Revisiting Design Criteria for Stormwater Treatment Systems, Part 3
Flow-through treatment swales & strips
This is the third of a four-part series examining design criteria for stormwater treatment systems. Basins and fine-media filters were the focus of the first two articles. In this article, the focus is on grass-covered, flow-through treatment swales and strips. Stormwater is treated as it passes through the swale. Infiltration may occur but is not relied upon to contribute to performance.
In the 25 years since US communities first began requiring post-development treatment of stormwater, local, regional, and state governments have published many manuals and handbooks identifying acceptable treatment systems (structural best management practices, or BMPs) and design criteria. As noted in the first part of this series, few of the criteria were initially supported by laboratory or field research, and engineers relied on their best professional judgment in choosing design criteria.
Moreover, treatment strategy is evolving from a focus on the general removal of pollutants using total suspended solids (TSS) as the surrogate for all pollutants to the more recent trend to consider specific pollutants. In certain situations, some states now emphasize removal of total phosphorus, total nitrogen, and dissolved metals. Design criteria can vary depending on which pollutants are targeted.
Terminology
With a flow-through swale, the bulk of the stormwater passes through and exits at the downslope end of the swale. A strip, a lateral swale in which water enters along its longitudinal length, is considered by many to be a viable BMP for roads without curbs. The term biofiltration is commonly used to identify swales and strips—e.g., biofilters and filter strips. However, the grass covering does not filter out the particulates. Sedimentation is the dominant removal process for sediment and attached pollutants. Swales and strips are more appropriately characterized as biosettlers: they are shallow settling basins (Minton 2002).
Flow-through swales are designed based on the peak flow of the design event. Several states, most particularly along the East Coast, design swales to capture and retain the design event volume. The bulk of the water exits either through underdrains or by infiltration. These types are referred to as wet and dry swales, respectively. Here the swale is, in effect, either a long, shallow, extended detention basin or an infiltration basin. Sometimes swales whose method of discharge is infiltration are referred to as infiltration or bioinfiltration swales. Some manuals specify 0.5 in/hr as a precondition for the use of a swale or strip, the same criterion used for infiltration basins and trenches. Significant infiltration as a precondition prevents the use of swales and strips with more restrictive soils, while the opposite is the case with flow-through swales. Alternative definitions for the same terminology can lead to confusion regarding suitable design criteria and observed performance. This article focuses solely on swales and strips in which the bulk, if not all of the stormwater, exits the downslope end.
With flow-through swales and strips, grass is critical to success. Grass must remain erect through all storms up to the peak of the design event with the water surface below the top of the grass. Erect grass provides high resistance to flow. The resistance increases the depth of the water in the grass while decreasing flow velocity. The outcome is settling of sediment and attached pollutants (Minton 2002). Table 1 provides design criteria for flow-through swales from representative manuals.
Determining the Width
Sizing a flow-through swale consists of two steps: determining the bottom width and then determining the length. The width reflects the peak flow rate up to which effective treatment is desired based upon a management goal of treating the majority—e.g., 90%—of the stormwater generated over time. However, the length determines the performance.
Several equations are available to calculate the appropriate bottom width of a channel or swale (Minton 2002). Commonly used is Manning's equation, presented at right.
Width appears twice on the right side of the equation: as part of A, the vertical cross-sectional area of the flow, and as part of R, the hydraulic radius. The mathematical form of the hydraulic radius differs with the cross-sectional shape. The most common shape is trapezoidal, expressed in Figure 1.
Solving for bottom width is an iterative process, easily done with a spreadsheet program. Known are the peak flow of the design event, the design resistance coefficient, and the slope. Water depth at the peak flow is either selected by the design engineer or specified by the manual. The proper selection of the resistance coefficient, discussed later in this article, is critical to ensure that the depth of water at the peak design flow remains below the top of the grass and that the grass remains essentially erect. The appropriate value for Manning's coefficient, n, is discussed later.
Determining the Length
Once the bottom width is determined, the length is determined. Early manuals in the 1980s specified a swale length of 200 feet (about 60 meters). The origin of this specification was the study of a freeway ditch not designed to treat stormwater (Wang et al. 1982). In this study, removal efficiency was evaluated along the length of the grassed ditch. By happenstance, about 80% removal occurred at 200 feet with little apparent additional removal beyond this distance. Controlled flow tests of field swales have found 65 meters (about 215 feet) and 75 meters (about 246 feet) effective (Kuo et al. 1999; Fletcher et al. 2001, respectively). Some recently published manuals use detention time to determine the length, but specify a minimum length of 100 feet (about 30 meters).
Using the nominal detention or hydraulic residence time to define the length was introduced (Washington 1992) based on Kahn et al. (1992). The length is determined with Equation 2. The nominal detention time in Equation 2 is at the peak flow of the design event.
The information required to determine the water velocity is known: peak of the design event, bottom width, and depth of water at the peak of the design event. The term nominal is used because the above equation ignores the effects of the cross-sectional area of the grass on flow velocity (Backstrom 2002) and longitudinal dispersion, and in turn the real detention time. The actual detention time is likely to be considerably less than the calculated nominal detention time. Values currently used for detention time and their origin are presented later in this article.
Resistance Coefficient: Manning's n
The specification of Manning's n, the resistance coefficient, establishes the bottom width of the swale: The greater the value selected, the greater the bottom width. As noted previously, to function the vegetation must remain erect up to the depth of the peak flow of the design event. As such, the resistance to flow is significant. For a channel whose sole purpose is to carry water, with prone grass at high flows, the coefficient is typically taken to be 0.03. Initially, manuals specified this value not recognizing the importance of erect grass. At this value, the swale is too narrow and the result is that the grass becomes prone at a flow rate considerably less than the peak of the design event. Velocities are too high, and sediments do not settle in the grass. This is likely the reason why studies of some swales found unsatisfactory performance. These swales were likely sized with low resistance coefficients. The grass became prone in the storms sampled. In any given swale or strip, Manning's coefficient likely increases with flow depth. The value of interest is at the point of maximum water depth, occurring at the peak of the design event.
Unknown until recently to stormwater engineers, field studies conducted in the 1940s by agricultural engineers found resistance coefficients are very high when the grass remains erect and the water surface is below the top of the grass, on the order of 0.25 to 0.50 (Minton 2002). Once the grass begins to lie down in the direction of the flow, the resistance decreases and flow velocities increase, reducing treatment effectiveness. The relationship between water depth, grass height, and resistance is illustrated in Figure 2. Note that as the water level rises the resistance increases to the point where the grass begins to lie down. From this point resistance decreases as the water level continues to rise.
Eventually the grass becomes prone where it offers the least resistance. The increase in resistance with water depth below the height of the grass is probably because the grass blades are thicker at their upper ends, offering greater resistance to flow. Note that the grass begins to bend over at a water depth considerably below the top of the grass. This will be discussed in more depth later in this article.
Currently, most manuals specify a value on the order of 0.20 to 0.25, apparently based on one study of a treatment swale (Kahn et al. 1992). These values are consistent with prior work by agricultural engineers. Laboratory studies found Manning's coefficient values of 0.20 to 0.50 (Johnson et al. 2003). The authors reaffirmed the classic relationship between the coefficient and the product of the hydraulic radius and flow velocity (Ree 1949), but for flows where the water is below the top of the grass. A recent study of 14 field swales found the coefficient to vary from 0.19 to 0.53 with a median value of 0.28 (Colwell 2001). Colwell recommended a design value of 0.30. Interestingly, the study found that the resistance coefficient was high even in swales with patchy grass cover. Some manuals specify two values for Manning's coefficient based on the recommendation of Kahn et al. (1992): mowed, 0.20, and unmowed, 0.24. However, it is not certain that the observed difference was real, given the limited number of tests and the inaccuracy of measurements of low flow as noted by Kahn et al. (1992). It should be noted that mowed in this context does not mean regular mowing but rather that the grass is cut occasionally, perhaps only once after the early spring surge of growth. In this context, the true distinction should be between tall and short grass, not mowed and unmowed. Regardless, given the findings of Colwell (2001), distinguishing between unmowed and occasionally mowed grass is likely pointless. As pointed out by Kahn et al. (1992), regularly mowing may result in increased grass-stem density or thickness in its lower stems (taller grass tends to thin due to shading), increasing the resistance to flow. This has been the apparent reasoning for specifying a greater value for strips (Table 1). However, it seems more appropriate to distinguish between swales and strips that are unmowed or occasionally mowed with those regularly mowed to relatively short heights. The latter, in which the resistance coefficient is likely much higher, tends to occur in retail commercial developments where property owners are concerned about the aesthetics of the landscaping.
Note that regardless of the value selected for Manning's coefficient, the total area of the swale remains the same, for the same design peak flow and longitudinal slope. Using a higher Manning's coefficient increases the width of the swale, but correspondingly decreases its length. Hence, selecting a higher value does not increase the land dedicated to the treatment system. However, with a high Manning's coefficient, the width of the swale increases to a point of impracticality. If the swale is too wide, the stormwater cannot effectively spread across the entire width. Recognizing this point, some manuals specify a maximum width (Table 1). Others specify that lateral flow spreaders be placed at 50-foot intervals along the swale. There are, however, no data indicating the effects of either criterion on the hydraulic efficiency of swales. Regardless, specifying a maximum width limits the drainage area that can be served by a swale. A value in the range of 0.20 to 0.30 is recommended based on current knowledge. Values toward the higher end of this range should be used if the grass is kept relatively short (e.g., 6 inches) and well maintained.
Maximum Water Depth
Few manuals specify a maximum water depth at the design peak flow. Therefore, design engineers might choose a substantial depth (e.g., 12 inches) to minimize the width. In such cases, the requirement for erect grass is not likely to be met, particularly if the grass is regularly mowed. It is prudent to specify a maximum depth (Table 1).
Furthermore, grass becomes prone at water depths less than the grass height. Some manuals specify a water depth at the design flow but not a corresponding grass height, although they might, in a maintenance section, recommend the height or height range to which the grass should be cut. The latter is usually greater than the former. But the relationship between the two is not stated in the manual. Perhaps the relationship should be stated more explicitly in design manuals. Failure to specify the grass height in relationship to the water depth at design flow may result in the grass lying down at flows less than the design flow. Studies (Samani and Kouwen 2002, Ree 1949) suggest that significant bending and decline of the resistance to flow occurs when the water depth is less than half the height of the grass (Figure 2). Therefore, both the maximum water depth and grass height should be specified, with the grass height twice the depth of the water depth at the peak of the design event.
Detention Time
The concept of detention time derives from one study (Kahn et al. 1992) of a properly designed and constructed treatment swale. Performance was evaluated at two sets of average detention times: one of 4.3 to 5.6 minutes, taken as 4.5 minutes, and the second set of 8.3 to 9.5 minutes, taken as 9 minutes. The detention times represented samples taken at 100 and 200 feet, respectively, of a 200-foot swale. It was concluded that there was no statistically significant difference in performance between the two detention times, although the data suggest that performance was poorer at the lower detention time. Kahn et al. (1992) suggested "that a residence time of 9 minutes is sufficient to assure good pollutant removals. A minimum hydraulic residence cannot be given with certainty, although it can be said that with residence times of 4.5 minutes, deterioration in performance is likely, especially during larger storms."
The conclusions of Kahn et al. (1992) were based on only six storms sampled at each length. Furthermore, they were not identical storms; that is, during each storm sampling occurred at either 100 or 200 feet, not both. Recognizing the considerable uncertainty inherent to the study, the State of Washington chose 9 minutes as the peak rather than the average flow of the design event. Although this decision appears overly conservative, the detention time does not vary substantially with flow rate. This is because the water depth—and therefore volume in the swale—rises and falls with the flow rate. The detention time remains relatively constant during the storm, given that the resistance to flow is less at lower water depths (Figure 1). Hence, calculating the length based on the average flow of the design event would produce essentially the same length of swale.
Some manuals have included the range of 5 to 9 minutes as the design criterion, based on the qualified conclusions of Kahn et al. (1992). However, if given a range, it is reasonable to expect the design engineer will select 5 minutes as it gives a shorter swale. Jurisdictions should identify one value for the detention time if they use this criterion.
The work of Yu et al. (2001) is reasonably consistent with Kahn et al. (1999). They tested swales with controlled flows at detention times of 5.5, 7, 10, and 18 minutes. Removal efficiencies for TSS were 48%, 70%, 67%, and 86% respectively. These are mass reductions. The role of infiltration was not defined. Backstrom (2002) developed a relationship between performance and detention time using laboratory and field swales. The relationship suggests that if high (about 90%) removal of particles down to 15 microns is desired, a detention time on the order of 8 minutes is necessary for swales with modest or no infiltration. Consistent removal of particles smaller than 15 microns requires a considerably larger detention time according to Backstrom (2002). A swale with high infiltration appears to require a lower detention time (Backstrom 2002) to obtain the desired performance.
The classic kinetic removal equation has been proposed for sizing swales (Fletcher et al. 2001). The weakness of this approach, however, is the selection of the "average" influent and apparent background (lowest possible) concentrations, and the kinetic removal rate constant. Furthermore, the kinetic rate constant likely varies with the influent concentration and hydraulic loading rate (Kadlec 2000), complicating the use of the concept. Nonetheless, the concept should continue to be evaluated in field or laboratory studies.
Although detention time has become the common criterion to specify length, the proper design criterion may be hydraulic loading rate (Mazer 1998). As noted previously, a swale or strip is a shallow settling basin. As such, performance during a storm is a function of the hydraulic loading rate (Minton 2002): effectively, the flow rate divided by the plan view (width times length) of the swale or strip. However, most studies do not include the information necessary to calculate the hydraulic loading rate and in turn its relationship to performance. One study of swales under controlled flows found the performance decreased with increasing hydraulic loading (Fletcher et al. 2001). Values ranged from about 0.04 to about 0.23 ft/ft2. Performance dropped from about 95% to about 75% at the upper end of the loading range.
Longitudinal Slope and Internal Flow Control
Concerns for flow convergence and erosion led early to the specification of the maximum longitudinal slope. Maximum slopes vary widely in manuals: 1% to 5%. Some manuals specify that swales not be used unless the slope is less than the maximum allowed. Other manuals allow their use if check dams are included. Check dams reduce velocities and the potential for flow convergence. They retain some water, which may infiltrate and increase performance (Kaighn and Yu 1996, Yousef et al. 1985). One study found convergence in swales with slopes greater than 1% unless check dams were present (Mazer 1998). Another (Colwell 2001) found sufficient flow convergence at slopes greater than 2.5% to result in channelization. Some manuals include lateral flow spreaders regardless of the slope. Flow spreaders differ from check dams in that the top elevation is just slightly above the invert of the swale, facilitating mowing. However, there are no data or visual observations indicating that flow spreaders (or check dams) prevent flow convergence. Certainly a flow spreader should be placed at the entrance of the swale to spread the water as it exits the drainage pipe.
Filter Strips
Table 1 indicates that design criteria differ greatly for strips. Many manuals do not allow filter strips as "standalone" treatment systems. They do not view filter strips as equivalent in performance to other treatment systems such as basins, swales, and filters. However, performance data suggest otherwise (Caltrans 2004a and 2004b), indicating effective removal of TSS and attached pollutants occurs within a few meters of the pavement edge. The issue is maintaining sheet flow into and through the strip. Attention to pavement edge design and construction is important. Curbs with curb cuts should be avoided as convergent flow occurs. Exposing the entire length of the strip to inflow is preferred. Manuals should specify the maximum width of pavement allowed to drain to the strip. Failure to do so could result in large lateral flows into the strip per lineal foot or meter, exacerbating tendencies toward concentration of flow and rutting of the strip.
Not well established is the appropriate specification of the maximum width of pavement that should drain to a strip. This is likely a function of rainfall intensity; that is, the maximum allowable pavement width should be lower for regions with high-intensity rainfall like the southeastern and south-central United States. One jurisdiction specifies the maximum lineal flow rate to the strip (Table 1). Rainfall intensity is also a factor to consider when establishing the maximum slope. Values differ greatly according to Table 1. Satisfactory performance has been found with slopes exceeding 15% (Caltrans 2004b).
Final Observations
Researchers have found performance to be highly variable, from relatively poor to above 80% of the TSS, meeting the more common performance goals of many manuals. In some cases, poor performance can be attributed to the test facility being a roadside ditch not specifically designed to treat stormwater. It is, however, not possible to relate the variation in performance to alternative design criteria (e.g., detention time or hydraulic loading rate) as most researchers have failed to include the necessary information. This has made a comparative analysis impossible. Key information missing depending on the study may be the swale width; length and/or longitudinal slope; condition of the grass; and key attributes of each sampled storm, such as flow rates, maximum water depth relative to grass height (whether the grass remained erect), and the amount of infiltration. Confusion can also occur because of the varying role of infiltration in each study, and whether the researcher is reporting reductions in concentrations or loadings. To arrive at the appropriate design criteria requires that researchers report more than just influent and effluent concentrations.
Of concern is the erosion of swales due to excessive shear forces, reducing performance. This can be avoided with dense grass and consideration to the hydraulic radius-velocity product. Procedures to consider this question are provided in some manuals. Johnson et al. (2003) recently updated this concept based on laboratory and fieldwork. Resuspension during extreme events of previously removed sediment and pollutants is of concern but remains undefined. Some manuals recommend or require the swale to be offline, bypassing flows greater than the treatment design event. Inadequate maintenance leads to degraded and patchy grass, channelization, and erosion. The performance of swales and strips is more sensitive to poor maintenance than that of basins.
The most extensive studies relating internal characteristics such as grass density and cover to flow and performance outcomes are Colwell (2001) and Mazer (1998). These studies found that grass can experience extended periods of submergence with modest declines in density and health. Neither the detention time nor the resistance coefficient could be correlated to the extent of density of grass cover, to species mix, or to a wide range of flow resistance values. However, another researcher (Backstrom 2002) found that TSS removal improved with denser grass.
Whether swales and strips remove dissolved pollutants remains an open question. Field performances provide differing results. There remains a conflict between the desire to have thick grass to achieve a suitable resistance to flow and to protect the system from erosion with the desire to have the stormwater contact the soil where dissolved pollutants are removed. Without infiltration, the removal of dissolved pollutants is likely minor. Fertilization appears to cause the export of dissolved phosphorus (Caltrans 2004a). Plant die-off may release nutrients. Dissolved pollutant removal will be further explored in the last article of this series.
An inherent limitation of flow-through swales may be that regardless of the design value selected for Manning's coefficient, the actual resistance for most constructed swales at the peak design flow differs from the design value, either higher or lower (Colwell 2001). If the actual value is greater than the design value, the grass becomes prone at flows less than the design flow. As a consequence, the management goal of treating, for example, 90% of the stormwater over time is not met. If the actual value is less than the design value, the depth of water at the design peak rate (and each flow rate) is less than expected and in turn the flow velocity is greater than expected. As a consequence, the detention time and therefore performance is less than expected.
Practical limitations on system width and grass height constrain the use of swales to small drainage areas of a few acres or less. They become less practical in regions with high intensity rainfall such as the southeastern and south-central United States. In such areas, the combined effect of a maximum bottom width and a substantial design peak flow is to limit use to small drainage areas. Swales and strips may not be appropriate for regions with cold winters. Treatment of snowmelt and spring storms may be severely compromised by the temporary degraded condition of the grass.
Summary
With respect to the removal of sediment and attached pollutants, flow-through swales and strips can provide an acceptable level of treatment as a standalone system. They should be able to meet the performance goal of most manuals, 80% removal of TSS. However, less confidence can be placed in the performance of swales in contrast to wet basins and sand filters, particularly if regular maintenance is not performed. Given the uncertainties expressed in this article, it is not unreasonable that some jurisdictions do not find them to be an acceptable treatment system unless significant infiltration is involved.
Satisfactory performance requires that the systems be properly sized. The specified Manning's coefficient should be at least 0.20, and possibly 0.30 where it is expected that the swale will be mowed frequently, as is the case with retail commercial developments. Maximum and minimum longitudinal slopes should be specified. The specification of the maximum slope absent check dams may differ with the climate, being less in areas with higher-intensity storms. The relationship between flow depth and grass height should be recognized in design and maintenance criteria. The manual should specify a maximum water depth at the peak of the design event and a minimum grass height. The latter should be at least twice the former. Most uncertain of the design criteria is length. Our understanding of the relationship between detention time and performance, and the effect of infiltration on this relationship, is poorly understood. A detention time of 9 or 10 minutes is reasonable.
Advertisement
Hydraulic loading rate may be a more appropriate design criterion, but more data on its relationship to performance are needed.
Swales and strips are not "biofilters," as the primary process of particulate pollutant removal is sedimentation. The terms biofiltration and filter should be dropped. The author recommends the terms vegetated swale and vegetated strip.
Author's Bio: Gary R. Minton, Ph.D., P.E., is an independent consultant on stormwater treatment with Resource Planning Associates. He is the author of the book Stormwater Treatment: Biological, Chemical, and Engineering Principles.
March-April 2005
Revisiting Design Criteria for Stormwater Treatment Systems, Part 3
Flow-through treatment swales & strips
This is the third of a four-part series examining design criteria for stormwater treatment systems. Basins and fine-media filters were the focus of the first two articles. In this article, the focus is on grass-covered, flow-through treatment swales and strips. Stormwater is treated as it passes through the swale. Infiltration may occur but is not relied upon to contribute to performance.
In the 25 years since US communities first began requiring post-development treatment of stormwater, local, regional, and state governments have published many manuals and handbooks identifying acceptable treatment systems (structural best management practices, or BMPs) and design criteria. As noted in the first part of this series, few of the criteria were initially supported by laboratory or field research, and engineers relied on their best professional judgment in choosing design criteria.
Moreover, treatment strategy is evolving from a focus on the general removal of pollutants using total suspended solids (TSS) as the surrogate for all pollutants to the more recent trend to consider specific pollutants. In certain situations, some states now emphasize removal of total phosphorus, total nitrogen, and dissolved metals. Design criteria can vary depending on which pollutants are targeted.
Terminology
With a flow-through swale, the bulk of the stormwater passes through and exits at the downslope end of the swale. A strip, a lateral swale in which water enters along its longitudinal length, is considered by many to be a viable BMP for roads without curbs. The term biofiltration is commonly used to identify swales and strips—e.g., biofilters and filter strips. However, the grass covering does not filter out the particulates. Sedimentation is the dominant removal process for sediment and attached pollutants. Swales and strips are more appropriately characterized as biosettlers: they are shallow settling basins (Minton 2002).
Flow-through swales are designed based on the peak flow of the design event. Several states, most particularly along the East Coast, design swales to capture and retain the design event volume. The bulk of the water exits either through underdrains or by infiltration. These types are referred to as wet and dry swales, respectively. Here the swale is, in effect, either a long, shallow, extended detention basin or an infiltration basin. Sometimes swales whose method of discharge is infiltration are referred to as infiltration or bioinfiltration swales. Some manuals specify 0.5 in/hr as a precondition for the use of a swale or strip, the same criterion used for infiltration basins and trenches. Significant infiltration as a precondition prevents the use of swales and strips with more restrictive soils, while the opposite is the case with flow-through swales. Alternative definitions for the same terminology can lead to confusion regarding suitable design criteria and observed performance. This article focuses solely on swales and strips in which the bulk, if not all of the stormwater, exits the downslope end.
With flow-through swales and strips, grass is critical to success. Grass must remain erect through all storms up to the peak of the design event with the water surface below the top of the grass. Erect grass provides high resistance to flow. The resistance increases the depth of the water in the grass while decreasing flow velocity. The outcome is settling of sediment and attached pollutants (Minton 2002). Table 1 provides design criteria for flow-through swales from representative manuals.
Determining the Width
Sizing a flow-through swale consists of two steps: determining the bottom width and then determining the length. The width reflects the peak flow rate up to which effective treatment is desired based upon a management goal of treating the majority—e.g., 90%—of the stormwater generated over time. However, the length determines the performance.
Several equations are available to calculate the appropriate bottom width of a channel or swale (Minton 2002). Commonly used is Manning's equation, presented at right.
Width appears twice on the right side of the equation: as part of A, the vertical cross-sectional area of the flow, and as part of R, the hydraulic radius. The mathematical form of the hydraulic radius differs with the cross-sectional shape. The most common shape is trapezoidal, expressed in Figure 1.
Solving for bottom width is an iterative process, easily done with a spreadsheet program. Known are the peak flow of the design event, the design resistance coefficient, and the slope. Water depth at the peak flow is either selected by the design engineer or specified by the manual. The proper selection of the resistance coefficient, discussed later in this article, is critical to ensure that the depth of water at the peak design flow remains below the top of the grass and that the grass remains essentially erect. The appropriate value for Manning's coefficient, n, is discussed later.
Determining the Length
Once the bottom width is determined, the length is determined. Early manuals in the 1980s specified a swale length of 200 feet (about 60 meters). The origin of this specification was the study of a freeway ditch not designed to treat stormwater (Wang et al. 1982). In this study, removal efficiency was evaluated along the length of the grassed ditch. By happenstance, about 80% removal occurred at 200 feet with little apparent additional removal beyond this distance. Controlled flow tests of field swales have found 65 meters (about 215 feet) and 75 meters (about 246 feet) effective (Kuo et al. 1999; Fletcher et al. 2001, respectively). Some recently published manuals use detention time to determine the length, but specify a minimum length of 100 feet (about 30 meters).
Using the nominal detention or hydraulic residence time to define the length was introduced (Washington 1992) based on Kahn et al. (1992). The length is determined with Equation 2. The nominal detention time in Equation 2 is at the peak flow of the design event.
The information required to determine the water velocity is known: peak of the design event, bottom width, and depth of water at the peak of the design event. The term nominal is used because the above equation ignores the effects of the cross-sectional area of the grass on flow velocity (Backstrom 2002) and longitudinal dispersion, and in turn the real detention time. The actual detention time is likely to be considerably less than the calculated nominal detention time. Values currently used for detention time and their origin are presented later in this article.
Resistance Coefficient: Manning's n
The specification of Manning's n, the resistance coefficient, establishes the bottom width of the swale: The greater the value selected, the greater the bottom width. As noted previously, to function the vegetation must remain erect up to the depth of the peak flow of the design event. As such, the resistance to flow is significant. For a channel whose sole purpose is to carry water, with prone grass at high flows, the coefficient is typically taken to be 0.03. Initially, manuals specified this value not recognizing the importance of erect grass. At this value, the swale is too narrow and the result is that the grass becomes prone at a flow rate considerably less than the peak of the design event. Velocities are too high, and sediments do not settle in the grass. This is likely the reason why studies of some swales found unsatisfactory performance. These swales were likely sized with low resistance coefficients. The grass became prone in the storms sampled. In any given swale or strip, Manning's coefficient likely increases with flow depth. The value of interest is at the point of maximum water depth, occurring at the peak of the design event.
Unknown until recently to stormwater engineers, field studies conducted in the 1940s by agricultural engineers found resistance coefficients are very high when the grass remains erect and the water surface is below the top of the grass, on the order of 0.25 to 0.50 (Minton 2002). Once the grass begins to lie down in the direction of the flow, the resistance decreases and flow velocities increase, reducing treatment effectiveness. The relationship between water depth, grass height, and resistance is illustrated in Figure 2. Note that as the water level rises the resistance increases to the point where the grass begins to lie down. From this point resistance decreases as the water level continues to rise.
Eventually the grass becomes prone where it offers the least resistance. The increase in resistance with water depth below the height of the grass is probably because the grass blades are thicker at their upper ends, offering greater resistance to flow. Note that the grass begins to bend over at a water depth considerably below the top of the grass. This will be discussed in more depth later in this article.
Currently, most manuals specify a value on the order of 0.20 to 0.25, apparently based on one study of a treatment swale (Kahn et al. 1992). These values are consistent with prior work by agricultural engineers. Laboratory studies found Manning's coefficient values of 0.20 to 0.50 (Johnson et al. 2003). The authors reaffirmed the classic relationship between the coefficient and the product of the hydraulic radius and flow velocity (Ree 1949), but for flows where the water is below the top of the grass. A recent study of 14 field swales found the coefficient to vary from 0.19 to 0.53 with a median value of 0.28 (Colwell 2001). Colwell recommended a design value of 0.30. Interestingly, the study found that the resistance coefficient was high even in swales with patchy grass cover. Some manuals specify two values for Manning's coefficient based on the recommendation of Kahn et al. (1992): mowed, 0.20, and unmowed, 0.24. However, it is not certain that the observed difference was real, given the limited number of tests and the inaccuracy of measurements of low flow as noted by Kahn et al. (1992). It should be noted that mowed in this context does not mean regular mowing but rather that the grass is cut occasionally, perhaps only once after the early spring surge of growth. In this context, the true distinction should be between tall and short grass, not mowed and unmowed. Regardless, given the findings of Colwell (2001), distinguishing between unmowed and occasionally mowed grass is likely pointless. As pointed out by Kahn et al. (1992), regularly mowing may result in increased grass-stem density or thickness in its lower stems (taller grass tends to thin due to shading), increasing the resistance to flow. This has been the apparent reasoning for specifying a greater value for strips (Table 1). However, it seems more appropriate to distinguish between swales and strips that are unmowed or occasionally mowed with those regularly mowed to relatively short heights. The latter, in which the resistance coefficient is likely much higher, tends to occur in retail commercial developments where property owners are concerned about the aesthetics of the landscaping.
Note that regardless of the value selected for Manning's coefficient, the total area of the swale remains the same, for the same design peak flow and longitudinal slope. Using a higher Manning's coefficient increases the width of the swale, but correspondingly decreases its length. Hence, selecting a higher value does not increase the land dedicated to the treatment system. However, with a high Manning's coefficient, the width of the swale increases to a point of impracticality. If the swale is too wide, the stormwater cannot effectively spread across the entire width. Recognizing this point, some manuals specify a maximum width (Table 1). Others specify that lateral flow spreaders be placed at 50-foot intervals along the swale. There are, however, no data indicating the effects of either criterion on the hydraulic efficiency of swales. Regardless, specifying a maximum width limits the drainage area that can be served by a swale. A value in the range of 0.20 to 0.30 is recommended based on current knowledge. Values toward the higher end of this range should be used if the grass is kept relatively short (e.g., 6 inches) and well maintained.
Maximum Water Depth
Few manuals specify a maximum water depth at the design peak flow. Therefore, design engineers might choose a substantial depth (e.g., 12 inches) to minimize the width. In such cases, the requirement for erect grass is not likely to be met, particularly if the grass is regularly mowed. It is prudent to specify a maximum depth (Table 1).
Furthermore, grass becomes prone at water depths less than the grass height. Some manuals specify a water depth at the design flow but not a corresponding grass height, although they might, in a maintenance section, recommend the height or height range to which the grass should be cut. The latter is usually greater than the former. But the relationship between the two is not stated in the manual. Perhaps the relationship should be stated more explicitly in design manuals. Failure to specify the grass height in relationship to the water depth at design flow may result in the grass lying down at flows less than the design flow. Studies (Samani and Kouwen 2002, Ree 1949) suggest that significant bending and decline of the resistance to flow occurs when the water depth is less than half the height of the grass (Figure 2). Therefore, both the maximum water depth and grass height should be specified, with the grass height twice the depth of the water depth at the peak of the design event.
Detention Time
The concept of detention time derives from one study (Kahn et al. 1992) of a properly designed and constructed treatment swale. Performance was evaluated at two sets of average detention times: one of 4.3 to 5.6 minutes, taken as 4.5 minutes, and the second set of 8.3 to 9.5 minutes, taken as 9 minutes. The detention times represented samples taken at 100 and 200 feet, respectively, of a 200-foot swale. It was concluded that there was no statistically significant difference in performance between the two detention times, although the data suggest that performance was poorer at the lower detention time. Kahn et al. (1992) suggested "that a residence time of 9 minutes is sufficient to assure good pollutant removals. A minimum hydraulic residence cannot be given with certainty, although it can be said that with residence times of 4.5 minutes, deterioration in performance is likely, especially during larger storms."
The conclusions of Kahn et al. (1992) were based on only six storms sampled at each length. Furthermore, they were not identical storms; that is, during each storm sampling occurred at either 100 or 200 feet, not both. Recognizing the considerable uncertainty inherent to the study, the State of Washington chose 9 minutes as the peak rather than the average flow of the design event. Although this decision appears overly conservative, the detention time does not vary substantially with flow rate. This is because the water depth—and therefore volume in the swale—rises and falls with the flow rate. The detention time remains relatively constant during the storm, given that the resistance to flow is less at lower water depths (Figure 1). Hence, calculating the length based on the average flow of the design event would produce essentially the same length of swale.
Some manuals have included the range of 5 to 9 minutes as the design criterion, based on the qualified conclusions of Kahn et al. (1992). However, if given a range, it is reasonable to expect the design engineer will select 5 minutes as it gives a shorter swale. Jurisdictions should identify one value for the detention time if they use this criterion.
The work of Yu et al. (2001) is reasonably consistent with Kahn et al. (1999). They tested swales with controlled flows at detention times of 5.5, 7, 10, and 18 minutes. Removal efficiencies for TSS were 48%, 70%, 67%, and 86% respectively. These are mass reductions. The role of infiltration was not defined. Backstrom (2002) developed a relationship between performance and detention time using laboratory and field swales. The relationship suggests that if high (about 90%) removal of particles down to 15 microns is desired, a detention time on the order of 8 minutes is necessary for swales with modest or no infiltration. Consistent removal of particles smaller than 15 microns requires a considerably larger detention time according to Backstrom (2002). A swale with high infiltration appears to require a lower detention time (Backstrom 2002) to obtain the desired performance.
The classic kinetic removal equation has been proposed for sizing swales (Fletcher et al. 2001). The weakness of this approach, however, is the selection of the "average" influent and apparent background (lowest possible) concentrations, and the kinetic removal rate constant. Furthermore, the kinetic rate constant likely varies with the influent concentration and hydraulic loading rate (Kadlec 2000), complicating the use of the concept. Nonetheless, the concept should continue to be evaluated in field or laboratory studies.
Although detention time has become the common criterion to specify length, the proper design criterion may be hydraulic loading rate (Mazer 1998). As noted previously, a swale or strip is a shallow settling basin. As such, performance during a storm is a function of the hydraulic loading rate (Minton 2002): effectively, the flow rate divided by the plan view (width times length) of the swale or strip. However, most studies do not include the information necessary to calculate the hydraulic loading rate and in turn its relationship to performance. One study of swales under controlled flows found the performance decreased with increasing hydraulic loading (Fletcher et al. 2001). Values ranged from about 0.04 to about 0.23 ft/ft2. Performance dropped from about 95% to about 75% at the upper end of the loading range.
Longitudinal Slope and Internal Flow Control
Concerns for flow convergence and erosion led early to the specification of the maximum longitudinal slope. Maximum slopes vary widely in manuals: 1% to 5%. Some manuals specify that swales not be used unless the slope is less than the maximum allowed. Other manuals allow their use if check dams are included. Check dams reduce velocities and the potential for flow convergence. They retain some water, which may infiltrate and increase performance (Kaighn and Yu 1996, Yousef et al. 1985). One study found convergence in swales with slopes greater than 1% unless check dams were present (Mazer 1998). Another (Colwell 2001) found sufficient flow convergence at slopes greater than 2.5% to result in channelization. Some manuals include lateral flow spreaders regardless of the slope. Flow spreaders differ from check dams in that the top elevation is just slightly above the invert of the swale, facilitating mowing. However, there are no data or visual observations indicating that flow spreaders (or check dams) prevent flow convergence. Certainly a flow spreader should be placed at the entrance of the swale to spread the water as it exits the drainage pipe.
Filter Strips
Table 1 indicates that design criteria differ greatly for strips. Many manuals do not allow filter strips as "standalone" treatment systems. They do not view filter strips as equivalent in performance to other treatment systems such as basins, swales, and filters. However, performance data suggest otherwise (Caltrans 2004a and 2004b), indicating effective removal of TSS and attached pollutants occurs within a few meters of the pavement edge. The issue is maintaining sheet flow into and through the strip. Attention to pavement edge design and construction is important. Curbs with curb cuts should be avoided as convergent flow occurs. Exposing the entire length of the strip to inflow is preferred. Manuals should specify the maximum width of pavement allowed to drain to the strip. Failure to do so could result in large lateral flows into the strip per lineal foot or meter, exacerbating tendencies toward concentration of flow and rutting of the strip.
Not well established is the appropriate specification of the maximum width of pavement that should drain to a strip. This is likely a function of rainfall intensity; that is, the maximum allowable pavement width should be lower for regions with high-intensity rainfall like the southeastern and south-central United States. One jurisdiction specifies the maximum lineal flow rate to the strip (Table 1). Rainfall intensity is also a factor to consider when establishing the maximum slope. Values differ greatly according to Table 1. Satisfactory performance has been found with slopes exceeding 15% (Caltrans 2004b).
Final Observations
Researchers have found performance to be highly variable, from relatively poor to above 80% of the TSS, meeting the more common performance goals of many manuals. In some cases, poor performance can be attributed to the test facility being a roadside ditch not specifically designed to treat stormwater. It is, however, not possible to relate the variation in performance to alternative design criteria (e.g., detention time or hydraulic loading rate) as most researchers have failed to include the necessary information. This has made a comparative analysis impossible. Key information missing depending on the study may be the swale width; length and/or longitudinal slope; condition of the grass; and key attributes of each sampled storm, such as flow rates, maximum water depth relative to grass height (whether the grass remained erect), and the amount of infiltration. Confusion can also occur because of the varying role of infiltration in each study, and whether the researcher is reporting reductions in concentrations or loadings. To arrive at the appropriate design criteria requires that researchers report more than just influent and effluent concentrations.
Of concern is the erosion of swales due to excessive shear forces, reducing performance. This can be avoided with dense grass and consideration to the hydraulic radius-velocity product. Procedures to consider this question are provided in some manuals. Johnson et al. (2003) recently updated this concept based on laboratory and fieldwork. Resuspension during extreme events of previously removed sediment and pollutants is of concern but remains undefined. Some manuals recommend or require the swale to be offline, bypassing flows greater than the treatment design event. Inadequate maintenance leads to degraded and patchy grass, channelization, and erosion. The performance of swales and strips is more sensitive to poor maintenance than that of basins.
The most extensive studies relating internal characteristics such as grass density and cover to flow and performance outcomes are Colwell (2001) and Mazer (1998). These studies found that grass can experience extended periods of submergence with modest declines in density and health. Neither the detention time nor the resistance coefficient could be correlated to the extent of density of grass cover, to species mix, or to a wide range of flow resistance values. However, another researcher (Backstrom 2002) found that TSS removal improved with denser grass.
Whether swales and strips remove dissolved pollutants remains an open question. Field performances provide differing results. There remains a conflict between the desire to have thick grass to achieve a suitable resistance to flow and to protect the system from erosion with the desire to have the stormwater contact the soil where dissolved pollutants are removed. Without infiltration, the removal of dissolved pollutants is likely minor. Fertilization appears to cause the export of dissolved phosphorus (Caltrans 2004a). Plant die-off may release nutrients. Dissolved pollutant removal will be further explored in the last article of this series.
An inherent limitation of flow-through swales may be that regardless of the design value selected for Manning's coefficient, the actual resistance for most constructed swales at the peak design flow differs from the design value, either higher or lower (Colwell 2001). If the actual value is greater than the design value, the grass becomes prone at flows less than the design flow. As a consequence, the management goal of treating, for example, 90% of the stormwater over time is not met. If the actual value is less than the design value, the depth of water at the design peak rate (and each flow rate) is less than expected and in turn the flow velocity is greater than expected. As a consequence, the detention time and therefore performance is less than expected.
Practical limitations on system width and grass height constrain the use of swales to small drainage areas of a few acres or less. They become less practical in regions with high intensity rainfall such as the southeastern and south-central United States. In such areas, the combined effect of a maximum bottom width and a substantial design peak flow is to limit use to small drainage areas. Swales and strips may not be appropriate for regions with cold winters. Treatment of snowmelt and spring storms may be severely compromised by the temporary degraded condition of the grass.
Summary
With respect to the removal of sediment and attached pollutants, flow-through swales and strips can provide an acceptable level of treatment as a standalone system. They should be able to meet the performance goal of most manuals, 80% removal of TSS. However, less confidence can be placed in the performance of swales in contrast to wet basins and sand filters, particularly if regular maintenance is not performed. Given the uncertainties expressed in this article, it is not unreasonable that some jurisdictions do not find them to be an acceptable treatment system unless significant infiltration is involved.
Satisfactory performance requires that the systems be properly sized. The specified Manning's coefficient should be at least 0.20, and possibly 0.30 where it is expected that the swale will be mowed frequently, as is the case with retail commercial developments. Maximum and minimum longitudinal slopes should be specified. The specification of the maximum slope absent check dams may differ with the climate, being less in areas with higher-intensity storms. The relationship between flow depth and grass height should be recognized in design and maintenance criteria. The manual should specify a maximum water depth at the peak of the design event and a minimum grass height. The latter should be at least twice the former. Most uncertain of the design criteria is length. Our understanding of the relationship between detention time and performance, and the effect of infiltration on this relationship, is poorly understood. A detention time of 9 or 10 minutes is reasonable.
Hydraulic loading rate may be a more appropriate design criterion, but more data on its relationship to performance are needed.
Swales and strips are not "biofilters," as the primary process of particulate pollutant removal is sedimentation. The terms biofiltration and filter should be dropped. The author recommends the terms vegetated swale and vegetated strip.