Evaluation of Infiltration Swale Media Using Small-Scale Testing Techniques and Its SWMM Modeling Considerations

Abstract

Impervious surfaces reduce natural infiltration, leading to increased runoff, erosion, and pollutant transport. The Alabama Department of Transportation (ALDOT) relies on implementing infiltration swales, a linear bioretention-based practice, along roadside drainage channels to reduce surface runoff. This study developed and constructed modified permeameters and infiltrometers to evaluate and optimize media used to construct infiltration swales. The average measured falling head infiltration rate of sandy topsoil used in the media matrix was 0.63 ft/day (0.19 m/day). A series of amended topsoil mixtures were tested to improve the infiltration rate of the media. In particular, the mixture of 80% topsoil and 20% pine bark fines (by weight) significantly improved the infiltration rates of the swale media. Through iterative testing, the F3 design with 6 in. (15.2 cm) mixture and 10 in. (25.4 cm) sand achieved up to 13.73 ft/day (4.18 m/day) of infiltration rate under constant head, far surpassing the infiltration rate of the current ALDOT design. SWMM bioretention cell models were developed to understand the swale infiltration process and revealed that the infiltration rates obtained from column tests were the saturated hydraulic conductivities of the soil layer when there was no other restriction on vertical flow. The simulated swale hydrological performance depends not only on variations in soil conductivity but also on other swale characteristics under field conditions. Findings from this research can be used to enhance the performance of infiltration-based stormwater practices.

1. Introduction

With rapid development, urbanization, and a significant increase in population, the expansion and modernization of road infrastructure constitute a critical foundation for national development and economic growth in the United States. Impervious roadway infrastructures have been shown to adversely affect the health of urban streams [1], leading to increased runoff volumes and higher peak flows that severely erode soil [2], as well as increasing pollutant loading into receiving water bodies [3].

Stormwater runoff is the excess rainfall that is not absorbed into the ground or lost through evapotranspiration and instead travels across land surfaces. This runoff volume is influenced by a variety of hydrologic losses, including soil infiltration, plant uptake, surface storage, and evaporation, which reduce the effective precipitation contributing to flow [4]. In developed areas, the widespread presence of impervious surfaces—such as pavements and rooftops—amplifies runoff volumes by limiting infiltration. This increase in surface flow can intensify flooding, accelerate erosion, diminish groundwater replenishment, and negatively affect downstream aquatic habitats [5].

Low-impact development (LID) practices, such as bioswales [6], are stormwater management methods used to mimic natural rainfall-runoff processes, like increasing the infiltration rate of the LID area to minimize and slow down surface runoff [7]. LID benefits include effective stormwater management, improved water quality, and groundwater recharge while enhancing urban green spaces to reduce the heat island effect. Implementing LID practices also lowers infrastructure costs, increases property value, and promotes public health and environmental sustainability [7]. There is a difference from conventional drainage systems that rely on rapid conveyance. LID focuses more on mimicking natural hydrological processes through localized infiltration, retention, and treatment of runoff [6]. In LID practices, topsoil as a part of the soil media mixture is commonly used for its significant role in regulating infiltration processes and improving surface water management efficiency [8,9]. Compared to natural soils, engineered media exhibit superior water retention capacity, highlighting their potential for multifunctional stormwater management applications [9].

In much of the existing literature—both academic and practical design guidance—swales are commonly identified as grass-lined channels. These grass swales are typically engineered conveyance features, constructed to defined geometric standards and stabilized with vegetation suited for managing surface runoff [4]. Dry swales or grass drainage swales are common roadside grass-covered channels that convey runoff. Wet swales are designed to detain stormwater runoff for a period of time. Bioswales may be planted with native vegetation in addition to grass. One of the primary purposes of grass swales is to reduce channel erosion and improve stormwater quality by capturing sediments, reducing total suspended sediment (TSS) and other pollutants, particularly in areas where concentrated runoff may cause erosion damage. While grass swales can promote some infiltration into the underlying native soils, their primary function is to convey surface runoff. Unlike engineered systems, traditional grass swales generally lack designed soil media layers beneath the vegetated surface, which limits their infiltration performance.

The Alabama Department of Transportation (ALDOT) is required by law to reduce runoff during and after roadway construction [10]. ALDOT’s Guideline for Operation (GFO 3-73) [10] states that designers should attempt to provide features and practices that cause post-development hydrology to mimic pre-development hydrology of the site to the maximum extent practicable for all small, frequent rain events, working within the constraints of the project, at all locations of discharge. To meet these goals, ALDOT has developed an infiltration swale design, which is a unique LID practice that is suitable for implementation along the roadside environment (Figure 1). The infiltration swales used by ALDOT, often categorized as a form of bioretention practice, differ from standard grass swales by incorporating an engineered media matrix (soil, sand, and gravel layers (Figure 1a)) beneath the vegetated channel bed. These systems are typically designed with features such as earthen check dams to retain and slow the flow of stormwater along the swale, enhancing both detention and infiltration functions (Figure 1b). The infiltration swale with engineered media linearly covers a part of the roadside drainage channel, and the remaining part of the channel is a conventional grass swale. By conveying runoff through the engineered subsurface media, water infiltrates into the subsoil profile, supporting groundwater recharge. Infiltration swales increase in roughness over traditional concrete channels to slow water velocity and help the contributing drainage basin mimic natural hydrologic cycles by facilitating processes such as infiltration, evapotranspiration, and surface runoff regulation. These characteristics make them an effective LID practice [11]. ALDOT infiltration swales have layered media (Figure 1), but many other bioretention cells have a homogenous soil mix as media, e.g., a mix of 75% loamy sand and 25% compost [12]. ALDOT’s infiltration swales along roadways are similar to infiltration swales built along a two-lane road in Munich, Germany, and field monitored by Eben et al. from 2022 to 2024 [13]. Their infiltration swales have a mineral mulch layer (7 cm), a soil media layer (30 cm), and a sand layer (20 cm), but do not have a gravel layer. The sand layer contains a seepage water collection system (a gravel-filled stainless-steel bin) to measure infiltrated runoff volume after pumping out. The surface layer has various types of vegetation [13], while the ALDOT infiltration swales are limited to grass sod. Infiltration swales in Germany have demonstrated strong potential for delivering multiple hydrological and ecological benefits, including peak flow attenuation, heavy metal retention, and habitat protection under compact urban settings [13].

The current design of an ALDOT infiltration swale includes 10–18 in. (25–46 cm) of sandy topsoil, underlain by at least 12 in. (30 cm) of sand, and at least 8 in. (20 cm) of washed #57 stone that is surrounded by separation geotextile so that sand cannot migrate into the voids among stones. An upper layer of topsoil on the surface of the channel is designed to support the growth of turf grass. The infiltration swale will also require regular maintenance by machine mowing, weed control, and debris removal. The maximum spacing between check dams is 100 ft (30 m) (Figure 1b), depending on the longitudinal slope of the grass channel. Some preliminary observations of the performance of infiltration swales installed at multiple ALDOT project sites have demonstrated varying levels of success. In some instances, these infiltration swales have been able to achieve adequate levels of infiltration. In a few cases, these facilities ponded water for prolonged periods, exceeding the 48 h design drawdown time following a rainfall event. There can be various explanations for this outcome that include (1) variations in material properties (porosity and infiltration rate) of the engineered media layers within swales, (2) unintended compaction of soil layers during construction, (3) natural or unintended mechanical compaction post construction, and/or (4) hydrological interactions created by shallow groundwater, etc. Therefore, a study was performed through the Auburn University Highway Research Center to understand the runoff reduction performance of the existing design of the ALDOT infiltration swales and to identify and propose an enhanced engineered media design promoting infiltration performance. First, we need to identify the limiting factor of the existing design that results in a large drawdown time or a slow infiltration rate. Normal bioretention cells are not placed in the drainage channels; therefore, they normally do not have a topsoil layer covered by grass as a part of the engineered media. Designers lack an understanding of the infiltration capacity of the topsoil layer and its impact on the overall infiltration capacity of the engineered media. This is the research gap in the field of LID designs that this study tried to address.

Small-scale column tests were conducted to first understand/quantify the infiltration capacity of the existing ALDOT infiltration swale design. Rapid small- or laboratory-scale column tests provide an efficient substitute for costly, time-intensive pilot-scale trials traditionally used to design full-scale experiments because they deliver reliable performance predictions, require only small volumes of water, and can be completed in a short time frame [14].

The constant-head test, as outlined in ASTM D2434-68 [15], is a standard laboratory procedure used to determine the hydraulic conductivity of granular soils under saturated conditions. This method involves maintaining a steady hydraulic head across a soil specimen in a permeameter to measure the ease with which water flows through the material [16]. Hydraulic conductivity—also referred to as the coefficient of permeability—describes a soil’s ability to transmit fluids and is typically measured using either constant-head or falling-head tests, depending on the soil type and conditions. These tests can be performed on both undisturbed and reconstituted soil samples to evaluate fluid-flow characteristics [17]. To determine the hydraulic conductivity of granular soils, a standard laboratory process is followed: the soil is prepared and placed within a permeameter, after which measurements are taken for several parameters—such as discharge flow rate, spacing between manometer ports, specimen cross-sectional area, test duration, and hydraulic head difference [18]. According to Darcy’s Law, the volumetric flow rate is directly proportional to the specimen’s hydraulic gradient, hydraulic conductivity, and cross-sectional area [19].

Numerical modeling of a bioretention cell could provide a better understanding of the swale infiltration mechanics and column tests. Various computational methods/models have been developed and applied to study LIDs as stormwater control measures [20]. The Storm Water Management Model (SWMM) [21], developed by the United States Environmental Protection Agency (USEPA), is widely used to simulate hydrological processes under various conditions [20,22,23]. The SWMM platform has been adapted to include modular components that simulate various LID practices, including bioretention systems [24], enabling the simulation and evaluation of ALDOT infiltration swale performance under various design and rainfall scenarios [21]. SWMM’s bioretention LID module includes three horizontal layers along its depth: a surface layer with berms to pond/hold/produce surface runoff, a soil layer to allow runoff to be retained or infiltrate downward, and a storage layer as a buffer to regulate the swale’s infiltration capacity. These three layers require various model input parameters that can be either measured, estimated, or calibrated. A comparative analysis between uncalibrated and calibrated watershed simulations can provide valuable insights into the predictive capabilities and inherent limitations of SWMM in modeling LID practices [25].

The primary objective of this study is to evaluate the current infiltration swale media design and to propose/optimize and test improved media designs using small-scale column testing techniques. Basic materials (topsoil, sand, mixtures, and stones) of the media were first tested using a series of permeability and infiltration rate column tests to identify the limiting factor of runoff control performance of the infiltration swale designs [26]. The efficiency of the infiltration swale media was then assessed through constant-head and falling-head infiltration rate tests. The SWMM model with LID was then developed and used to understand the infiltration swale’s internal mechanisms for reducing surface runoff. The study explored how measured permeability or infiltration rates of topsoil, sand, and mixture from the column tests can be utilized to appropriately specify the SWMM LID parameters to predict/simulate the ALDOT infiltration swale’s runoff-reduction performance under design and historical long-term rainfall.

2. Materials and Methods

To determine the permeability or hydraulic conductivity of a soil sample, a standard permeameter [15] is required. ALDOT infiltration swales typically extend 4 to 5 ft (1.2 to 1.5 m) below the grass channel bottom. To understand its infiltration performance, a standard permeameter could not be used; therefore, a series of columns were designed and constructed to perform tests that measure permeability or infiltration rates of single uniform media and composite or layered engineered media (Figure 1a). A comprehensive description of the apparatus construction, testing protocols, and methodological framework employed in investigating the infiltration swale media is provided below. The primary objective of this portion of the study was to conduct a rigorous evaluation of the permeability and infiltration rates of diverse infiltration media configurations. This also involved comprehensively examining material properties, including particle size gradation distribution, density, porosity, and their response to consolidation and compaction.

2.1. Apparatus Design and Construction

The modified permeameters (Figure 2) and clear infiltrometers (Figure 3) were designed to facilitate the execution of permeability constant-head tests and falling and constant infiltration rate tests on a small-scale basis. A total of 18 permeameters were constructed and supported by a 10 × 1.2 × 4 ft (3.0 × 0.4 × 1.2 m; L × W × H) wooden frame. Each permeameter’s core was constructed from a 6 in. (15.2 cm) diameter, 3 ft (0.9 m) long schedule 40 PVC core, secured to the frame using two 6 in. (15.2 cm) stainless-steel hose clamps (Figure 2a). Three manometers (Figure 2a,b) were used in each permeameter to allow for measurements of water heads at different points in the sample, which were used to calculate the hydraulic gradient, and sealed with silicone to ensure water tightness of the plastic hose connector. To facilitate water head measurement, a metric measuring tape was affixed adjacent to each permeameter (Figure 2a). Additionally, to prevent the intrusion of sample materials into the manometers, geotextile fabric was affixed to the inner ends of the connectors.

To contain the water head column over the sample during the test, a 6 in. (15.2 cm) diameter PVC pipe extension was affixed to the top of the permeameter core using a 6 in. (15.2 cm) rubber coupling. To minimize material loss due to water flow, geotextile pieces were secured at the bottom of the core with clamps. Water was introduced from the top of the PVC extension pipe and connected to the laboratory faucet through a hose. To maintain a constant 2 ft (0.6 m) water head, a drainage pipe with a diameter of 0.5 in. (1.27 cm) was set in the system and the connection method was the same as that used for the manometer and permeameter system. Water flowing to this drain, as well as the water flowing out through the samples, was collected in plastic totes, as illustrated in Figure 2a.

The structure of the clear infiltrometers consists of six units, each securely mounted on a 4.6 × 1.2 × 4.0 ft (1.40 × 0.37 × 1.22 m; L × W × H) wooden frame (Figure 3a). The core of each infiltrometer was constructed from 6 in. (15.2 cm)-diameter clear plastic tubing, with a thickness of 5/6 in. (2.12 cm) and extending to a length of 3 ft (0.91 m), affixed to the structure using two 6 in. (12.7 cm) stainless-steel hose clamps to ensure robust attachment to the wooden structure (Figure 3a). Given that these plastic cylinders were relatively less resistant and more flexible compared to PVC pipes, it became necessary to reinforce them at four key points with 6 in. (12.7 cm)-diameter PVC rings: at the top (for rubber coupler connection), the base (for geotextile or hardware cloth attachment), and both clamp points (Figure 3a).

The infiltrometers were designed to accommodate materials filled up to their maximum height of 3.0 ft (0.91 m). To effectively contain the water head column above the samples, a 6 in. (15.2 cm) PVC pipe extension was carefully attached to the top of the infiltrometer core using a 6 in. (15.2 cm) rubber coupler. To keep the materials inside the column and allow water to flow, it was attached at the bottom of the clear column with a clamp, geotextile sheeting, or stainless-steel wire mesh, depending on the matrix design under evaluation. This ensured the confinement of materials within the column while allowing water to flow freely. For the purpose of simultaneously supplying water to all six clear columns, an irrigation system was constructed. This system consisted of six 0.75 in. (1.91 cm) ball valves interconnected with PVC pipe and associated components. To maintain the water column constant during the constant-head infiltration rate tests, a 0.5 in. (1.3 cm)-diameter drain connected to a 0.5 in. (1.3 cm) clear hose through a 0.5 in. (1.3 cm) PVC adapter was installed in the 6 in. (15.2 cm) PVC extension (Figure 3a). Figure 3a also shows two clear infiltrometers prior to the installation of the 6 in (15.2 cm) rubber coupler and PVC extension.

2.2. Testing Procedures

To assess the water infiltration capacity of materials and matrices composed of multiple layers used in engineered infiltration swales, three distinct tests were conducted: (1) permeability constant-head tests, (2) constant-head infiltration rate tests, and (3) falling-head infiltration rate tests.

The constant-head test method using modified permeability columns differs from the standard method (ASTM D2434-68) [15] in two main aspects. First, the customary pair of porous disks was modified: the top disk was omitted and the bottom disk was replaced by a geotextile layer, which retains the sample while permitting water flow. Second, the spring mechanism used in the standard test to apply a 5.0–10.0 lbm (2.27–4.54 kg) load was omitted, allowing the test to replicate the swale’s field conditions better and capture consolidation effects relevant to in situ material behavior.

The constant-head test involved installing a geotextile at the base of the permeameter core, followed by sequential placement and compaction or consolidation of material layers, including #57 stone, pea gravel, field sand, and topsoil to achieve target density. A 6 in. (15.2 cm) rubber coupler and PVC extension were attached atop the core, and a porous circular sponge was placed on the sample surface to prevent water impact during initial saturation. Water was then gradually introduced until a 6 in. (15.2 cm) water column was established, at which point the sponge was removed, and until a 2 ft (0.6 m) water column was established and maintained at this constant head when water started to flow out through the top drainage pipe. Once steady outflow from the sample indicated full saturation, outflow volume, manometer readings, and water temperature were recorded. The coefficient of permeability (k) was calculated at the test temperature and subsequently corrected to 20 °C (68 °F).

The coefficient of permeability was calculated by applying Darcy’s Law (Equation (1)):

k = QL/(Ath)

(1)

where k is the coefficient of permeability at the test temperature, Q is the volume of water discharged over the time period t, L is the distance between manometers, A is the cross-sectional area of the specimen, t is the total time of discharge, and h is the difference in the water head on the manometers.

Finally, the coefficient of permeability, k, was corrected to that for 20 °C (68 °F) using Equation (2):

k (20 °C) = k μ/μ(20 °C),

(2)

where k (20 °C) is the coefficient of permeability at 20 °C, k is the coefficient of permeability at the test temperature, and μ and μ(20 °C) are water viscosity at the test temperature and 20 °C, respectively. The intrinsic permeability is a property of the porous medium (e.g., topsoil, sand, etc.) and depends only on the soil’s physical characteristics, like pore size and tortuosity [27]. Soil’s hydraulic conductivity depends on the properties of the soil (i.e., intrinsic permeability) and the fluid (e.g., viscosity and density) passing through the soil. When the coefficient of permeability k is determined using Darcy’s equation under soil saturation, it is the saturated hydraulic conductivity of the soil.

The constant-head infiltration rate tests involved installing a geotextile layer or stainless-steel wire mesh at the bottom of the infiltrometer core, followed by placement and compaction or consolidation of material layers to achieve the target density. A 6 in. (15.2 cm) rubber coupler and PVC extension were affixed at the top and a porous circular sponge was placed on the sample surface to protect against water impact. Water was slowly introduced until a 6 in. (15.2 cm) water column was established, after which the sponge was removed. A constant 2.0 ft (0.6 m) water column (i.e., 24.5 lbf at 6 in. diameter soil surface) was applied to saturate the sample first, indicated by steady discharge from the column bottom. Following saturation, the volume of water discharged over hourly intervals was measured to calculate infiltration rates over 6 h.

The falling-head infiltration rate tests followed the same procedures described for the constant-head infiltration rate tests. However, after the sample saturation, infiltrated water was replenished to 2.0 ft (0.6 m) and then we stopped adding additional water and the falling-head test began. Periodic measurements of water height and elapsed time were recorded until the entire 2.0 ft (0.6 m) column was infiltrated.

Materials were compacted using two methods to achieve target densities. The first, mechanical, compaction employed a custom-built wooden manual rammer with a disc-shaped head designed to fit within the permeameters and infiltrometers. Samples were divided into sublayers, each compacted sequentially until the target density was reached, ensuring uniform compaction throughout the sample. The second method utilized consolidation by a 1.0 ft (0.30 m) water column (i.e., 12.3 lbf at 6 in. diameter soil surface) placed over the material within the permeameter, infiltrometer, or infiltration chamber. Target density was attained when the entire water column infiltrated the material. To prevent surface disturbance from direct water impact, a circular sponge was positioned beneath the water column in permeameters and infiltrometers.

2.3. Infiltration Swale Media Design

The media design of the current ALDOT infiltration swale is shown in Figure 1a and consists of three layers: a bottom layer of #57 stone enveloped in non-woven geotextile, an intermediate layer of field sand, and a top layer of topsoil. To improve the infiltration capacity of the swale, alternative designs (Figure 4) were also proposed and explored, incorporating pea gravel and pine bark fines as additional material components.

Design A shown in Figure 4a is based on the current ALDOT infiltration swale design (Figure 1), and the dry swale design from the Georgia Department of Transportation (GDOT) is similar to Design A. Design B (Figure 4b) replaced topsoil with amended soil at a mixture of 80% topsoil and 20% pine bark fines by weight (or 50% by volume for topsoil and pine bark fines). Literature showed that incorporating porous material into the topsoil layer may enhance infiltration capacity, as demonstrated by Jiang et al. in 2019 [28]. Pine bark fines selected for the column experiments were sourced from a commercially available product, Evergreen All-Purpose Top Soil from Lowe’s (https://www.lowes.com/pd/Evergreen-1-cu-ft-Organic-Top-Soil/999911449 (accessed on 1 May 2025)). The material is also available in bulk and was ultimately sourced for the full-scale testing experiments [29,30]. Design B (Figure 4b) consisted of a 10.0 in. (25.4 cm) amended topsoil layer, a 12.0 in. (30.5 cm) field sand layer, and an 8.0 in. (20.3 cm) #57 stone layer. Both Design A and B had a geotextile layer (Figure 3a,b) to separate sand and stone to prevent sand from getting into the voids of the stone (mimicking the ALDOT infiltration swale design, Figure 1a). Design F (Figure 4c) replaces the geotextile between the field sand and the stones by a layer of pea gravel [e.g., 1, 2, or 6 in. (2.54, 5.08, or 15.2 cm) thicknesses] and replaces the geotextile at the column bottom with the stainless-steel wire mesh from Design B. Design F3 (Figure 4d) includes grass above amended soil with specific thickness for each layer of Design F [6 in. (15.2 cm) amended topsoil, 10 in. (25.4 cm) field sand, 6 in. (15.2 cm) pea gravel, and 9 in. (22.9 cm) #57 stones].

For each design, combinations of varying thickness in different layers were also tested [26]. Design C [26], similar to Design B, was comprised of a 6.0 in. (15.2 cm) amended topsoil layer, a 16.0 in. (40.6 cm) field sand layer, and an 8.0 in. (20.2 cm) geotextile-wrapped #57 stone layer. Design D [26], similar to Design F, included a 6.0 in. (15.2 cm) amended topsoil layer, a 15 in. (38.1 cm) field sand layer, a 1.0 in. (2.5 cm) pea gravel layer, and an 8.0 in. (20.3 cm) #57 stone layer not wrapped in geotextile. Design E [26], similar to Design F, consisted of a 6.0 in. (15.2 cm) layer of amended topsoil, a 4.0 in. (10.2 cm) layer of pea gravel, and an 18.0 in. (45.7 cm) layer of #57 stone not wrapped in geotextile.

2.4. SWMM Infiltration Swale Modeling

The USEPA Storm Water Management Model (SWMM) is a widely used dynamic simulation tool for modeling hydrologic and hydraulic processes in urban and rural catchments. It supports both short-term event-based and long-term continuous simulations of runoff generation, transport, and pollutant dynamics [21,31]. The model operates using a link-node framework, where sub-catchments generate runoff routed through nodes (e.g., junctions and storage units) and links (e.g., pipes and channels). SWMM also enables integration of low-impact development (LID) practices to mimic natural hydrology and reduce surface runoff through infiltration and storage.

Among the eight LID controls modeled by SWMM, the bioretention cell is most compatible with the experimental design of the ALDOT and GDOT infiltration swales. Unlike typical LIDs that intercept runoff from impervious surfaces, these swales are situated in drainage channels that receive runoff from both pervious and impervious areas. Structurally, the bioretention cell module, comprising surface, soil, and storage layers with or without underdrain, each defined by distinct hydraulic and physical parameters, closely reflects the vertical stratification of the tested infiltration systems, making it the most appropriate choice for simulating column and swale performance in SWMM.

For column tests (no evapotranspiration in each layer), SWMM uses three governing (differential) Equations (3)–(5) to model and solve three LID variables [21]: surface depth d₁, soil layer moisture content θ₂, and depth of water in the storage layer d₃,

$${\displaystyle \frac{\partial d_1}{\partial t}}=q_0-f_1-q_1\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\mathrm{S}\mathrm{u}\mathrm{r}\mathrm{f}\mathrm{a}\mathrm{c}\mathrm{e} \mathrm{l}\mathrm{a}\mathrm{y}\mathrm{e}\mathrm{r}$$

(3)

$$D_2{\displaystyle \frac{\partial {\theta }_2}{\partial t}}=f_1-f_2\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\mathrm{S}\mathrm{o}\mathrm{i}\mathrm{l} \mathrm{L}\mathrm{a}\mathrm{y}\mathrm{e}\mathrm{r}$$

(4)

$${\varphi }_3{\displaystyle \frac{{\partial d}_3}{\partial t}}=f_2-f_3\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\mathrm{S}\mathrm{t}\mathrm{o}\mathrm{r}\mathrm{a}\mathrm{g}\mathrm{e} \mathrm{L}\mathrm{a}\mathrm{y}\mathrm{e}\mathrm{r}$$

(5)

where θ₂ is soil layer moisture content (volume of water/total volume of soil), i is precipitation rate falling directly on the surface layer (ft/s), q₀ = inflow to the surface layer from runoff captured from other areas (ft/s), q₁ = surface layer runoff or overflow rate (ft/s), f₁ is infiltration rate of surface water into the soil layer (ft/s), f₂ = percolation rate of water through the soil layer into the storage layer (ft/sec), f₃ is exfiltration rate of water from the storage layer into native soil (ft/s), ϕ₂ = porosity (void volume/total volume) of the soil layer, ϕ₃ = void fraction of the storage layer (void volume/total volume), D₁ = freeboard height for surface ponding (ft), D₂ = thickness of the soil layer (ft), and D₃ = thickness of the storage layer (ft).

The surface layer serves as a temporary detention zone for stormwater and is controlled primarily by the berm height D₁, which defines the maximum allowable depth of ponding, e.g., 2 ft (0.6 m) for column tests (Figure 2 and Figure 3). Under falling-head conditions, the surface depth d₁ decreases from D₁ to 0 over time. Under constant-head conditions, the surface depth d₁ maintains a constant D₁ over time. Other surface parameters, such as vegetation volume fraction, surface roughness, and slope, are omitted or set to zero, as they are not relevant in the column tests.

Beneath the surface, the soil layer is an engineered media for water retention and infiltration. Its properties were specified based on properties of the soil media used in the column tests, including layer thickness, porosity, field capacity, wilting point, saturated hydraulic conductivity, suction head, and conductivity slope. These parameters directly influence the vertical infiltration rate and water redistribution in the soil layer and are critical in replicating the temporal response of the column test system. The infiltration of surface water into the soil layer, f₁ (ft/s), is modeled with the Green–Ampt equation (an optional infiltration method in SWMM):

$$f_1=K_s\left(1+{\displaystyle \frac{({\varphi }_2-{\theta }_{20})(d_1+{\psi }_2)}{F}}\right)$$

(6)

where K_s = the soil’s saturated hydraulic conductivity (ft/s), θ₂₀ = moisture content at the top of the soil layer (fraction), ψ₂ = suction head at the infiltration wetting front formed in the soil (ft), and F = cumulative infiltration volume per unit area over a storm event (ft). The rate f₁ is limited by the amount of empty pore space available [(ϕ₂ − θ₂)D₂/Δt] plus the volume removed by drainage (i.e., percolation f₂). When the soil layer is saturated (i.e., θ₂ = ϕ₂), f₁ is limited by the percolation rate, f₂, which is equal to the saturated hydraulic conductivity K_2s.

The rate of percolation (e.g., drainage) of water through the soil layer into the storage layer below it, f₂, is modeled using Darcy’s Law in the same manner used in SWMM’s existing groundwater module [21].

$$f_2=\left\{\begin{array}{c}{K}_{2s}exp\left(-HCO\left({\varphi }_2-{\theta }_2\right)\right), {\theta }_2>{\theta }_{FC}\\ 0, { \theta }_2\le {\theta }_{FC}\end{array}\right.$$

(7)

HCO is a decay constant derived from moisture retention curve data that describes how conductivity decreases with decreasing moisture content.

The storage layer below the soil layer has gravel with porosity ϕ₃, e.g., at the bottom of the columns (Figure 2 and Figure 3). Key parameters include thickness D₃, void ratio, and, and seepage rate. Notably, the seepage rate, representing the percolation of water into underlying native soil, is a crucial boundary condition in the LID model. To replicate the fully drained base of the column, the seepage rate can be set to equal to or higher than the soil’s saturated hydraulic conductivity, which is the maximum percolation rate.

A bioretention cell is modeled with one soil layer by SWMM; however, an ALDOT infiltration swale design has a topsoil or mixture layer and a field sand layer. The effective or equivalent saturated hydraulic conductivity (K_s) of the SWMM soil layer, comprising two media layers, can be calculated using Equation (8) based on Darcy’s law for each layer,

$$k_s={\displaystyle \frac{h_1+h_2}{\frac{h_1}{K_{s1}}+\frac{h_2}{K_{s2}}}}$$

(8)

where h₁ and h₂ are layer thickness and K_s1 and K_s2 are the saturated hydraulic conductivity for media layers 1 and 2, respectively.

3. Results and Discussion

Performance of the infiltration swale media was investigated through structured data collection and analysis using small-scale column experiments (Figure 2 and Figure 3). The evaluation involved tracking multiple performance indicators: (the coefficient of) permeability, constant- and falling-head infiltration rates, and changes in material elevation due to settlement. The modified constant head permeability tests on swale materials and the ALDOT and GDOT swale designs were conducted first. Field sands with varying compaction were tested up to 9 and 72 h to assess density effects on permeability. To enhance the functionality of the surface layer, infiltrometer testing under constant- and falling-head conditions was applied to native and amended topsoil samples, followed by tests on engineered media, including modified ALDOT variants.

3.1. Material Properties

Understanding the relationship between material behavior and swale infiltration performance involved conducting ASTM D698-12 Proctor compaction tests [32], porosity assessments, and bulk density measurements, and gradation size distribution analyses were conducted in accordance with ASTM guidelines. Conducting these tests helped confirm that the media adhered to ALDOT guidelines for infiltration swale applications.

Table 1. Results of bulk density and porosity measurements.

Material	Bulk Density		Porosity 2
	Loose Sample	After Consolidation
Topsoil	88.8 lb/ft3 (1.42 g/cm3)	96.8 lb/ft3 (1.55 g/cm3)	43%
#57 stone	98.6 lb/ft3 (1.58 g/cm3)	N/A	46%
Pea gravel	101.1 lb/ft3 (1.62 g/cm3)	N/A	41%
Field sand	93.6 lb/ft3 (1.50 g/cm3)	See Figure 5	33%
Pine bark fine	22.2 lb/ft3 (0.36 g/cm3)	N/A	73.5%
Amended topsoil 1	61.2 lb/ft3 (0.98 g/cm3)	68.7 lb/ft3 (1.10 g/cm3)	58%

Note: ¹—Amendment of the topsoil was achieved using 80% native topsoil and 20% organic pine bark fines by weight (both 50% by volume). ²—Porosity after consolidation.

displays the corresponding porosity and bulk density values. Because porosity affects how water moves through a medium, it is a significant variable in modeling infiltration behavior within the SWMM LID module. In addition to topsoil, field sand, and #57 stones, bulk density and porosity of peak gravel, pine bark fines and an amended topsoil used in alternative media designs (Figure 4) were also determined (Table 1). The pine bark fine has very large porosity (73.5%) and leads to a relatively large porosity (58%) for amended topsoil, but the field sand used in this study has relatively small porosity (33%) after consolidation. Rawls et al. [33] reviewed various studies and suggested that the effective porosity of sand is 0.417 and ranges from 0.354 to 0.480.

Materials with higher porosity and lower dry density generally allow water to flow more easily, while lower porosity and higher bulk density are typically associated with reduced permeability. Based on the data presented in Table 1, the topsoil sample is expected to have greater permeability than field sand due to its higher porosity and lower density. However, permeability is influenced by more than just pore quantity and compaction; the size, shape, and spatial arrangement of pore spaces also play a substantial role in determining hydraulic conductivity [17], including the interparticle attractions that lead to soil aggregation and binding [34]. The field sand’s bulk density changes with compaction and consolidation, which is reflected in changes of compaction degrees in Figure 5 for some cases.

Particle Size Distribution

Figure 6 shows the particle size distribution tests performed for topsoil, field sand, #57 stone (9–25 mm), and pea gravel (2–25 mm), which are basic components of the engineered media of the infiltration swale designs (Figure 4). The results validated compliance with ALDOT’s present material standards [35]. ALDOT’s topsoil (Section 650 [35]) for highway construction follows ASTM D 5268 and has 2–20% by weight of organic material, 10–90% by weight of sand content (0.05–2 mm, from the Soil Science Society of America), and 10–90% by weight of silt (0.002–0.05 mm) and clay (<0.002 mm) content in a portion of sample passing through a 10 (2 mm) sieve. The sandy topsoil (Figure 1) used in this study has approximately 88% sand content and less than 2% silty content (Figure 6). It should also be noted that ALDOT does not have any specifications on sandy topsoil. The particle size distribution analysis shows that topsoil contains more fine particles and demonstrates a more uniform gradation compared to field sand. This finer and more cohesive texture contributes to the topsoil’s lower permeability, even though it possesses higher porosity than the sand sample [26].

3.2. Modified Laboratory Testing for Constant Head Permeability

Modified constant head permeability tests were initially performed on ALDOT-specified media, including topsoil, field sand, and #57 stone (see Figure 1a and Figure 4a). To explore alternative design options (Figure 4), pea gravel was also evaluated for its permeability characteristics. Additional testing was conducted on field sand compacted to varying densities and monitored over extended durations to assess how consolidation and density affect permeability. The test was further applied to swale media configurations representative of both ALDOT and Georgia DOT designs to compare their hydraulic conductivity performance.

3.2.1. Permeability Evaluation of Infiltration Swale Media

Uncompacted (loose) samples of #57 stone, field sand, topsoil, field sand, and pea gravel were evaluated using the constant-head permeameter setup (Figure 2a) at a controlled temperature of 20 °C. As shown in the results (

Table 2. Modified permeability constant-head test results.

Materials	Permeability, k, at 20 °C
	in./min	cm/min	m/day
Topsoil	0.016	0.04	0.58
Field sand	1.56	3.95	557.06
#57 stone	2403.03	6103.76	87,894.29
Pea gravel	215.31	546.98	7875.27

), topsoil emerged as the most restrictive layer in the ALDOT infiltration swale design, even though its permeability is greater than 1.0 ft/day (0.30 m/day), the minimum required infiltration rate specified in the LID Manual of Alabama [36]. Its measured permeability was approximately 1% of that of field sand, indicating that it functions as the primary limiting layer for infiltration in the current configuration. Permeabilities of #57 stone and pea gravel are very large (Table 2), so the SWMM model does not use stone’s permeability as a model input parameter for the storage layer and assumes water moves downward freely through the storage layer without any resistance.

Based on ASTM D 5268, ALDOT’s topsoil could have quite different particle size distributions, e.g., 30% silt and clay and 60% sand. In comparison to the topsoil used for this study in Figure 5 (~88% sand), other topsoil can have quite different permeability, much smaller than 0.58 m/day (1.9 ft/day) (Table 2), with more fine silt and clay. This may not be important for various highway construction projects, but smaller permeability is crucial to infiltration-based LIDs. Using topsoil with different permeabilities can result in quite variable performance of the infiltration swale in controlling surface runoff and retaining/storing inflow.

3.2.2. Evaluation of Field Sand Permeability Across Varying Densities

To examine how varying density affects the permeability of field sand, modified constant-head permeability tests were carried out on 11 separate sand columns, each measuring 3.0 ft (0.91 m) in height. During testing, hourly readings were collected for temperature, discharge, and hydraulic head at manometers 1 and 3 (Figure 2b), which were used to compute the permeability coefficient (k). For each sand sample, a time-based permeability curve was developed by plotting hourly permeability values, as shown in Figure 5. The curves are labeled by the compaction degree of each sample, followed by its identifier (S1 through S11, Figure 5). Compaction degree refers to the ratio of the sample’s density to the maximum/optimum dry density, which in this study is 109.5 lbm/ft³ (1.75 g/cm³) [26]. S1, S2, S4, S5, and S7 were mechanically compacted samples, S3 and S6 were loose samples, and samples S8 to S11 were consolidated with 1 ft (0.3 m) of water. Figure 5 shows that sand’s permeability generally reduces with compaction, e.g., loose sample S3 has a permeability of 77.0 m/day (252.4 ft/day), and 97.0% compacted sample S2 has a permeability of 6.9 m/day (22.6 ft/day) at the start of a series of tests. The loose sample S3 decreases its permeability to 42.9 m/day (140.7 ft/day) (56.7% reduction) and a compaction degree of 83.1% at the end of 9 h; however, sample S2 had a permeability of 6.3 m/day (20.7 ft/day) at 9 h, with a small change in permeability. The other three mechanically compacted samples (S4, S5, and S7) had compaction degrees from 85.4% to 92.3% and 13.3% to 17.1% reduction in permeability at the end of tests (9 h). Soil permeability depends on not only compaction or consolidation but also on pore connectivity and forming micro-channels during the test, e.g., S4 with 86.6% compaction degree has higher permeability than S5 with 85.4% compaction degree and S3 (initially loose sample) after 5 h (83.1% compaction degree at 9 h).

For samples S8 to S11, when placed without any mechanical compaction and only subjected to a 1 ft (0.3 m) flowing water column, the final density after 9 h of the modified constant-head test is 85.5% of its optimum density, with a 1.10–3.29 m/day (3.6–10.8 ft/day) reduction in permeability, ranging from 30.4 to 39.9 m/day (99.7–130.9 ft/day) [26]. Rawls et al. [33] suggest 2.89 m/day (9.5 ft/day) of saturated hydraulic conductivity for sand with Green–Ampt equation applications. In the field, engineered media inside the swale could undergo similar consolidation or compaction phenomenon due to ponding water and machine mowing. Consequently, if the sand is loosely installed without compaction, consolidation/compaction over time could lead the sand to achieve a density of 85.5%. Therefore, in subsequent tests, field sand was consolidated with water after being placed in the infiltrometers to attain the 85.5% compaction degree, corresponding to 93.6 lbm/ft³ (1.50 g/cm³).

Two field sand samples, initially at densities of 88.1% and 91.8% of the optimum density, underwent a 72 h modified constant-head permeability test to evaluate the effects of consolidation on them, as shown in Figure 7.

The field sand sample compacted to 88.1% of its maximum dry density had an initial permeability of 52.6 m/day (172.6 ft/day). Over a 72 h period, this value declined to 31.7 m/day (104.0 ft/day)—a reduction of approximately 39%. During the same timeframe, the sample’s density increased slightly from 88.1% to 89.5%. In another test, the sample with an initial compaction level of 91.8% showed an initial permeability of 32.0 m/day (105.0 ft/day), which dropped to 19.0 m/day (62.3 ft/day) after 72 h, corresponding to a 41% decrease. Over the test duration, the density rose from 91.8% to 92.3%. The results from the extended duration tests indicate that exposing the material to a sustained 2 ft (0.6 m) hydraulic head leads to noticeable consolidation, which in turn reduces the material’s ability to transmit water. This decline in infiltration capacity is attributed to compaction occurring as water flows through the media under elevated head conditions. For the two samples tested, the average decrease in permeability was approximately 40%. This finding is particularly relevant to the construction and field performance of infiltration swales, as real-world installation and maintenance often involve varying degrees of compaction that can affect long-term hydraulic behavior.

3.2.3. Comparative Permeability Testing of GDOT and ALDOT Infiltration Swale Designs

Five samples, representative of the ALDOT infiltration swale design, and two samples, representative of the GDOT infiltration swale design, underwent the modified constant-head permeability tests. The configuration of all seven samples, along with the corresponding test results, is detailed in

Table 3. Modified permeability tests and results for ALDOT and GDOT designs.

Design	Materials
	Topsoil Layer Height in. (cm)	Field Sand Layer Height in. (cm)	#57 Stone Layer Height in. (cm)	Permeability, k (20 °C) ft/day (m/day)
ALDOT 1	9.4 (24)	14.2 (36)	9.4 (24)	2.28 (0.69)
ALDOT 2	11.8 (30)	12.6 (32)	8.7 (22)	1.80 (0.55)
ALDOT 3	8.3 (21)	16.5 (42)	7.9 (20)	1.56 (0.48)
ALDOT 4	8.3 (21)	16.5 (42)	8.3 (21)	0.48 (0.14)
ALDOT 5	10.6 (27)	15.0 (38)	7.5 (19)	0.24 (0.07)
GDOT 1	22.4 (57)	1.6 (4)	9.1 (23)	0.12 (0.04)
GDOT 2	22.0 (56)	2.4 (6)	8.7 (22)	0.24 (0.07)

. These swales have different thicknesses for three horizontal layers, but the total depth is about 33 in. (84 cm) for all samples. Permeability of the ALDOT and GDOT designs ranged from 0.001 to 0.019 in./cm (0.002 to 0.050 cm/min), which is similar permeability of topsoil and much smaller than sand’s permeability (Table 2) and, therefore, these results confirmed again that the low permeability of topsoil is the limiting factor to control/affect the overall infiltration rate of the ALDOT infiltration swale design. To improve the infiltration performance of the ALDOT swale design (Figure 4a), the permeability of the soil layer must be improved; e.g., by incorporating porous material into topsoil, as demonstrated by Jiang et al. in 2019 [28].

3.3. Infiltration Rate Tests

3.3.1. Infiltration Rates of Infiltration Swale Materials

Topsoil and amended topsoil (a topsoil mixture amended with pine bark fines), as basic infiltration swale materials, were first evaluated using falling-head infiltration tests. Three replicas of 6 in. (15.2 cm) loose topsoil samples were tested three times for each, using falling-head (2 ft or 0.6 m) infiltration rate columns (Figure 3,

Table 4. Measured infiltration rates of topsoil under falling-head conditions.

Topsoil Sample	Falling-Head Test			Average	Overall Average
	Test 1	Test 2	Test 3
Sample 1	0.76 ft/day (0.23 m/day)	0.35 ft/day (0.11 m/day)	0.27 ft/day (0.08 m/day)	0.46 ft/day (0.14 m/day)	0.63 ± 0.36 ft/day (0.19 ± 0.11 m/day)
Sample 2	0.86 ft/day (0.26 m/day)	0.41 ft/day (0.12 m/day)	0.28 ft/day (0.09 m/day)	0.52 ft/day (0.16 m/day)
Sample 3	1.39 ft/day (0.42 m/day)	0.94 ft/day (0.29 m/day)	0.39 ft/day (0.11 m/day)	0.91 ft/day (0.28 m/day)

). The measured topsoil’s infiltration rates ranged from 0.27 ft/day (0.08 m/day) to 1.39 ft/day (0.42 m/day). The average measured infiltration rate of topsoil was 0.63 ft/day (0.19 m/day), with a standard deviation of 0.36 ft/day (0.11 m/day). Except for Test 1 of Sample 3, all other tests had measured infiltration rates below the minimum standard of 1.0 ft/day (0.30 m/day) defined by the Alabama LID Manual [36]. Repeated testing on the same sample led to a further decline in infiltration due to consolidation (Table 4); e.g., the infiltration rate of Sample 3 reduced from 0.42 to 0.11 m/day.

Due to the low permeability of topsoil, it was amended by adding pine bark fines. This is similar to incorporating porous material into the topsoil layer [28]. Eben et al. [13] also used two engineering media or mixture for their infiltration swales: one had 32% (by volume) natural topsoil, 8% green waste compost, 45% quartz sand (0–2 mm), and another had 70% quartz sand (0–2 mm) and 15% green waste compost. Also, both medias had 15% brick sand (0–3 mm). Stinshoff et al. [9] studied three types of soil media, which are mixtures of lawn soil (natural topsoil), silica sand or brick sand, and compost, using plant containers and collected stormwater runoff to test their water quality performance for infiltration swales, but no results on infiltration capacity were reported. In this study, ten of the mixture samples were prepared using varying weight-based ratios of topsoil and pine bark fines. For comparison, one control sample contained pure topsoil, while another included only pine bark fines. Summary information for all samples—including their uniform height of 6 in. (15.2 cm)—along with the corresponding infiltration rates measured through repeated falling-head testing, is presented in

Table 5. Falling-head infiltration rate results of the mixtures.

Sample Composition by Weight		Infiltration Rate, ft/day (m/day)
Topsoil (%)	Pine Bark Fines (%)	First Test	Second Test	Third Test	Average
100	0	1.00 (0.30)	0.57 (0.17)	0.31(0.09)	0.63 (0.19)
95	5	0.87 (0.27)	0.55 (0.17)	0.87 (0.27)	0.76 (0.23)
93	7	0.96 (0.29)	1.67 (0.51)	0.03 (0.01)	0.89 (0.27)
90	10	0.92 (0.28)	0.87 (0.27)	1.63 (0.50)	1.14 (0.35)
85	15	1.50 (0.45)	2.32 (0.71)	3.29 (1.00)	2.37 (0.72)
80 1	20	5.70 (1.73)	3.40 (1.04)	7.70 (2.35)	5.60 (1.71)
75	25	14.3 (4.35)	17.0 (5.19)	21.3 (6.50)	17.5 (5.35)
70	30	12.9 (3.94)	30.6 (9.34)	35.1 (10.7)	26.2 (7.99)
60	40	45.0 (13.7)	15.7 (4.77)	16.3 (4.96)	25.6 (7.81)
50	50	222 (67.2)	411.4 (125.4)	320 (97.5)	318 (96.8)
25	75	262 (79.8)	320.0 (97.54)	411 (125.4)	331 (101)
0	100	2160 (658.4)	1440 (438.9)	1920 (585.2)	1840 (560.8)

Note: ¹—the mixture of 80% topsoil and 20% pine bark fines was selected for several alternative design configurations evaluated in this study (see Figure 4b–d).

.

Test results showed a clear relationship between pine bark content and infiltration performance—mixtures with higher proportions of pine bark fines exhibited improved infiltration rates. Among the various blends, the composition containing 80% topsoil and 20% pine bark fines (by weight) achieved an average rate of 5.60 ft/day (1.71 m/day), which is approximately 8.9 times greater than that of unamended topsoil (Table 4), which infiltrated at 0.63 ft/day (0.19 m/day). The addition of pine bark fines to topsoil increases infiltration primarily by altering the physical structure of the soil matrix. Pine bark fines are coarse, fibrous organic particles that increase macroporosity in the mixture, improving both the volume and connectivity of void spaces. This structure allows water to move more freely through the media by reducing resistance typically caused by smaller, less connected pores. Due to this substantial enhancement, the 80/20 mix was selected for incorporation into several alternative design configurations evaluated in this study (see Figure 4b–d). Unless stated otherwise, all future references to “amended topsoil” correspond to this 80:20 topsoil-to-pine bark fines mixture, which was 50% by volume for each material. It should be noted that the infiltration rate of the mixture is still much less than the measured permeability of field sand (Figure 5 and Figure 6, 28.8–77.0 m/day when the compaction degree is less than 90%). The mixture in this study is based on 80% sandy topsoil with ~88% sand content (Figure 6) used for all testing. If another topsoil (e.g., with less sand content and more silt and clay) is used, the mixture may need more than 20% pine bark fines and less than 80% topsoil to achieve 5.60 ft/day (1.71 m/day) of the falling-head infiltration rate. Table 5 also clearly shows that when pine bark fines are 20% more by weight, the infiltration rate increases rapidly; therefore, when another topsoil is used, the increase in pine bark fines is necessary and can still reach a higher infiltration rate for the swales with the mixture layer.

3.3.2. Infiltration Testing of Swale Design Alternatives Using the Falling-Head Method

Out of 15 engineered media designs, three samples (

Table 6. Infiltration rate results for infiltration swale designs tested.

Description	Designs 2	Measured Infiltration Rates (ft/day)		Estimated ks (ft/day)
		Falling Head	Constant Head
10″ topsoil, 12″ sand, 9″ #57 stone	A_geo_geo	0.31 ± 0.14 (0.16–0.54)	N/A 3	1.81 ± 0.706 (0.59–3.02)
10″ topsoil, 12″ sand, 8″ #57 stone	A-1G_geo_swm	0.62 ± 0.28 (0.34–1.18)	0.46 ± 0.06 (0.40–0.56)	1.81 ± 0.706 (0.59–3.02)
10″ topsoil, 12″ sand, 9″ #57 stone	A*_geo_geo_con	0.49 ± 0.31 (0.23–1.29)	1.73 ± 0.45 (1.16–2.31)	1.81 ± 0.706 (0.59–3.02)
10″ topsoil, 12″ sand, 9.5″ #57 stone	A*_geo_geo_con_grass	0.31 ± 0.07 (0.24–0.43)	0.91 ± 0.08 (0.79–1.04)	1.81 ± 0.706 (0.59–3.02)
10″ mixture 12″ sand, 8″ #57 stone	B_geo_geo	2.25 ± 1.94 (0.33–6.46)	N/A	11.51 ± 2.465 (7.19–15.82)
10″ mixture 12″ sand, 9.5″ #57 stone	B*_geo_geo_con	1.10 ± 0.64 (0.46–2.25)	5.38 ± 1.23 (3.46–7.69)	11.51 ± 2.465 (7.19–15.82)
6″ mixture 16″ sand, 8″ #57 stone	C_geo_geo	1.32 ± 0.36 (0.86–1.94)	N/A	17.93 ± 3.616 (11.43–24.40)
6″ mixture 15″ sand, 1″ peag 1, 8″ #57 stone	D_pea1_geo	0.92 ± 0.17 (0.67–1.23)	N/A	17.24 ± 3.500 (10.96–23.49)
6″ mixture, 4″ peag, 18″ #57 stone	E_pea4_geo	1.60 ± 1.01 (0.45–3.30)	N/A	5.55 ± 1.253 (3.4–7.7)
10″ mixture 12″ sand, 6″ peag, 8″ #57 stone	F_pea6_swm	5.99 ± 2.72 (2.26–11.08)	7.66 ± 1.97 (4.80–9.45)	11.51 ± 2.465 (7.19–15.82)
10″ mixture 12″ sand, 6″ peag, 4″ #57 stone	F*_pea6_swm_con	1.26 ± 0.46 (0.73–2.03)	5.31 ± 0.76 (4.18–6.43)	11.51 ± 2.465 (7.19–15.82)
6″ mixture 16″ sand, 6″ peag, 7″ #57 stone	F1_pea6_swm	1.11 ± 0.16 (0.89–1.35)	4.75 ± 1.36 (3.32–7.41)	17.93 ± 3.616 (11.43–24.40)
8″ mixture 14″ sand, 6″ peag, 7″ #57 stone	F2_pea6_swm	1.58 ± 0.38 (1.17–2.17)	6.73 ± 1.39 (5.22–8.82)	14.02 ± 2.934 (8.82–19.20)
6″ mixture 10″ sand, 6″ peag, 9″ #57 stone	F3*_pea6_swm_con	2.24 ± 0.31 (1.94–2.98)	5.75 ± 0.89 (4.52–7.48)	13.65 ± 2.866 (8.58–18.70)
6″ mixture 10″ sand, 6″ peag, 9″ #57 stone	F3*_pea6_swm_con_grass	11.66 ± 5.69 (3.52–24.26)	13.73 ± 4.78 (7.48–21.31)	13.65 ± 2.866 (8.58–18.70)

Note: ¹ “peag” stands for pea gravel; ² first “geo”—geotextile between sand/mixture and gravel layers and second “geo”—geotextile at the column bottom; “swm”—stainless-steel wire mesh at the column bottom; “con”—consider topsoil’s or mixture’s sample consolidation under 2 ft (61 cm) of water head; “pea*”—replace geotextile between sand/mixture and gravel layers with * (1, 4, and 6) inches of pea gravel; and “grass”—grass sod above the topsoil or the mixture (amended topsoil); ³—“N/A” stands for no data were collected or no tests were performed.

) experienced triplicate falling-head tests to assess infiltration performance [26]. “Description” in Table 6 lists the engineered media composition and the thickness of each layer, e.g., Design C has 6 in. (15.2 cm) of the mixture, 16 in. (40.6 cm) of field sand, and 8 in. (20.3 cm) of #57 stone. “Designs” provides alternative swale design names and two to four other features for the design. The first design element involves a geotextile barrier that separates the field sand from the coarse stone base, the second feature is the material at the column bottom, the third feature is whether the topsoil/mixture’s sample consolidation was considered, and the fourth feature is whether there is grass above the soil layer. The feature “geo” stands for geotextile, “swm” for the stainless-steel wire mesh at the column bottom, “pea*” stands for replacing geotextile with * (1, 4, and 6) inches of pea gravel, “con” considers the sample’s consolidation under 1 ft (30.5 cm) of water head, and “grass” for grass sod above the topsoil or the mixture (amended topsoil) (Figure 3d). For measured infiltration rates and estimated k_s in Table 6, the mean value is followed by the plus/minus sign and the standard deviation, and numbers inside brackets are the minimum and maximum values used to provide the range of measurements or estimates.

Each swale design has 30–36 in. (76.2–91.4 cm) total column height, and excluding topsoil or mixture and sand from the top, the last or bottom layer is #57 stone (Figure 4); e.g., both Designs B and C have 8 in. (20.3 cm) of #57 stone because there is a total of 22 in. (55.9 cm) of mixture and sand, which remain consistent with the ALDOT standard design. Some designs include pea gravel between sand and stone layers. The “Description” and “Design” columns in Table 6 provide detailed information on each swale design. For measured infiltration rates, average values with corresponding standard deviations were calculated and reported in Table 6, including the minimum and maximum values from nine measured values of each design (three tests for each of three samples).

The structural difference in the swale’s engineered media design directly contributed to the observed variations in infiltration performance. Notably, Design B, which solely replaced the original topsoil with amended topsoil while maintaining the remaining ALDOT configuration, achieved an average infiltration rate (2.25 ft/day or 0.69 m/day) exceeding the traditional ALDOT design (0.31 ft/day or 0.09 m/day) by a factor of 7.26. This substantial improvement highlights the critical role of surface media composition. The results of Design B from nine falling-head infiltration tests suggest that modifying only the topsoil layer with a more permeable amended mixture can significantly enhance the overall infiltration capacity of the swale system, even without altering the underlying structural layers.

As shown in Table 6, removing the geotextile layer at the base of the ALDOT swale—an approach used in the A-1G configuration—resulted in a twofold increase in average infiltration rate under falling-head testing. The rate improved from 0.31 ft/day to 0.62 ft/day (0.09 to 0.18 m/day). Design F shares the same engineered media setup as Design B (see Description), but substitutes the geotextile wrap around the #57 stone (geo_geo) with pea gravel at a depth of 6 in. (15.2 cm) and a stainless-steel wire mesh at the bottom (Figure 3c, referred to as pea6_swm). This modification significantly enhanced infiltration performance, with Design F reaching 5.99 ft/day (1.83 m/day) compared to the 2.25 ft/day (0.69 m/day) observed in Design B—an improvement by a factor of 2.66.

The amended topsoil was applied to a uniform thickness of 6 in. (15.2 cm) in Designs C through E (Figure 4); however, the other materials vary in composition and configuration. Specifically, Designs C and D include field sand layers of 16 in. (40.6 cm) and 15 in. (38.1 cm), respectively, while Design E does not incorporate a sand layer. All three designs include a geotextile layer at the base, consistent with the configuration in Design B. However, in Designs D and E, the geotextile above the stone was replaced with a 6 in. (15.2 cm) layer of pea gravel, which serves to separate the field sand from the underlying #57 stone and prevent migration of fine particles into the drainage layer. These structural variations were implemented to evaluate the influence of individual material layers and interfaces on overall infiltration performance and system stability. Designs C to E have similar infiltration rates, which could be influenced by the geotextile at the column bottom.

In relation to Design F and its variations F1 and F2, they shared similarities with Design B but varied thickness of the mixture and sand. These configurations also incorporated a 6.0 in. (15.2 cm) layer of pea gravel positioned to act as a separator between the intermediate and drainage layers, substituting the geotextile used in Design B. A stainless-steel wire mesh was additionally installed at the base to reinforce the system. Design F had a significant increase (19.32 times larger) in the average infiltration rate in comparison to the traditional ALDOT design (0.31 ft/day or 0.09 m/day). Designs F1 and F2 had 3.6 and 5.1 times larger average infiltration rates than the traditional ALDOT design. All the above tests did not consider consolidation effects.

For Designs A, B, F, and F3, the sample consolidation by a 1 ft (0.3 m) water column over about 10 h was considered; i.e., the final density of the topsoil or mixture layer and its column height were updated and they were named A*, B*, F*, and F3* in Table 6. The 10 in. (25.4 cm) topsoil (Design A) and the 6 in. or 10 in. (15.2 cm to 25.4 cm) mixture (Design F3 or F) layer had 7–8% and 17–21% reduction in height after consolidation, which results in decreasing porosity. With the consideration of consolidation, the average infiltration rates of B* and F* were reduced from 2.25 to 1.10 ft/day (0.69 to 0.34 m/day) and from 5.99 to 1.26 ft/day (1.83 to 0.38 m/day) (Table 6), respectively, for the falling-head tests. For ALDOT Design (A*), the falling-head infiltration rate was slightly increased from 0.36 to 0.49 ft/day (0.11 to 0.15 m/day).

Finally, three samples of ALDOT Design A and F3 Design with grass above topsoil or the mixture were subjected to three constant-head infiltration rate tests, and then to an additional series of triplicated falling head tests to assess infiltration behavior further. ALDOT Design A with grass reduced performance on average infiltration rate, 0.49 to 0.31 ft/day (0.15 to 0.09 m/day) for the falling head tests and 1.73 to 0.91 ft/day (0.53 to 0.28 m/day) for the constant-head tests with consideration of consolidation. Under falling-head conditions, the Design F3 configuration with grass cover demonstrated an average infiltration rate of 11.66 ft/day (3.55 m/day), which represents a 37.6-fold improvement over the ALDOT + Grass design, whose rate was 0.31 ft/day (Table 6). A comparison between the F3 + Grass and the standard F3 configuration (without vegetative cover) revealed notable differences in infiltration performance. The grass-covered version exhibited a 2.4-fold increase in constant-head test results and a 5.2-fold improvement under falling-head conditions (Table 6). This enhanced performance is likely due to improved media stability. In the absence of grass, particles of pine bark fines in the upper layer tend to detach and float when exposed to water, potentially disrupting pore structure and reducing flow capacity during testing [11]. A few inches (centimeters) of mulch have been widely used for surface materials of rainfall gardens and bioretention cells to retain moisture, suppress weeds, and prevent soil erosion, but it is reported to be relatively easily washed away by surface runoff due to its low density, which is similar to what happened on the amended topsoil during column tests. With turfgrass cover above the amended topsoil, this infiltration swale has not only higher infiltration rates (Table 6) but also less maintenance (avoiding washing away), better overall design, and performance enhancement in comparison to the existing ALDOT infiltration swale design. However, an in-depth analysis in combination with soil–vegetation–water interactions (such as the improvement in pore structure by root systems) was not conducted for these small-scale column tests.

3.3.3. Constant Head Infiltration Rate Tests

A series of triplicated falling- and constant-head tests were first conducted on Designs A-1G, A*, B*, and all F configurations, with constant-head testing performed subsequently to assess infiltration behavior (Table 6). Except for Design A-1G, the constant-head infiltration rate tests for all other Designs had 1.2–4.9 times (3.2 ± 1.27) larger average infiltration rates in comparison to the corresponding falling-head results. The constant-head tests showed that Designs with 6–10 in. (15.2–25.4 cm) mixture (B* and all F variations) yielded average infiltration rates 10.2–29.8 times (15.3 ± 6.26) higher than Design A-1G with 10 in. (25.4 cm) of topsoil. This further proves the great benefit of replacing topsoil with the mixture (Table 5, 50% by volume for pine bark fines and topsoil). Constant-head testing indicated that the F3 + Grass configuration achieved an average infiltration rate of 13.73 ft/day (4.18 m/day), which is approximately 15.1 times greater than the rate recorded for the ALDOT + Grass design.

3.3.4. Comparison with Infiltration Rates in Other Studies

Many bioretention applications use an engineered media mix (not a layered media in Figure 1 and Figure 4), e.g., a mix of 60–70% construction sand, 15–25% topsoil, and 15–25% organic matter (e.g., Minnesota DOT Grade 2 compost), and it can achieve 2–8 ft/day (0.6–2.4 m/day) of saturated hydraulic conductivity [12], which is lower than the one for the F3 design with grass configuration. Le Coustumer et al. [37] evaluated the long-term performance of biofilter systems built in Australia and found large variability of measured hydraulic conductivities using infiltrometers, 40% being below the recommended range of 3.9–15.5 ft/day (1.2–4.7 m/day), 43% within it, and 17% above that is larger than the average infiltration rate of F3 design with grass.

Thompson et al. [38] performed column experiments to measure the saturated hydraulic conductivities K_s of bioretention mixtures. A flow-through column apparatus, like the infiltration testing column (Figure 3) used in this study, contains 1 ft (30.5 cm) of water (constant head of ponding water), 1 ft (30.5 cm) of soil mixture, and a small layer of gravel along its depth to determine the infiltration rate. Eleven mixtures were made and blended by hand to an even consistency using three components: 0% or 20% (volumetric) sandy or silt loam soil, 30–70% coarse sand (0.5–1.0 mm or 0.02–0.04 in. diameter), and 20–70% organic compost (>99% with a particle size of less than 19 mm). Measured average infiltration rates ranged from 68.2 to 140 ft/day (20.8 to 42.8 m/day), which is significantly greater than the infiltration rate of native soil in the region [38]. The mixtures composed solely of sand and compost exhibited the highest K_s values (36.1–42.8 m/day or 118.5–140.4 ft/day). The mixtures with the addition of sandy soil had measured average infiltration rates ranging from 29.9 to 41.8 m/day (98.1–137.1 ft/day) and silt loam soil from 20.8 to 33.8 m/day (68.1–110.9 ft/day). These mixtures’ infiltration rates are much larger than the design infiltration rates of silt loam and sandy soil (0.08–0.3 m/day) in the region [38,39]. This also indicate that using a layer of native topsoil becomes the limiting factor to control or reduce the infiltration rate of the layered-media swales (Figure 1). Compared to the tests conducted by Thompson et al. [38], the falling-head infiltration rates of the amended topsoil (a mixture of sandy topsoil and pine bark fines) also increased with pine bark fine percentage by weight. The infiltration rates increased slowly from 0.63 ft/day (0.19 m/day) to 2.37 ft/day (0.72 m/day) (Table 5) when the pine bark content was below 15%. Once the pine bark content exceeded 25%, infiltration rates increased dramatically (from 5.4 to 560.8 m/day or 17.7 to 1840 ft/day) for 15.2 cm (6 in.) of the mixture under 61.0 cm (2 ft) of water head. However, infiltration rates for the layered media (Table 6) cannot be directly compared with the mixture’s infiltration rates in Table 5 and in the Thompson et al. study.

The observed infiltration rate for 100% uncompacted sand was 46.6 m/day (152.8 ft/day), which is used as a threshold reference by Thompson et al. [38] and is about the same as the loose sand sample S3 after 5 h (Figure 5). Infiltration rates of 11 mixtures tested with the addition of compost and with/without two types of soil were less than the threshold reference. The saturated hydraulic conductivity of Bonneau fine sand (89.3% sand and 10.6% silt) was determined as 2.47 m/day (8.1 ft/day) [40] for the top 23 cm (0.75 ft) of the soil in the study area investigated by Gregory et al. [41], which is much smaller than the measured permeability of the loose sand in this study and about the same order of magnitude as for 97% compacted sand S2 (Figure 5). The saturated hydraulic conductivity for undisturbed (non-compacted) natural wooded lots in the area ranged from 0.8 to 24.6 m/day (2.6–80.7 ft/day) [41] with averages from 9.0 to 15.2 m/day (29.5–49.9 ft/day). Their saturated hydraulic conductivities decreased by about 3–5 times due to the compaction [41]. The saturated hydraulic conductivities of silty sand and pure sand were determined to be 0.05 and 2.76 m/day (0.2–9.1 ft/day) by Hussian et al. [42]. They added biochar with different particle sizes (5–15% by weight) to these two types of sand to form amended soils and quantified the impact of biochar’s particle sizes on the saturated hydraulic conductivity of amended soil. Adding biochar with larger particle sizes (e.g., 0.425–2 mm or 0.017–0.079 in., 2–4.75 mm or 0.079–0.187 in.) increased saturated hydraulic conductivity up to 84% (from 0.05 to 0.094 m/day or 0.16 to 0.31 ft/day) [42], which is a much smaller increase compared to adding pine bark fines in this study (Table 5). Adding biochar with fine particle sizes (<0.425 mm or 0.017 in.) made the saturated hydraulic conductivity of the amended soil smaller than that of the corresponding sand.

3.4. Modeling Infiltration Swales and Challenges

To better understand, interpret, and validate the column test results, an SWMM model of a bioretention cell was configured with a sub-catchment that was 100% impervious and entirely drained through the LID control, ensuring that all precipitation was routed through the bioretention cell system. A constant rainfall intensity was applied to simulate a steady water supply, analogous to the inflow condition of the constant-head infiltration test. The three models or case studies were set up to mimic Designs A, B, and F (Table 6) with specific soil and storage layer properties (height and porosity) from measured data (Table 1 and Table 6). The soil’s field capacity of 0.17, wilting point of 0.06, and conductivity slope of 30 were used [43]. The soil’s saturated hydraulic conductivity was set to a value of 5.99 ft/day or 3.00 in./h (1.83 m/day) for cases 1 and 2 (Figure 8 and Figure 9, Designs using the mixture) and 0.62 ft/day (0.19 m/day) for case 3 (Figure 9, ALDOT Design using topsoil). The seepage rate of 2.25 ft/day (1.13 in./h or 0.69 m/day) was used for cases 1 and 3 (Figure 8 and Figure 10), which mimics geotextile at the column, and ≥3.00 in./h (6.00 ft/day or 1.83 m/day) for case 2 (Figure 8), which mimics the stainless-steel wire mesh. In the SWMM models, the berm height of the surface layer was set to 2 ft (0.6 m) to replicate the ponding depth used in the column tests.

3.4.1. Swale’s Infiltration Process Simulated by SWMM

The LID’s initial soil saturation of 10% is assumed in the SWMM models (modeling cases 1, 2, and 3 in Figure 8, Figure 9 and Figure 10), and the initial values of soil moisture (volumetric water content) and storage level (Figure 8a, Figure 9a and Figure 10a) are greater than zero. When the inflow to LID continues, the surface infiltration makes the soil moisture increase until the soil layer is fully saturated at 1.67 h (1:40) due to large infiltration rates 3–9.5 in./h or 6–19 ft/day (1.8–5.8 m/day, Figure 8b). In Figure 8b and Figure 9b, during the wetting phase, when the soil layer is still unsaturated, many pore spaces exist between the sand and soil particles, enabling a high potential infiltration capacity. As a result, the initial infiltration rate spikes, reaching its maximum due to the smallest hydraulic resistance. As saturation progresses within the soil layer, surface ponding begins, and the infiltration rate decreases according to the Green–Ampt infiltration equation (Equation (6)), which accounts for the diminishing soil moisture deficit $({\varphi }_2-{\theta }_{20})$ and wetting front advancement (F in Equation (6) increases). The surface water depth linearly increases from 0 at time t = 0 and reaches its maximum depth of 2 ft (0.6 m) at about 2.17 h (2:10) because of a large constant inflow; e.g., 16 in./h (32 ft/day or 9.8 m/day) used for the case studies, which is greater than the largest surface infiltration rate 9.5 in./h (19 ft/day or 5.8 m/day) (Figure 8b and Figure 9b).

For natural rainfall with smaller intensities (i.e., less than the soil’s infiltration capacity), the surface ponding does not start at the beginning of the rainfall event, but it could have a significant delay. The saturation process occurs sequentially: the soil layer saturates first, followed by the onset of surface ponding, which increases until the maximum surface depth of 2 ft (0.6 m) is reached. Eventually, the storage layer also becomes fully saturated at 3.33 h.

The constant rainfall or inflow stops at 8.0 h for cases 1 and 2 in Figure 8 and Figure 9; therefore, the SWMM model mimics the constant-head infiltration tests from 2.17 to 8.0 h when the surface depth remains at 2 ft (0.6 m). When the inflow continues until the LID system reaches full saturation, i.e., both soil and storage layers are saturated, the soil moisture, storage level, and surface depth all reach and maintain their respective maximum values, at 3.33 h shown in Figure 8a. Figure 8b shows, from 2.17 to 3.33 h, both the surface infiltration rate and the soil percolation rate are equal to the soil’s saturated hydraulic conductivity (3.00 in./h or 1.83 m/day); however, from 3.33 to 8.0 h, they both drop and maintain at 1.13 in./h (0.69 m/day), which is the seepage rate (mimicking the geotextile). For the constant-head infiltration tests, following saturation or steady flow from the column bottom, the volume of water discharged over hourly intervals is measured to calculate infiltration rates over 6 h; therefore, the calculated infiltration rate measures the seepage rate through geotextile for Design B. The SWMM case 1 modeling, presented in Figure 8, shows that the seepage rate, if restricting or limiting the vertical flow through the infiltration swale system, does not give/present the soil’s saturated hydraulic conductivity, which is 3.00 in./h (1.83 m/day) used in this SWMM modeling study.

When the rainfall/inflow input is terminated at 8:0 h and the surface depth is at 2 ft (0.6 m), it initiates a drainage phase, utilizing the water remaining in the surface layer as well as the outflow from the LID system. Infiltration, soil percolation, exfiltration (seepage rate), and declining surface water depth are reported in Figure 8 and Figure 9. This post-rainfall response effectively simulates/mimics the falling-head infiltration tests discussed in previous sections, where the declining surface water level governs the observed infiltration behavior (Figure 8 and Figure 9). Figure 8 shows that, under the falling-head condition (8–12 h), the surface infiltration, soil percolation rate, and seepage rate are the same, as the seepage rate restricts the vertical flow. For the falling head infiltration rate tests, periodic measurements of water height and elapsed time (i.e., declining surface water depth curve in Figure 8a) are recorded until the entire 2 ft (0.6 m) column is infiltrated for determining the infiltration rate. Again, this measured/determined falling-head infiltration rate gives the seepage rate, not the soil’s saturated hydraulic conductivity.

Figure 8. SWMM simulated (a) surface depth, storage level, and soil moisture (volumetric water content); (b) surface infiltration rate, soil percolation, and exfiltration of a bioretention cell with soil hydraulic conductivity of 3.00 in./h (1.83 m/day) and soil seepage rate of 1.13 in./h (0.69 m/day) (1 in. = 2.54 cm, 1 in./h = 0.61 m/day).

Figure 8. SWMM simulated (a) surface depth, storage level, and soil moisture (volumetric water content); (b) surface infiltration rate, soil percolation, and exfiltration of a bioretention cell with soil hydraulic conductivity of 3.00 in./h (1.83 m/day) and soil seepage rate of 1.13 in./h (0.69 m/day) (1 in. = 2.54 cm, 1 in./h = 0.61 m/day).

Figure 9 presents an alternative simulation or swale design scenario represented for free drainage conditions, or the absence of a geotextile layer (replacing it with the stainless-steel wire mesh), on the hydrologic behavior of the bioretention system. For modeling case 2 (mimicking Design F hydrologic behavior), the seepage rate of the storage layer was set equal to or greater than the saturated hydraulic conductivity of the soil layer, i.e., 5.99 ft/day (1.83 m/day or 3.00 in./h). This configuration assumes that the underlying native soil is highly permeable and imposes no restriction on drainage above it, thereby emulating a condition where water freely exits the storage layer without accumulation, as Figure 8a always shows zero values on the storage level.

The soil wetting process in Figure 9a is the same as case 1 in Figure 8a (before 1.67 h). When the soil layer is saturated and the surface ponding depth reaches the 2 ft (24 in. or 0.6 m) maximum berm height at 2.17 h (2:10), the surface infiltration rate, the soil percolation rate, and the exfiltration rate are all equal to the soil’s saturated hydraulic conductivity (3.00 in./h or 1.83 m/day, Figure 9a) from 2.17 to 8.0 h, as representing the constant-head infiltration rate tests. From 8.0 to 16.0 h, it mimics the falling-head infiltration rate tests with three identical downward vertical flow rates (5.99 ft/day or 3.00 in./h or 1.83 m/day). This behavior is directly attributable to the elevated seepage rate, which always exceeds or matches the percolation rate, thus preventing any accumulation within the gravel layer. Therefore, case 2 (or Design F) SWMM modeling results indicate that both the constant-head and falling-head infiltration rates would give the soil’s saturated hydraulic conductivity, especially when 6 in. (15.2 cm) of pea gravel and the wire mesh replace geotextile at the top and the bottom of the gravel (#57 stone) layer to remove potential vertical flow restrictions. That is also the reason why cases 1 and 2 SWMM models use 5.99 ft/day (3.00 in./h or 1.83 m/day) as the soil’s saturated hydraulic conductivity, based on the measured infiltration rate for Design F in Table 6.

Figure 9. SWMM simulated (a) surface depth, storage level, and volumetric soil moisture; (b) surface infiltration rate, soil percolation, and exfiltration of a bioretention cell with soil hydraulic conductivity of 3.00 in./h and soil seepage rate of ≥3.00 in./h (5.99 ft/day or 1.83 m/day) (1 in. = 2.54 cm, 1 in./h = 0.61 m/day).

Figure 9. SWMM simulated (a) surface depth, storage level, and volumetric soil moisture; (b) surface infiltration rate, soil percolation, and exfiltration of a bioretention cell with soil hydraulic conductivity of 3.00 in./h and soil seepage rate of ≥3.00 in./h (5.99 ft/day or 1.83 m/day) (1 in. = 2.54 cm, 1 in./h = 0.61 m/day).

Figure 10 presents SWMM modeling results (case 3) to mimic ALDOT infiltration swale or Design A in Table 6 when the soil’s saturated hydraulic conductivity is set at 0.62 ft/day (0.19 m/day) and the seepage rate at 2.25 ft/day (0.69 m/day). This configuration assumes that the soil’s infiltration rate is smaller than the geotextile’s seepage rate. Surface infiltration starts at 3.36 in./h (2.05 m/day) and decreases to 0.31 in./h (0.19 m/day) at 6.42 h when the soil is saturated (takes a long time to saturate with smaller K_s). The seepage rate is 1.13 in./h (0.69 m/day) at t = 0 to drain the initial water storage to zero, then increases to the soil’s saturated hydraulic conductivity when the soil moisture is large enough for saturation. Figure 10 is similar to Figure 8, where the measured infiltration rate from the falling-head test would give the soil’s saturated hydraulic conductivity, as the geotextile seepage is large enough, resulting in no restriction to small vertical flow through the soil layer with topsoil. Due to the small hydraulic conductivity, the falling head process (t > 12 h) is very slow, and Figure 10 only shows a small portion of it. The constant head surface depth of 2 ft (0.6 m) ranges from 1.67 to 12 h (Figure 10a), but the infiltration rate changes with time before point A (Figure 10b). If the data collection from the constant-head tests starts at 1.67 h, measured infiltration rates do not reflect the soil’s saturated hydraulic conductivity. If the data collection starts at point A (6.42 h), it gives the hydraulic conductivity. When the column tests use a very large inflow rate to fill, it reduces the time reaching 2 ft (0.6 m) waterhead (e.g., 1.67 h) and may have minimal change to the timing of point A because of the small infiltration rate. It means it must wait for several hours before collecting data (water discharge from the column bottom) for the constant head tests.

It is worth noting that the SWMM LID module has three layers and cannot represent the effect of geotextile on vertical flow between the field sand and the storage layer, such as Designs A, A-1G, A*, B, C, and B*. It does not show the impact of grass on the surface infiltration, even though SWMM LID does consider that the dense grass reduces the available volume for the surface ponding. Also, SWMM assumes/models a single soil moisture content value (θ₂ in Equation (4)) and is an approximation or represents average moisture content in the soil layer, as it can change along the soil depth.

Figure 10. SWMM simulated (a) surface depth, storage level, and volumetric soil moisture; (b) surface infiltration rate, soil percolation, and exfiltration of a bioretention cell with soil hydraulic conductivity of 0.31 in./h (0.62 ft/day or 0.19 m/day) and soil seepage rate of 1.13 in./h (2.25 ft/day or 0.69 m/day) (1 in. = 2.54 cm, 1 in./h = 0.61 m/day).

Figure 10. SWMM simulated (a) surface depth, storage level, and volumetric soil moisture; (b) surface infiltration rate, soil percolation, and exfiltration of a bioretention cell with soil hydraulic conductivity of 0.31 in./h (0.62 ft/day or 0.19 m/day) and soil seepage rate of 1.13 in./h (2.25 ft/day or 0.69 m/day) (1 in. = 2.54 cm, 1 in./h = 0.61 m/day).

3.4.2. Comparison Between Measured and Estimated Infiltration Rates

The challenge for the SWMM modeling of the ALDOT infiltration designs is how to select/specify the soil layer’s saturated hydraulic conductivity, as the soil layer contains a layer of topsoil or the mixture at the surface and a field sand layer below (Figure 4). One method is to use Equation (8) to determine the equivalent saturated hydraulic conductivity K_s. For each column test (Table 6), the height of each media layer (topsoil/mixture and sand), h₁ and h₂ in Equation (8), is given. If one can specify each layer’s conductivity (K_s1 and K_s2), then K_s of the two-layer media can be calculated and used as the SWMM soil layer’s saturated conductivity to perform the modeling of an infiltration swale design.

For topsoil and the mixture (50% pine bark fines and topsoil by volume), Table 4 and Table 5 list measured infiltration rates of 6″ (15.2 cm) loose samples under 2 ft (0.6 m) of falling head, which can be considered as their saturated conductivity, by assuming that geotextile at the falling-head columns under 2 ft (0.6 m) of water head does not restrict the vertical flow. For field sand, Figure 5 and Figure 6 give measured permeability under the constant head, which are the saturated hydraulic conductivities (derived from Darcy’s Equation (1)) under different degrees of compaction at different hours. These measured conductivities have relatively large ranges, so instead of using their average measured conductivities as values of K_s1 and K_s2 to determine K_s of the two-layer media, potential ranges of K_s for each swale-medium design in Table 6 are calculated using 100 possible values (uniform distributions) of K_s1 and K_s2 within their measured ranges. The minimum, maximum, mean, and standard deviation values of these 10,000 possible K_s values are calculated and reported in Table 6 as estimated K_s of the soil layer. The estimated K_s may or may not be the infiltration rate of a swale medium design that includes other components (geotextile or wire mesh and gravel or grass), depending on how other components affect the vertical flow. For example, the estimated K_s of the soil layer for Designs A, A-1G, and A* are the same, but these designs have different measured infiltration rates. Based on Figure 10 (assuming the flow through the geotextile is larger than through the topsoil), measured infiltration rates for Designs A, A-1G, and A* give saturated hydraulic conductivity of the topsoil–sand layer and have some variations (ranging from 0.16 to 1.28 ft/day or 0.05 to 0.39 m/day) but in the range of the estimated K_s (0.59–3.02 ft/day or 0.18–0.92 m/day).

Designs B to F3 (including their variations) in Table 6 have 6 to 10 in. (15.2–25.4 cm) of the mixture and improved the overall infiltration rate of the engineered medium with three–four layers. Except for Design F3 with grass, the estimated Ks of the soil layer are generally larger than the measured overall infiltration rate of the column tests. Based on Figure 8, measured infiltration rates from Design F variations with pea gravel and wire mesh correspond to the soil layer’s saturated hydraulic conductivity or estimated Ks. For F3 design with grass, estimated Ks ranges from 8.58 to 18.7 ft/day (2.62–5.70 m/day) and agrees well with measured saturated hydraulic conductivities or infiltration rates under both falling-head at 3.52–24.25 ft/day (1.07–7.39 m/day) and constant-head tests at 7.48–21.31 ft/day (2.28–6.60 m/day) (Table 6).

If the soil is saturated, the seepage rate is large enough and the surface infiltration rate predicted by the SWMM LID model is constant and equal to the soil’s saturated hydraulic conductivity. If this is true, the infiltrated water versus time would be a straight line (linear). These data points in Figure 11 for the three samples of Design A (ALDOT swale, Table 6) show three non-linear correlations and fit well with the third-order polynomial functions. Infiltration rates determined between two data points in sequence also decreased with time; for example, the infiltration rate of Sample 3 was 4.13 ft/day (1.26 m/day) at the beginning and decreased to 0.906 ft/day (0.276 m/day) at the end of the test (1.14 days). The average infiltration rate was reported as 1.75 ft/day, which is 2 ft (0.6 m) of water divided by 1.14 days (total time for water to infiltrate through the media). The average infiltration rates for Samples 1 and 2 were 0.88 and 0.84 ft/day (0.27 and 0.26 m/day) for the same media configuration, but they are about half of the infiltration rate of Sample 3. The measured data (water height versus time) for a falling-head test of Design A show that the calculated infiltration rate from the data varies with time, e.g., reflecting head changes, but the SWMM LID module with simplification gives a constant infiltration rate (Figure 9 and Figure 10).

3.4.3. Swale Performance Sensitive to Estimated Hydraulic Conductivities

Four ALDOT infiltration swales were constructed in drainage channels along a roadway project in southern Alabama, and this project site is used to examine LID performance with respect to the estimated soil’s hydraulic conductivity. The drainage area includes 1.55 acres (6273 m²) of road, 4.06 acres (16,430 m²) of undeveloped area, 0.044 acres (178 m²) of roof areas, 0.019 acres (77 m²) of driveway, and 0.092 acres (372 m²) of channels [30]. The seepage rate to native soil is 0.43 in./h (0.26 m/day) for the hydrologic soil group B [30]. The bottom width of the drainage channels is 6 ft (1.8 m), which is the width of the four infiltration swales. The total length of the four infiltration swales is 670 ft (204 m); therefore, the total surface area of the LIDs is 4020 ft² (0.09228 acres or 373 m² for SWMM input). The height of these ALDOT infiltration swales is 4 ft (48 in. or 1.22 m) deep. Swale designs A, B, and F (Table 6) are tested with several potential estimated hydraulic conductivities, and swale results from SWMM models are summarized in

Table 7. Swale runoff-control performance at different conductivities of the soil layer.

Name	Conductivity ft/day (m/day)	Runoff ft3 (m3)	Runoff/Inflow (%)	Infiltration ft3 (m3)	Percolated ft3 (m3)	Max Storage Level, in. (cm)	Water into Native Soil, in. (cm)
Design A	0.32 (0.10)	17,072 (483)	80.0%	3514 (100)	1425 (40)	2.5 (6.4)	5.4 (13.8)
	0.48 (0.15)	16,592 (470)	77.8%	4580 (130)	2491 (71)	2.5 (6.4)	8.6 (21.9)
	0.64 (0.20)	16,155 (458)	75.7%	5177 (147)	3351 (95)	2.5 (6.4)	11.2 (28.4)
	0.80 (0.24)	15,724 (445)	73.7%	5608 (159)	3938 (112)	2.5 (6.4)	13.0 (32.9)
Design B	0.50 (0.15)	16,304 (462)	76.4%	4947 (140)	2400 (68)	2.5 (6.4)	8.4 (21.2)
	1.00(0.30)	14,993 (425)	70.3%	6340 (180)	4326 (123)	3.3 (8.3)	14.1 (35.8)
	1.50 (0.46)	13,728 (389)	64.4%	7604 (215)	5752 (163)	13.7 (34.9)	16.7 (42.5)
	2.00 (0.61)	12,499 (354)	58.6%	8833 (250)	7078 (200)	23.1 (58.8)	17.1 (43.4)
	2.50 (0.76)	11,384 (322)	53.4%	9949 (282)	8178 (232)	26.0 (66.0)	17.3 (43.9)
	3.00 (0.91)	10,693 (303)	50.1%	10,640 (301)	8806 (249)	26.0 (66.0)	17.4 (44.2)
	3.50 (1.07)	10,675 (302)	50.0%	10,658 (302)	8855 (251)	26.0 (66.0)	17.4 (44.3)
	4.00 (1.22)	10,666 (302)	50.0%	10,667 (302)	8890 (252)	26.0 (66.0)	17.5 (44.4)
	4.50 (1.37)	10,660 (302)	50.0%	10,672 (302)	8919 (253)	26.0 (66.0)	17.5 (44.4)
	5.00 (1.52)	10,654 (302)	49.9%	10,678 (302)	8945 (253)	26.0 (66.0)	17.5 (44.5)
	5.50 (1.68)	10,649 (302)	49.9%	10,683 (303)	8968 (254)	26.0 (66.0)	17.5 (44.5)
	6.00 (1.83)	10,645 (301)	49.9%	10,688 (303)	8989 (255)	26.0 (66.0)	17.5 (44.5)
	6.50 (1.98)	10,640 (301)	49.9%	10,692 (303)	9009 (255)	26.0 (66.0)	17.6 (44.6)
Design F	2.50 (0.76)	11,384 (322)	53.4%	9949 (282)	8164 (231)	26.0 (66.0)	17.3 (43.8)
	5.00 (1.52)	10,758 (305)	50.4%	10,574 (299)	8841 (250)	26.0 (66.0)	17.5 (44.4)
	7.50 (2.29)	10,737 (304)	50.3%	10,596 (300)	8940 (253)	26.0 (66.0)	17.5 (44.6)
	9.00 (2.74)	10,726 (304)	50.3%	10,606 (300)	8984 (254)	26.0 (66.0)	17.6 (44.6)
	11.50 (3.51)	10,712 (303)	50.2%	10,620 (301)	9043 (256)	26.0 (66.0)	17.6 (44.7)

.

The SWMM models were under 2.4 in. (6.1 cm) NRCS Type III rainfall over 24 h for a 48 h simulation. The rainfall used is the 95th percentile rainfall for southern Alabama for the ALDOT LID designs [10]. Simulated runoff volume is 21,332 ft³ (604 m³) from the drainage area into the infiltration swale. For Design A with topsoil, the estimated conductivities used for SWMM models increase from 0.32 ft/day to 0.80 ft/day (0.10 m/day to 0.24 m/day). For Design B utilizing 10 in. (25.4 cm) amended topsoil, the estimated saturated hydraulic conductivities in the SWMM models increase from 0.50 ft/day to 6.50 ft/day (i.e., from 0.15 m/day to 1.98 m/day). Similarly, for Design F, the conductivity increases from 2.5 ft/day to 11.5 ft/day (i.e., from 0.76 m/day to 3.51 m/day).

As shown in Table 7, runoff volume (flowing from the swale surface to the downstream outlet) linearly decreases with increasing conductivity below 3 ft/day (0.91 m/day), i.e., infiltration linearly increases. However, once the conductivity reaches 0.91 m/day, further increases lead to almost no reduction in runoff volume or no increase in surface infiltration. The runoff volume is about 50% of the inflow as the conductivity is over 0.91 m/day and the surface berm height is 6 in. (15.2 cm) for all these simulations in Table 7. This is because, after the soil layer becomes saturated, the infiltration rate equals the saturated conductivity until the storage layer also becomes saturated, at which point the infiltration rate declines to the underlying seepage rate, as shown in Figure 8, Figure 10 and Figure 12. In this process, higher infiltration rates shorten the time required for the storage layer to reach saturation, resulting in only minor differences in total infiltration volume, as illustrated in Figure 12b (infiltration volumes, the areas under all different curves, are about the same).

Table 7 further confirms this trend. When conductivity is greater than 2.5 ft/day (0.76 m/day), the maximum storage level reaches 26 in. (66.0 cm), the full height of the storage layer, indicating that the storage layer becomes saturated under these conditions. Figure 12a shows detailed results for the case run when the soil’s conductivity is 1.5 in./h (0.91 m/day or 3 ft/day) under NRCS III rainfall. Before the soil becomes saturated at 12 h, rainfall losses through surface infiltration, percolation, and exfiltration (2nd y axis in Figure 12a) are limited by rainfall or runoff inflow. After the soil is saturated, surface infiltration is the same as percolation from the soil to the storage layer.

The infiltration rate constant and falling head tests (Table 6) show that the SWMM modeling faces a challenge of selecting appropriate soil’s saturated hydraulic conductivity. The sensitivity study shows that the swale performance in controlling the runoff is not sensitive to the soil’s conductivity when above 3 ft/day (0.91 m/day). The recommended optimal design F3 (Table 6) has a conductivity greater than 3.52 ft/day (1.07 m/day), so selecting the conductivity may not be an issue for the SWMM modeling on F3 swale. The sensitivity analysis (Table 7 and Figure 12b) also shows that the swale’s performance in controlling the surface runoff depends on not only the soil’s saturated conductivity but also the swale’s other characteristics, e.g., soil porosity and height, storage layer’s porosity and height, and native soil’s seepage rate. All these characteristics can affect when both the soil and storage layers become saturated, and the infiltration rate is limited by the seepage rate. This topic should be further investigated, as this study involves a simple sensitivity analysis for one study site.

4. Summary and Conclusions

ALDOT infiltration swales along the roadways mimic linear bioretention cells along drainage channels and with complete grass coverage at the surface layer (Figure 1). In this study, a series of modified permeameters (Figure 2) and infiltration rate testing columns (Figure 3) were designed and constructed. Permeabilities and infiltration rates of basic swale materials and various infiltration swale designs were determined using these small-scale test columns. SWMM models were developed and simulations were performed to understand the swale’s infiltration process and column test results. The following conclusions can be drawn from the study:

• The permeability of the topsoil used in this study (~88% sand content) as the top layer in the existing ALDOT infiltration swale design was measured as 0.58 m/day (1.90 ft/day), only about 1% of the field sand’s permeability (Table 2). The average measured infiltration rate of topsoil under falling head tests was 0.63 ft/day (0.19 m/day). Low permeability and infiltration rate of topsoil is the limiting factor of the low permeabilities (0.02–0.69 m/day or 0.07–2.26 ft/day) of several ALDOT and GDOT swale designs (Table 3). The topsoil based on ASTM D 5268 could contain 10–90% silt and clay and could even have much lower permeability than the topsoil used in this study; therefore, using a topsoil layer in the infiltration swale designs is not recommended and can potentially greatly reduce the infiltration capacity/rate of the swales.
• An amended topsoil mixture, consisting of pine bark fines and topsoil, was proposed to replace the topsoil layer. Infiltration rates of the mixtures with 5–75% of pine bark fines by weight were determined and ranged from 0.23 to 7.81 m/day (0.75 to 25.62 ft/day). An amended topsoil with 20% pine bark fines and 80% topsoil by weight (or 50% by volume for each) was used for new infiltration swale designs, and its average infiltration rate is 1.71 m/day (5.60 ft/day), which is much larger than the ALDOT swale’s drainage rate of 1 ft/day (0.3 m/day).
• There were 15 different infiltration swale designs (Table 7, including existing ALDOT design) that were tested using 2 ft (0.6 m) constant-head and falling-head columns. Even replacing the geotextile at the column bottom with stainless-steel wire mesh and replacing the geotextile between sand and #57 stones with 6 in. (15.2 cm) of pea gravel increased the swale infiltration performance, while using 6–10 in. (15.2–25.4 cm) of amended topsoil significantly increased the swale infiltration rates (Table 6).
• The F3 design with grass, consisting of 6 in. (15.2 cm) of amended topsoil, 10 in. (25.4 cm) sand, 6 in. (15.2 cm) peak gravel, and 9 in. (22.9 cm) #57 stone, has average infiltration rates of 11.66 ft/day (3.55 m/day) under falling-head tests and 13.73 ft/day (4.18 m/day) under constant-head tests, which are 37.6 and 15.1 times larger, respectively, than the existing ALDOT design with grass under the same tests (Table 6).
• SWMM modeling of the column test cases reveal that measured infiltration rates from constant-head and falling-head column tests give the soil’s saturated hydraulic conductivities if the pea gravel layer between sand and stone layers and the wire mesh at the column bottom do not restrict the vertical flow; otherwise, they indicate the overall average infiltration rates for the particular swale design.
• Sensitivity of the surface runoff from a southern Alabama project site shows that the infiltration capacity of the swale design linearly increases with the estimated saturated hydraulic conductivity below 0.91 m/day (1.5 in./h). Once the saturated hydraulic conductivity exceeds 0.91 m/day (1.5 in./h), the runoff volume stabilizes at approximately 50% of the total inflow volume. The recommended optimal design F3 (Table 6) has a conductivity greater than 3.52 ft/day (1.07 m/day), so selecting the conductivity may not be an issue for the SWMM modeling on F3 swale.

The F3 design is a proposed optimal infiltration design and is subject to field-scale tests at the next step. Avoiding the use of the topsoil layer but using the amended topsoil is essential to the infiltration swales used in the drainage channels of the roadways.