Underperformance Assessment Framework for Bioinfiltration Systems

While SMP performance has been evaluated in regards to a variety of metrics including nutrient and metals removal, solids removal, bacteria retention, runoff volume retention and reduction, peak flow reduction, delay in flow peak, and groundwater mounding (Ahiablame et al. 2012; Hatt et al. 2007; Davis 2008; Catalano de Sousa et al. 2016; Jefferson et al. 2017; Erickson et al. 2018; William et al. 2019), government compliance standards require SMPs to be evaluated by drain-down times, infiltration rates, and maintenance frequency (PADEP and Bureau of Watershed Management 2006; PWD 2020; Minnesota Pollution Control Agency and Minnesota Stormwater Steering Committee 2008; Wadzuk et al. 2021). Here, we define underperformance in line with government compliance standards, as the inability of an SMP to effectively infiltrate stormwater, resulting in consistently ponded water or insufficient drainage.

Sustained ponding may be the result of inadequate infiltration due to factors such as a restrictive soil layer or high groundwater. Abnormally high inflow rates may also contribute to inadequate drain-down times. A restrictive layer at the surface can result from compaction, high inflow of fine material which may result from nearby construction, or accumulation of fines, sediment, or mulch over time (i.e., clogging) (Asleson et al. 2009; Brown and Hunt 2010; Emerson et al. 2010; Olson et al. 2013; Reddi et al. 2000). An undetected restrictive layer under the bioinfiltration systems could also impede infiltration. Insufficient or inadequate maintenance practices often exacerbate these conditions leading to the low performance of such systems. Untangling the complex, sometimes overlapping drivers of underperformance is challenging but necessary for extending the lifespan of sustainable infrastructure.

Bioinfiltration systems rely on the ability of the media and underlying soil to effectively infiltrate water over time. Thus, subsurface investigations and characterization of the basin media and underlying soils are critical to the design. Field infiltration testing typically is a required component of design and construction. However, results from commonly used spot-infiltration, or in-situ, test methods can be highly variable (Ebrahimian et al. 2020) and are affected not only by variability in soil properties but also by factors such as water temperature (Braga et al. 2007; Emerson and Traver 2008; Jaynes 1990), macropores from plant roots and macroorganisms (e.g., worms) (Le Coustumer et al. 2009; Fassman-Beck et al. 2015), and water quality. Variations in soil texture, plasticity, and dry density significantly impact the saturated hydraulic conductivity ($K_{sat}$) and the ability to store and drain water (e.g., field capacity). While soil properties have been thoroughly studied and related to infiltration in the literature, they can be difficult to measure and confirm in the field due to the destructive and laborious nature of soil sampling and laboratory testing.

Multiple projects have shown the high cost associated with overcoming initial soil issues due to construction (Shafique and Kim 2017), resulting in a range of programs targeted at improving SMP construction (Lai et al. 2006; Jayasooriya and Ng 2014). Common issues arising during construction include soil compaction due to heavy equipment or foot traffic, improper or heterogeneous soil placement, or use of inappropriate media (e.g., Pitt et al. 2008; Brown and Hunt 2010). Compaction or clogging of all bioinfiltration systems soil during or after construction can cause significant reductions in $K_{sat}$ (e.g., by $>1–2$ orders of magnitude), resulting in infiltration rates below regulatory requirements (e.g., $0.4–10\mathrm{in.}/\mathrm{h}$; PWD 2020).

There is a growing interest in understanding and quantifying the impact of sediment deposition on SMP performance (Siriwardene et al. 2007; Jenkins et al. 2010; Le Coustumer et al. 2012). Bioinfiltration systems over time can be heavily impacted by erosion or deposition of significant amounts of sediment (USEPA 1999; Delgrosso et al. 2019). While large flows may introduce litter that requires additional maintenance, both small and large flows transport fine-grained particles. Many studies have shown the ability of SMPs to remove total suspended solids (TSS) from stormwater flows (Shafiquea and Kim 2017). Fine particles within the TSS have the potential to clog the SMP media pores, reducing the $K_{sat}$ of the soil (Langergraber et al. 2003; Asleson et al. 2009; Virahsawmy et al. 2014).

Inversely, erosion when present has adverse impacts on SMP performance. Surface destabilization, mulch removal, bare spots, cave-ins, and associated structural deficiencies are all potential negative effects of erosion on bioinfiltration systems (DelGrosso et al. 2019). The prevalence of these issues and their impact on performance has made erosion a key indicator in SMP inspections and maintenance (Environmental Services Division, Dept. of Environmental Resources, and The Prince George’s County, Maryland 2007; WSU and PSP 2012; Portland Bureau of Environmental Services 2016; DPWES and MSMD 2017).

Two potential issues arise associated with inflow parameters: inadequate/excessive flow or excessive sediments. Excessive inflows can be highly erosive (DelGrosso et al. 2019). Excessive influent TSS concentrations entering SMP systems have been found to adversely affect their long-term performance (William et al. 2019). Excessive sedimentation impacts the surface dynamics, and ultimately performance, of the SMP, as discussed. Often, issues due to inflows occur during construction activities when the system is temporally altered from the design to facilitate construction or when there is substantial earthwork in the area (Shafique and Kim 2017; DelGrosso et al. 2019). Subsurface inflow from high groundwater tables may also affect infiltration performance (Zhang and Chui 2019).

Bioinfiltration sites rely on infiltration into the underlying soils below the amended media. Regulatory agencies often require that underdrains be provided for bioinfiltration and bioretention SMPs. These underdrains may be capped to facilitate infiltration into native soils (e.g., PWD 2020). However, if underperformance is observed, drilling an orifice into the underdrain cap or uncapping of the underdrain may be an initial attempt to facilitate faster drain down if the issue is due to impeded infiltration into the native soil. Underdrains may be required in SMPs when underlying soils have low $K_{sat}$ [e.g., $<13\mathrm{mm}/\mathrm{h}$ ($0.5\mathrm{in.}/\mathrm{h}$)] (Sileshi et al. 2018). Utilizing the underdrain essentially allows the bioinfiltration system to function more similarly to a bioretention system if needed.

Underdrain pipes are used to collect water from the bottom of SMPs and deliver the collected water to a sewer system or to a river system during storm events (Guo 2012; Hunt et al. 2012). One type of design has an underdrain pipe slightly sloped toward the outlet of the SMP (i.e., entrance elevation is lower) to facilitate debris removal utilizing gravity. Not all underdrains perform equally. The underdrain pipe can have various pressure heads at the outlet and from water in the rock bed storage and may vary in slope, perforation, or pipe size. Such differences increase the variability of performance of the underdrain pipe and pose challenges to analyses (Tu and Traver 2019). It is difficult to accurately size underdrains to obtain desired flow capacities without inhibiting the use of the natural infiltration of the underlying soils (Hunt and Lord 2006; Sileshi et al. 2018). In addition, the sand/stone filter layer and geotextile fabric commonly used around the underdrains may be susceptible to clogging due to sediments and biomass (Bouwer 2002). If the underdrain becomes clogged, it may need to be cleaned out or replaced. Additional actions may include installing longer underdrains to increase surface area or adding a second underdrain to provide redundancy.

For field infiltration rates of the media, the arithmetic and geometric means of $K_{sat}$ normalized to 20°C ($K_{sat20}$) for each test season for each site are given in Table 1. The geometric means were consistently more conservative (lower) than the arithmetic means. All arithmetic means were above the PWD minimum infiltration requirement of $1.02\mathrm{cm}/\mathrm{h}$ ($0.4\mathrm{in.}/\mathrm{h}$) (PWD 2020) and the slightly greater PADEP infiltration rate recommendation of $1.27\mathrm{cm}/\mathrm{h}$ ($0.5\mathrm{in.}/\mathrm{h}$) (PADEP and Bureau of Watershed Management 2006). Based on the arithmetic means, there should not have been the extended periods of ponding that were observed in the field. In terms of geometric means, only two of the four geometric means for Site 1 and one of the four geometric means for Site 2 met the PWD requirement. One of the geometric means for Site 1 and none of the geometric means for Site 2 met the PADEP recommendation. These data paired with field observations suggested the geometric mean values were more representative for this study. The weighted average approach (32% arithmetic mean $+68\%$ geometric mean; Weiss and Gulliver 2015) was also applied and resulted in values still above the PWD requirements.

It is important to note that methods such as a synthetic drawdown test, simulated runoff test, or recession rate monitoring may provide a more representative estimate of the infiltration rate for the site as a whole, compared to calculating a representative value from multiple infiltration tests. However, such larger-scale measurements would not allow for identifying specific problematic areas that may be contributing to underperformance. Thus, spot-infiltration tests are still an important assessment method within the UAF to provide more targeted rehabilitation recommendations (e.g., only a portion of the basin may require rehabilitation). As with all field testing, the UAF user should apply site-specific engineering judgment as to which infiltration test techniques and analysis methods are most appropriate for the number of tests, size of the site, overall range of $K_{sat}$ values across the site, and constraints of the project. Since both single-ring and double-ring infiltration tests methods have been shown to be accurate predictors of SCM performance (Zukowski et al. 2016), and the results of infiltration testing can be highly spatially variable, it is recommended that the UAF user choose the test method that would allow them to perform the most tests within the budget and time constraints. Even though in this study the results obtained from using the geometric mean were determined to be the most representative based on corresponding field observations, the UAF user is encouraged to confirm how the analysis method impacts the conclusions from the infiltration testing before making rehabilitation recommendations based on the results. Calculations to determine the arithmetic mean, geometric mean, and weighted average (Weiss and Gulliver 2015) are simple to perform and should always be compared with field ponding observations and engineering judgment. The bioinfiltration media was classified based on the USDA and USCS classification schemes. Based on USDA classification, all the collected media samples from Sites 1 and 2 were categorized either as sand, loamy sand, or sandy loam. Texture results (percent sand, silt, and clay) were within recommended guidelines by the PADEP and Bureau of Watershed Management (2006) and the PWD (2020). Based on the USCS classification, which also considers soil plasticity, the Site 1 media samples classified as either well-graded sand with gravel (SW), poorly graded sand with gravel (SP), well-graded sand with silt and gravel (SW-SM), or silty sand (SM). For Site 2 media, soil samples classified according to USCS were either silty sand with gravel (SM), well-graded sand with silt and gravel (SW-SM), poorly graded sand with silt and gravel (SP-SM), or silty sand (SM). Thus, based on the results of texture and particle-size distribution, the media at both bioinfiltration sites should have exhibited acceptable infiltration rates if constructed properly.

Samples collected from the underlying subgrade at Site 1 were all classified as loamy sand and sandy loam per USDA. For Site 2, samples from the underlying subgrade were classified as sand or loamy sand. The USCS soil classification results showed most (16 of 17) native soil samples from Site 1 classified as silty SW (SM), and one was classified as silty sand (SM). For Site 2, USCS soil classification results showed three subgrade soil samples classified as silty sand (SM) and one as silty clayey sand (SC-SM). Literature and municipal guideline recommendations suggest rain gardens will operate well with sandy and silty subgrade soils (e.g., PADEP and Bureau of Watershed Management 2006; PWD 2020). The subgrade soil type at site 1 is likely not causing infiltration performance issues.

The PWD (2020) Manual recommends using uncompacted sand, loamy sand, or sandy loam soil for use in rain gardens and SMPs. The density for typical uncompacted loamy sand soil may range from 1.0 to $1.6\mathrm{g}/{\mathrm{cm}}^3$, while the density for compacted loamy sand soil may exceed $1.8\mathrm{g}/{\mathrm{cm}}^3$ (PADEP and Bureau of Watershed Management 2006). For both Sites 1 and 2, the dry density of the soil media was determined from undisturbed soil core samples taken from the surface to a depth of 45.7 cm (18 in.). The dry density of the media varied from 1.0 to $1.14\mathrm{g}/{\mathrm{cm}}^3$ for Site 1 and from 1.10 to $1.59\mathrm{g}/{\mathrm{cm}}^3$ for Site 2, falling within recommended ranges from PADEP and Bureau of Watershed Management (2006). The density of the underlying subgrade soils at both sites ranged from 1.37 to $1.84\mathrm{g}/{\mathrm{cm}}^3$, although quality undisturbed cores were difficult to obtain. Therefore, compaction of rain garden media soil is likely not impacting the performance at Sites 1 and 2. Compaction of subgrade soil may impact performance at Sites 1 and 2.

Spatial correlations were observed between mulch and sediment deposition depth, slopes, aggradation, low infiltration, and ponded areas for Sites 1 and 2, as shown in Fig. 6. In panel D of Fig. 6, the upstream areas of Site 1 to the left of the figure have mulch thicknesses below the recommended depths of 2.6 to 7.6 cm (1–3 in.); this trend reverses downstream toward the outlet structure, seeing as the deepest mulch and sediment deposition is near the inlet. In Fig. 6 panel D for Site 2, generally, the mulch is moving downstream within the system, similarly to Site 1. Areas upstream, especially upstream of both inlets, show mulch depths less than the recommended mulch thickness of 2.6 to 7.6 cm (1 to 3 in.) (PADEP and Bureau of Watershed Management 2006; PWD 2020).

Other possible causes of underperformance include inflow or underdrain failure. At Site 1, the contributing drainage area had changed, leading up to underperformance due to alterations on the highway deck, especially through the movement of jersey barriers. It was considered that this could be routing additional flow to the site. Inflow was analyzed by comparing inflow values calculated using the rational method to observe inflow measurements. The observed flow was well below the estimated flow. The difference suggests that the site was not receiving inflow beyond its design, but may be receiving less flow than designed.

After sustained ponding was observed at the sites and they were identified as underperforming, the underdrains were opened in an attempt to reduce drain-down times. However, sustained ponding continued to occur even after opening the underdrains. Visual inspections from opening the end caps of the underdrain in Site 1 showed no obvious soil at the end of the underdrain. Further inspection with a drain scope camera revealed deposits of sediment throughout the length of the underdrain, but not to an extent that would suggest clogging. It is possible that soil accumulation in the underdrain was affecting performance. Further work is required to estimate the impact of the soil deposits. Assessment of the SMP will be completed following rehabilitation to resolve the UAF analysis.

Soil samples collected from case study Sites 3 and 4 from the SMPs in Delaware were analyzed in the laboratory at Villanova University. The Site 3 media had a lower percentage of fines compared to the native soils collected at 0.7 m below the basin surface. The Atterberg limits provided insight on the type of fines that each soil contained. The media used at both sites contained a significant clay fraction based on the measure PI of 16 for Site 3 and 13 for Site 4. (Table 2). The grain-size distribution and the Atterberg limits were used to classify the soils according to USCS and USDA. Based on the USDA classification alone, which indicated the media was sandy loam, the media would have been considered appropriate for use in a bioinfiltration system. However, additional consideration of soil plasticity (i.e., using USCS classification) clearly reveals the correlation between the failed sections of the SMPs and the presence of clayey media, as shown in Table 2.

From field observations, the $K_{sat}$ of the native soils was noticeably greater than that of the engineered soils, such that the use of the already in-situ soils may have been preferable to importing and placing engineered media. The sites were not built with underdrains, and shallow groundwater levels were not observed in the area. Additional investigation of the sites was not performed after the use of inappropriate media was identified as a root cause of underperformance.

Although spot-infiltration tests could not be performed at Site 5, the recession rate could still be determined from the pressure transducer measurements. The recession rate yields the drain time after a rain event and can be considered representative of the entire system, mitigating measurement variability due to small-scale heterogeneities often observed for spot-infiltration tests (Ebrahimian et al. 2020). Collected data for rainfall data, pond level data, and the outlet height are compared in Fig. 7.

From the data in Fig. 7, the average water level recession rate over the period from July 2018 to October 2018 was only $0.10\mathrm{cm}/\mathrm{h}$ ($0.0040\mathrm{in.}/\mathrm{h}$), This low rate is consistent with the observed sustained ponding and falls well below recommended infiltration rates for a bioinfiltration system [$0.4–0.5\mathrm{in.}/\mathrm{h}$ for the amended media, per PADEP and Bureau of Watershed Management (2006) and PWD (2020)].

Laboratory analysis of the soil samples collected from the upper 50.8 cm (20 in.) of the basin indicated that the soil was generally classified as sand or loamy sand per USDA soil classification. The USCS soils classification results indicated the soil was silty sand (SM) or silty SW (SP-SM). Thus, based on measurement of both the full particle-size distribution and the soil plasticity, the media retrieved from the upper 50.8 cm (20 in.) at multiple locations within the site was appropriate for use in a bioinfiltration system.

However, attempts to collect samples from deeper than 50.8 cm (20 in.) using the hand auger were unsuccessful due to encountering a very stiff layer. A small test hole was dug as documented by Fig. 8, indicating the potential presence of a thin, fine-grained restrictive layer below a portion of the infiltration media. However, further exploration and sampling could not be performed due to difficulty in fully pumping out the basin. A final assessment of the SMP will be completed following rehabilitation to complete the loop for the UAF analysis.

SMP media and subgrade soils beneath Site 1 are within USDA soil classification recommendations provided by the PADEP and Bureau of Watershed Management (2006) and PWD (2020) Stormwater BMP manuals, as well as USCS soil classification recommendations from literature, are likely not contributing to the infiltration performance issues present at either site. The density of the rain garden media in Site 1 is well below the upper limit for the dry density of uncompacted loamy sand of $1.6\mathrm{g}/{\mathrm{cm}}^3$ and is therefore unlikely compacted and a cause of system underperformance. Compaction of native soils beneath Site 1 could not be ruled out as a possible contributor to the infiltration underperformance.

Spatial analysis showed areas draining steep side slopes were found to correlate with areas of aggradation of soil and areas with a thick mulch layer, suggesting that over time mulch and soil had migrated down the steep side slopes and accumulated up to depths of 21 cm (8 in.), as corroborated by field observations. Mulch had also migrated downslope within the basin. Visual observations of the splash aggregate pad in Site 1 included large aggregate displacement, suggesting that design flows for the system had been exceeded. The sediment deposition near the inlet was the result of high energy flow entering through the inlet perpendicularly to the main flow path. This had forced a geomorphic change within the SMP, resulting in the creation of a cutoff point by shifting the deposition culmination from the steep slopes, ultimately altering the rain garden configuration. The cutoff forces isolation of the upstream ponded area from the downstream draining area. While downstream may have good infiltration potential, water is retained upstream, causing persistent ponding in the SMP. To address this, rehabilitation recommendations include regrading the slopes to correct the steep slopes and to remove, and replace, as needed, media and mulch to correct for media and mulch accumulation within the system. Substituting domed risers for underdrain cleanout caps is also recommended to improve drainage in areas where ponding occurs.

Although there was not as clear of a consistently ponded area in Site 2 as there was in Site 1, ponding times exceeding design ponding times were observed at the site. Similar to Site 1, rain garden media and subgrade soil types of Site 2 are likely not contributing to the infiltration performance issues, seeing as they are within USDA and USCS soil classification recommendations. The majority of the media samples for Site 2 were also below the upper limit for the recommended dry density of uncompacted loamy sand (with the exception of one sample which slightly exceeded this range, but is still not considered compacted); therefore, compaction of the Site 2 media is likely not a problem in this system. Similar to Site 1, compaction of underlying soils beneath Site 2 could not be ruled out as a possible contributor to the underperformance.

Spatial trends suggest depositional areas occur near steep side slopes, although to a lesser extent than Site 1 that exhibits more deposition than in Site 2. Although it is apparent that mulch is moving downstream within the system, mulch depths do not exceed the recommended depth range and are unlikely an issue in this system. Unlike Site 1, there is no restriction in the flow path. The use and configuration of double inlets in Site 2 seem to improve the flow path, contrasting the use and configuration of a single inlet in Site 1. It is likely that each of these factors individually is not the cause of infiltration underperformance, but the combination of them. Recommendations include regrading the slopes in this system, as well as removing rain garden media in areas of media accumulation, summarized in Table 3.

Lab characterization of samples from these sites suggests that the overuse of compost that degraded to high plasticity materials is the likely culprit for the underperformance of these systems (Table 3). The rehabilitation recommendation for Sites 3 and 4 was to replace or amend the soil media to media that better facilitates infiltration of water or to introduce an underdrain to the system. Notably, the soil mixture composition was technically compliant to local regulations for bioinfiltration media, further supporting the need for more robust soil specification approaches that also consider soil plasticity. Through systematic analysis of the basin slopes, sediment and mulch accumulation, and inflow designs, these items were determined to not be negatively impacting the infiltration in these two systems. To remediate the underperformance, the developers of Sites 3 and 4 introduced trench drains for rehabilitation. Since implementation, infiltration and function of these sites have increased. Further study is needed on the impact of incorporating trench drains into a system as a form of remediation and whether it should be adopted into design standards and construction techniques.

For Site 5, infiltration testing and deeper undisturbed core sampling could not be performed adequately due to the persistent ponding at the site. However, the collection of soil cores in the upper 50.8 cm (20 in.) indicated that the media generally classified as sandy loam. Thus, the use of inappropriate soil media was ruled out as the cause of underperformance. Difficulty advancing the hand auger beyond the 50.8 cm (20 in.) depth, combined with observations from staff present during the initial construction, supports the potential for an unexpected restrictive layer (e.g., due to construction dust/debris accumulation, or a dense clayey layer not discovered during the initial investigation) existing at 50.8 cm (20 in.) depth or deeper that has been inhibiting infiltration. Rehabilitation options for the presence of a subsurface restrictive layer are more challenging than the previously described case studies. At the time of writing, infiltration at the site had not yet been restored and restoration actions were still being explored. In addition, a high or fluctuating groundwater table could not be ruled out as a potential contributor to the underperformance. Notably, there are limited possible remedies for groundwater intrusion into bioinfiltration sites that have already been constructed. Issues associated with high groundwater are best managed if prevented altogether through a thorough site investigation and proper site selection. If groundwater intrusion is observed during construction, swift corrective actions (e.g., postbid construction revisions) can sometimes be made before construction is completed.

The opinions presented in this publication are those of the authors and do not necessarily express the opinions of the PennDOT. References in this report to any commercial product, process, or service, or the use of any trade, firm, or corporation name are for general informational purposes only and do not constitute an endorsement or certification of any kind by the authors. This project is a research initiative of the Villanova Center for Resilient Water Systems.