Comparative performance of activated alumina and iron gravel media for phosphorus removal at the bench and prototype scales

Introduction

Eutrophication continues to impair freshwater and coastal ecosystems through harmful algal blooms, oxygen depletion, and loss of ecosystem services (Anderson et al., 2002; Robertson and Saad, 2011). Phosphorus loading frequently drives eutrophication in freshwater systems, and agricultural non-point sources remain major contributors despite long standing point source controls (Robertson and Saad, 2011). Dissolved phosphorus (DP) presents a persistent management challenge because DP is immediately bioavailable and can export chronically from legacy P soils even when particulate transport declines (Penn et al., 2014; Penn et al., 2017).

Best management practices (BMPs) for agricultural watersheds include in-field strategies and edge-of-field practices such as buffers, wetlands, and sedimentation basins. Many structural BMPs are optimized for settling and filtration and therefore preferentially remove suspended sediment and particulate phosphorus, whereas DP removal is often limited and event dependent under tile dominated hydrology (Holly et al., 2026). Long term monitoring of a constructed pond wetland system treating agricultural runoff and tile drainage reported substantial reductions in suspended sediment and particulate phosphorus but modest reductions in DP, with the weakest DP reductions during tile dominated events (Holly et al., 2026).

Phosphorus removal structures (PRS) provide an edge-of-field BMP class designed to remove DP from runoff, tile drainage, or ditch flow by directing water through phosphorus sorption materials (PSMs) contained in a replaceable media bed (Penn et al., 2017). Scaling PRS from laboratory to field remains challenging because field designs must simultaneously treat high flow events and maintain sufficient contact time, while avoiding premature hydraulic failure from media clogging (Penn, 2021; Tabatabaeefar et al., 2025). Clogging is therefore not only an operational nuisance but a core performance limiter because pore constriction and head loss reduce treated flow fraction and shorten effective lifespan, which can decouple bench removal estimates from field outcomes (Claveau-Mallet and Comeau, 2020; Tabatabaeefar et al., 2025).

Activated alumina and iron-based media represent widely used PSM families with distinct removal mechanisms and practical constraints. Activated alumina removes phosphate primarily through surface complexation on alumina hydroxyl sites (Roy et al., 2021). Field deployment of activated alumina reactors on agricultural tile outlets demonstrated structural viability across seasons in a cold temperate climate but documented limited adsorption of dissolved P fractions, emphasizing the need for design refinement and media evaluation under realistic drainage conditions (Husk et al., 2024). Iron based filtration has achieved substantial phosphate removal in tile drainage field monitoring, while still requiring routine and non-routine maintenance to sustain flow and capture performance (Erickson et al., 2017).

Hydraulic orientation influences contact efficiency, drainage reliability, head loss development, and clogging risk, and PRS guidance explicitly discusses tradeoffs between top down and bottom-up flow regimes while balancing retention time with peak flow conveyance (Penn and Bowen, 2017). Downflow PRS configurations are typically selected where sufficient hydraulic head enables gravity driven top-to-bottom drainage with straightforward inlet distribution and underdrain collection, whereas up flow configurations are often used where limited relief makes downflow impractical or where reduced susceptibility to clogging driven flow loss is desired, provided target retention time is maintained (Khambhammettu et al., 2007; Penn and Bowen, 2017; NRCS, 2025; Tabatabaeefar et al., 2025). Filtration literature indicates that up flow operation can be less susceptible to clogging driven reductions in treatment flow, whereas dissolved constituent removal depends primarily on achieving the target contact time and associated retention conditions, which remain central design criteria for either flow orientation (Khambhammettu et al., 2007; Penn and Bowen, 2017; Penn et al., 2017).

A comparative gap remains in PRS literature: few studies directly compare activated alumina and iron gravel media across bench and prototype scales while simultaneously evaluating how gravity up flow versus gravity downflow and clogging trajectories affect removal and predictability. The present study addresses this gap by comparing activated alumina and iron gravel media across bench and prototype by quantifying how gravity driven up flow and downflow configurations influence dissolved phosphorus removal and lab to field predictability.

Materials and methods

Field site and influent source

The field site was a constructed pond–wetland system treating runoff and tile drainage from an 8.5 ha mixed agricultural basin in northeastern Wisconsin, managed under a rotation of alfalfa, corn silage, and rye cover crops with no-till practices implemented after 2019 (Holly et al., 2026). The system consisted of a primary sedimentation basin, two surface-flow wetlands, and a secondary basin designed in accordance with NRCS Wisconsin Conservation Practice Standard 350 and instrumented for long-term monitoring of flow, water quality, and sediment accumulation (Holly et al., 2026). Agricultural runoff collected from the sedimentation basin was used as influent for laboratory and, prototype testing. For laboratory columns and prototype trials, runoff was spiked to 0.5 mg L^-1 dissolved phosphorus using KH₂PO₄ to represent the upper range of dissolved phosphorus concentrations observed during site monitoring (Holly et al., 2026).

Filter media and experimental overview

Conventionally used PRS media (i.e., activated alumina (AA400G 14 × 28 (0.6–1.4 mm); Delta Adsorbents, Roselle, IL, United States) and an iron/gravel blend were evaluated across laboratory, prototype, and pilot scales. The iron/gravel medium was prepared as a mixture of 8% iron aggregate (Zero-valent iron, CC-1107, Connelly-GPM, Inc. Chicago, IL, United States) and pea gravel (Mastercraft® pea gravel by MM Concrete, Eau Claire, WI, United States; nominal 6.4–9.5 mm), consistent with prior edge-of-field drainage structures that used 8% steel shavings (by media weight) with pea gravel to maintain infiltration while providing phosphorus sorption (Buseman et al., 2024). Manufacturer particle size distribution, bulk density, and typical elemental analysis for activated alumina and iron are provided in Supplementary Table S1. Testing examined the effects of media type, hydraulic retention time, scale, and flow orientation (bottom-up vs. top-down) on phosphorus removal. Retention times of 1 and 10 min were selected to represent high-flow, low-footprint systems requiring frequent replacement and longer-retention, larger-footprint systems with reduced maintenance, respectively. Iron/gravel media were assigned a longer hydraulic retention time (10 min) than activated alumina (1 min) because laboratory results and prior studies indicate that phosphate removal by metallic iron systems is strongly contact-time dependent and remains minimal under short-retention (1 min) conditions (Xue et al., 2018; Konadu-Amoah et al., 2022).

Flow-through column experiments

Laboratory flow-through columns were operated in triplicate to quantify media-specific phosphorus removal under controlled hydraulic conditions. Columns consisted of 1.8 cm-diameter PVC pipes with a total length of 60 cm and were packed with 7.6 cm of filter media (35–36 g iron/gravel or 17–18 g activated alumina), bounded by 2 cm layers of glass beads on each side and separated by glass wool to promote uniform flow distribution. Columns were operated under bottom-up flow to minimize inlet clogging. Influent runoff was spiked to 0.5 mg L^-1 P using KH₂PO₄. Hydraulic retention times of 1 and 10 min were achieved by adjusting flow rates (9 and 0.9 mL min^-1, respectively) using a peristaltic pump, calculated as:where Q is the flow rate (mL min^-1), V is the bulk volume of media in column (mL), n_e is the effective porosity (unitless, 0.5), and τ is the hydraulic retention time (min). Influent and effluent samples were collected every 24 h for activated alumina and every 40 h for iron/gravel.

Q = V · n e τ

Prototype-scale experiments

Prototype testing was conducted using box-type (1:10 linear scale) and column-type (1:1000 volumetric scale) systems to evaluate scale-dependent hydraulics and flow orientation effects. Scaling targeted retention time and flow regime (HRT-based design), while acknowledging that hydraulic distribution and solids exposure were not strictly dynamically similar between bench columns and manifold-fed prototypes. Box prototypes were constructed by proportionally reducing full-scale system dimensions, while column prototypes retained pilot-scale media depths with reduced cross-sectional area and flow. Prototypes were designed for a flow rate of 204 mL min^-1, corresponding to a full-scale design flow of 204 L min^-1, and were configured for 1- and 10-min retention times depending on media type.

Box prototypes were fabricated from 12.7 mm PVC sheets with top and bottom manifolds constructed from 12.7 mm PVC pipes perforated with 1.6 mm holes arranged in three rows spaced 25.4 mm apart. Column prototypes were constructed from clear PVC pipes with internal diameters of 63.5 mm (1 min) and 152.4 mm (10 min). Prototypes were operated under top-down and bottom-up flow, except for iron/gravel column prototypes, which were operated only under bottom-up flow due to reduced performance measured in initial top down trails. Sump pumps were used to intermittently agitate sediment to simulate elevated turbidity during runoff events. Media layering and use of pea gravel varied by configuration and are detailed in Supplementary Figures S1, S2; Supplementary Table S2. Prototypes received agricultural runoff spiked with KH₂PO₄ (concentration adjusted to achieve 0.5 mg L^-1 P) and influent and effluent samples were collected every 24 h until reaching a cumulative influent loading of 65 mg P kg^-1 media or a 20% removal endpoint (instantaneous percent removal). For each prototype configuration, calendar duration, cumulative treated volume, and estimated pore volumes processed are reported in Supplementary Table S3. Influent and effluent pH were spot-checked during prototype operation and are reported in Supplementary Table S4.

Statistical analysis

Laboratory column experiments were conducted in triplicate for each media/retention-time condition, and results are reported as mean ± standard deviation (SD) (n = 3). A two-sided Welch’s t-test was used to compare cumulative phosphorus removed between activated alumina (1 min) and iron/gravel (10 min) because Welch’s test does not assume equal variances and is appropriate for small sample sizes. The analysis assumes independent replicate columns and approximately normal replicate-to-replicate variation. Uncertainty around replicate means is additionally summarized using 95% confidence intervals (CI) computed from the t distribution (df = 2) as mean ± t₀.₉₇₅,₂ · SD/√n. Inferential statistics were applied only to replicated laboratory column comparisons (n = 3); prototype configuration results (n = 1) are reported descriptively as observed ranges because prototype trials were not replicated.

Results and discussion

Laboratory flow-through column performance

Activated alumina achieved substantially greater P removal than iron/gravel media under the tested conditions, with cumulative removal of 3,756 ± 75 mg P kg^-1 at 1-min retention versus 33.1 ± 2.2 mg P kg^-1 at 10-min retention for iron/gravel (Figure 1A); corresponding 95% CIs were 3,570 to 3,942 mg P kg^-1 and 27.6–38.6 mg P kg^-1, respectively. A two-sided Welch test confirmed the difference (p = 1.35 × 10⁻⁴),with low replicate variability (CV: 2.0% and 6.6%), supporting consistent performance for triplicate columns. The iron/gravel mixture was less effective at a 1-min retention time and was therefore not advanced to prototype testing, while activated alumina was not evaluated at 10 min because its intended application targets short-RT, compact systems rather than long-contact designs. Breakthrough behavior reinforced the separation in longevity: iron/gravel declined to 32% removal by 68 mg P kg^-1, whereas activated alumina maintained >89% through 470 mg P kg^-1 before declining to 18% by 7,310 mg P kg^-1 under continued loading (Figure 1B). Short contact time supports activated alumina in higher-flow designs and a smaller footprint.

Wen et al. (2020) evaluated iron-filings media in unsaturated downflow columns (10-cm ID PVC; three 30-cm sections) fed with distilled water spiked to total phoshorus (TP) = 0.3–0.7 mg L^-1 at 8 mL min^-1; this corresponds to a nominal empty-bed contact time of ∼295 min per 30-cm section EBCT = π D 2 4 L Q, which is longer than the 10-min retention time used here. Under these conditions, Wen et al. (2020) reported overall TP removals of 54%–83% for IFGEM-1 (3.8% iron filings + sand, v/v) and 46%–62% for IFGEM-2 (5% iron filings + sand/tire crumb/clay, v/v), indicating that iron-filings media can sustain 50%–80% TP removal under controlled, low-rate column operation. Differences between studies align with longer contact time and lower influent pH in Wen et al. (2020) relative to the higher pH of agricultural runoff used in the current study. Supplementary Table S4 in Mao and Yue (2016) reports inlet pH = 6.76 to 7.10 across the three influent levels. Prototype spot measurements confirmed influent pH in the circumneutral-to-alkaline range (8.2–8.6; Supplementary Table S4), which is consistent with slower iron-based phosphate removal kinetics reported at higher pH; however, pH was not measured at every sampling interval and is therefore treated as a contributing condition rather than a quantified driver in this study. Surface-complexation kinetic modeling for hydrous ferric oxides shows adsorption rate constants decreased from 0.205 to 0.0386 mM^-1 min^-1 (81% lower) as pH increased from 6 to 8, implying 4-to-5-fold slower kinetics at higher pH (Mao and Yue, 2016). Therefore, lower influent pH should enhance iron/gravel P removal, making this media best suited to runoff from acid-prone, low-buffering coarse-textured (sandy) soils.

Activated alumina AA400G had high initial measured removal (95% at 0% and 69% at 1,000 mg P kg^-1) but declined to 14% at 10,000 mg P kg^-1, with 3,835 mg P kg^-1 removed by the 20% endpoint (42% cumulative) and was comparable to other activated alumina flow through column trials (Table 1). Activated-alumina datasets used in the P-TRAP software (Penn et al., 2021) span wide condition-dependent ranges across particle size (3 mm F-200 vs. 0.6–1.4 mm AA400G/AAFS50), retention time (0.28–2 min), and influent P (0.2–0.5 mg L^-1) (Table 1). Performance differences reflect media formulation and test conditions, with longer retention time, iron coating, particle size, and influent P concentration shifting removal trajectories and cumulative capacity. AAFS50/AA400G is the preferred choice for short-RT systems when sustained removal and higher cumulative capacity are priorities, provided runoff has low turbidity and suspended solids. F-200 (3 mm) supports higher permeability and lower headloss than finer alumina grades, but achieving comparable treatment typically requires longer retention time or a larger footprint.

TABLE 1. Phosphorus removal from AA400G and previously tested activated alumina (Penn et al., 2021).

Prototype-scale effects of hydraulic orientation and scale

Activated-alumina prototype results indicate that hydraulic non-idealities and solids handling, not sorption capacity alone, controlled performance (Figure 2A). Prototype underperformance reflects hydraulic non-idealities (maldistribution, bypass, headloss) that shorten effective contact time unless inlet distribution and solids control scale with the media. Prototypes exhausted more rapidly (20% removal at 493 to 1,190 mg P kg^-1) than lab-scale columns because bypass and headloss limited effective contact. In contrast, glass wool and glass beads in the lab columns likely acted as a prefiltration and flow-distribution layer, reducing surface deposition on the reactive media and increasing treatment resilience. Therefore, for small particle size reactive media (0.6–1.4 mm), treatment potential is limited by clogging rather than sorption capacity, with prototype-scale exhaustion occurring at 493 to 1,190 mg P kg^-1, representing a 3 to 8 fold reduction relative to the 3,757 mg P kg^-1 achieved in lab-scale columns. Phosphorus reactive systems designed for backwashing or coupled with settling basins and flocculation could increase treatment resilience of media.

Orientation effects were clearest in the box prototypes: the bottom-up box remained at 31.5% removal at 2,230 mg P kg^-1, whereas the top-down box declined to 13.1% by 1,858 mg P kg^-1 (and was already 16.2% by 1,397 mg P kg^-1), indicating earlier loss of effective contact under top-down operation. Column prototypes showed a smaller and less consistent separation: the bottom-up column decreased from 78% to 20% by 1,192 mg P kg^-1, while the top-down column decreased from 81% to 9% by 1,776 mg P kg^-1. Box prototypes exhibited a stronger top vs. bottom flow contrast due to their larger cross-section and manifold-driven inflow increased sensitivity to inlet-zone solids accumulation and flow maldistribution. Under top-down flow, suspended solids accumulated at the upper media surface, rapidly reducing effective contact and promoting bypass across the box footprint, whereas bottom-up flow dispersed solids over a larger internal volume, delaying surface sealing and sustaining higher effective contact. For activated alumina and other small-particle-size reactive media, box-scale reactive structures should prioritize bottom-up flow with uniform inflow distribution.

Relative to the lab column, the bottom-up prototype column performed as well as or better, achieving approximately 1.2 to 1.5 fold higher P removal across the 20–80 mg P kg^-1 loading range (50%–70% removal versus 30%–40%). Improved performance is attributed to the larger column diameter, which reduced wall effects and promoted more uniform flow. Although iron/gravel showed greater persistence and a clearer orientation effect in column tests, the bottom-up prototype box exhibited lower and more variable removal, indicating that scale-up introduced hydraulic non-uniformity and preferential flow that limited effective media utilization. Activated alumina performed better in the bottom-up prototype box because its particle size supported more uniform flow and sustained contact at box scale, whereas iron/gravel performance remained constrained by reaction kinetics and hydraulic limitations, underscoring the need for longer retention times and improved inlet distribution in iron-based designs. Field studies demonstrate that iron-based media can achieve high phosphorus removal when deployed in blind inlets treating surface runoff; for example, Buseman et al. (2024) reported 83% dissolved reactive P and 62% total P removal using steel shavings mixed with pea gravel in a farmed pothole. However, those results were based on only five paired samples, with limited flow monitoring, no reported retention time, and clogging after two inundation events, constraining interpretation of long-term performance. Headloss evolution and solids profiles were not measured; therefore, clogging and bypass are presented as supported interpretations based on observed performance trajectories and configuration differences, and direct hydraulic verification is identified as a priority for future work.

Cost analysis of filter media

A media-only cost analysis was conducted using the best-performing prototype configurations: activated alumina (AA400G) in the bottom-up prototype box and iron/gravel in the bottom-up prototype column. This media-only analysis is a screening comparison intended to illustrate sensitivity to performance endpoints; it excludes capital, hydraulics, labor, transport, maintenance, and replacement logistics. Activated alumina achieved approximately 1,170 mg P kg^-1 media removed, and iron/gravel achieved approximately 98 mg P kg^-1 mixed media removed, with cumulative removal for both media calculated as the area under the prototype removal curves to the 20% instantaneous removal endpoint. Supplementary Table S5 reports cumulative removal and media-only cost sensitivity at 10%, 20%, and 30% instantaneous removal thresholds.

Media prices were $4.21 kg^-1 for activated alumina and $0.89 kg^-1 for iron ($0.40 lb^-1); iron cost was applied only to the 8% iron fraction of the gravel mixture, resulting in an effective iron cost of $0.071 kg^-1 mixed media. On this basis, the estimated media-only cost was $3,600 kg^-1 P removed for activated alumina and $730 kg^-1 P removed for iron/gravel Lower media-only cost can be achieved with iron-based mixtures when evaluated to a conservative removal cutoff; however, performance depends strongly on adequate contact time and sustained hydraulic functionality. Activated alumina is better suited for compact designs that incorporate backwashing or regeneration, whereas iron–gravel media are most appropriate for systems tolerating lower performance with larger footprints where media replacement is feasible. Completed media-only estimates are scenario- and site-specific (influent chemistry, solids loading, hydraulics, and replacement logistics) and should not be interpreted as universally transferable costs across PRS installations.

Conclusion

Laboratory column testing demonstrated substantially higher phosphorus removal capacity for activated alumina than for iron/gravel media, but prototype-scale performance was strongly constrained by hydraulic non-idealities and clogging. Prototype systems reduced activated alumina treatment capacity by 3 to8-fold relative to laboratory columns, whereas iron/gravel maintained more persistent removal under bottom-up column operation when sufficient contact time was achieved. Media-only cost analysis indicated lower cost per unit phosphorus removed for iron/gravel when evaluated to a conservative 20% cutoff, while activated alumina provided markedly higher removal capacity per unit mass and favored compact designs. Design selection should balance footprint and hydraulic resilience; the observed patterns suggest that pretreatment or hydraulic maintenance (e.g., backwashing/regeneration) may improve activated alumina performance and that iron/gravel may suit larger footprints where media replacement is feasible, but these are design hypotheses requiring dedicated testing. Several limitations apply, including short prototype run durations, simplified solids loading, and exclusion of construction, transport, and replacement costs from the economic comparison. Future studies should prioritize long-term field deployments that quantify treated flow, retention time, and hydraulic evolution while integrating pretreatment or regeneration strategies to improve prototype-scale resilience.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fenve.2026.1866629/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Dominika Dąbrowska, University of Silesia in Katowice, Poland

Reviewed by: Mahesh Balasaheb Chougule, DKTE Society’s Textile and Engineering Institute, India

Haribansha Timalsina, University of Illinois at Urbana-Champaign, United States