Using Bioretention Retrofits to Meet Virginia ’ s New Stormwater Management Regulations : A Case Study

Virginia’s new stormwater regulations involve the use of the Runoff Reduction Method (RRM), a methodology to estimate a volume reduction in predicted runoff. Regulations require that for downstream erosion control, the product of the peak flow rate and runoff volume (Q*RV) from one-year storm events in the post-development condition be reduced to less than pre-development Q*RV. This study models different bioretention sizing scenarios in a developed watershed in Blacksburg, Virginia to determine the performance at both the sub-watershed and watershed levels. In addition, models of “optimal” bioretention cells, sized to meet the RRM for each sub-watershed, are evaluated. A direct relationship is observed between the size of the cell required to meet the RRM and the sub-watershed’s developed Natural Resources Conservation Service (NRCS) curve number, and a sizing analysis is provided. Modeling shows that the required size of “optimal” cells for many sub-watersheds exceeds conventional bioretention designs. Upon applying the RRM for all sub-watersheds, the resulting hydrograph at the watershed outlet more closely resembles the pre-development hydrograph than existing development.


Introduction
Low Impact Development (LID) is a design methodology that seeks to restore a developed site's hydrologic response to a storm to its pre-development condition (Prince George's County, 1999).Bioretention, a common LID practice, accepts runoff, allows the water to pond on top of it, and then lets water percolate through its engineered soil media to either the underlying soil or an underdrain.In Virginia, bioretention cells with an underdrain are referred to as "bioretention filters", and those without underdrains are called "bioretention basins" (DCR, 2011).Bioretention cells often utilize an outlet structure or overflow weir to allow any water in excess of the intended treatment volume that enters the cell to be routed in an efficient manner to a desired location downstream.Retrofitting urbanized areas with LID and Best Management Practice (BMP) technologies is an effective way of reducing runoff in a watershed (Damodaram et al., 2010).Although many BMPs exist that could be used in urban stormwater infrastructure retrofits, bioretention is a practice that has increasingly become attractive to designers.This study strictly focuses on the volume reduction benefits of bioretention; however, other reasons that make it a widely used practice are its high removal efficiency of nutrients and pollutants and creation of canopy and wildlife habitat for small species in urban settings.
In September 2011, the Commonwealth of Virginia's (VA) Department of Conservation and Recreation (DCR) made substantial revisions to the Virginia stormwater management regulations.Since then, authority has been transferred to the Virginia Department of Environmental Quality (DEQ), which has taken the lead in development and implementation of the regulations.These regulations were divided into two main categories: quantity and quality of stormwater runoff (DEQ, 2013).Although improvements to storm water quality and its ultimate effect on the Chesapeake Bay were a huge driving force in development of these regulations, this study deals exclusively with the stormwater quantity aspect of the regulations; specifically, those dealing with channel erosion.
The stormwater quantity regulations have changed significantly with the recent revisions, which previously required the peak developed flow rates from the 2-and 10-year storm events to be returned to the pre-development flow rates (DCR, 1999).In addressing stormwater runoff quantity, the new regulations consider channel protection and flood protection as the two primary components of interest.Discharge requirements are based on the type and condition of the receiving channel.When discharging into a natural conveyance system, for instance, a primary channel protection criterion requires comparison of the 1-year, 24-hour storm event's peak runoff rate and total runoff volume for both pre-and post-development conditions (VA, 2011).
To minimize erosion, the new channel protection requirements use a method that is unique to Virginia.Equation 1 (VA, 2011) is used for channel protection calculations when discharging to a natural channel.The purpose of the equation is to calculate the maximum allowable peak flow rate for the developed condition during the 1-year storm event (Q Dev ).Rearranging the equation by multiplying both sides of Equation 1 by the developed runoff volume (RV Dev ), yields Equation 2, where the peak flow rates (Q) are multiplied by the volumes (RV) of flow for the 1-year storm event for both the pre-and post-development conditions.This product is used as the basis for analysis in the rest of this study and referred to as Q*RV.Note that in Equation 2, the developed Q*RV must be less than or equal to 80% (I.F.) of the pre-development Q*RV for sites greater than 0.4 hectares, which constitutes all of the sites in this study.The Q*RV value seems to be unique to Virginia regulations, as none of the surrounding states have regulations that incorporate this value.However, several large municipalities in Northern Virginia, including Fairfax and Prince William Counties, have begun using the similar channel and flood protection criteria to the new Virginia regulations that include the Q*RV value (Fairfax County, 2014 andPrince William County, 2014).Rolband and Graziano (2012), who describe this method as the "Energy Balance" method, aided in the method's development with VA's DCR.For flood protection, calculation of the 10-year, 24-hour runoff is required to ensure that the post-development peak flow rate is at, or below, the pre-development peak (VA, 2011).
In an attempt to estimate stormwater volume reductions through BMPs, the new regulations use a tool called the Virginia Runoff Reduction Method (RRM).The RRM is used to adjust post-development curve numbers downstream of proposed BMPs.Although the use of a particular hydrologic method for calculations is not explicitly required in the regulations, the integrated computation of the curve number adjustment practically forces design engineers in Virginia to use NRCS methodology for site design, without regard to the size of the contributing drainage area.The strategies for this method were developed for Virginia by the Center for Watershed Protection (CWP) and the Chesapeake Stormwater Network (CSN) in an attempt to better emulate pre-development hydrologic conditions on the developed site (CWP & CSN, 2008) and estimate the effects of BMPs in series.A review of the Virginia Runoff Reduction Spreadsheet shows that it incorporates a number of BMPs with varying runoff reduction and pollutant removal capabilities.One of the most efficient BMPs is bioretention.Brown and Hunt (2010) stated that bioretention improves both water quality and quantity aspects.Due to both water quality and quantity requirements in Virginia's regulations and bioretention's benefits in both of these aspects, it is likely to become more frequently implemented.
Many studies have demonstrated that bioretention is an effective means of stormwater management for both quantity and quality, especially at the site scale.Bioretention is especially effective for less intense, more frequent storm events (Davis, 2008;James and Dymond, 2012).The results of the installation of two bioretention cells in a Maryland parking lot suggest that bioretention can greatly reduce the volume of runoff, lower peak flow rates, and increase lag times (Davis, 2008).Bioretention retrofits are becoming more popular due to their hydrologic benefits.Winston et al. (2013) found a substantial reduction in runoff volume can be achieved in a developed watershed through the addition of bioretention cells along the roads, permeable pavement parking stalls, and a tree filter device.A retrofit bioretention cell installed in the Stroubles Creek watershed in Blacksburg, Virginia was shown to reduce the average peak and volume of runoff by over 90%, even though its surface area is only 2% of the drainage area, which is below the recommended and commonly used percentages (DeBusk and Wynn, 2011).However, there were very few large, intense storms studied due to the timing of the monitoring.
The location of bioretention in a watershed is critical for maximizing its efficiency.James and Dymond (2012) found that bioretention is more efficient when it is treating large impervious areas, than when it is treating areas that have a higher percentage of pervious cover.Gilroy and McCuen (2009) had similar conclusions, and also determined that installing BMPs in series compounds their effects.Proper sizing, maintenance, and construction practices are also critical to the performance of bioretention and, if designed correctly, can result in a practice that reduces both the peaks and volumes of flows leaving a site (Brown and Hunt, 2010).Li et al. (2009) studied four locations with bioretention cells in Maryland and North Carolina and found that cells with larger storage volumes, either through a larger cell area or deeper media depths, more closely replicated pre-development conditions, even for larger storms, by reducing peak flow rates, reducing outflow volumes, and promoting more infiltration.Although these studies provide insight on the functionality of bioretention cells, they do not go as far as determining what size facilities would be required to meet water quantity regulations for their respective jurisdictions.
The purpose of this study is to provide insight into bioretention sizes required to meet peak reductions as required by the channel protection criteria in the new Virginia stormwater regulations.The study is performed on a watershed in the Town of Blacksburg, Virginia for which a calibrated rainfall-runoff model was developed.
Using several different modeling scenarios, various sizes of bioretention cells are modeled to simulate their retrofitted installation throughout the watershed.In addition, the "optimal" scenario is found for each sub-watershed within the watershed, so that it can meet the channel protection criteria of the RRM.Furthermore, effects of the RRM is studied at the watershed outlet for the scenario when all sub-watersheds within the watershed are meeting the requirements outlined for channel protection.

Method
The watershed modeled in this study is the "North Stroubles" watershed in Blacksburg.It is a 192-hectare watershed consisting of residential, commercial, industrial, institutional, and open space land uses in the headwaters of Stroubles Creek, a tributary of the New River.There is a flow sensor within the stream at the outlet of the watershed, near Webb Street, that is owned and monitored by the Town of Blacksburg.The flow sensor used is an Acoustic Doppler velocimeter, which measures the felocity of the flow and the flow area (as a function of stage).Upstream of the flow sensor, the watershed has been delineated into 41 sub-watersheds, or catchments, ranging from just over 0.4 hectares up to approximately 14 hectares, as shown in Figure 1.The catchments were delineated based on key points of interest, such as ponds or intersections of major conveyances.
In addition to the flow sensor, there is a rain gauge less than a mile outside of the boundary of the watershed.Flow and rainfall data measured by these devices were used for model calibration.
Virginia regulations require the volumetric sizing of bioretention facilities to be based on a composite weighted runoff coefficient which incorporates impervious, turf, and forested components of runoff.However, Virginia regulations do not require that bioretention facilities be designed to fully meet the requirements of the downstream erosion protection requirement of the regulations since it is realized that multiple BMPs may be required on site to achieve the various quantity and quality improvement goals.Therefore, strict adherence to the Virginia sizing methodology will not ensure that bioretention sized by the Virginia method will meet the downstream erosion protection goals examined in this study.Because of this, a simplified sizing methodology using a percentage of contributing drainage area has been used throughout.This technique is similar to previous Virginia methods which have correlated sizing of bioretention surface area with the percentage of upstream impervious area.This study examines a number of modeling scenarios which are summarized here for clarification and further described below.

•
The existing conditions scenario is used as a base for other modeled scenarios

•
The pre-development scenario is used for comparison at the watershed scale and to determine target values for design of each catchment's "optimal design" bioretention cell.

•
There are four different "performance" scenarios that evaluate the performance of bioretention cells with surface areas sized as a percentage (3%, 5%, 7%, and 10%) of the contributing drainage areas.As mentioned previously, the bioretention basins are sized strictly based on upstream drainage area and are not strictly related to the impervious percentage in the sub-watershed.Because this study is focused on volumetric improvement and not water quality removal efficiency, it is believed by the authors that varying the percentage of the entire contributing drainage area may be more appropriate than varying the percentage of a single land cover (upstream impervious area).

•
There are also two "optimal design" scenarios where the area percentage is adjusted such that the flow leaving the bioretention cell exactly meets the channel protection requirements of the RRM. www.ccsen

Existin
In order t modeled a are presen such as ma Since the cross-section of the cells remained uniform in the model, the variation of cell surface areas provides the means for directly calculating the changes in cell storage volume.Due to the large area of some of the sub-watersheds modeled, the cells associated with these large drainage areas are much larger than typical bioretention cells.However, it is assumed that these large cells can represent a distributed network of cells located throughout each of the sub-watersheds.Elliott et al. (2009) determined that it is acceptable to aggregate a network of bioretention cells for modeling purposes, and Gilroy and McCuen (2009) found that BMPs in series have their effects compounded regardless of the distance between them.Although the bioretention cells are designed as retrofits, they can also be implemented upon the initial development of the land.
By keeping the same vertical structure of each cell and changing the surface area, the volume of each cell is changed in a consistent manner.Since each cell only receives flow from a single sub-watershed, they were sized based on a percentage of the area of their respective sub-watershed.The four consistent percentages used for sizing the surface areas of the cells were 3%, 5%, 7%, and 10% of the sub-watershed's area.

"Optimal" Models
Along with the sizing scenarios based solely on the percentage of the sub-watershed's drainage areas, two "optimal" scenarios were tested.The design of the "optimal" cells was achieved by adjusting the surface area for each cell until the calculated Q*RV leaving the cell for the 1-year storm event equaled 80% of the related pre-development value.The 80% value was chosen to meet the RRM's channel protection requirements.
The first of these two scenarios maintained the typical cell's cross-section used in all other non-optimal scenarios.In the second optimal scenario, the engineered soil media depth was increased from 91 to 122 cm.This scenario using cells of increased depth would represent an urban area where space is limited and constructing a deeper cell would be desired.

Flood Protection Analysis
In the RRM, the flood protection requirements call for reducing the peak flow rate from the 10-year storm event in the developed condition back to, or below, the pre-development peak.The model was run with the 10-year storm event for the pre-development scenario, the existing conditions scenario, and the 91-cm optimal scenario.Existing stormwater management ponds are present in the watershed and affected the flood protection analysis at the watershed outlet, but did not affect any analysis at the sub-watershed scale because the bioretention was modeled to be upstream of the pond, if present in the sub-watershed.

Results
Model scenarios were run for the 24-hour, 1-and 10-year return frequency NRCS design storm events for Blacksburg, Virginia.Rainfall depth for each storm was obtained from NOAA Atlas 14, Volume 2 (Bonnin et al., 2004) partial duration series.These values were 5.8 cm and 10.4 cm for the 24-hour storm events for return frequencies of 1 and 10 years, respectively.
For each sub-watershed, as well as the watershed as a whole, model results were obtained for each scenario and compared to the pre-development values with respect to the peak flow rate, the volume of flow, and the Q*RV.These values were plotted against the calibrated CN of the sub-watershed's developed condition, as shown in Figures 4 and 5. Generally as the CN of the sub-watershed increased, the peak flow rate and volume of flow increased, and therefore the peak multiplied by the volume, increased as well.Also, as expected, as the area and volume of bioretention installed in each sub-watershed increased, the peak and volume of flow decreased.
When compared to the pre-development peak for the 1-year storm event, the sub-watersheds in the existing condition model (0% bioretention) produced peak flows between 2 and 10 times higher, as shown in Figure 4a.
The 3% scenario includes one sub-watershed which had a low CN and was brought below the pre-development peak, and the 5% scenario includes five sub-watersheds achieving that reduction.Almost half of the 41 sub-watersheds in the 7% scenario had peaks at or below the pre-development value, and all of the watersheds in the 10% scenario had peak flows below the pre-development peak.
Meeting the pre-development values for volume was less successful.None of the 3%, only 1 of the 5%, and only 4 of the 7% sub-watersheds met the pre-development threshold (Figure 4b).Only about one-third of the sub-watersheds in the 10% scenario released less total flow than the pre-development scenario.Note that the storage volume in the 10% scenario was so large that it resulted in no flow leaving the bioretention cell for several of the sub-watersheds.

Discussion
This study considered the channel protection criteria of the new Virginia stormwater regulations, how bioretention could be used to meet these new regulations, and the effects that implementing bioretention to meet the new regulations would have on a larger watershed scale.Bioretention cells, located at the outfall of the sub-watersheds that make up a larger, calibrated watershed model, were simulated in variable sizes as discussed previously.Resulting models yielded information regarding the effect of bioretention and overall watershed response as outlined below.

Bioretention
The installation of bioretention cells can result in the developed hydrology more closely mimicking the pre-development hydrology for both the site-and overall watershed-scales.All sizes of bioretention retrofits that were modeled showed decreased peak flows and volumes of flows from the developed, existing condition sub-watersheds.However, when larger percent area cells (7 and 10%) were modeled, the flows leaving some of the cells were very small, or non-existent, which could have hydrologic ramifications on downstream receiving waters.

Optimal Sizing
The area, and resulting volume, of bioretention required to meet the RRM is directly related to the difference between the CN of the developed condition and that of the pre-development condition.As this difference increases, larger cells are needed to retain the larger amount of flow volume.The resulting sizes of the bioretention cells needed for the new standard of 80% of the Q*RV are typically larger in area than those traditionally seen in practice.Due to large differences between the developed and pre-development condition, modeling indicates that some sub-watersheds require cell areas to be more than 10% of the drainage area, which is infeasible.However, these large single cells were modeled in this manner to simplify the model and, in reality, depict a distributed network of cells in the sub-watershed.As concluded in Tillinghast et al. (2012), it may be unreasonable to attempt to mimic the pre-development conditions of an intensely-developed watershed through retrofits for several reasons.The sub-watersheds with the most development, and therefore requiring larger bioretention cells, would need to have available land area to accommodate the large cells.However, based on preliminary observation, this amount of open space would not be available in many of the sub-watersheds in the area of study.In addition, the available land may not be in a location that permits the directing of runoff to the cells.Other forms of LID and BMP techniques would likely be required in this situation.In an urbanized area, the combined use of underground sand filters and underground detention can help to achieve the required water quality and quantity metrics, while limiting the impact on the surface area requirements.

Depth Effects
If available surface area is a major issue, modeling revealed that increasing the engineered soil media depth from 91 cm to 122 cm is a valid option.A consistent 18% reduction in area of the bioretention was shown for this change in design for meeting the regulations.This reduction was consistent across all of the sub-watersheds.Also, increasing the depth of the cells seems to result in slightly more volume attenuation, and slightly less peak attenuation.Therefore, this could be taken into account if regulations apply to either the volume or peak, but not necessarily their product.

Watershed-Scale Effects
Finally, the retrofit of bioretention in a watershed with the channel protection criteria results in a watershed that has hydrologic characteristics closely approaching the pre-development.When this method is applied at individual sites throughout a watershed, the modeled results at the watershed outlet are shown to result in a peak lower than the pre-development, but with prolonged flows that are somewhat higher than the pre-development levels, as shown in Figure 7.This is a substantial improvement over previous stormwater management methods that have resulted in higher peaks with longer, higher receding limbs.However, implementation of this method requires much more space for the distributed network of smaller facilities.
The sizing of bioretention cells is critical to their performance.If they are sized too small, there is little channel protection benefit in their installation, and if they are too big, the decrease in outflow for lower recurrence interval storms can be so small that it could affect the nature of the receiving waters.When sizing cells for channel protection goals, the required size is directly related to the CN of the contributing drainage area.However, the feasibility of the space requirements for meeting some of these goals, especially in a retrofit environment, is questionable.One possible way to overcome limitations in available area is increasing the depth of the bioretention cell, which corresponds with a decrease in the required surface area.With the RRM applied throughout the study watershed, the hydrograph at the watershed outlet mimicked the pre-development hydrograph much closer than the hydrograph of traditional stormwater management techniques.
. (Improvement Factor) = 0.8 for sites > 0.4 hectares or 0.9 for sites ≤ 0.4 hectares Q Dev = peak flow rate for the developed condition 1-year storm (m 3 /s) Q Pre = peak flow rate for the pre-developed condition 1-year storm (m 3 /s) RV Dev = volume of runoff -developed condition using RRM for the 1-year storm (cm) RV Pre = volume of runoff -pre-developed condition for the 1-year storm(cm) Figure 4

Figure
Figure 7. R