IntroductionInfiltration trenches are stormwater control measures (SCMs) consisting of a trench filled with a highly permeable media, typically gravel or coarse stone aggregate, wrapped with a geotextile liner. They are used in urban and ultraurban areas to reduce runoff volume by exfiltrating collected runoff into the surrounding subsoil. Although infiltration trenches can operate for multiple years with little or no observed reduction in exfiltration capacity (Warnaars et al. 1999; Toran and Jedrzejczyk 2017), they have also been shown to experience significant reductions in their ability to exfiltrate water from the trench within the first year of operation (Emerson et al. 2010; Brown and Borst 2015). In addition, infiltration trenches that operated with little reduction in performance for the first 3 years of operation (Warnaars et al. 1999) showed a more pronounced reduction in exfiltration ability when reexamined after 15 years of operation (Bergman et al. 2011). Reduction in exfiltration ability has been attributed to suspended solids present in the stormwater runoff entering the trench and forming a depositional layer at the filter media–soil interface, thus decreasing the hydraulic conductivity at the interface (Reddi et al. 2000; Dechesne et al. 2005; Siriwardene et al. 2007; Emerson et al. 2010). A major cause of clogging at the filter media–soil interface has been attributed to finer-grained (<6 μm) particles present in the stormwater runoff (Siriwardene et al. 2007) as well as fine-grained particles washed off the trench aggregate itself (Brown and Borst 2015). The geotextile typically used for separation and filter purposes around the trench has been shown to be subject to blinding (i.e., filter cake buildup) on the upstream side, internal clogging by finer particles, and downstream chemical clogging like calcite crust formation (Veylon et al. 2016), all processes that could negatively affect stormwater exfiltration from the trench.If sediments present in stormwater runoff are responsible for decreases in exfiltration rates, as was indicated by the previous studies, the rate at which solids enter the infiltration trench should influence how long an infiltration trench can operate as designed. The ratio of impervious drainage area to infiltration trench bottom area ranged from 40:1 to 160:1 for infiltration trenches that showed a significant reduction in performance within the first year of operation (Emerson et al. 2010; Brown and Borst 2015), whereas the ratio of impervious drainage area to infiltration trench bottom area was lower (23:1) for infiltration trenches that only experienced slight clogging after 3 years of operation (Warnaars et al. 1999). This indicates that the stormwater loading rate may have influenced the observed differences in length of time until significant loss of performance was observed. Stormwater runoff from roads and parking lots also tends to have higher total suspended solids (TSS) concentrations than roof runoff (Charters et al. 2021). The two systems that experienced decreased exfiltration rates after 1 year of operation received runoff from either a parking deck (Emerson et al. 2010) or roadway (Brown and Burst 2015), whereas the source of stormwater for the system that only had minimal performance loss after 3 years was from roof runoff (Warnaars et al. 1999; Bergman et al. 2011). Therefore, the concentrations of solids in the runoff source also may have played a role in length of operation. Whereas stormwater loading rate and source may play a role in how long infiltration trenches can operate, all three of the previous systems eventually showed a decrease in performance over time (Emerson et al. 2010; Bergman et al. 2011; Brown and Borst 2015), indicating the need to better understand the specific failure mechanisms so that future infiltration trenches can be designed for long-term performance.The research presented herein is a follow-up study of one of the previously mentioned infiltration trenches that experienced diminished exfiltration performance over its first 3 years of operation (Emerson et al. 2010). This infiltration trench did not have a pretreatment system and was purposely undersized to more rapidly understand how a SCM ages. This present study examined the exfiltration performance over time beyond the first 3 years of operation. In addition, results from the excavation of the infiltration trench after 11 years of operation and subsequent forensic analysis of the infiltration matrix, geotextile, influent piping, and sediment under the trench will be discussed. The goal of the infiltration trench forensic analysis was to determine the major causes that led to the system’s failure. The forensic analysis was split into four parts. The first part was a grain size analysis of sediments collected from multiple locations inside the trench, as well as soils from just underneath and outside the infiltration trench. The second part of the forensic analysis involved in-situ infiltration tests on the bottom of the trench. The third part examined the degree of geotextile clogging by estimating the density, pore size distribution, and porosity of the exhumed geotextile. The final part of the forensic analysis involved quantifying the organic matter content and phosphate and metals accumulation in the sediment deposited in the trench over the 11 years of operation. Determination of the source(s) of failure would lead to design improvements for future infiltration systems with the goal of improved longevity as well as decreased maintenance over the lifetime of infiltration-based SCMs.Materials and MethodsInfiltration Trench DescriptionThe infiltration trench in this study was a retrofit constructed in July 2004 at Villanova University in southeastern Pennsylvania to treat runoff from a 0.19 ha, 100% impervious watershed (i.e., half of the top deck of a parking garage). The infiltration trench was 1.8 m deep, 3.9 m long, and 3.0 m wide, lined with a nonwoven geotextile and filled with crushed stone (Fig. 1), yielding an initial effective storage capacity of 5.7 m3, which is equivalent to approximately 3.0 mm of runoff depth over the drainage area. The impervious drainage area to infiltration trench bottom area was approximately 160:1, which is greater than the current design guidance of 5:1 (PADEP 2006). The site was purposely undersized with no pretreatment to artificially accelerate longevity-related processes and exacerbate pollutant loading (Emerson et al. 2010). Water from the parking deck first traveled through the monitoring bench to trap large debris (e.g., leaves) before entering the trench via a 0.3-m-diameter horizontal inflow manifold pipe located below the geotextile cover and at the top layer of the rock aggregate. The corrugated plastic inflow manifold pipe (Advanced Drainage Systems, Hilliard, Ohio) stretched the length of the trench and had multiple rows of circular perforations (diameter ∼5–10 mm) along the bottom half of the pipe that allowed the water to drain from the pipe down through the rock aggregate and into the native soil below. A 15-cm overflow pipe was located approximately 1.6 m above the bottom of the trench (which corresponds to 0.2 m below ground surface). Pervious pavers, placed at the ground surface (i.e., 1.8 m above the trench bottom), acted as a secondary overflow device when the system reached its maximum capacity. The soil was augured to a depth of 3 m (1.2 m below the trench bottom) during construction. No groundwater was present, and no soil mottling was observed at this depth.Water Quantity MonitoringThe infiltration trench and watershed area were continuously monitored for rainfall and trench water depth from July 2004 to December 2008 at 1-min intervals. The monitoring system was placed back online for all of 2010 and again from 2014 to 2015, recording data at 5-min intervals. Rainfall was measured by an American Sigma Model 2149 tipping bucket rain gage (Hach Company, Loveland, Colorado) located in the drainage area that measured 0.25 mm of rainfall per bucket tip at 0.5% accuracy for intensities up to 13 mm/h. Rainfall events were considered once measured rainfall was greater than 1.3 mm. Based on Driscoll et al. (1989), a period of at least 6 hours between measurable rainfalls was used to distinguish a rainfall event.A pressure transducer (INW PS9800, Seametrics, Kent, Washington) measured water depth within the stone bed. Accuracy of the pressure transducer was validated quarterly, and drift was less than 1.3 mm/year. The system was considered to overflow when the trench water depth reached 1.6 m. The water depth measurements were used to quantify the recession rate (i.e., how quickly the water depth in the infiltration trench decreased). Because exfiltration is substantial through both the sides and bottom of the trench (Barraud et al. 2014; Emerson et al. 2010; Gonzalez-Merchan et al. 2012), the recession rate was used as an indicator of performance. Recession rates were only calculated after precipitation ended so that the only change in water level in the trench was exfiltration. Recession rates in the infiltration trench were determined for different depths: 0–0.3, 0.3–0.6, 0.6–0.9, and 0.9–1.2 m. For multiple peak storms, a depth increment may have several infiltration rates that are averaged. Emerson et al. (2010) showed a seasonal trend in the recession rate at all depths each year, with July tending to have the highest recession rates associated with the warmest temperatures (Braga et al. 2007). Therefore, July of each year was used for illustrative purposes to compare the evolution of recession rates over the study period, although similar year-to-year trends were observed for other months.ExcavationThe infiltration trench was removed during October and November of 2015. This removal included the overlaying soil, stone aggregate, and the geotextile on top of the infiltration trench as well as the 0.3-m-diameter horizontal inflow pipe. Care was taken to not disturb or damage the geotextile along the walls and on the bottom of the 1.8-m-deep infiltration trench. Samples of the sediment that accumulated within the infiltration trench, geotextile samples from the bottom and side of the trench, and soil samples from just outside the geotextile were collected during the excavation process for analysis.Soil Classification and Organic Matter Content AnalysisA grain size analysis (ASTM 2007) was performed on the collected sediment samples. The soil was classified using the USDA Soil Classification System, and median grain sizes were interpolated from the grain size distribution curves. Duplicate grain size distribution analyses were performed for sediment samples from inside the trench. One grain size distribution analysis was performed for soil samples collected at each location immediately outside the infiltration trench geotextile liner. The soils collected immediately outside the geotextile liner were split into three subsections with respect to their sample locations (i.e., the bottom of the trench and two different depths along the side wall). The organic matter content of oven-dried (105°C) sediment and soil samples was quantified by loss due to ignition (550°C) (APHA 1995).In-Situ Infiltration Test (Modified Philip Dunne)The modified Philip Dunne (MPD) in-situ infiltration test was performed on the bottom of the trench just after the infiltration trench was excavated in 2015. The MPD method assumes that the water plume from the infiltrometer is spherical and uses falling-head infiltration techniques (Munoz-Carpena et al. 2002). The range of hydraulic conductivities calculable with the MPD is ∼0.0036 to ∼36 cm/h. The procedure involves driving a 10.2-cm–inner diameter infiltrometer into the soil ∼5 cm and then filling the infiltrometer with water to a height of 0.6 m. Time measurements were taken when the water level dropped to 0.3 m (tmed) and when the water was completely infiltrated into the subsoil (tmax). A total of four tests were performed at different locations at the bottom of the trench. The geotextile had been removed prior to testing. Eqs. (1) and (2) were used to approximate the saturated hydraulic conductivity (Ksat; Gulliver et al. 2010). The results for each test represented the saturated hydraulic conductivity at that specific location (1) τmax=0.73(tmaxtmed)−1.1258,if tmaxtmed<5.4(2) where τmax = dimensionless max time; tmax = time required for the water to infiltrate from the tube (min); tmed = time required for half of the water column to infiltrate (min); RTube = inner radius of infiltrometer (cm); and Ksat = saturated hydraulic conductivity (cm/s).Geotextile AnalysisA geotextile sample collected from the bottom of the trench was analyzed using the capillary flow test (ASTM 2014) to determine the sample’s pore size characteristics (i.e., permeability and porosity). This location was chosen for analysis because visual observation indicated that the geotextile at the bottom of the trench, which appeared uniformly caked with sediment, had more sediment accumulation compared to the geotextile on the vertical side walls of the trench. The geotextile on the vertical side wall appeared to have more sediment accumulation near the bottom of the trench and no visual signs of sediment accumulation near the top of the trench. A geotextile sample collected from the wall of the infiltration trench 0.3 m (1 ft) below the ground surface (i.e., an area of the trench that experienced minimal exfiltration and visually had minimal sediment accumulation) was also analyzed for comparison purposes. This sample with no visual signs of sediment accumulation was washed with tap water to remove any sediment that may have accumulated over time using a low-pressure hose and was considered a clean geotextile sample for the purpose of this study. Care was taken during the cleaning process to ensure that the fibers within the geotextile were not destroyed or deformed. One 10×10-cm section of the bottom geotextile sample and one 10×10-cm section of the clean sample were tested to determine their thickness and mass/unit area according to ASTM standards D5199 (ASTM 2001) and D5261 (ASTM 2010), respectively. Each section was rinsed with water several times from both sides prior to being dried at room temperature and divided into four 5×5-cm samples. The bubble point (O98), which is the particle size corresponding to 2% passing, and pore size distribution were then measured with a Geo Pore Pro (GPP-101A, Porous Materials, Ithaca, New York) using each of the four 5×5-cm samples of the bottom and clean geotextiles.The degree of clogging (λ) is defined as the mass of soil particles in the geotextile voids (Ms) divided by the mass of geotextile fibers (Mf) [Eq. (3)] (3) Furthermore, Eq. (4) shows the relationship between the porosity of a clean geotextile (n) and its thickness (Giroud and Perfetti 1977) (4) where μ=mass/unit area of geosynthetic; ρ = density of the geosynthetic material; and t = thickness of geosynthetic.The porosity of partially clogged geotextiles (n′) can be determined using Eq. (5) (5) n′=1−(1−n)*(1+ρfρs)where n = porosity of clean geotextile; ρf = density of clean geotextile; and ρs = density of soil inside the exhumed geotextile.Phosphate and Metals ExtractionWeak acid extraction experiments were performed on soil and sediment samples from four different locations to quantify the phosphate, copper, cadmium, chromium, lead, and zinc accumulation in trench sediments and the surrounding soils. The phosphate and metals extractions were performed separately and based on Komlos and Traver (2012) and Bates (2014), respectively. The extraction procedure involved combining 3 to 4 g of each sample (wet weight) with 15 mL of 0.5 N hydrochloric acid (for phosphate extraction) or 25 mL of 0.5 M nitric acid (for metals extraction) into a 50-mL plastic centrifuge tube. All tubes were rotated at 100 rpm for 24 h. The tubes were then allowed to settle for 15 min. The supernatant was filtered using a 0.45-μm syringe filter (Fischer Scientific) and stored at 4°C until the phosphate or metals analysis was performed as described subsequently. Phosphate concentrations were measured with a Systea EasyChem Discrete Analyzer using United States Environmental Protection Agency (USEPA) Method 365.1. Cadmium, chromium, copper, and lead concentrations were measured using a Perkin Elmer model AAnalyst 800 Graphite Furnace system according to USEPA Method 7010 (USEPA 2007a) . Zinc was analyzed using a Perkin Elmer AAnalyst 800 Flame Atomic Absorption Spectrometer according to USEPA Method 7000b (USEPA 2007b). The dry weight of each sample was quantified on the same day the extraction experiment was initiated by weighing a subsample from each location before and after drying at 103°C for about 24 h (n=3 for the dry weight analysis). Statistical analysis was performed using a two-sample (unequal variance) t-test; data sets were considered significantly different if the p-values were lower than 0.05.Results and DiscussionInfiltration Trench Operation (2004–2015)The recession rates during July in the first year of operation (8.9±3.3 cm/h, n=12; Fig. 2) for the bottom portion of the trench (0–0.3 m) were close to the tested hydraulic conductivity of the surrounding soil (7.1 cm/h, Emerson et al. 2010). The average recession rate was approximately an order of magnitude lower during July of the second year of operation (0.7 cm/h± 0.3 cm/h, n=7) and remained low for the following decade with the average July recession rates between 0.2 and 0.7 cm/h for 2005–2008, 2010, and 2014). Although there was substantial variation in observed recession rates in 2004 (max=14.4 cm/h and min=3.1 cm/h), which was partially due to antecedent moisture conditions, there was relatively small variation in recession rate for 2005 onward. The order of magnitude–lower average recession rates after the first year of operation combined with relatively small variation in the recession rate for 2005 onwards indicates clogging within the infiltration trench. The average recession rate decrease of at least an order of magnitude that was observed from the bottom depth range (0–0.3 m) of the trench within the first 3 years of operation was also observed for the other depth ranges (0.3–0.6, 0.6–0.9, and 0.9–1.2 m) during this same time period (Table S1). In addition to recession rate decrease after the first year, diminished performance was also quantified by an increase in occurrence of overflow events within the first 3 years of operation (Emerson et al. 2010) as well as visual observation of the trench unable to handle the full extent of the parking deck runoff during storm events (Fig. 2 insert). Piezometer measurements in the trench during the final 9 months of operation indicated that not only did water levels continue to regularly exceed that of the overflow pipe invert (i.e., 1.6 m) but also that the trench was unable to fully drain between rain events for the final 5 months of operation (Fig. 3).Infiltration Trench ExcavationThe aged infiltration trench was taken offline and excavated 11 years after installation. Excavation revealed that the horizontal 0.3-m-diameter perforated influent manifold pipe located just below the geotextile cover, at the top layer of the rock aggregate, had sediment accumulation that filled a third of the pipe [Fig. 4(a)]. Clogging of the manifold pipe perforations, as well as reduced capacity caused by sediment accumulation in the pipe, could have contributed to the observed backup of stormwater upstream of the influent manifold pipe (Fig. 2 insert). However, stormwater completely filling the trench during the months prior to excavation (Fig. 3) and significant sediment accumulation visually observed on the rock aggregate [Figs. 4(b and c)] and the geotextile liner [Fig. 4(d)] indicates that stormwater was still able enter the trench, even in its later stages of operation.Grain Size DistributionParticle size distribution curves (Fig. S1) were used to determine the gravel (>2 mm), sand (between 0.05 and 2 mm), silt (between 0.002 and 0.05 mm), and clay (<0.002 mm) content in the sediment collected at multiple locations throughout the infiltration trench during excavation. Silt-sized particles made up a significant percentage of the sediment that accumulated in the inflow manifold pipe [Fig. 4(a)] and on the infiltration media [Fig. 4(b)] (52% and 33%, respectively), whereas the soil collected at various locations outside of the trench geotextile contained less than 15% silt (Table 1). Therefore, results indicate that silt made up a significant fraction of the sediment present in the parking deck runoff. The percentage of clay-sized particles in soil just outside the geotextile increased with depth with the soil collected from just below the geotextile at the bottom of the trench containing the highest percentage of clay-sized particles. The clay content in all samples collected in and adjacent to the infiltration trench during the 2015 forensic analysis was approximately an order of magnitude higher than the clay content of the background soil (Table 1) indicating that clay-sized particles were present in the stormwater runoff and that clay-sized particles from inside the infiltration trench did migrate through the trench geotextile on the sides and bottom of the trench. USDA classification indicates that sediment from the perforated inflow pipe [Fig. 4(a)] was a silt loam, whereas sediment that accumulated on the infiltration media [Fig. 4(b)] was a sandy loam. The soil collected from just outside the side wall of the geotextile or just below the bottom geotextile was classified as a gravelly sandy loam or gravelly sandy clay loam, respectively. The median particle size (D50) of the sediment collected from the influent manifold pipe (26.4±1.3 μm, Table 1) was lower than previously reported D50 values for stormwater runoff from parking lots (46–54 μm; Roseen et al. 2011; Selbig and Bannerman 2011). The accumulation of larger-sized particles (i.e., grit) was visually observed at locations in the stormwater conveyance system between the parking deck and the infiltration trench (i.e., upstream of the inflow manifold pipe). The removal of these larger-sized particles upstream of the inflow manifold pipe may explain the lower D50 values in the inflow manifold pipe compared to previously reported values from parking lots.Table 1. Median particle size (D50), grain size distribution, and USDA soil classification of samples from in and around the infiltration trench. Unless stated otherwise, all samples were collected during the excavation of the infiltration trench in 2015 (i.e., 11 years after installation). Values are the average of multiple replicates ± standard deviation (when n>1)Table 1. Median particle size (D50), grain size distribution, and USDA soil classification of samples from in and around the infiltration trench. Unless stated otherwise, all samples were collected during the excavation of the infiltration trench in 2015 (i.e., 11 years after installation). Values are the average of multiple replicates ± standard deviation (when n>1)Soil sampleD50 (μm)Gravel (%)Sand (%)Silt (%)Clay (%)USDA classificationInflow pipe sedimenta26.41.032.351.515.2Silt loam(n=2)(±1.3)(±0.4)(±1.1)(±0.8)(±0.7)Trench sedimenta55.72.652.133.411.9Sandy loam(n=2)(±4.4)(±0.5)(±3.8)(±4.1)(±0.2)Soil, side wall 0.3-m depthb (n=1)30725.151.913.59.6Gravelly sandy loamSoil, side wall 1.8-m depthb (n=1)24922.214.171.1243.2Gravelly sandy loamSoil, below trenchb (n=1)20216.055.410.318.3Gravelly sandy clay loamBackground soil from 2005 test pit (n=1)85128.5126.96.36.199Gravelly loamy sandSaturated Hydraulic ConductivityModified Philip Dunne infiltration tests were performed at four locations on the bottom of the infiltration trench after the rock aggregate and geotextile were removed, resulting in field saturated hydraulic conductivities (Ksat) that ranged from 0.44 to 3.7 cm/h (Table S2). Of the four tests, the lowest saturated hydraulic conductivity (0.44 cm/h) was measured nearest to where the parking deck runoff entered the trench, and the saturated hydraulic conductivities increased the farther from where the parking deck runoff entered the trench (Fig. 5). Lower saturated hydraulic conductivities closer to the stormwater inlet suggest that sediment load was not evenly distributed throughout the trench for all storm events. The field saturated hydraulic conductivity measurements from the bottom of the trench (0.44–3.7 cm/h) were only 6% to 52% of the hydraulic conductivity of the native soil (7.1 cm/h; Emerson et al. 2010), indicating permeability reduction throughout the bottom of the trench over the 11-year period. Following the guidance for subsurface drainage (surrounding soil >50% fines) from the American Association of State Highway and Transpiration Officials (AASHTO), a typical geotextile would have an apparent opening size (AOS) of 180 to 220 μm. The AOS is an indication of the approximate largest particle that would pass through the geotextile (Koerner 2012). Considering the particle size of fine-grained soils (<75 μm or #200 sieve), clay and silt migration through the geotextile and filter cake buildup at the geotextile–subsoil interface can be anticipated. The grain size distribution analysis indicated that the clay content in soil collected below the trench was an order of magnitude higher than the native soil, although the silt content was comparable (Table 1). This indicates that there was migration of a significant number of clay-sized particles (<2 μm) through the geotextile. The lack of comparable silt-sized (2–75 μm) particle migration through the geotextile indicates either (1) clogging throughout the geotextile by particles smaller than ∼200 μm that are able to enter the geotextile pores or (2) clogging of only the top surface of the geotextile by silt-sized and larger particles that would lead to blinding and an eventual reduction in permittivity.Investigation of Geotextile CloggingA geotextile sample collected from the bottom of the infiltration trench after 11 years of operation was compared to a clean geotextile sample for a variety of geotextile parameters (mass/area, thickness, bubble point, and degree of clogging) to determine the extent of geotextile clogging. Results showed that the mass per area of the geotextile sample from the bottom of the trench was greater than the mass per area of the clean geotextile (Table 2), indicating sediment accumulation in the geotextile at the bottom of the trench. The geotextile from the bottom of the trench was also thicker than the clean geotextile (Table 2). In addition, the calculated porosity of the geotextile from the bottom of the trench (n′=0.80) was less than the porosity of the clean geotextile (n=0.87), indicating that sediment accumulation in the geotextile at the trench bottom resulted in a decrease in porosity. A porosity of 0.80 is still a relatively high porosity. Another technique used to evaluate the pore-size distributions of geotextiles is the bubble point method (Bhatia and Smith 1995). Particle size distribution curves from the capillary flow analysis (Figs. S2 and S3) showed that the bubble point (O98) range (137.5–200 μm) for the geotextile from the bottom of the trench was less than the O98 range for the clean geotextile (220–232 μm), indicating that sediment accumulation decreased the pore size of the geotextile at the bottom of the trench. Another parameter obtained from the capillary flow test was the degree of clogging (λ), which could range from 0.2 to 15 (where 15 represents a severely clogged geotextile) (Palmeira and Gardoni 2000). The degree of clogging for the geotextile from the bottom of the trench (λ=2.47) was found to be at the lower end of this scale. Although the bubble point values (O98) and porosity decreased due to the accumulation of soil particles inside the geotextile, a relatively low degree of clogging and a relatively high porosity indicate that the internal structure of the geotextile at the bottom of the trench was not clogged. Therefore, internal clogging of the filter fabric was not found to be one of the major causes for the reduction in recession rates observed in Fig. 2. The other mechanism besides internal clogging that can cause overall permeability reduction related to the geotextile is sediment accumulation on top of the geotextile that can cause blinding of the pore openings at the filter media–geotextile boundary. Visual observations of sediment accumulation on top of the geotextile [Fig. 4(d)] and the lack of evidence to support internal clogging indicate that the decrease in geotextile permittivity was caused by blinding.Table 2. Physical properties from the geotextile analysis for the clean geotextile sample and the geotextile sample from the bottom (i.e., floor) of the infiltration trenchTable 2. Physical properties from the geotextile analysis for the clean geotextile sample and the geotextile sample from the bottom (i.e., floor) of the infiltration trenchGeotextileMass/unit area (g/m2)Thickness (mm)Bubble point, O98 (μm)Clean geotextile (n=4)175–1880.98–1.23220–232Geotextile from trench bottom (n=4)559–6501.08–1.62137.5–200Phosphate and Metals Accumulation in Trench Sediments and Surrounding SoilWeak acid extractions were performed to quantify phosphate and metals sorption to sediment located in the inflow pipe of the infiltration trench and in soil surrounding the trench (i.e., native soil just outside the geotextile on the sides and underneath the geotextile) after 11 years of infiltration trench operation. Soil samples collected directly outside the trench geotextile 0.9 m below the surface and underneath the bottom of the trench (i.e., 1.8 m below the surface) had average phosphate concentrations between 50 and 75 mg of PO4−3-P per kg of dry soil (Table 3). However, an average concentration of more than 1,200 mg of PO4−3-P per kg of dry soil was measured in soil samples located just outside the geotextile of the infiltration trench that were collected 0.3 m below the surface. A possible explanation for the high phosphate concentration in soil near the surface is the usage of fertilizers at the infiltration trench site. Fertilizers contain high concentrations of phosphorus (P) to stimulate early plant growth and hasten maturity (Busman et al. 1997). The phosphate concentrations measured in soil just outside the infiltration trench at depths of 0.9 and 1.8 m (Table 3) were comparable to phosphate concentrations measured in native soil collected near another SCM on Villanova University’s campus (45±27 mg/kg dry soil, n=5, Komlos and Traver 2012), indicating that phosphate concentrations just outside the infiltration trench at depths below 0.9 m were at or near background levels. The low PO4−3-P concentration in sediments from the infiltration trench inflow pipe (1.2±0.3 mg/kg dry soil, n=3) compared to the soil just outside the infiltration trench, as well as decreasing sorbed phosphate concentrations with depth in soils just outside the infiltration trench, corroborates the findings of Batroney et al. (2010) that the stormwater runoff from the parking deck contained low concentrations of phosphate.Table 3. Phosphate and metals extractions from (1) native soil samples just outside the geotextile at 0.3 and 0.9 m below ground, (2) native soil just below the geotextile at the bottom of the infiltration trench, and (3) sediment samples collected from the horizontal influent manifold pipe. Values are the average (±standard deviation) of either duplicate or triplicate soil samplesa collected from each locationTable 3. Phosphate and metals extractions from (1) native soil samples just outside the geotextile at 0.3 and 0.9 m below ground, (2) native soil just below the geotextile at the bottom of the infiltration trench, and (3) sediment samples collected from the horizontal influent manifold pipe. Values are the average (±standard deviation) of either duplicate or triplicate soil samplesa collected from each locationSample locationPO43−-P (mg/kg)Copper (mg/kg)Cadmium (mg/kg)Chromium (mg/kg)Zinc (mg/kg)Lead (mg/kg)Soil, side wall 0.3-m depth126111.80.1012.429.03.2(±27.4)(±0.4)(±0.01)(±1.9)(±0.7)(±0.3)Soil, side wall 0.9-m depth74.5188.8.131.52.23.2(±13.7)(±1.2)(±0.02)(±0.5)(±5.6)(±0.3)Soil, below trench50.920.30.474.837.35.4(±3.7)(±4.4)(±0.03)(±0.3)(±2.2)(±1.5)Inflow pipe sediments1.2132.30.233.088071.4(±0.3)(±2.8)(±0.0)(±0.0)(±258)(±1.2)Samples from the same locations used for the phosphate analysis were also analyzed for the accumulation of five metals (copper, cadmium, chromium, zinc, and lead) (Table 3). Results show that copper, lead, and zinc average concentrations were an order of magnitude higher in the inflow pipe sediments (132±2.8, 71.4±1.2, and 880±260 mg/kg, respectively; n=2) compared to the average of all soil samples collected from the side and bottom of the trench (14.6±4.9, 3.94±1.4, and 38.5±9.3 mg/kg, respectively; n=9). The relatively high average copper, lead, and zinc concentrations in the inflow pipe sediments compared to soil samples just outside the trench indicates that the runoff from the parking deck contained a high concentration of these metals compared to the native soil. Copper and zinc concentrations in the soil below the trench (20.3±4.4 and 37.3±2.2 mg/kg, respectively; n=3) were higher than (p-value <0.04) copper and zinc concentrations in soil from the side of the trench 0.3 m below the surface (11.8±0.4 and 29.0±0.7 mg/kg, respectively; n=3), indicating that some copper and zinc present in the stormwater migrated through the trench geotextile and accumulated on the soil below the trench. Parking lot runoff has been shown to be a major contributor of total zinc and copper to the urban stormwater (Steuer et al. 1997). Even though heavy metal concentrations in the runoff from a parking deck can differ widely according to their use, higher zinc and copper concentrations can be caused by drip losses and tire and brake wear due to braking and steering activity (Huber et al. 2016). It is interesting to note that elevated lead concentrations were higher (p-value <0.005) in the inflow pipe sediment (71.4±1.2 mg/kg, n=2) compared to the surrounding soil (3.94±1.4 mg/kg, n=9), even though lead was not consistently found in the runoff from this parking deck (Batroney et al. 2010), indicating that lead can accumulate in infiltration trench sediments even though lead concentrations in the stormwater runoff were not consistently above detection. Copper, zinc, and lead accumulation in infiltration trench sediments also provides evidence of an infiltration trench’s ability to remove metals from stormwater runoff. In addition, examination of the soil below the trench suggests that native soil below the trench provided additional metals removal. Although elevated average copper, zinc, and lead concentrations were observed in inflow pipe sediments (132, 880, and 71 mg/kg, respectively), the levels were still below the Pennsylvania Department of Environmental Protection clean fill concentrations limits (CFCL) of 8,100, 12,000, and 450 mg/kg for copper, zinc, and lead, respectively (25 PA Code Chapter 250 Appendix A).Organic Content of Inflow Pipe SedimentThe sediment that accumulated in the inflow pipe had an organic matter content of 290±19 g/kg of dry mass (or 29.0%±1.9%, n=3), which is above the range previously reported for sediments from other infiltration basins (66–223 g/kg of dry mass) (Durand et al. 2005;Murakami et al. 2008;Coulon et al. 2013). One explanation for the higher-than-reported organic matter content was the removal of low organic matter content sediment from the stormwater runoff prior to entering the infiltration trench. Larger-sized particles (i.e., grit) that can accumulate in stormwater conveyance systems has been shown to have a relatively low organic matter content (13 g/kg of dry mass) (Komlos et al. 2018). As mentioned previously, the accumulation of grit-like material was visually observed at locations in the stormwater conveyance system between the parking deck and the infiltration trench (e.g., upstream of the inflow weir) and may explain the concentration of smaller-sized particles (with higher organic matter content) in the inflow pipe of the infiltration trench.The source of the organic matter in the parking deck sediment is unknown. The parking deck was located on the second floor of the parking garage with no overhanging tree cover, so debris from trees and vegetation (i.e., leaves, acorns, etc.) was assumed to be negligible. However, the presence of localized atmospheric deposition of natural organic matter (i.e., pollen) is a possible source of the organic matter. The inflow pipe sediment also contained elevated levels of metals typically associated with vehicle use (copper, lead, and zinc; Table 3), which indicates that the organic matter in the stormwater runoff could, at least in part, also be due to anthropogenic sources (e.g., tire wear, oil, gasoline). Additional research is needed to determine the exact source(s) of the organic matter from the parking deck runoff.Recommended Infiltration Trench Design Modifications based on Forensic AnalysisThis infiltration trench forensic analysis documented how sediments present in stormwater runoff can negatively impact infiltration trench performance. Therefore, infiltration trenches should be designed to minimize sediment buildup and subsequent clogging over time. Suggested infiltration trench design modifications based on the forensic analysis include stormwater pretreatment to prevent sediment from entering the trench. Gravitational settling is a stormwater pretreatment technology used to remove particulates and stormwater pretreatment systems based on gravitational settling tend to target larger particles (i.e., >80 μm) (e.g., Minnesota Stormwater Manual 2021). However, the median particle size (D50) of the sediment that accumulated in the influent manifold of the trench was 26.4 μm (Table 1), and 82% of these particles were less than 80 μm (Fig. S1), suggesting that pretreatment based on gravitational settling would not have prevented the accumulation of the sediment present in the parking deck runoff that caused this infiltration trench to fail. Stormwater containing sediment with relatively small particle sizes is not unique to this study. Stormwater runoff from parking lots (Fowler et al. 2009; Selbig and Bannerman 2011), institutional roofs (Selbig and Bannerman 2011), arterial streets (Drapper et al. 2000; Selbig and Bannerman 2011), collector streets (Selbig and Bannerman 2011), and mixed-use sources (Anta et al. 2006; Selbig and Bannerman 2011) have been reported to contain sediment with median particle sizes less than 100 μm. Based on these results, it is recommended that another technology besides gravitational settling pretreatment be used to minimize sediment buildup in infiltration trench SCMs. Additional research is needed to explore the potential of other technologies such as surface filtration, which uses screens with typical pore openings of 20–35 μm (Metcalf and Eddy et al. 2003) to remove the sediments that cause clogging in infiltration trench SCMs.The 3.9-m-long, 3.0-m-wide, and 1.8-m-deep infiltration trench evaluated in this study was filled with ∼35 metric tons of rock aggregate. Adequately removing the sediment buildup within the rock aggregate [Figs. 4(b and c)] or replacing the clogged geotextile on the bottom of the trench [Fig. 4(d)] would have been difficult without use of heavy equipment (i.e., backhoe) to first remove the rock aggregate. Therefore, another suggested design modification to enhance long-term performance is to replace the stone aggregate with plastic modular stormwater systems as the infiltration trench matrix, which would allow for maintenance of the infiltration surface. A plastic modular stormwater infiltration system would be significantly lighter than a rock aggregate matrix and can be equipped with inspection ports or clean-outs to enable routine cleaning. Inspection ports or clean-outs are also recommended for the distribution pipes in areas where hydrodynamic changes may result in settling or filtering of stormwater sediment, such as those observed in the influent manifold pipe of the infiltration trench [Fig. 4(a)].References Anta, J., E. Peña, J. Suárez, and J. Cagiao. 2006. “A BMP selection process based on the granulometry of runoff solids in a separate urban catchment.” Water SA 32 (3): 419–428. https://doi.org/10.4314/wsa.v32i3.5268. APHA (American Public Health Association. 1995. Standard methods for the examination of water and wastewater, edited A. D. Eaton, L. S. Clesceri, and A. E. Greenberg. Washington, DC: APHA. ASTM. 2001. Standard test method for measuring the nominal thickness of geosynthetics. D5199-01. West Conshohocken, PA: ASTM. ASTM. 2007. Standard test methods for particle size analysis of soils. D422-63. West Conshohocken, PA: ASTM. ASTM. 2010. Standard test method for measuring mass per unit area of geotextiles. D5261-10. West Conshohocken, PA: ASTM. ASTM. 2014. Standard test method for pore size characteristics of geotextiles by capillary flow test. D6767-16. West Conshohocken, PA: ASTM. Barraud, S., C. Gonzalez-Merchan, N. Nascimento, P. Moura, and A. Silva. 2014. “A method for evaluating the evolution of clogging: Application to the Pampulha Campus infiltration system (Brazil).” Water Sci. Technol. 69 (6): 1241–1248. https://doi.org/10.2166/wst.2013.819. Bates, S. A. 2014. “Long term performance of a bioinfiltration rain garden with respect to metals removal.” Master’s thesis, Dept. of Civil and Environmental Engineering, Villanova Univ. Bergman, M., M. R. Hedegaard, M. F. Petersen, P. Binning, O. Mark, and P. S. Mikkelsen. 2011. “Evaluation of two stormwater infiltration trenches in central Copenhagen after 15 years of operation.” Water Sci. Technol. 63 (10): 2279–2286. https://doi.org/10.2166/wst.2011.158. Bhatia, S. K., and J. L. Smith. 1995. “Application of bubble point method to the characterization of the pore-size distribution of geotextiles.” Geotech. Test. J. 18 (1): 94–105. https://doi.org/10.1520/GTJ10125J. Busman, L., J. Lamb, G. Randall, G. Rehm, and M. Schmitt. 1997. The nature of phosphorus in soils. Minneapolis, MN: Univ. of Minnesota. Charters, F. J., T. A. Cochrane, and A. D. O’Sullivan. 2021. “The influence of urban surface type and characteristics on runoff water quality.” Sci. Total Environ. 755 (Feb): 142470. https://doi.org/10.1016/j.scitotenv.2020.142470. Coulon, A., A. El-Mufleh, P. Cannavo, L. Vidal-Beaudet, B. Bechet, and S. Charpentier. 2013. “Specific stability of organic matter in a stormwater infiltration basin.” J. Soils Sediments 13 (3): 508–518. https://doi.org/10.1007/s11368-012-0549-2. Driscoll, E. D., G. E. Palhegyi, E. W. Strecker, and P. E. Shelley. 1989. Analysis of storm event characteristics for selected rainfall gages throughout the United States. Washington, DC: EPA. Durand, C., V. Ruban, and A. Ambles. 2005. “Characterization of complex organic matter present in contaminated sediments from water retention ponds.” J. Anal. Appl. Pyrolysis 73 (1): 17–28. https://doi.org/10.1016/j.jaap.2004.09.001. Emerson, C., B. Wadzuk, and R. Traver. 2010. “Hydraulic evolution and total suspended solids capture of an infiltration trench.” Hydrol. Processes 24 (8): 1008–1014. https://doi.org/10.1002/hyp.7539. Fowler, G. D., R. M. Roseen, T. P. Ballestero, Q. Guo, and J. Houle. 2009. “Sediment monitoring bias by autosampler in comparison with whole volume sampling for parking lot runoff.” In Proc., World Environmental and Water Resources Congress, 1514–1522. Reston, VA: ASCE. Giroud, J. P., and J. Perfetti. 1977. “Classification des textiles et mesure de leurs propriétés en vue de leur utilisation en géotechnique.” In Proc., Int. Conf. on the Use of Fabrics in Geotechnics, 345–352. Marne-la-Vallee, France: Ecole Nationale des Ponts et Chaussees. Gonzalez-Merchan, C., S. Barraud, S. Le Coustumer, and T. Fletcher. 2012. “Monitoring of clogging evolution in the stormwater infiltration system and determinant factors.” Eur. J. Environ. Civ. Eng. 16 (1): 34–47. https://doi.org/10.1080/19648189.2012.682457. Gulliver, J. S., A. J. Erickson, and P. T. Weiss. 2010. Stormwater treatment: Assessment and maintenance. Minneapolis, MN: Univ. of Minnesota. Huber, M., A. Welker, and B. Helmreich. 2016. “Critical review of heavy metal pollution of traffic area runoff: Occurrence, influencing factors, and partitioning.” Sci. Total Environ. 541 (Jan): 895–919. https://doi.org/10.1016/j.scitotenv.2015.09.033. Koerner, R. M. 2012. Designing with geosynthetics. 6th ed. 140–145. Bloomington, IN: Xlibris Corporation. Komlos, J., K. Vacca, and B. M. Wadzuk. 2018. “Phosphate retention in a constructed stormwater wetland with low total suspended solids concentrations.” J. Sustainable Water Built. Environ. 4 (1): 04017017. https://doi.org/10.1061/JSWBAY.0000838. Metcalf, E., G. Tchobanoglous, F. L. Burton, and H. D. Stensel. 2003. Wastewater engineering: Treatment and reuse. 4th ed. New York: McGraw-Hill. Minnesota Stormwater Manual. 2021. “Technical basis for pretreatment sizing for basins and filter strips.” Accessed August 23, 2021. http://stormwater.pca.state.mn.us. Munoz-Carpena, R., C. M. Regalado, J. Alvarez-Benedi, and F. Bartoli. 2002. “Field evaluation of the new Philip-Dunne permeameter for measuring saturated hydraulic conductivity.” Soil Sci. 167 (1): 9–24. https://doi.org/10.1097/00010694-200201000-00002. PADEP (Pennsylvania Department of Environmental Protection). 2006. “Pennsylvania stormwater best management practices manual.” In Bureau of stormwater management, division of waterways, wetlands and erosion control. Harrisburg, PA: PADEP. Palmeira, E. M., and M. G. Gardoni. 2000. “The influence of partial clogging and pressure on the behavior of geotextiles in drainage systems.” Geosynth. Int. 7 (4): 403–431. https://doi.org/10.1680/gein.7.0178. Roseen, R. M., T. P. Ballestero, G. D. Fowler, Q. Guo, and J. Houle. 2011. “Sediment monitoring bias by automatic sampler in comparison with large volume sampling for parking lot runoff.” J. Irrig. Drain. Eng. 137 (4): 251–257. https://doi.org/10.1061/(ASCE)IR.1943-4774.0000168. Selbig, W. R., and R. T. Bannerman. 2011. Characterizing the size distribution of particles in urban stormwater by use of fixed-point sample-collection methods. USGS Open-File Rep. No. 2011-1052. Washington, DC: USGS. https://doi.org/10.3133/ofr20111052. Siriwardene, N. R., A. Deletic, and T. D. Fletcher. 2007. “Clogging of stormwater gravel infiltration systems and filters: Insights from a laboratory study.” Water Res. 41 (7): 1433–1440. https://doi.org/10.1016/j.watres.2006.12.040. Steuer, J., W. Selbig, N. Hornewer, and J. Prey. 1997. Sources of contamination in an urban basin in Marquette, Michigan and an analysis of concentrations, loads, and data quality. Water-Resources Investigations Rep. No. 97-4242. Washington, DC: USGS. https://doi.org/10.3133/wri974242. Toran, L., and C. Jedrzejczyk. 2017. “Water level monitoring to assess the effectiveness of stormwater infiltration trenches.” Environ. Eng. Geosci. 23 (2): 113–124. https://doi.org/10.2113/gseegeosci.23.2.113. USEPA. 2007a. SW-847 test method 7000b: Flame atomic absorption spectrophotometry. Washington, DC: USEPA. USEPA. 2007b. SW-847 test method 7010: Graphite furnace atomic absorption spectrophotometry. Washington, DC: USEPA. Veylon, G., G. Stoltz, P. Meriaux, Y. H. Faure, and N. Touze-Foltz. 2016. “Performance of geotextile filters after 18 years’ service in drainage trenches.” Geotext. Geomembr. 44 (4): 515–533. https://doi.org/10.1016/j.geotexmem.2016.02.002. Warnaars, E., A. V. Larsen, P. Jacobsen, and P. S. Mikkelsen. 1999. “Hydrologic behaviour of stormwater infiltration trenches in a central urban area during 2 3/4 years of operation.” Water Sci. Technol. 39 (2): 217–224. https://doi.org/10.2166/wst.1999.0121.