higher logP compounds), whereas silt, fine particles, higher pH, higher water content, and curvature coefficient are negatively associated. Practical recommendation: increase the proportion of sand in the infiltration medium to enhance adsorption capacity, lower pH, and raise organic matter, while preserving vegetation and microbial biodiversity to leverage biofilm and phytostabilization/phytoextraction processes. Small-scale vegetated sedimentation and infiltration ponds along roads can cost-effectively manage runoff pollution at the source. Future research directions: incorporate direct assessments of microbial communities and biofilm roles; test the proposed sand-centered management strategy across diverse sites and soil types; evaluate controlled pH adjustments; and extend analyses to additional pollutant classes and long-term performance/aging effects.

• The presence and roles of small organisms and microorganisms in the soil could not be studied due to lack of technical means.
• Due to the small number of samples per condition, nonparametric statistics (Wilcoxon) were used; this and limited replication constrain inference strength.
• Water content and pH exhibited limited variability across soil layers at the site, limiting assessment of their potential effects.