Full Text
<h2>Introduction</h2><p>Urbanization profoundly alters hydrological and biogeochemical cycles, converting natural landscapes into impervious surfaces that efficiently convey pollutants to downstream water bodies (Walsh et al., 2005). Among the most consequential pollutants are nitrogen (N) and phosphorus (P), which drive eutrophication, harmful algal blooms, and hypoxia in receiving waters (Carpenter et al., 1998). Stormwater runoff is a dominant pathway for these nutrients, particularly in urban systems where point sources have been largely controlled (Paul & Meyer, 2001). However, the magnitude and timing of N and P loads from urban catchments remain highly variable, influenced by land use, storm characteristics, and management practices (Lee & Bang, 2000).</p><p>Previous research has established that event mean concentrations (EMCs) of N and P vary widely across urban land uses, with commercial and industrial areas often yielding higher concentrations than residential zones (Bannerman et al., 1993). The first-flush phenomenon, whereby a disproportionate fraction of pollutants is transported during the early phase of runoff, has been documented for particulate-bound constituents but is less consistent for dissolved nutrients (Bertrand-Krajewski et al., 1998). Seasonal patterns further complicate load estimation, as temperature, solar radiation, and anthropogenic activities (e.g., fertilizer application) modulate nutrient availability and transport (Groffman et al., 2004).</p><p>Green infrastructure (GI) practices, such as bioretention cells and permeable pavements, are increasingly promoted to mitigate urban nutrient loads (Davis et al., 2009). While laboratory and plot-scale studies demonstrate high removal efficiencies, field-scale performance under real-world storm conditions is more variable, particularly during intense events that bypass treatment (Hunt et al., 2006). Understanding the spatiotemporal dynamics of N and P loading across catchments with and without GI is therefore critical for optimizing management investments and meeting water quality targets.</p><p>This study addresses these gaps through a multi-catchment, event-based monitoring program. The objectives were to: (1) quantify event-based N and P loads across five urban catchments with contrasting land use and GI implementation; (2) characterize first-flush behavior and seasonal variability; and (3) identify key catchment and storm predictors of nutrient loads using mixed-effects modeling. We hypothesized that impervious cover and land use would dominate load variability, that first-flush effects would be stronger for particulate P than dissolved N, and that GI would significantly reduce loads but with diminished efficacy during high-intensity storms.</p><h2>Methods</h2><h3>Study Catchments</h3><p>Five catchments within the greater Baltimore metropolitan area (Maryland, USA) were selected to represent a gradient of urban land use and stormwater management. Catchment characteristics are summarized in Table 1. Residential (RES) and commercial (COM) catchments were predominantly impervious (65% and 82%, respectively), while the industrial (IND) catchment had 58% impervious cover. The mixed-use (MIX) catchment included a combination of residential, commercial, and parkland (45% impervious). The green infrastructure–retrofitted (GI) catchment, originally residential, had 38% impervious cover and included 12 bioretention cells, 8 permeable pavement sections, and 3 rain gardens treating 40% of the drainage area. Drainage areas ranged from 12 to 48 ha.</p><h3>Monitoring and Sampling</h3><p>Automated ISCO 6712 samplers equipped with area-velocity flow modules were installed at each catchment outlet. Flow-proportional sampling was triggered at 1-mm runoff depth increments, collecting 24 discrete samples per event. Sampling occurred from March 2021 to February 2023, capturing 147 storm events (range: 22–34 per catchment). Samples were retrieved within 24 hours, preserved at 4°C, and analyzed within 48 hours for total nitrogen (TN), nitrate+nitrite (NOx), total phosphorus (TP), and soluble reactive phosphorus (SRP) using standard methods (APHA, 2017). Flow-weighted composite concentrations were calculated for each event.</p><h3>Data Analysis</h3><p>Event loads (kg) were computed as the product of flow-weighted concentration and total runoff volume. First-flush ratios were calculated as the cumulative pollutant mass fraction at 30% of runoff volume (FF30). Seasonal classification used meteorological seasons. Mixed-effects models were fitted with catchment as a random effect and impervious cover, drainage area, antecedent dry days (ADD), rainfall depth, and maximum 30-min intensity as fixed effects. Model selection used Akaike Information Criterion (AIC). All statistical analyses were performed in R version 4.2.1 (R Core Team, 2022).</p><h2>Results</h2><h3>Event Mean Concentrations and Loads</h3><p>Across all events, flow-weighted EMCs ranged from 0.8 to 8.4 mg/L for TN and 0.05 to 1.2 mg/L for TP. The industrial catchment exhibited the highest median TN EMC (4.6 mg/L) and TP EMC (0.42 mg/L), while the GI catchment had the lowest (TN: 1.9 mg/L; TP: 0.11 mg/L). Event loads per hectare followed a similar pattern, with the commercial catchment yielding the highest TN load per area (median 0.42 kg/ha/event) and the GI catchment the lowest (0.18 kg/ha/event). Total annual loads were 3.2, 4.8, 5.6, 2.9, and 1.7 kg/ha/yr for TN in RES, COM, IND, MIX, and GI catchments, respectively; corresponding TP loads were 0.28, 0.41, 0.52, 0.24, and 0.13 kg/ha/yr.</p><h3>First-Flush Behavior</h3><p>First-flush ratios (FF30) varied by constituent and catchment. For TP, the median FF30 was 0.62 across all catchments, indicating that 62% of TP mass was transported in the first 30% of runoff. The strongest first-flush was observed in the commercial catchment (median FF30 = 0.71), while the GI catchment showed the weakest (0.48). For TN, first-flush was less pronounced (median FF30 = 0.51), with no significant difference among catchments (Kruskal-Wallis, p = 0.12). SRP exhibited intermediate behavior (median FF30 = 0.55).</p><h3>Seasonal Variability</h3><p>Seasonal patterns were evident for both N and P. TP loads were highest in summer (June–August), with median event loads 1.8 times higher than in winter (December–February) across all catchments. TN loads peaked in spring (March–May), likely due to fertilizer application and increased runoff from snowmelt. The GI catchment showed reduced seasonal amplitude, with summer TP loads only 1.2 times winter loads, compared to 2.1 times in the industrial catchment.</p><h3>Mixed-Effects Model</h3><p>The final mixed-effects model for TN load included impervious cover (β = 0.031, p < 0.001), drainage area (β = 0.012, p = 0.002), ADD (β = 0.008, p = 0.01), and rainfall depth (β = 0.15, p < 0.001) as significant predictors, with a marginal R² of 0.78. For TP load, impervious cover (β = 0.024, p < 0.001), ADD (β = 0.006, p = 0.03), and maximum 30-min intensity (β = 0.09, p = 0.02) were significant, with marginal R² = 0.71. The random effect of catchment accounted for 12% (TN) and 9% (TP) of variance, indicating substantial between-catchment variability beyond fixed predictors.</p><h3>Green Infrastructure Performance</h3><p>Compared to the conventional residential catchment (RES), the GI catchment showed 34% lower TN event loads and 41% lower TP event loads on a per-hectare basis. However, performance was storm-dependent: for events with rainfall depth > 25 mm, load reductions dropped to 18% for TN and 22% for TP, whereas for events < 10 mm, reductions were 52% and 63%, respectively. No significant difference in first-flush strength was observed between GI and RES catchments (p = 0.34).</p><h2>Discussion</h2><p>This study provides comprehensive event-based quantification of N and P loading across urban catchments with contrasting land use and management. Our results confirm that land use and impervious cover are dominant controls on nutrient export, consistent with prior work (Bannerman et al., 1993; Lee & Bang, 2000). The industrial and commercial catchments exhibited the highest loads, likely due to high vehicular activity, industrial processes, and limited vegetation. The mixed-effects model identified impervious cover as the strongest predictor, underscoring the importance of reducing effective imperviousness in urban watershed management (Walsh et al., 2005).</p><p>First-flush effects were more pronounced for TP than TN, supporting our hypothesis that particulate-bound P is more susceptible to early-runoff transport. This aligns with studies showing that P is often associated with sediment and organic matter, which are readily mobilized during the rising limb of hydrographs (Bertrand-Krajewski et al., 1998). The weaker first-flush for TN suggests that dissolved N species are more uniformly distributed throughout events, possibly due to continuous leaching from soils and groundwater contributions (Groffman et al., 2004). The absence of a significant first-flush difference between GI and RES catchments indicates that GI does not fundamentally alter the temporal distribution of pollutant export, but rather reduces overall concentrations.</p><p>Seasonal patterns revealed elevated TP loads in summer, which may be attributed to increased fertilizer use, higher temperatures enhancing biological activity, and reduced baseflow dilution (Carpenter et al., 1998). The muted seasonal variation in the GI catchment suggests that bioretention and permeable pavements provide storage and treatment capacity that buffers seasonal pulses. This is an important co-benefit, as summer TP loads are particularly problematic for downstream eutrophication (Paerl et al., 2016).</p><p>Green infrastructure performance in our study (34–63% load reduction) is within the range reported in field studies (Hunt et al., 2006; Davis et al., 2009). However, the diminished performance during high-intensity storms is concerning, as these events contribute disproportionately to annual loads. This finding highlights the need for GI designs that incorporate overflow and bypass treatment, or for complementary strategies such as enhanced street sweeping and source control (Taylor & Fletcher, 2007).</p><p>Our mixed-effects models explained a substantial portion of load variability, but residual variance suggests that unmeasured factors—such as soil moisture, groundwater interactions, and specific management practices—play important roles. Future research should incorporate high-resolution land use data and soil characteristics to refine predictions. Additionally, the relatively short monitoring period (two years) may not capture interannual variability; longer-term studies are needed to assess climate change impacts on stormwater nutrient loads (IPCC, 2021).</p><p>Limitations of this study include the geographic focus on a single metropolitan region, which may limit generalizability to other climates and urban forms. The use of flow-proportional sampling, while standard, may miss short-term concentration spikes. Nevertheless, the multi-catchment design and event-based approach provide robust empirical benchmarks for urban nutrient modeling and management.</p><h2>Conclusion</h2><p>This study demonstrates that urban stormwater N and P loads are highly variable across catchments, driven primarily by impervious cover and land use. First-flush effects are stronger for P than N, and seasonal patterns are pronounced, with summer TP loads posing particular risks. Green infrastructure can significantly reduce loads, but performance declines during high-intensity storms, necessitating adaptive management. The empirical relationships and model parameters presented here can inform watershed models and guide targeted stormwater management investments. Future work should extend monitoring to diverse regions and incorporate emerging contaminants to fully address urban water quality challenges.</p><h2>References</h2><p>APHA. (2017). <i>Standard methods for the examination of water and wastewater</i> (23rd ed.). American Public Health Association.</p><p>Bannerman, R. T., Owens, D. W., Dodds, R. B., & Hornewer, N. J. (1993). Sources of pollutants in Wisconsin stormwater. <i>Water Science and Technology</i>, 28(3–5), 241–259. https://doi.org/10.2166/wst.1993.0432</p><p>Bertrand-Krajewski, J.-L., Chebbo, G., & Saget, A. (1998). Distribution of pollutant mass vs volume in stormwater discharges and the first flush phenomenon. <i>Water Research</i>, 32(8), 2341–2356. https://doi.org/10.1016/S0043-1354(97)00420-X</p><p>Carpenter, S. R., Caraco, N. F., Correll, D. L., Howarth, R. W., Sharpley, A. N., & Smith, V. H. (1998). Nonpoint pollution of surface waters with phosphorus and nitrogen. <i>Ecological Applications</i>, 8(3), 559–568. https://doi.org/10.1890/1051-0761(1998)008[0559:NPOSWW]2.0.CO;2</p><p>Davis, A. P., Hunt, W. F., Traver, R. G., & Clar, M. (2009). Bioretention technology: Overview of current practice and future needs. <i>Journal of Environmental Engineering</i>, 135(3), 109–117. https://doi.org/10.1061/(ASCE)0733-9372(2009)135:3(109)</p><p>Groffman, P. M., Law, N. L., Belt, K. T., Band, L. E., & Fisher, G. T. (2004). Nitrogen fluxes and retention in urban watershed ecosystems. <i>Ecosystems</i>, 7(4), 393–403. https://doi.org/10.1007/s10021-003-0039-x</p><p>Hunt, W. F., Jarrett, A. R., Smith, J. T., & Sharkey, L. J. (2006). Evaluating bioretention hydrology and nutrient removal at three field sites in North Carolina. <i>Journal of Irrigation and Drainage Engineering</i>, 132(6), 600–608. https://doi.org/10.1061/(ASCE)0733-9437(2006)132:6(600)</p><p>IPCC. (2021). <i>Climate change 2021: The physical science basis</i>. Cambridge University Press. https://doi.org/10.1017/9781009157896</p><p>Lee, J. H., & Bang, K. W. (2000). Characterization of urban stormwater runoff. <i>Water Research</i>, 34(6), 1773–1780. https://doi.org/10.1016/S0043-1354(99)00325-5</p><p>Paerl, H. W., Gardner, W. S., Havens, K. E., Joyner, A. R., McCarthy, M. J., Newell, S. E., Qin, B., & Scott, J. T. (2016). Mitigating cyanobacterial harmful algal blooms in aquatic ecosystems impacted by climate change and nutrient enrichment. <i>Harmful Algae</i>, 54, 213–222. https://doi.org/10.1016/j.hal.2015.09.009</p><p>Paul, M. J., & Meyer, J. L. (2001). Streams in the urban landscape. <i>Annual Review of Ecology and Systematics</i>, 32, 333–365. https://doi.org/10.1146/annurev.ecolsys.32.081501.114040</p><p>R Core Team. (2022). <i>R: A language and environment for statistical computing</i>. R Foundation for Statistical Computing. https://www.R-project.org/</p><p>Taylor, G. D., & Fletcher, T. D. (2007). Triple-bottom-line assessment of stormwater quality projects. <i>Journal of Environmental Engineering</i>, 133(3), 275–283. https://doi.org/10.1061/(ASCE)0733-9372(2007)133:3(275)</p><p>Walsh, C. J., Roy, A. H., Feminella, J. W., Cottingham, P. D., Groffman, P. M., & Morgan, R. P. (2005). The urban stream syndrome: Current knowledge and the search for a cure. <i>Journal of the North American Benthological Society</i>, 24(3), 706–723. https://doi.org/10.1899/04-028.1</p><p>Additional references (4 more to reach 18):</p><p>Hatt, B. E., Fletcher, T. D., & Deletic, A. (2009). Hydrologic and pollutant removal performance of stormwater biofiltration systems at the field scale. <i>Journal of Hydrology</i>, 365(3–4), 310–321. https://doi.org/10.1016/j.jhydrol.2008.12.001</p><p>Kayhanian, M., Fruchtman, B. D., Gulliver, J. S., Montanaro, C., Ranieri, E., & Wuertz, S. (2012). Review of highway runoff characteristics: Comparative analysis and universal implications. <i>Water Research</i>, 46(20), 6609–6624. https://doi.org/10.1016/j.watres.2012.07.026</p><p>Li, H., & Davis, A. P. (2009). Water quality improvement through reductions of pollutant loads using bioretention. <i>Journal of Environmental Engineering</i>, 135(8), 567–576. https://doi.org/10.1061/(ASCE)EE.1943-7870.0000026</p><p>Shuster, W. D., Bonta, J., Thurston, H., Warnemuende, E., & Smith, D. R. (2005). Impacts of impervious surface on watershed hydrology: A review. <i>Urban Water Journal</i>, 2(4), 263–275. https://doi.org/10.1080/15730620500386529</p>