Developing Multiple Lines of Evidence to Decrease Drainage-to-Surface Area Ratio for Effective Stormwater Control Sizing Using Bioretention

Abstract

Bioretention units were constructed at the US Environmental Protection Agency’s Edison Environmental Center to evaluate drainage-to-surface runoff ratio for sizing of bioretention stormwater controls. Three sizes of hydraulically isolated bioretention units were tested in duplicate with changes in aspect ratio of length from inlet wall by doubling successive length from smallest (3.7 m) to largest (14.9 m) while width remained the same (7.1 m). The watershed areas were nominally the same, resulting in watershed-to-surface area ratios of 5.5:1 for largest duplicate units, 11:1 for the middle units, and 22:1 for the smallest. Each unit was instrumented for continuous monitoring with water content reflectometers (WCRs) and thermistors with data collected since November 2009. The bioretention units were filled with planting media initially comprising 90% sand and 10% sphagnum peat moss by volume and approximately 99% and 1%, respectively, by weight. These units were then planted between May and November of 2010 with a variety of native grasses, perennials, shrubs, and trees that were tolerant to inundation, drought and salt. In late 2012, a survey of the shrubs planted in these bioretention units was performed. The published results of the combined analyses of moisture content, rainfall, and size of shrubs indicated that the smaller units had superior shrub growth due to the more frequent saturation of the root zone as measured by WCR, while the plants in the largest units, particularly away from front wall where runoff entered, potentially relied on direct rainfall only. Starting in 2017, additional monitoring was performed in these units, including chemistry analysis by loss on ignition and total phosphorus of the engineered planting media and an additional survey of the plants. As in the previous study, plants did better in the medium (11:1) and small (22:1) bioretention units than in the largest units (5.5:1), and there was greater buildup of organic matter and phosphorus in the smaller units. One species of grass that dominated the two largest bioretention units away from the inlet was drought tolerant, which indicated that plants in these units relied on rainfall rather than stormwater runoff. Oversized units did not completely use the stromwater control volume, and many of the other original plantings grew slower or were less widespread in comparison to plantings in that smaller units that flooded more frequently and achieved greater growth.

Practical Applications:

Defining the size of stormwater controls can be difficult because there are often multiple objectives imposed on the final design of these structures, including safety and flooding. Results presented here would indicate that if the objective is to create a bioretention area with healthy vegetation, undersized controls may be acceptable because undersized infiltrating controls will have healthier plantings and infiltrate throughout the storm. For municipalities, this means that rights of way previously thought to be too small to use for infiltrative stormwater controls may be converted to such a purpose. This does not free municipalities from stormwater systems that address flooding and safety design objectives, but demonstrates that increasing plantings in the municipal right of way could help to address stormwater as well as other objectives, like greenhouse gas emissions, urban heat island reduction, and clean air. Distributed bioretention controls that capture part or all the runoff of the smaller, most frequent rainfall events should be incorporated throughout municipalities and into their overall stormwater control systems. If clogging by runoff is a concern, roof runoff may be more appropriate for bioretention, or other measures such as sediment capture or increased maintenance may need to be performed.

Introduction

For urban applications of stormwater control, size is often a constraint due to cost and limited land availability. Whole life costs for bioretention are three times higher than other stormwater controls (with the exception of infiltration trench) and can range from $100 to $400/m² (2004 cost) due to curbing, storm drains, and underdrains, as opposed to residential rain gardens which averaged between $3 to $40/m² (WEF and ASCE/EWRI 2012); bioretention costs typically include excavation, geotextile fabric, gravel layer, mulch, project management, overflow structure, permitting and construction inspection, engineering, land acquisition, and contingency (WEF and ASCE/EWRI 2012), in addition to backfill (engineered media) and plants. Other costs for bioretention built in a municipal or transportation authority right of way (ROW) may include maintenance and protection of traffic (MPT), and may entail safety training for personnel. Engineered media for bioretention is typically specified by municipality or state guidance documents and varies accordingly, but usually consists primarily of sands with lower proportions of silt and clay, emphasizing hydraulic conductivity, and organic amendments (e.g., compost, peat moss), to improve pollutant removal (Tirpak et al. 2021).

In 2008, the United States Environmental Protection Agency (USEPA) moved forward with the design and construction of a permeable parking lot and adjacent bioretention stormwater controls at the Edison Environmental Center (EEC) for the purposes of research and demonstration of low-impact development (LID) and green infrastructure (GI). At the time the bioretention units at the EEC were being planned, regulatory guidance for bioretention design throughout the United States varied greatly, from 3% to 43% of the associated drainage area to surface control area based on rainfall characteristics, soil type, and other factors (Stander et al. 2010). Instead of constructing a common gallery for a single bioretention area located at the southern end of the permeable parking lot, as initially planned, an experimental design using differently sized and hydraulically isolated bioretention units was designed. The initial objectives of the experimental design (Stander et al. 2010) were to (1) quantify the hydrologic performance of the bioretention receiving parking lot and roof runoff over time, (2) test several ratios of surface drainage to bioretention area in terms of hydrologic performance, and (3) evaluate the ability of in-situ monitoring devices to detect soil moisture in in engineered media and underlying soil.

An observation that some of the plants were not doing well—potentially due to water-deficit stress—led to a survey of shrub stem and trunk diameters and heights in 2012. Statistical analysis revealed differences in shrub growth based on the size of bioretention unit, with the smaller units having greater growth habits for bayberry than the larger units (Brown et al. 2015); distance from the inlet was also negatively correlated with bayberry and winterberry growth, particularly for the larger units. These results were linked to observed water availability for plants (i.e., soil moisture), which was far less in the largest units, particularly away from inlet. Soil moisture measurements indicated that stormwater runoff most often infiltrated near the inlets minimizing surface runoff in the bioretention units, and plants farthest from the inlet (i.e., in the largest units), only relied on rainfall, which put plants in an irrigation deficit during warmer months (Brown et al. 2015). Other researchers (Tirpak et al. 2018; Tu et al. 2020) noted conflicting goals of providing means for healthy vegetation versus prioritizing functioning of stormwater controls.

The 2012 plant survey was limited to bushes and was supported by WCR data and an evaluation of reference crop evapotranspiration analysis, but only really identified the largest units as being oversized (Brown et al. 2015). Could further metrics be assessed for the bioretention units constructed at the EEC and more importantly, as time went on, were the earlier conclusions still supported by a wider range of observations and could more be said about increasing the drainage-to-surface area ratio? Beginning in 2017, media samples were analyzed for organic matter (OM) and phosphorus content. A second plant survey was also conducted during 2018, which included grasses and trees as well as previously surveyed shrubs. The purpose was to show that plant, OM, and total phosphorous (TP) observations supported the earlier conclusions (Brown et al. 2015) regarding better plant growth due to increased saturation of the media in the smaller bioretention units and greater plant growth closer to the front of the units where the runoff inlets were located. If this was the case, the experimental design of evaluating different sizes of bioretention units would support increasing the drainage-area to surface-area ratio. The media sample evaluation and second plant study would also show the effects over time, which would address the first objective of the experimental design (Stander et al. 2010).

In their study of soil building properties of 10 bioretention systems, Ayers and Kangas (2018) noted several confounding factors and variations in their assessments, which included amount of OM, root biomass, plant cover, litter/mulch layer, organism populations, distance to natural areas, and time since construction. They did not have access to specifications of original planting media or maintenance activities. They referenced Stevens and Walker (1970), who indicated that many studies failed to control for all soil-forming factors besides time, but such studies are still useful for qualitative comparisons. However, Stevens and Walker (1970) lamented the inability to control studies occurring across different regions and compare studies that used a range of analysis methods. The current study offers substantial control over soil forming properties in the bioretention units, as the engineered media specifications were well defined and are the same in each unit; the planting palette is also the same, distance from other landscaped and nonlandscaped growth areas is similar, each unit is hydraulically isolated, and contributing drainage area to each unit is nominally the same.

Of note also, the Ayers and Kanga (2018) study did not address water quality directly. Monitored pollutant inputs to the bioretention units included total organic carbon (TOC) and phosphorus with samples of asphalt runoff from the parking lot (144 samples during 40 events) and rainfall (110 samples during 47 events) collected between October 2010 and August 2017 (Razzaghmanesh and Borst 2019a). The mean TOC for runoff to the bioretention units was 9.4 mg/L with a median of 6.1 mg/L, and the mean total orthophosphate (PO₄) concentration was 0.080 mg/L with a median of 0.047 mg/L. The mean TOC in rainfall was 2.4 mg/L with a median of 1.8 mg/L, and the mean PO₄ concentration was 0.065 mg/L with a median of 0.026 mg/L. Roof runoff concentration from January 2010 through October 2012 observed for TOC (21 events, mean 1.9 mg/L, and median 1.4 · mg/L) and PO₄ (25 events, mean 0.059 mg/L, and median 0.038 mg/L) were similar to rainfall (O’Connor and Amin 2015). During 28 events over the period from January 26, 2010 to June 4, 2013, the suspended solids concentration (SSC) of runoff to the rain gardens was 103 mg/L. By contrast, 28 roof runoff samples had a mean SSC of 9.0 mg/L (O’Connor and Amin 2015). The actual pollutant removals of runoff by the infiltrating engineered media was not quantified beyond the observation of OM by loss on ignition (LOI) and TP.

Methods

The site had six bioretention units. Each was 7.1-m wide, and all were tested in duplicate of three different sizes by changing the aspect ratio of length from the front (northern wall with inlets) by doubling successive length from smallest (3.7 m) to largest (14.9 m) to produce three surface areas of 26.3 (units 1 and 6), 52.5 (units 2 and 5), and 105.1 m² (units 3 and 4), respectively. Drainage areas for each unit were nominally 571 m², composed of nearby roof area (465 m²) and asphalt parking row and driving lane (105 m²) (Brown et al. 2015), resulting in drainage to bioretention surface aspect ratios of 22:1, 11:1, and 5.5:1.

At the time of design and construction, only the middle bioretention unit (7.5-m long) had a drainage-to-surface area ratio at 11:1, which was similar to the New Jersey Department of Environmental Protection (NJDEP 2007) design guidance for bioretention, with a design example drainage-area to surface-area ratio of 10.4. Bioretentions in New Jersey are designed only to address the Water Quality Design Storm (New Jersey Administrative Code 2020) defined as “1.25 inches of rain falling nonuniformly in a 2-h storm event”). In comparison, other stormwater controls are required to address the 2-, 10- and 100-year design storm (NJDEP 2021b).

The planting media was 0.86 m deep, comprised of 90% sand and 10% sphagnum peat moss by volume (Stander et al. 2010); OM was 1.1% (±0.2%). Quality Material Engineered Sand and Soils a Division of Geo. Schofield Co. Inc. (Bound Brook, New Jersey) mixed and delivered 670.76 t of engineered media at a cost of $23,337.68 and unitized cost of $34.79/m³. Planting media was placed over a 0.1-m layer of 25-mm gravel that was supplied by a construction contractor (without unitized cost information). Two separate layers of a nonwoven double needle punched geotextile liner (Propex Geosynthetics, Chattanooga, Tennessee) at unitized cost of $ 1.07/m² separated the gravel from the underlying soil and the engineered planting media from the gravel. Hardwood mulch at a unitized cost of $29.13/m³ was raked onto the top at a depth of 0.05 m (2 in). Each bioretention unit was hydraulically isolated along its perimeter by high-density polyethylene (HDPE) sheeting that extended 0.1 m into the native soils (Brown et al. 2015).

Fig. 1 shows one-half of an early planting plan and a complementary half of the final hand-drawn planting diagram with location of plants and electronic monitoring instrumentation. The six units were instrumented for continuous monitoring with an array of water content reflectometers (WCRs) to measure soil moisture and thermistors in the planting media at approximately 0.4 m below the surface. These instruments were placed 2.9 m from the front where the inlets were located, and 2.9 m from the rear of each cell. See Stander et al. (2013) and Brown et al. (2015) for further details. Solinst pressure transducers recording at 10-min intervals were deployed on August 18, 2011, on an L-bracket attached to the piezometer cluster above the media surface and below the top of the sidewalls of each bioretention unit. Data were collected through July 29, 2013. The bioretention units had a nominal freeboard of 0.20 m (8.0 in.) (Brown et al. 2015).

Fig. 1.

Planting diagrams with location of instrumentation.

Three types of shrubs were planted in May and June 2010: Myrica pensylvanica (northern bayberry) (M in Fig. 1); Vaccinium corymbosum (northern highbush blueberry) (V); and Ilex verticillata (winterberry holly) (I); and a tree, Amelanchier laevis (Allegheny serviceberry) (T). Herbaceous flowering perennials were also planted at this time and included Echinacea purpurea (L.) Moench (purple cone flower) (EP), Helianthus salicifolius (willow leaf sunflower) (Ht), Iris versicolor (northern blue flag) (IV), Juncus effusus (softrush), Liatris spicata (blazing star) (Ls), Lobelia cardinalis (cardinal flower) (Lc), Oenothera speciosa Siskiyou (Siskiyou evening primrose) (Hh–substituted item), Penstemon digitalis Nutt. ex Sims (foxglove beardtongue) (Pd), Rudbeckia fulgida var. sullivantii Goldsturm (sold as “black-eyed susan”) (R), Symphyotrichum novae-angliae (formerly Aster novae-angliae) (New England aster) (A), and Zizia aurea (golden Alexander) (Z). A variety of grasses and sedges, including 100 plugs each of Panicum amarum Ell. (bitter panicum; beach grass), Andropogon gerardii (big bluestem), Sorghastrum nutans L. (Indiangrass), Schoenoplectus pungens (common three square) and 80 pots of virgatum (heavy metal switch grass), were planted in November 2010. Total cost for plants used in these units was $5,785.

In the monitoring system, the OM of the planting media from the bioretention units was measured by an alternative method of LOI (NRP 2011), which burns off OM from 5 to 10 g of dried media in crucibles, and is calculated by the following equation:

$$\begin{gathered} OM(\%)=(\text{weight at}{105}^{{}_{°}}\mathrm{C}-\text{weight at}{360}^{{}_{°}}\mathrm{C} \\ \times 100)∕\text{weight at}{105}^{{}_{°}}\mathrm{C} \end{gathered}$$

Hand-troweled samples were taken along the centerline (the grassed area in the left side of Fig. 1, avoiding bushes and trees) of each bioretention unit from front to back at intervals of 0.3 m, starting 0.15 m from the front and at a depth of 0.15 m in 2017 for units 4, 5, and 6, and in 2018 in units 1, 2, and 3. Samples were also collected from the smallest units, 1 and 6, at intervals of 0.5, 1.5, 2.3, and 3.0 m from the front while samples from the medium units 2 and 5 were collected at intervals 0.5, 2.7, 4.3, and 6.4 m. These additional samples were collected closer to shrubs (as shown in Fig. 1) in two rows approximately 0.6 m from each sidewall and at depths of 0.15, 0.3, and 0.6 m; the upper depth samples were collected by hand trowel, while only the deepest was collected by hand auguring. The first set of these samples was collected along the western side wall in 2018, and the second set along the eastern side wall was collected in 2019.

TP extraction of the planting media, ~2.5 g each, was also performed on a subset of samples collected for LOI analysis. This subset, representing 38 locations, comprised 4 samples from bioretention unit 4, 18 from unit 5, and 16 from unit 6. Furthermore, unit 5 had 12 samples at 0.15-m depth while unit 6 had 10, and each of these units had three samples at depths of 0.3 and 0.6 m to observe if there were changes in TP concentrations with depth. Because the median rainfall pH was <6 (Razzaghmanesh and Borst 2019a), the extraction method most suitable was Mehlich 3 (NRP 2011). TP was analyzed per USEPA method 365.1 (USEPA 1993). One archive sample of the original engineering media mixture from 2009 was analyzed in duplicate.

The plant survey conducted in 2018 measured trunk or largest stem diameter at breast height (DBH) (1.4 m; 4.5 ft) with diameter tape. The previous survey (Brown et al. 2015) measured the three widest diameters of shrubs with digital vernier calipers for stems or branches, while the newer analysis only assessed that largest stem diameter at DBH for multistemmed plants. Attaining the height for DBH (1.4 m) was used as a threshold for measuring diameter for shrubs and service berries due to the substantial growth of plants in the intervening 5.5 years between surveys. Heights of all other woody vegetation were measured if they attained or exceeded the threshold height for DBH measurement.

The DBH of bayberries were used to calculate basal area (BA = π(DBH)2/4) (Environmental Laboratory 1987) for comparative statistical analysis between both surveys. Two bayberries in bioretention unit 2, obscured during the 2018 survey due to overgrowth of grasses, were subsequently measured February 2020.

Similarly, diameter of grasses can be measured near the base of the stand (Environmental Laboratory 1987) and can be used to calculate basal cover area. Of the four grass species originally planted in the bioretention units, only Indiangrass was measured with diameter tape at approximately 0.11–0.15 m above the media surface, which provided the best consistency for the existing stands of grass. Because Indiangrass is a warm-weather plant, these diameter tape measurements were made in the first 3 weeks of the survey from June 19 through July 12 of 2018 to minimize complications of this grass species growing during summer season. Grass measurements included a second species, bitter panicum, although the growth pattern of this species with open clumps or stands, and spreading by rhizomes (USDA 2006), was not conducive to diameter tape measurements at the base; therefore, area-wide measurements, where applicable, were made.

Statistical and Data Analysis Methods

Rudimentary summary statistics were performed using Microsoft Excel. OM (%) data from each depth were tested for normality by Shapiro–Wilk histogram analysis, and a nonparametric analysis by multiple comparisons Kruskal–Wallis test was performed in Statistica 9.1 (StatSoft 2010). Regression analysis was performed in Statistica 9.1 for remaining analysis on OM (%). One-way analysis of variance (ANOVA) and multiple analysis of variance (MANOVA) were used to test if there were significant differences (p < 0.05) for bayberry BA (StatSoft 2010). Plant survey graphical analyses of position in relation to bioretention unit borders were performed in AutoCAD (Autodesk AutoCAD LT 2015) and Microsoft Excel.

Results

We have invoices of material and plant costs. Construction labor was not itemized; it was complicated by two contractors performing work, and may not have represented typical bioretention construction due to atypical construction labor cost of the impermeable barrier between units and the instrumentation for monitoring in each unit. The total cost of construction material (other than gravel) and plants in each unit is contained in Table 1. Planting costs were divided between units based on surface area except for service berry ($65) planted in each unit and bayberry ($10.75) with one in smallest units, three in medium units, and 11 in largest units. Plants were purchased in 2010, while other materials were purchased in 2009. The largest percentage cost of material was the engineered media.

Table 1.

Cost of construction materials and plants for bioretention units

Bioretention unit size	Engineered media	Mulch	Geotextile	Plants	Total	Percentage cost of engineered media (%)
Smallest	$1,440	$38	$57	$438	$1,970	73
Medium	$2,890	$78	$114	$822	$3,910	74
Largest	$5,780	$154	$228	$1,633	$7,798	74

The OM in a bioretention system is expected to increase over time in the uppermost layer (Ayers and Kanga 2018). The greatest buildup of OM was nearest the surface at the 15-cm sampled depth (Fig. 2). A Shapiro–Wilk W test indicated the OM (%) at all three depths was nonparametric. A nonparametric analysis by multiple comparisons Kruskal–Wallis test was statistically significant (p = 0.0033), with statistical differences of OM (%) between 15- and 60-cm depth, while 30-cm depth was not significantly different from the other depths. These results confirm buildup of OM in the upper layer.

Fig. 2.

Organic matter (%) of planting media samples from bioretention units 1, 2, 5, and 6 at three separate depths.

Within each unit, OM (%) at 15-cm depth at the same distance from the front where runoff enters units were averaged, and regression analysis was performed. Only the regression of bioretention unit 6 had a clear trend for increasing slope (R² = 0.64) as the maximum OM (%) came near the far end of the unit. Other units had poor regressions (R² < 0.2), although bioretention units 5, 2, and 1 had positive slopes while bioretention units 3 and 4 had negative slopes. Data for bioretention units 1, 2, 5, and 6 were averaged with values rising to a maximum OM (%) at about 3 m from the front. A regression that included OM (%) over 2% at a distance greater than 3 m is shown in Fig. 3. Unit 2 had an outlier measurement, which may have been a result of the influent pipe distributing roof runoff past the first sampling point and potentially pushed litter back toward the northern wall rather than further into the unit; if the first data point near the influent is omitted, then the regression R² improves from 0.34 to 0.57.

Fig. 3.

Regression of normalized organic matter (%) values that peak near 3 m from influent.

Two extreme OM (%) data points are of note, and both were removed from statistical analysis. A value of 7.5% was observed in bioretention unit 3 at 0.15 m from front, which would have further skewed data. Unlike bioretention unit 4, bioretention unit 3 had pipe that delivered roof runoff to a point further into the unit than at the wall; discharge may have pushed surface OM (surface litter) back toward front wall. The next OM (%) value in line at 0.3 m was observed to be 2.60%. An OM (%) value of 15.4% was observed in bioretention unit 6 at 3 m from the front, which would have also skewed slope analysis; this sample was collected adjacent to the trunk of one of the bayberries, and accumulation of litter may have contributed to the buildup of surface OM.

TP concentrations had a R = 0.78 correlation with OM (%) for all 38 coevaluated samples. Fig. 4 shows box plots of the TP concentration in bioretention units 4, 5, and 6 and at depth for units 5 and 6. There is an obvious buildup of TP concentration in the uppermost layer at the depth of 0.15 m in units 5 and 6; the media for unit 4 were only tested at 0.15 m due to lack of observed variation for OM (%) for the largest units. For comparative purposes, the median concentration of bioretention unit 4 at 0.15 m and other depths for units 5 and 6 are close to archive media sample collected in 2009 during construction, which had a 0.20-mg/L concentration, indicating little or no change. Unit 6 had the highest TP concentration by far, as was the case for OM (%), and the correlation (R) between TP and OM (%) was 0.97 in this unit.

Fig. 4.

Box plot of total phosphorus concentration grouped by bioretention unit and depth.

The TP concentration at 0.15-m depth in bioretention unit 5 had a maximum concentration of 0.79 mg/L at 3.2 m from the front while unit 6 had a maximum concentration of 0.95 mg/L at 2.6 m. As for the treatment for OM (%) in Fig. 3, averaging TP concentrations at the same distance in bioretention units 5 and 6 and performing a regression for 9 points through the maximum TP concentration at 3.2 m resulted in a regression coefficient R² of 0.50; similarly, as for OM (%), omitting the first TP data point near the entrance also improved the regression coefficient (R² = 0.57).

We conducted a second plant survey in 2018. Woody vegetations attaining 1.4 m for measurement (DBH) are shown in Fig. 5 with defined diameters of either 2 m (bayberry) or 1 m (rest). Fig. 5 also shows that BA of individual stands of Indiangrass dominate the front and smaller bioretention units, while measurable coverage areas of bitter panicum dominate the larger units, 3 and 4, away from the influent. Bitter panicum or beach grass was also observed in units 2, 5, and 6, but in quantities insufficient to quantify, and was not observed in unit 1.

Fig. 5.

Plan of bioretention units with 2018 plant survey results.

Fig. 6 shows the height of woody vegetation to scale, with diameter measurements to scale for bayberry and serviceberry. Fig. 6 also shows unplanted, observed trees, two of which were black cherry and a third was identified as common mulberry. Most trees and shrubs that attained height for DBH measurement (1.4 m) were within the first 7.5 m from front, which coincides with distance to the back wall of the medium-sized bioretention unit. The extent of unplanted trees [Wissler et al. (2020) describes unmowed, dry detention basin as “overgrown”] is around 3.8 m from inlet wall, which is slightly beyond the distance to the back wall of the smallest unit (3.7 m) and observed maxima (3 m) and 2% range (3.2 m) for OM (%) and TP maximum concentration (3.2 m) in bioretention unit 5 and regression analysis. Fig. 6 also shows increasing cost per length of construction material and original plantings as distance from the front increases (line fit through total cost of each unit size listed in Table 1).

Fig. 6.

Woody vegetation distance from front and associated material and plant costs.

Only three bayberries were measured in the back half of the largest bioretention units, that is, beyond the length to the back wall (7.5 m) of medium-sized units 2 and 5. During the 2018 survey, there were fewer bayberries observed at DBH in bioretention unit 3 (7 fewer) and more in unit 4 (3 more) and unit 5 (1 more). Two bayberries in bioretention unit 2 were obscured during the 2018 survey, but were later identified after vegetation was cut back to allow access for the last set of planting media sampling performed during September 2019; these two bayberries were measured in February 2020. In general, bayberry plants that attained DBH indicated increases in size (one-way ANOVA) for both BA [F_(1,36) = 16.9, p = 0.00022] and height [F_(1,36) = 36.2, p > 0.0001]. Bayberry also appeared to be spreading by rhizomatous growth (Snell 2019), as there was an abundance of smaller plants and shoots in the understory. Only four blueberry heights exceeded height for DBH (1.4 m); no winterberry plants were observed at any height.

MANOVA of bayberry height (cm) and BA (cm²) versus categorical variables of bioretention unit size and compass location (east vs. west) previously indicated statistically significant differences (p ≪ 0.05) with larger mean plant height and BA for smaller units and greater growth on the west (Brown et al. 2015). The same analysis of the 2018 survey results also had statistically significant differences (p ≪ 0.05) for smaller units and greater growth for the western units; however, when the additional data of bayberries in bioretention unit 2 from 2020 were added and MANOVA performed, no statistical difference was observed.

The two bayberries in the front of bioretention unit 2 not surveyed in 2018 had stunted growth. During winter snowstorms, snow was plowed toward the bioretention units to clear the parking lot and often piled up in the front of the units, which may have impacted some woody vegetation. Of note, one of the bayberries in bioretention unit 2 had evidence of a broken-off main stem of this typically multistemmed shrub. The average changes in height and diameter for the two plants measured in 2020 since the 2012 study were 5.8% and 26%, respectively. In comparison, the average changes of height and diameter for the other bayberry in unit 2 and the one bayberry in unit 1 were 21% and 140%, respectively. Pruning and shaping of bayberry are noted to reduce plant vigor (Dickerson 2002).

During the previous survey (2012) after three growing seasons, all but one bayberry had attained 75% of expected maximum height of 2.74 m (observed at 10 years) (USDA 2013) in bioretention units 1, 2, and 6, while none in units 3, 4, and 5 had attained this height (Brown et al. 2015). Bayberry are expected to have a “slow” growth rate (Gilman and Watson 1994). During the recent plant survey (2018), six bayberry plants either attained or exceeded the expected maximum height of 2.74 m, with four of the six in the small or medium bioretention units, representing half (4 of 8) the bayberry planted in these units, while only two (14% of original planting) exceeded 2.74 m in the large units, and these were 4.1 and 4.45 m from the front. No bayberry was observed at or greater than the expected maximum height beyond 6.6 m, which was well within the distance to back wall of the medium bioretention unit (i.e., 7.5 m) as shown in Fig. 6.

One-way ANOVA of the plant cover area of individual Indiangrass stands did not show any difference among units. However, the area of Indiangrass per bioretention unit is presented in the Fig. 7 bar chart, which shows more cover in the smaller units than the largest units. Also presented is the summation of plant cover area per surface area of bioretention unit as a percentage. While the medium-sized units have the largest summation of plant cover, the smallest units have the largest percentage summation per unit area for Indiangrass. The mean percentage of the events achieving saturation as measured by WCR per bioretention unit size, as derived from Brown et al. (2015), is also shown, and indicates the smallest units achieved saturation much more frequently than the largest units. On the other hand, bitter panicum dominates approximately 50% of the surface area in the largest units away from the front.

Fig. 7.

Summation of Indiangrass cover and cover area per bioinfiltration surface area with mean percentage of the events achieving saturation.

We deployed pressure transducers in the bioretention units August 18, 2011, due to an already wet month, the sixth wettest since 1895 even before Hurricane Irene (Robinson 2011). Hurricane Irene crossed New Jersey as Tropical Storm Irene August 27 and 28, 2011, making August 2011 the wettest month on record. All six bioretention units flooded and overflowed during Tropical Storm Irene, as 221.8 mm (8.73 in.) of rainfall was recorded over 24 h at the EEC, exceeding the estimated National Oceanic and Atmospheric Administration mean projection total of 219.5 mm (8.64 in.) for a 100-year recurrence interval 24 -h event (O’Connor and Amin 2015).

Another exceptional rainfall event, 69.2 mm with a peak 30-min rainfall intensity of 8.6 mm, occurred a little over a week later ending on September 8, 2011, and led to five of the six bioretention units overtopping (only unit 4 did not overtop). Evaluating the pressure transducer data through July 29, 2013, indicated there were only five more overtopping events all from the smallest bioretention units, four of which overtopped unit 1 and one of which overtopped unit 6. As noted in Brown and Borst (2014), there was more runoff flow on the western side of the parking lot directed toward the bioretention units, including unit 1, than on the eastern side of the parking lot where unit 6 was located. The event, when both bioretention cells 1 and 6 overtopped and other bioretention cells did not, occurred on June 13, 2013; this rainfall event accumulated 35.3 mm and had a peak five-minute intensity of 5.3 mm/h. Overtopping in bioretention unit 6 was for two 10-min interval recordings, while unit 1 was one recording. The short duration of the observed overtopping is due to the underlying soils beneath the bioretention units, which had mean infiltration rates (from Fig. 9, Stander et al. 2013) that exceeded 100 cm/h as measured by double-ring infiltrometers after excavation and prior to sidewall construction, backfilling, and planting.

Discussion

Several metrics were used to determine effects that drainage-to-surface area ratio had on plants in the bioretention stormwater controls and the accumulation of OM and phosphorus near the surface. Table 2 shows metric versus distance from front and associated material and planting cost. As seen, soil-building processes maximize within the distance to back wall of smallest unit, and most plant metrics would be confined within the wall of the medium unit. As the soil-building maxima for both OM (%) and TP concentration are derived from averages from both the smallest and medium units, this process does not appear to be a result of a limiting factor of the back wall of the smallest unit, as the medium unit wall extends an additional 3.7-m further. A lone bayberry attained the DBH threshold near the back of unit 4, and the plant that dominates the back of units 3 and 4 is bitter panicum, or beach grass, which is drought tolerant (USDA 2006) and starts its dominance at 4.1 m from front.

Table 2.

Summary of extent of metrics as measured from front of units

Metric	Bioretention unit size (number)	Distance (m) from front	Material and plant cost per distance from front
Maximum normalized phosphorus concentration (mg/L)	Small (6) and medium (5)	2.6	$1,382 a
Maximum normalized LOI (%) value	Small (1 and 6) and medium (2 and 5)	2.9	$1,538 a
Distance to back wall	Small (1 and 6)	3.7	$1,970
Last unplanted tree or succession	Large (4)	3.8	$2,007
Beginning of dominance of bitter panicum	Large (3 and 4)	4.1	$2,163
Last measurable stand of Indiangrass	Medium (5)	5.8	$3,049
Last blueberry to attain DBH	Large (4)	6.2	$3,258
Last bayberryb to attain height of 2.74 m	Medium (5)	6.6	$3,466
Distance to back wall	Medium (2 and 5)	7.4	$3,910
Last bush (bayberry) measured at DBH	Large (4)	13.3	$6,957
Distance to back wall	Large (3 and 4)	14.9	$7,798

^aBased on regression equation of 521.09 × (distance from front) + 26.885; values calculated based on distance less than 3.7 m length to back wall of smallest bioretention units italicized as estimates based on lower than observed construction material and plant costs; bold indicates the size and cost of the six units as opposed to other metrics (e.g., Last measurable stand of Indian grass).

^bExpected bayberry maximum height observed at 10 years (USDA 2013).

If one assumes that the influent runoff spreads out evenly across the surface before infiltrating, one would have expected similar soil moisture between front and back, similar growth rates of plants, equitable distribution of grasses, and accumulation of OM and TP throughout bioretention units. Instead, there is an observed gradient of moisture content of planting media (Brown et al. 2015), varying growth rates of plants based on distance from where the stormwater runoff enters, first observed by Brown et al. (2015) and expanded here, and concentrated soil-forming processes toward the front. Fig. 8(a) shows the limits of a surface runoff event in the preplanted bioretention units, and in the closest (western side) three units, nearly all of unit 1 is covered while one-half of unit 2 and only one-quarter of unit 3 are covered. Fig. 8(b) shows the Indiangrass (from the eastern side) in full bloom after plant survey was performed near the front of the units where runoff enters.

Fig. 8.

Photos of bioretention units: (a) from west side, unplanted units during rainfall event in April 2010; and (b) from east side, Indiangrass in full bloom August 2018. (Images courtesy of the US Environmental Protection Agency.)

Media sampling for OM (%) and TP concentration indicated that there is high correlation between OM (%) and TP. The highest correlation was in the smallest unit at 0.15 m below surface, and the smaller bioretention units were accumulating more OM and phosphorus than the largest units, 3 and 4. The smaller units are inundated more frequently, leading to greater surface flows and plant growth. Surface flow may bring OM to the back of bioretention units 1 and 6, but this trend appears in bioretention units 2 and 5 as well, near the halfway point between the front and back. The largest units do not build up OM in this way. As designed, the larger cells with increased surface area available for drainage were expected to have reduced depth and duration of ponding (Stander et al. 2010). Brown et al. (2015) showed that the small and medium-sized units had greater periods of saturation for both front and back WCRs than the largest units, which were observed to reach saturation only once during Hurricane Irene and which had peak hourly rainfall intensity of 44.4 mm/h (O’Connor and Amin 2015). Other than two exceptional events, in an almost 2-year period the smallest units had a mean of 2.5 overtopping events or a 1.25 annual recurrence interval. This return period is less frequent than 2-, 10-, and 100-year design storm required for other stormwater controls (NJDEP 2021b) (even with the additional runoff on western side included). Brown and Hunt (2012) observed that undersized bioretention controls with large infiltration capacity were able to attain 90% annual water capture. It should be noted that only Tropical Storm Irene was a large enough event to discharge overflowing water from the bioretention units through a 460-mm (18-in.)-diameter pipe connected to the existing stormwater system (O’Connor and Amin 2015).

According to the United States Department of Agriculture, Indiangrass “grows best in deep, well-drained floodplain soils” (USDA 1991) while bitter panicum “can tolerate the harsh environment of the dune system …, storm surges, occasional inundation, high temperatures, low soil moisture and fertility … It is adapted to very dry, sterile sites and can flourish on fertile, well drained soils … does not perform well in shade” (USDA 2006). The USDA description of Indiangrass supports the observations of Indiangrass in the front near the inlet for runoff of all bioretention units and having the greatest surface coverage area per unit surface area in the smallest bioretention units (Fig. 7) due to more frequent inundation. The bitter panicum is observed dominating the back part of the larger bioretention units 3 and 4, where there are fewer plants and less plant cover—i.e., smaller observed basal area of bayberry and lower heights leading to less shade and more open space with drier conditions (Brown et al. 2015).

Measures of greater OM (%) and TP concentration in the smaller units coincided with greater plant density in these units. Ayer and Kangas (2018) noted that plant litter accumulates on the bioretention surface, contributing to the buildup of OM in the media. These measures of OM and phosphorus comparatively may also point to greater soil health and biological activity as microfauna (<0.1 mm in size, e.g., bacteria and protozoa) and mesofauna (0.1–2 mm in size, e.g., arthropods) break down OM (USDA n.a.) and are involved with nutrient cycling (e.g., phosphorus mineralization). Swift and Bignell (2001) detailed indirect methods to assess decomposer communities and soil biodiversity to assess microbial biomass using low-resolution chemical methods to determine OM and nutrients (including phosphorus Melich-3 extraction); however, they compared lysed versus unlysed control samples (lysis of cells breaks down bacteria cell membrane), which was not performed here.

While specific results are dependent on configuration, media composition, and runoff, generally, greater drainage-to-surface area ratios could increase performance of bioretention stormwater controls (i.e., plant and soil health and accumulation of pollutants). The implications for the bioretention units under study is that, with this configuration and materials, units with longer lengths from inlet (5.5:1 drainage-to-surface area ratio) have infrequent inundation and limited, if any, surface flow affecting plant growth (Brown et al. 2015); the cost of materials would also indicate a diminishing return on investment.

Results imply that the watershed-to-surface area ratios should be no less than 10:1 based on the function of middle-sized bioretention units (nominal 11:1 ratio). The smallest units (22:1) suggest that a 20:1 drainage-to-surface area ratio is functionally applicable and may be more appropriate for design. As noted in the Methods section, the middle unit with a ratio of drainage surface area of 11:1 most closely matched the NJDEP (2007) Stormwater Best Management Practice Manual, with suggested drainage-to-surface area ratio of 10.4 to 1, at the time of construction in 2009. Since that time, the NJDEP manual chapter for bioretention design has been updated twice. In the first chapter revision (NJDEP 2016), the example calculation for a bioretention unit for a nominal design storm had a 11.6:1 ratio of drainage-to-bioretention surface area, which once again closely matched the middle-sized bioretention units with 11:1 drainage-to-surface area ratio. However, the most recent update (NJDEP 2021a) has a 21.25:1 drainage-to-surface area ratio for an example of bioretention planters, which more closely matches the design of the smallest EEC bioretention unit with a drainage-to-surface area ratio of 22:1.

The confining layer of the native soil infiltrated slower than engineered planting media but still infiltrated quite quickly—i.e., greater than 95% double ring infiltrometer tests exceeded 60 cm/h as measured before bioretention construction (Stander et al. 2013), therefore drawdown within 72 h [upper limit (NJDEP 2021b)] was not a concern. WCRs indicated (Brown et al. 2015) full saturation was attained in the smallest units was more frequent than in the largest unit, but not every saturation event resulted in overtopping the sidewalls, as the WCR were measuring soil moisture below surface. The smallest units were more prone to overflow and water discharged over the HDPE sidewalls, but this was less than a 2-year recurrence interval barring exceptional events.

Of note, an upper limit on the drainage-to-surface area ratio was not determined during the course of this study. This is due to the side-by-side configuration for all six bioretention units, which used impervious HDPE paneling to prevent sidewall exfiltration to side wall depth of about 1 m (3.5 ft). The potential for additional subsurface drainage due to sidewall exfiltration for the media tested could increase the maximum functional watershed to surface area ratio. In studies where subsurface sidewall exfiltration was not restricted (Lee et al. 2015; Razzaghmanesh and Borst 2019b), sidewall exfiltration was shown to be the predominant component. Additionally, infiltrating during storms is not typically incorporated into most stormwater control design but is a component of the control processes observed (Brown and Hunt 2012; Brown et al. 2015). However, there could be other limitations and concerns to increasing infiltration in urban areas based on local geology [e.g., karst geology (Clar et al. 2004a), high groundwater table or aquitards], infrastructure (e.g., subways, utilities), or buildings (e.g., basement flooding), none of which were concerns for the bioretention units of the EEC.

The current study varied drainage-area to surface-area of stormwater controls, a surrogate for water availability, while planting media and plant species were the same in each unit. In a study of tree health in various stormwater controls (Tirpak et al. 2018), bioretention media composition was a critical factor, as was selection of species capable of tolerating the bioretention system and urban environments, while recommendations for further study included water availability between different media.

Overdesign of green infrastructure may lead to water stress of plants due to insufficient runoff getting to plants, resulting in plants relying on rainfall alone, as was observed with shrubs away from the inlet in the largest bioretention units (Brown et al. 2015) and bitter panicum’s dominance over other grasses. During a 4-year period, Tu et al. (2020) monitored tree trench systems and concluded that the planted trees were dependent on direct rainfall only; water levels in the infiltration beds did not supply water to trees planted in soil pits as planned due to overdesigned infiltration capacity, which left the infiltration component hydraulically isolated from the tree pits.

The bioretention units at the EEC were not actively maintained. This allowed bayberry to spread and take over habitat of other planted shrubs. Two grass species now dominate respective zones: Indiangrass in areas of more frequent inundation and bitter panicum in areas of infrequent inundation. Also, three unplanted trees were observed growing in the bioretention units. The accumulation of litter did not interfere with infiltration in EEC bioretention units. Recently, Wissler et al. (2020) noted that unmaintained stormwater dry detention basins that allowed vegetation (including trees) to grow provided water quality improvement such as sediment and phosphorus reduction, while increasing infiltration and not significantly reducing surface water storage. Clogging was not an observed problem for the EEC bioretention units over a 12-year period; however, the majority of the runoff (81.5%) was from the roof, which had approximately 10 times lower SSC than the asphalt parking surface, as mentioned earlier.

Deer were seen in the bioretention units and are suspected of the demise of the winterberry, as the remaining blueberry showed evidence of herbivory predation. One serviceberry was planted in each unit and all six were surveyed in December 2012, but results were not published as there was severe damage from antler rubbing by deer, which might skew any growth analysis between units. Only four serviceberry plants were surveyed in 2018, with serviceberry in bioretention units 5 and 6 having stunted vertical growth and fewer or damaged multistems due to continued deer damage.

Snow removal from the adjacent permeable pavement parking lot led to snow piles in the bioretention units impacting the bayberry, as a damaged (broken) trunk branch was observed in bioretention unit 2. As discussed earlier, stunted plants in bioretention unit 2 (2) and 5 (1) impacted repeating MANOVA results of the first study. The MANOVA statistical analysis for the 2012 bayberry survey indicated greater BA in the smaller units with a statistical power exceeding 0.90 (at alpha = 0.05), while the 2018 survey also indicated greater BA in the smaller units but with reduced observed statistical power of 0.72 (at alpha = 0.05); the 2012 statistical analysis at DBH included BA values of zero while the 2018 survey did not. The initial survey in 2012 occurred during a period of more open plant growth and included measuring DBH at 0.6 m and at the base as well, while the 2018 survey, which represented a period of overgrowth, was limited to bayberry plants attaining DBH and did not include damaged plants as evidenced by bayberries revealed in unit 2 and surveyed in 2020.

Construction of bioretention units not only excavates material but typically refills excavation with higher-cost engineered media. Bioretention areas are typically limited to a drainage area of 1 ha (2.5 acres) (NJDEP 2021a; WEF and ASCE/EWRI 2012); however, the EPA design manual (Clar et al. 2004b) recommended limiting drainage area to 0.4 ha (1 acre). While standalone stormwater controls may increase cost (WEF and ASCE/EWRI 2012), Clar et al. (2004b) indicated that a smaller-scale approach reduces the extent of hydrologic alterations for a site, and makes managing other impacts easier, more effective, and less costly. Planning and mobilizing for installing several smaller bioretention units for smaller drainage areas may reduce construction time and materials and time for MPT (i.e., road closures or lane restriction), as when built in municipal ROW. While bioretention is more suited for the smaller more frequent events, to address the increased frequencies of larger and more intense storms associated with climate change, economies of scale would favor larger flood control projects, as is case for wet and dry basins having lower unit cost as size increases (WEF and ASCE/EWRI 2012). “Ultimately, each region or municipality will need to identify its watershed and water resources protection goals and objectives and select the approach or combination of approaches that are appropriate to meet these goals” (Clar et al. 2004a).

Of the four items tracked for construction cost, engineered media comprised 74% of cost (cost of the gravel or other amendment and drainage pipes would reduce this, but generally this will remain a significant percentage of cost). The EEC bioretention units used a standard 90/10 sand and sphagnum peat moss mix. The current NJDEP specifications (NJDEP 2021a) specifies planting bed material to consist of 85% to 95% sand by volume, with no more than 25% fine sand, 15% or less silt, and 2%–5% clay with 3%–7% OM by weight. As noted by Tirpak et al. (2021), specified mixtures will vary by municipality or other authority. Ayers and Kangas (2018), in their study of 10 bioretention units, indicated that the observed accumulation of OM in the top 10 cm may improve pollutant removal performance and is similar to naturally occurring “A” soil horizon; they recommended confining OM amendments to the top layer where it is most available to plants and soil organisms. Both Ayer and Kangas (2018) and Tirpak et al. (2021) mention the potential for OM amendments to leach pollutants (i.e., nutrients), so Ayers and Kangas (2018) recommended minimizing OM amendments in lower layers in the bioretention media profile away from where needed for plant growth, to reduce nutrient leaching.

Conclusions

Multiple lines of evidence were evaluated to determine how the sizing of bioretention affects plantings within the boundaries due to increased water availability and soil forming in the root zone. As the length of the units designed here increased, there were diminishing returns on the investment in the construction materials and plants. Results of the 2018 plant survey confirm the results of the 2012 plant survey, that the plants do better in the smaller bioretention units, which have comparatively larger drainage-to-surface area ratios than the largest units.

The taller bayberries were mostly observed in the smaller bioretention units. Bayberry plants not subject to damage by piling up of snow were much bigger in the medium and smaller bioretention units than the largest bioretention units; and, as in the earlier 2012 bayberry survey (Brown et al. 2015), greater plant size was associated with being within the first 6 m from the inlet regardless of bioretention units.

OM is accumulating in the top layer of the media in the bioretention units with the smaller surface areas. Regression analysis indicates that OM (%) is accumulating in the media at depth of 0.15 m to a maximum at 3.2 m from the front (where water enters) for the smaller surface area bioretention units, while there is a lack of (or even negative) distance effect in the largest bioretention units. TP accumulation correlates with OM (%) and also has a pattern of maximizing at 3.2 m which is inside the back wall of the smallest unit (3.6 m). These results would suggest using the lower end of recommended OM amendments in local guidance, as properly sized bioretention units will accumulate OM in the upper layer of the media, which will support plant life and potentially remove more pollutants.

The current results support previous observations of the plant sizes in the bioretention units and observations of soil moisture. The largest units most likely continue to rely on surface rainfall, which limits plant health of shrubs and promotes drought-tolerant grasses, while OM and phosphorus in the upper root zone (0.15-m deep) has been increasing in smaller units and is most likely due more frequent inundation resulting in greater plant health, litter formation, and better pollutant removal.

More broadly, results indicate that oversized bioretention stormwater controls attain full volume control infrequently and, during smaller storms, do not disperse water throughout the surface area or root zone sufficiently to provide water for plantings; this indicates diminishing returns on the investment of materials used in the construction of such oversized units. Smaller units have lower material and associated construction costs, attain full control volumes more frequently, and during smaller storms attain a level of water moisture that provides for plantings.

Due to the lack of clogging, bioretention systems would appear appropriate for accepting larger quantities of roof runoff for infiltrating into the ground. If clogging of a bioretention system is a concern due to inputs from parking lot or street runoff, then solutions could include more frequent bioretention surface maintenance, inclusion of designed sediment capture at inlet in addition to bioretention surface area, or choosing alternate technology to manage stormwater.

Data Availability Statement

Some or all data, models, or code generated or used during the study are available in a repository online (DOI: 10.23719/1524656) in accordance with EPA ORD data retention policies.