Estimation of Rain Garden Field Hydraulic Conductivity Based on Spot Infiltration Tests

Abstract

Rain gardens are increasingly being used to control stormwater. Infiltration is a key component of volume control. Thus, determining the infiltration rate or field saturated hydraulic conductivity (K_sat) of rain gardens is critical to their continued successful operation. Designers and inspectors of rain gardens need to rapidly and efficiently determine the field K_sat. Prior research has found that single-ring infiltrometers (Princeton Hydro, Trenton, NJ, USA) can reliably be used to determine the infiltration rates of soils. The question often posed by designers and inspectors is “how many spot-infiltration tests are needed to sufficiently characterize the infiltration capacity of a rain garden?” Five rain gardens, varying in size from 62 to 429 m², were analyzed for this study. Three different spot infiltration methods were used: single-ring (Princeton Hydro, Trenton, NJ, USA) (least sophisticated and expensive), modified Philip–Dunne (Villanova University, Villanova, PA, USA), and SATURO (METER Group, Pullman, WA, USA) (most sophisticated and expensive). These rain gardens also had been instrumented to capture the recession rates during either natural or artificial ponding events. The linear portion of the recession curve obtained during ponding events was used to provide the rain-garden-wide K_sat. It was found that the geometric mean of six spot infiltration tests provided a reliable K_sat value similar to that found by the recession rate, which best represents the value of K_sat for the entire rain garden. This indicates that an inspector can reliably determine the infiltration capacity of a rain garden in less than a day.

1. Introduction

As urbanization and land development continue to generate more impervious surfaces within hydrologic regions, it is becoming increasingly necessary to combat the negative effects of human activity on stormwater runoff [1]. The hydrologic cycle is highly sensitive to increases in impervious area, and this sensitivity manifests itself in several ways, including but not limited to degradation of stormwater quality, reduction in groundwater recharge, and increases in runoff volume and flow rates (e.g., [2,3]). An increase in stormwater runoff flows, coupled with deteriorating quality of that water, is a particularly potent combination which takes its toll on the receiving waterways that serve as ultimate destinations for stormwater [4]. As a result of this trend, the need for responsible and context-sensitive stormwater management has never been greater.

In response to this challenge, green stormwater infrastructure (GSI) systems, also known as stormwater control measures (SCMs), are being widely implemented to manage stormwater from developed areas [5]. The use of SCMs is driven by regulations founded on decades of observations and research [6]. Green or vegetated SCMs allow for natural runoff remediation via infiltration through soil media, as well as groundwater recharge [7,8]. In many cases, further volume reduction potential may also be achieved through evapotranspiration (ET) [9,10]. Bioretention/bioinfiltration basins, or rain gardens, have been established as one of the best engineered solutions for land developers to address stormwater management in a way that optimizes the cost–benefit while responsibly managing environmental concerns [11].

Infiltration is the primary mechanism controlling runoff volume reduction in rain gardens. The rate at which water infiltrates into and then flows through the soil is a function of numerous factors, including the soil characteristics and hydraulic properties, water properties, rainfall intensity, and other soil-surface factors [12]. However, the field saturated hydraulic conductivity of the soil, K_sat, is commonly considered the most important parameter influencing the rain garden infiltration rate. Depending on the regulatory agency, SCMs are typically required to remove ponded water within 48 to 96 h. The official volume of stormwater managed that is credited to a rain garden design is, in most cases, determined through infiltration capacity and soil media storage. Thus, estimating the drain-down time and capacity through measurement of K_sat in a cost-effective manner is critical to design and predicting system performance [13].

When water enters a rain garden, the water ponds in the rain garden’s bowl. One can record the change in depth over time using pressure transducers, bubblers, or simply a staff gauge. The recession of water over time can then be plotted and the linear portion of the recession limb over the entire rain garden area provides a rain-garden-wide estimate of the field K_sat [14]. This value includes all variabilities caused by changes in soil type, compaction, and vegetation across the rain garden. Although measuring the recession rate across an entire rain garden provides the most accurate value for the field K_sat, most rain gardens are not instrumented to collect the necessary data to analyze recession rates. Instead, spot-infiltration-testing methods are more commonly used to determine field K_sat values for SCMs. At the laboratory scale, the variability of K_sat is largely controlled by pore size and tortuosity. At the field scale, the variability of K_sat is dependent upon pore size and tortuosity as well as macro-scale variabilities such as compaction, changes in soil type that could occur as a result of construction or from the deposition of particles during storm events, and the distribution of vegetation and mulch [15,16]. Due to the high spatial variability of K_sat throughout a rain garden, multiple spot infiltration tests are typically required to determine a representative value of K_sat for the basin [17]. To accommodate the need for testing multiple locations, small-diameter spot-infiltration-testing methods have been further developed over the last two decades to make field testing more efficient and reduce water volumes required for testing. Three such methods are the single-ring infiltrometer (Princeton Hydro, Trenton, NJ, USA) [15], modified-Philip Dunne (MPD) infiltrometer (Villanova University, Villanova, PA, USA) [18,19], and SATURO dual head infiltrometer (METER Group, Pullman, WA, USA).

As rain gardens become increasingly pervasive in the developed landscape, these field measurement tools will play a critical role in allowing engineers to understand the function and performance of these sites. While using these spot infiltration tests can help engineers account for spatial variability of soil at a site, this raises an important question: how many tests are required to represent the true, or at least the acceptable, infiltration characteristics of a site? Although previous research studies have included conclusions regarding an acceptable minimum number of test locations [20], the general recommendations in the literature vary widely and are generally specific to the site and test method.

Thus, to further address this question, five different rain gardens in the Philadelphia, PA region were tested using three different small-diameter infiltration test devices, and the data obtained were compared to observed ponding recession rates to contextualize the results in terms of actual rain garden performance. The geometric mean of multiple infiltration tests was used in the comparison. A combinatorial, probability-based analysis was employed using a geometric mean average for the infiltration test results, and statistical metrics were applied to the data.

2. Materials and Methods

Spatial variability of soil media is frequently discovered even over relatively small areas [21]. Such factors as macropores, root structures, localized compaction, and unexpected soil mixtures can cause this variability. This inherent spatial variability means that multiple spot tests are needed to accurately represent the infiltration characteristics of any feature. One must then consider the reliability of each of those tests and how to consolidate or average those values [22]. While numerous methods for spatial analysis exist [23,24], previous studies have shown that statistical analyses utilizing the geometric mean of multiple infiltration test values can be used to represent the total site with reasonable accuracy and conservatism [21,25]. The study described herein considered three different spot infiltration test methods in a statistical analysis of five different rain gardens to determine the appropriate number of infiltration tests required to reasonably represent performance.

2.1. Recession Rates

The field K_sat was approximated using the linear portion of the ponded recession rate of a rain garden [14]. When the bowl of a rain garden is full, either after a rain event or after filling the rain garden bowl with water for a simulated runoff test, a bubbler or pressure transducer is used to measure the depth of the water over time. In this method, the ponded recession rate is measured, and the linear portion of the curve is delineated. This linear portion is when the soil is assumed to be nearly saturated, and thus, the hydraulic conductivity is dominated by gravimetric forces rather than the tension. This value, K_sat-pond, approximates the field of the entire rain garden, accounting for all variability since the entire surface of the rain garden is utilized.

2.2. Spot Infiltration Tests

Three different spot infiltration test methods were used to estimate the field hydraulic conductivity, K_sat-spot, throughout this work: single-ring infiltrometer (Princeton Hydro, Trenton, NJ, USA), modified Philip–Dunne infiltrometer (Villanova University, Villanova, PA, USA), and the SATURO dual head infiltrometer (METER Group, Pullman, WA, USA) (Figure 1). At each rain garden studied, a series of spot infiltration measurements were obtained at various locations avoiding vegetation or stones that would prevent the ability of the device to be inserted into the earth. The single-ring infiltrometer (Princeton Hydro, Trenton, NJ, USA) and modified Philip–Dunne (Villanova University, Villanova, PA, USA) values were corrected for temperature, since these tests are performed in the field and the viscosity of water depends on temperature. The following temperature correction is applied to the field measured K_sat [26,27]:

$$K_{sat, C}=K_{sat, T} ({\mu }_T/{\mu }_C)$$

(1)

where K_sat,C is the hydraulic conductivity corrected to the desired temperature, K_sat,T is the measured hydraulic conductivity, μ_C is the viscosity of water at the desired corrected temperature, and μ_T is the viscosity of water at the measured temperature. The SATURO device (METER Group, Pullman, WA, USA) automatically corrects for temperature, so further correction was not necessary.

2.2.1. Single-Ring Infiltrometer

The single-ring infiltrometer (Figure 1a) used in this study was manufactured by Princeton Hydro, Trenton, NJ, USA and designed for adequate precision within the range of K_sat of 7.1 × 10⁻⁷ m/s to 7.1 × 10⁻⁵ m/s. This range is acceptable for the soils typically encountered in rain gardens. This was the least sophisticated and least expensive device used in this study. The single-ring infiltrometer (Princeton Hydro, Trenton, NJ, USA) is a steel ring 15.2 cm in diameter and 15.2 cm long with a beveled cutting edge [15]. It is driven down halfway into the soil, with 7.6 cm exposed and 7.6 cm under the ground surface. The device measures infiltration and correlates a field K_sat-spot value through a constant-head approach of filling the ring to the top with water and allowing a drop of 2.5 cm before refilling and repeating this procedure.

During a single-ring test, a filter fabric is placed in the ring prior to adding water to prevent water from excessively disturbing soil on the infiltration surface. This filter fabric is removed once the single-ring infiltrometer is full. After a water level drop of 2.5 cm, the time is recorded with a stopwatch, and the ring is filled again. The ring is continuously filled to ensure negligible head difference between readings, and this procedure is repeated until the 2.5 cm drop time stabilizes, which signals saturated conditions in the soil. It is at this point that the infiltration rate can be correlated to the field K_sat-spot [15].

2.2.2. Modified Philip–Dunne Infiltrometer

The modified Philip–Dunne infiltrometer (Villanova University, Villanova, PA, USA) (Figure 1b) is a 10 cm diameter tube with a steel driving ring that is inserted to a minimum depth of 5 cm [19]. The tube is approximately 61 cm tall. The calculable range of K_sat-spot using the modified Philip–Dunne (Villanova University, Villanova, PA, USA) is around 1 × 10⁻⁸ m/s to 1 × 10⁻⁴ m/s. This technique utilizes a falling-head approach, as the tube is filled to the 61 cm mark and topped with an evaporation-preventing cap with a small hole to relieve pressure and allow the water to move downward [19,29]. The time required for half of the water column to infiltrate is recorded along with the time required to empty the tube, and the ratio of these times is used to calculate a field K_sat-spot value as described in ASTM D8152 [18] and summarized below.

The K_sat-spot value is measured with the modifed Philip–Dunne (Villanova University, Villanova, PA, USA) by first determining the dimensionless parametric value, τ_max, as follows [19]:

$${\mathsf{\tau }}_{max }=0.73\left({\displaystyle \frac{t_{max}}{t_{med}}}\right)-1.1258;\mathrm{w}\mathrm{h}\mathrm{e}\mathrm{r}\mathrm{e} {\displaystyle \frac{t_{max}}{t_{med}}}<5.4$$

(2)

where t_max is the time required for the water to completely infiltrate from the tube, and t_med is the time required for the half of the water column to infiltrate. If the value of t_max/t_med exceeds 5.4, the test is invalid, and the hydraulic conductivity value calculated with this τ_max value is not accurate. If τ_max is less than 5.4, K_sat is calculated as follows:

$$K_{sat}={\displaystyle \frac{{\mathsf{\pi }}^2\times r\times {\mathsf{\tau }}_{max}}{16\times t_{max}}}$$

(3)

where r is the radius of the infiltrometer tube (5 cm), τ_max is the parametric value, t_max is the time required for the water to completely infiltrate from the tube in seconds, and K_sat is the saturated hydraulic conductivity in cm/s [19].

2.2.3. SATURO Infiltrometer

The SATURO device (Figure 1c) manufactured by Meter Group, Pullman, WA, USA is a fully automated instrument for measuring field K_sat-spot. Using a multi-pressure head analysis to simplify corrections for three-dimensional flow from a single-ring infiltrometer (Princeton Hydro, Trenton, NJ, USA), the SATURO device (METER Group, Pullman, WA, USA) can make quick measurements of hydraulic conductivity without the need for postprocessing. It is the most expensive and most sophisticated device used in this study.

The device setup consists of a central control unit, a 5 cm insertion ring, infiltrometer head, collapsible water tank, and three tubes for air and water circulation through the control unit and infiltrometer. Before a test is begun, the insertion ring is driven into the soil by placing a metal plate on top of it and using a mallet to drive it to full depth. The infiltrometer head is then attached, and the sensor cable is connected to the control unit, while the water tank is connected via a tube to the control unit where flow is regulated and pumped out to the infiltrometer. The device can measure infiltration accurately in the range 1 × 10⁻⁸ m/s to 3 × 10⁻⁴ m/s.

The SATURO’s (METER Group, Pullman, WA, USA) on-board computer incorporates a robust theoretical approach to estimate hydraulic conductivity. In most single-ring infiltrometers, there may exist a tendency to overestimate K_sat due to lateral divergence of flow from capillarity of dry soils and ponding in the ring. The SATURO (METER Group, Pullman, WA, USA) corrects for this by incorporating the two-ponding head approach [30]. This approach accounts for soil capillarity, depth of ponding, ring radius, and depth of insertion ring to calculate hydraulic conductivity.

2.3. Conditional Probability Analysis

The data analysis was approached under the assumption that the “answer” (ponded recession rate) was unknown, and each infiltration rate value was obtained one at a time. Thus, if 10 infiltration tests were performed on a site (here referred to as the unique values Test 1, Test 2, Test 3, and so on, or Tests 1–10 for the whole set), the procedure begins by assuming only one test was performed, and then Tests 1–10 were evaluated in relation to the measured ponded recession rate (K_sat-pond). Then, it was assumed that two tests had been performed, and the geometric mean of those two tests was compared to K_sat-pond. Every combination of two tests from the field of Tests 1–10, such that the geometric mean of (Test 1, Test 2), (Test 1, Test 3), (Test 1, Test 4), and so on, were evaluated until every combination was included in the analysis. This process was repeated for three tests, then four, and so on until every combination of tests was averaged via the geometric mean and plotted on a graph. The total number of combinations for each stage of analysis is determined by Equation (4):

$$C\left(n,r\right)={\displaystyle \frac{n!}{r!\left(n-r\right)!}}$$

(4)

where C = the total number of combinations for the analysis; n = the total number of tests performed at the site; and r = the number of tests being considered at the stage of analysis (r ≤ n).

It is apparent from the mathematical structure of the analytical procedure that the geometric mean values for these tests converge to a smaller range of values as the r value increases until r = n, where the full dataset was used. Using this feature of the analysis in combination with knowing the actual ponded recession rate of the site being studied allows for the determination of an acceptable level of deviation for average infiltration test values against the measured ponded recession. This can then be used to determine the minimum tests required at a site for an adequate convergence.

To maintain a level of experimental consistency with this procedure, the infiltration tests were performed in the spring and summer so that air temperature (and thus infiltrating water viscosity) and vegetation maturity would match in both testing and measured recession. While it was not required that this specifically be a warm-season procedure, most tests occurred over the spring–summer period for ease of outdoor access and abundance of recent rainfall that allowed for ponding conditions.

2.4. Site Descriptions

Five rain gardens were included in this analysis (

Table 1. Summary of rain garden characteristics.

Site	Surface Area (m2)	Hydraulic Loading Ratio	Size of Design Storm (cm)	USCS Classification of Soil	USDA Classification of Soil
VURG1	235	10:1	2.54	SM Silty sand	Loamy sand
VURG2	92	0.87:1	2.54	SP Poorly graded sand	Sand
USRG	80	8.6:1	3.61	SM Silty sand	Sandy loam
DSRG	62	13:1	4.24	SM Silty sand	Sandy loam
SMP A	429	10:1	2.54	SW-SM Well graded sand with silt	Loamy sand/sandy loam

). All the rain gardens had a ponded recession rate dataset to allow a comparison to the spot infiltration test measurements. The rain gardens varied in size and hydraulic loading ratio, which is a comparison of the directly connected impervious area to the area of the rain garden.

2.4.1. Villanova University Rain Garden One (VURG1)

Villanova University Rain Garden One (VURG1) is located on Villanova University’s campus (Figure 2). It was constructed in 2001 as a retrofit of an existing traffic island and has been monitored for research purposes since 2003 (e.g., [7,8,14]). During construction, the site was excavated to a depth of 1.8 m, and the existing soil was mixed with sand to an engineered depth of 1.2 m. Before and after mixing, site soils were classified as USDA sandy loam.

The site has been instrumented with several sensors over the years to measure performance, including a pressure transducer (Campbell Scientific CS451, Logan, UT, USA) and bubbler (OTT HydroMET CBS, Kempton, Germany) to measure the water level in the rain garden bowl. These devices were used to determine the SCM field ponded recession value to which the spot infiltration test results would be compared. The recession rate used to calculate K_sat-pond in this paper was determined from the bubbler.

2.4.2. Villanova University Rain Garden Two (VURG2)

Villanova University Rain Garden Two (VURG2) is located near the main entrance to Fedigan Hall, a student residence hall on Villanova’s campus (Figure 3a). The rain garden was constructed in 2009 and was able to be located right against the front of the building because Fedigan Hall does not have a basement, negating seepage concerns from the rain garden’s presence. The rain garden receives runoff from the building’s roof.

VURG2 was designed with an engineered soil media of USDA sand to a depth of 0.46 m before interfacing with uncompacted native soils, which are silty sands (Figure 3). The runoff can infiltrate directly into the native soil and recharge the groundwater onsite. VURG2 was instrumented with a pressure transducer (Campbell Scientific CS451, Logan, UT, USA) to measure ponding levels in the rain garden pond during storm events so that K_sat-pond could be calculated. Compared to VURG1, VURG2 has an extremely conservative loading ratio and is significantly smaller in terms of surface area (Table 1).

2.4.3. Upstream and Downstream Rain Gardens at the Philadelphia Zoo (USRG and DSRG)

The Philadelphia Zoo rain gardens comprise two rain gardens connected by a grass swale, resulting in a “treatment train” type of stormwater management system (Figure 4 and Figure 5) [32]. This system receives inflow from the impervious pavement of Girard Avenue between the 3800 and 3900 blocks in Philadelphia, PA. Influent flow enters the upstream rain garden and ponds to 17 cm before overflowing through a V-notch weir into the grass swale. The grass swale can pond up to 13 cm before overflowing through a second V-notch weir into the downstream rain garden, where it may pond to 38 cm before overflowing into a rock bed.

USRG and DSRG both consist of 61 cm of engineered soil media sandy loam underlain with native soil (USCS silty sands). There are no liners or underdrains in the system, so infiltration to groundwater is promoted in both rain gardens and the swale. Ponding levels during storm events are measured using pressure transducers. Both HOBO (Onset, Bourne, MA, USA) and CS451 (Campbell Scientific, Logan, UT, USA) pressure transducers were used to measure the depth of water over time after the bowl was full to calculate K_sat-pond.

2.4.4. PennDOT I-95 Stormwater Management Practice A (SMP A)

The Pennsylvania Department of Transportation (PennDOT) I-95 rain garden site included in this study (SMP A) is located along Richmond Street in Philadelphia, PA (Figure 6). The site is a vegetated bioswale installed alongside the highway to manage and treat the stormwater runoff from the highway deck. The bioswale was designed and constructed with an underdrain, which had been capped to facilitate infiltration. As a result, the system is currently performing as a rain garden.

SMP A is approximately 146 m long and 8 m wide with three culvert inlets and two outlets that overflow into Philadelphia’s combined sewer system. Its design surface area is 429 m². The rain garden was constructed with a 61 cm thick layer of amended soil (sandy loam/loamy sand) as the engineered media for infiltration and storage with side slopes of 2 to 3 horizontal to 1 vertical. SMP A was instrumented with pressure transducers (Campbell Scientific CS451, Logan, UT, USA) in the rain garden bowl to measure ponding depth during storm events so that K_sat-pond could be calculated [34].

3. Results

At each of the five locations, between six and ten spot infiltration tests were performed (

Table 2. Summary of data collected.

Site	Number of Spot Infiltration Tests	Season and Year of Testing	Testing Equipment Used	Geometric Mean from All Spot Tests, Ksat-spot (m/s)	Ponded Recession Rate, Ksat-pond (m/s)
VURG1	7	Spring 2016	Single-Ring	2.63 × 10−6	1.90 × 10−6
	6	Summer 2019	SATURO	2.83 × 10−5
VURG2	10	Spring 2017	MPD	1.06 × 10−6	1.06 × 10−6
USRG	10	Fall 2017	SATURO	2.48 × 10−5	4.72 × 10−5
DSRG	8	Fall 2017	SATURO	1.87 × 10−5	4.22 × 10−5
SMP A	8	Summer 2018	SATURO	1.06 × 10−5	1.46 × 10−5

). At one location, VURG1, both single-ring (Princeton Hydro, Trenton, NJ, USA) and SATURO (METER Group, Pullman, WA, USA) infiltration tests were performed. Every attempt was made to evenly distribute the locations of the testing; however, vegetation, rocks, and pebbles needed to be avoided. Since the rain garden soils were designed to allow the ponded water to be emptied within a similar time frame, it is to be expected that the values are similar across all sites.

The conditional probability approach is best represented graphically to illustrate the data trends that develop upon incorporating additional combinations into the analysis. To demonstrate the process, it will be beneficial to follow the development of a single site’s conditional probability graph from infiltration test values to graph generation. This approach will then be applied to all other data sets.

The seven spot values obtained from the single-ring device (Princeton Hydro, Trenton, NJ, USA) were organized into classes ranging from “1 Test” to “7 Tests”, and the geometric mean values for all combinations of these tests were generated. These geometric mean values obtained for every possible combination of sample tests were then plotted on a graph and compared to the ponded recession rate, which represents the K_sat for the whole rain garden (Figure 7). As expected, more spot infiltration tests better capture the inherent variability of the rain garden soils. This is apparent in the decrease in the spread of values as more combinations are considered (e.g., decreasing coefficient of variation). In this case, performing seven single-ring infiltrometer (Princeton Hydro, Trenton, NJ, USA) tests resulted in a value of K_sat-spot of 2.63 × 10⁻⁶ m/s, which is close to the overall K_sat-pond value of 1.90 × 10⁻⁶ m/s. Six spot values were obtained using the SATURO (METER Group, Pullman, WA, USA) at VURG1 (Figure 8). The SATURO (METER Group, Pullman, WA, USA) slightly over predicted K_sat; the geometric mean of six spot tests was K_sat-spot of 2.83 × 10⁻⁵ m/s.

Ten spot infiltration tests were performed on VURG2 using the modified Philip–Dunne infiltrometer (Villanova University, Villanova, PA, USA) (Figure 9). This was the only site that utilized this device. The geometric mean of all ten infiltration tests, K_sat-spot, was identical to K_sat-pond for this site (1.06 × 10⁻⁶ m/s).

Ten spot infiltration tests were performed on the USRG using the SATURO (METER Group, Pullman, WA, USA) infiltrometer (Figure 10). The spot infiltration tests slightly underpredicted the overall infiltration rates at this site, with the geometric mean of all ten infiltration tests, K_sat-spot = 2.48 × 10⁻⁵ m/s as compared to K_sat-pond for this site, which was 4.72 × 10⁻⁵ m/s.

Eight spot infiltration tests were performed on the DSRG using the SATURO infiltrometer (METER Group, Pullman, WA, USA) (Figure 11). The spot infiltration tests also underpredicted the overall infiltration rates at this site, with the geometric mean of all eight infiltration tests, K_sat-spot = 1.87 × 10⁻⁵ m/s, as compared to K_sat-pond for this site, which was 4.22 × 10⁻⁵ m/s.

Eight spot infiltration tests were performed on SMP A using the SATURO infiltrometer (METER Group, Pullman, WA, USA) (Figure 12). The spot infiltration tests slightly underpredicted the overall infiltration rates at this site, with the geometric mean of all eight infiltration tests, K_sat-spot = 1.06 × 10⁻⁵ m/s, as compared to K_sat-pond for this site, which was 1.46 × 10⁻⁵ m/s.

4. Discussion

The spot infiltration values from the single-ring (Princeton Hydro, Trenton, NJ, USA) (Figure 7) and modified Philip–Dunne infiltrometers (Villanova University, Villanova, PA, USA) (Figure 9) converge close to or on the K_sat-ponded value. The results from the SATURO (METER Group, Pullman, WA, USA) did not consistently over- or under-predict the K_sat-pond at three of the sites (Figure 8, Figure 10 and Figure 11), while at one of the sites (Figure 12), the SATURO (METER Group, Pullman, WA, USA) very closely predicted the K_sat-pond. Considering that the SATURO infiltrometer (METER Group, Pullman, WA, USA) was used at four sites, and the single-ring (Princeton Hydro, Trenton, NJ, USA) and modified Philip–Dunne infiltrometers (Villanova University, Villanova, PA, USA) were only used at one site each, it may be a coincidence that the single-ring (Princeton Hydro, Trenton, NJ, USA) and modified Philip–Dunne infiltrometers (Villanova University, Villanova, PA, USA) converge on the K_sat-pond value for the sites at which they were used. All methods produced an acceptable level of accuracy (Table 2) when all spot infiltration tests were considered.

Additional spot infiltration tests result in a more consistent estimate of the infiltration rate of a rain garden (Figure 7, Figure 8, Figure 9, Figure 10, Figure 11 and Figure 12). This consistency is described by a decreasing coefficient of variation for each site examined (Figure 13). In any soil testing, variability is to be expected, so a designer or inspector will need to decide how much uncertainty they are comfortable accepting. If a coefficient of variation of 0.40 is acceptable to a designer or inspector, all sites and methods achieved this value after six spot tests regardless of method. With the exception of SMP A, for a given number of tests, the SATURO infiltrometer (METER Group, Pullman, WA, USA) yielded a lower coefficient of variation than the single-ring (Princeton Hydro, Trenton, NJ, USA) or modified Philip–Dunne infiltrometers (Villanova University, Villanova, PA, USA).

5. Conclusions

Between six and ten spot infiltration tests were performed at five different rain gardens in the Philadelphia, PA area. Three different methods of spot infiltration testing were utilized: single-ring (Princeton Hydro, Trenton, NJ, USA) (least sophisticated and expensive), modified Philip–Dunne (Villanova University, Villanova, PA, USA), and SATURO (METER Group, Pullman, WA, USA) (most sophisticated and expensive). Spot infiltration tests are commonly performed as part of the design and inspection process for rain gardens. Designers and inspectors are often left wondering how many spot infiltration tests are enough to adequately describe the overall infiltration capacity of a rain garden. The number of tests performed is often limited by how much time a technician can spend at each site. While more is better, it is not often practical. If designers and inspectors are willing to accept a coefficient of variability of 0.4, then six spot infiltration tests are sufficient to characterize a site. The results of this study can provide designers and engineers with a rationale for selecting the number of spot infiltration tests that should be performed to quantify the infiltration capacity of a rain garden.