The Case for a Water Quality Capture Volume for Stormwater BMPs

There still is confusion among practitioners and regulators as to whether a water-quality capture volume (WQCV) is a part of stormwater runoff treatment facilities (i.e., best management practices, or BMPs). Some amount of WQCV is integral to any BMP that removes significant portions of pollutants from stormwater runoff and those that help mitigate the hydrologic changes caused by urbanization. BMPs with WQCV differ from flow-through BMPs, which do not mitigate the effects of increased stormwater runoff peaks and volumes that result from urbanization. Flow-through BMPs are primarily used to remove gross pollutants consisting of floating trash and coarse sediment, but, for the most part, do not remove fine sediment and associated pollutants such as bacteria and dissolved constituents in significant amounts.

A WQCV is a part of the following types of BMPs:

• As storage in an extended detention basin (i.e., dry basin) (EDB)
• As surcharge storage above the permanent pool of a retention (i.e., wet) pond (RP)
• As surcharge storage above the permanent pool of a wetland basin (WB)
• As storage above or upstream of a media filter (MF)
• As storage above or upstream of a rain garden (RG, often called a bioretention cell)

Unless some amount of buffering volume (i.e., WQCV) is provided, stormwater runoff rates from urban catchments are often too high to enter the MF’s or the RG’s surface, resulting in a very large percentage of the runoff bypassing them. This buffering volume is needed to reduce the flow rate over the infiltration/filter basin surfaces to the rates that will permit the desired amount of runoff to actually go through the filter or to be infiltrated into the ground, thus limiting the amount of runoff bypassing these types of installations.

The challenge for all BMPs, then, is to provide a cost-effective WQCV vessel (i.e., basin) that is adequate for the task, but to not to oversize this volume. If undersized, the receiving waters will not benefit from the mitigation of urban runoff flow rates needed to protect their aquatic life and geomorphology. On the other hand, oversizing will result in waste of valuable land and fiscal resources. The most accurate way to size this WQCV vessel for a given catchment is to perform continuous simulation of runoff using long-term rainfall data, and then route this surface runoff through a storage basin (i.e., a WQCV vessel). The size of this vessel depends in part upon its drain time (defined as the time it takes to empty a brim-full WQCV when there is no additional runoff), which also defines the average discharge flow rate though the basin’s outlet or underdrain or through infiltration into the ground.

Separating Rainfall Data Series Into Individual Storms
In 1989 Driscoll, Palhegyi, Strecker, and Shelley (Driscoll et al. 1989) submitted to USEPA a draft report titled Analysis of Storm Events Characteristics for Selected Rainfall Gauges Throughout the United States. It contains maps of the United States showing a variety of characteristic parameters of rainstorms such as mean depth, mean storm duration, and mean number of storms per year. Their analysis concluded that a dry period of six hours between storms was sufficient and most statistically defensible to define a new storm. They also concluded that storms less than 0.1 inch (2.5 millimeters) in total depth produced virtually no runoff from urban areas and excluded all such storms from the data set of storms. However, that did not end the debate on what dry period defines a new storm, and there are advocates for different durations that range from three hours to 24 hours.

After reviewing the 1989 Driscoll report, we concurred with its recommendation of six hours for separating any continuous rainfall data into a record of individual storms. We also concurred with the need to filter out storms that produce no runoff, since most records throughout the United States show that almost half of all individual storms fall into this “no runoff” category. Storms that produce total precipitation depth of less than 0.10 inch (2.5 millimeters) are generally of no concern when sizing the BMPs. Also, depending on the region of the country, one may want to exclude precipitation data from certain periods of the year from this analysis. For example, winter months have precipitation in the form of snow, which often does not produce runoff until spring and at much different rates and quantities than would be generated from rainfall. Regardless of what protocols are used to keep or exclude individual storms, it is necessary to identify and tag each incremental precipitation depth in the data set with a storm number for later use in determining the WQCV based on the number of storm events it will capture in total.

One may ask, why separate storms when determining an appropriate WQCV for use in sizing BMPs? With continuous simulation of runoff, separating the data set into individual storms and then filtering out inconsequential events is not needed. Continuous runoff volume simulation tells us what fraction of the total runoff volume is captured and what fraction of total runoff exceeds the WQCV vessel’s size for the period of rainfall record analyzed. However, if we want to know the number of storm events and percentages that are captured in total, we need to have a data set that identifies individual storms with overflows/bypasses.

After many years of observing the impacts of urban runoff on receiving water aquatic life, we have concluded that the fraction of storms captured is much more important than the fraction of total runoff volume captured. This is because the total volume captured can be heavily weighted by a relatively few very large events. These large storms create significant drainage and flooding problems, but it would be prohibitively expensive to capture and treat the runoff. At the same time, this does not mean that runoff from larger events will not receive some treatment, because some of this runoff from larger storms, ranging from most to only a fraction, is also captured and treated or infiltrated by a BMP with a WQCV.

Estimating Surface Runoff
Surface runoff can be estimated in a number of ways, ranging from simple to quite complex. Although the more complex methods generally require the use of more input parameters, there is no assurance that this produces more accurate results because of the need to make many assumptions on the exact values of these parameters. We looked at two of these methods, namely the Rational Method and the Horton’s Infiltration Model.

The Rational Method is simple and uses a runoff coefficient to convert rainfall depth to a runoff volume. It merely says that after the initial retention loss is satisfied, runoff occurs and is equal to the product of rainfall and the runoff coefficient.

Horton’s Infiltration Model requires the hydrologist to separate the effective impervious area from the pervious area. For the impervious area, runoff begins to occur once the depression losses are satisfied, after which runoff volume is equal to rainfall. For pervious areas, the hydrologist needs to account for the initial and final infiltration rates and estimate how rapidly the initial infiltration rate decays to its final value. Once the rainfall rate (intensity) exceeds infiltration rate, pervious depressions begin to fill. Runoff occurs after the infiltration and depression losses are exceeded. After rainfall ends, drying of all surfaces begins and the initial infiltration losses and retention storage losses begin to be reclaimed. How much of the initial infiltration losses and retention losses are reclaimed depends on the drying time that the hydrologist deems to be appropriate, which typically can range from one to 14 days.

For continuous simulation, hourly (and sometimes 15-minute) precipitation data (available in comma-separated value format) for all regions of the United States can be obtained from the National Climatic Data Center (NCDC). Hourly data are typically available for much longer periods (most often exceeding 20 years or more), but these data need to be scrutinized for completeness and for questionable rainfall depths before use. The data that are found to be acceptable are used in a continuous simulation model to estimate surface runoff volume for each time step. Figure 1 illustrates how the rainfall, losses, and runoff accumulate for two successive storms. For the second, smaller storm in this illustration, the degradation of runoff loss was partially recovered by the drying time.

Next, the runoff data series is processed through a storage model (i.e., the WQCV vessel) to determine what fraction of the runoff from each storm is actually captured and what fraction bypasses it. During this process, a record is also kept of the number of storms that are fully captured by the WQCV vessel.

WQCV Drain Time
Defining the rate at which the WQCV is to be completely drained (i.e., emptied) is very important. This can be defined as the time it takes the brimful WQCV vessel to completely empty, namely its drain time. Drain time is defined by the person designing the facility based on the type of BMP being used and the effects on receiving waters that need to be mitigated. For example, the drain time for WQCV in an EDB typically ranges from 24 to 72 hours, depending on local requirements. Drain times for the surcharge WQCV of a RP typically range from 1 to 12 hours, with the lower value providing virtually no mitigation of flow increases due to urbanization.

Drain times for MFs and RGs depend on the infiltration rates of the filter or growth media and/or the capacity of the underdrains. When new, the infiltration rates for a MF and for the sandy growth media of a RG is very high, sometimes exceeding 24 inches per hour. But, as these facilities age and fine sediment accumulates on their surfaces, infiltration rates can drop to 0.5 inch per hour or less as shown in Figure 2. For practical reasons, and allowing for some degradation in surface infiltration rates, the authors suggest using a 24-hour drain time to size the WQCV vessel above an MF and a 12-hour drain time above an RG. The surface infiltration rate of the RG is expected to degrade more slowly than that of a media filter because the plant roots tend to reopen some of the clogged pores in the media. The drain time affects the required volume, which in turn determines the surface area these facilities need, because their surface area is found by dividing the volume of the WQCV vessel by its design depth.

Routing Runoff Through the WQCV Vessel
After the runoff volume for each time step in the filtered rainfall data set has been defined, it is routed through a series of WQCV vessels of increasing volume increments at a preselected drain time (Figure 3). Water discharges out of the WQCV vessel through an outlet pipe or underdrain, through infiltration into the soil, or by a combination of infiltration and underdrains.

Figure 4 illustrates how the surface runoff from two sequential storms described in Figure 3 is captured. At the start, the entire WQCV_max is available to capture runoff. Because this is a large storm, the runoff totally fills the WQCV vessel during the first hour. Runoff that follows after the first hour overflows or bypasses the vessel. At the end of the first storm, the vessel continues to empty and regain its WQCV capacity. When the second much smaller storm begins six hours later, the vessel has regained part of its original capacity and is able to intercept the runoff from the first three hours, but not the all of the runoff volume during the fourth hour, at which time a small amount of overflow or bypass occurs. The overflow/bypass volume is added to the cumulative overflow/bypass volume account, and both storms are added to the total count of storms that resulted in overflows or bypasses. Had the second storm occurred 12 hours after the first one, there would have been enough WQCV capacity in the vessel to fully capture it.

Continuous simulation of the rainfall-runoff process and how it affects the amount of water stored in the WQCV_max vessel gives us a complete picture of how the WQCV BMP functions over an extended period of time. Long-term averages of hydrologic and hydraulic performance can then be used to assess the effects on water quality and geomorphic effects on the receiving waters.

Finding the Point of Diminishing Returns
The concept of finding the point of diminishing returns in the sizing of BMPs and their WQCVs was suggested by Urbonas, Guo, and Tucker in 1990 (Urbonas et al. 1990), which was based on protocols developed in Germany (Pechter 1978). This was followed by Guo in the development of software to find this point, which was referred to at that time as the point of maximized WQCV (Guo 1992). This concept was further developed and tested by Guo and Urbonas using continuous rainfall data at a wide variety of locations in the United States (Guo and Urbonas 2002).

Since then, continuous simulation procedures were refined and the authors reexamined this topic. They developed simple software for the Urban Watersheds Research Institute (UWRI) to permit efficient analysis of the rainfall-runoff process using NCDC rainfall data, and to generate an array of WQCVs ranging from no capture to practically 100% capture of all runoff volumes and runoff events. An important feature built into this software is a procedure that identifies where the point of diminishing returns (i.e., maximized WQCV) occurs. Up to that point, incremental increases in WQCV vessel size result in corresponding favorable returns in terms of incremental increases in the fraction of the total volumes of runoff and the total numbers of storms captured. Beyond that point, incremental increases in WQCV result in rapidly diminishing returns in the fraction of total volumes and number of storms captured. Figure 5 illustrates this phenomenon, which was found to occur at all rain gage records examined at a number of locations in United States.

In order to find this point of diminishing returns, all incremental values of WQCV are normalized by dividing each of them by the WQCV that fully captures all of the runoff volumes or storm events generated by the rainfall-runoff model. A few very large events can dominate the runoff volume and WQCV averages, thereby skewing the results. This can be somewhat mitigated by defining the largest WQCV value that excludes these outlier runoff events, say the 99.5 percentile WQCV.

Water Quality Capture Optimization and Statistics Model (WQ-COSM)
To help with the computations of the WQCV vessel’s volume, the Water Quality Capture Optimization Statistical Model (WQ-COSM) was developed. It is a cooperative effort of the Urban Watersheds Research Institute, the Urban Drainage and Flood Control District, and the Civil Engineering Department of the University of Colorado Denver (all organizations located in Colorado). It is available as freeware from the first two agencies listed above. This program replaces a DOS-based program called PondRisk (Guo 1992).

WQ-COSM is a Windows-based computer program that uses recorded rainfall data from the NCDC operated by NOAA, and information developed by the user about the urban catchment’s hydrologic parameters. It has a modern user interface and computes stormwater runoff volumes using continuous runoff simulation by either of two user-selected methods. It then produces a list of increasing WQCV vessel sizes (i.e., volumes) and also produces the maximized WQCV for any type of structural BMP that uses a WQCV.

WQ-COSM is implemented as two programs, a user interface and the math engine. The user interface collects information from the user, generates properly formatted input files for the math engine, and displays the results after the math engine has successfully processed the information in the input file. What follows is a description of what the program does to provide WQCV information of the user to help him or her size a BMP. It can be used to size the WQCV vessel/basin for any catchment in the United States (or anywhere in the world for that matter). The WQCV selected from the output is then used to design the stormwater treatment or infiltration BMP being used.

Performing the necessary calculations for the sizing and optimization of a WQCV for any BMP can be a daunting task, even with the use of spreadsheets. Once WQ-COSM freeware is downloaded and installed, this task is easy to perform. It has two continuous runoff volume simulation options (i.e., Rational Method and a Horton’s Infiltration’s Model, same as the one used in the EPA SWMM 5.0 model). It uses 15- and 60-minute comma-separated NCDC rainfall data, for which the user can exclude specific seasons (i.e., snow season) or specific periods (periods of rainfall data with errors or periods of missing data) from being analyzed.

The user needs to specify the consecutive dry period in days for the program to identify new storms (default=6 hours), the drying period in days to recover Horton’s initial infiltration and depression losses (default=3 days), and the drain time for the WQCV. In addition, the user can specify the lowest value of total rainstorm depth to process to filter out non-runoff-producing storms (default=0.08 inch), and also the upper WQCV percentile to exclude large outlier storms (default=99.5%) for this analysis.

There are three output options:

1. Water quality capture volume and statistical summary that contains rainfall and runoff statistics and a complete list of increasing WQCV values, including the maximized WQCV, with information on how much total runoff volume and number of storms each increment captures. This information is provided in two ways. One is based on the Runoff Volume Capture Ratio and the other on the Event Capture Ratio.
2. Precipitation and runoff statistical summary
3. Complete record of storms and their total rainfall depths, durations, separations, etc.

All output is in an HTML format that can be viewed by any Internet browser and can be copied and pasted into Microsoft Excel, Microsoft Word, and other types of files. In addition, the user can request to generate an electronic file in a comma-separated format that can also be imported into Excel or other software.

Example of an Application
As an example for this article, we examined a 50% impervious catchment with clayey soils in the Denver, CO, region (i.e., most common condition there) using 60 years of data from the Denver Rain Gage NCDC 1-hour data collected between August 2, 1948, and December 28, 2009. We used a 6-hour period of no precipitation to separate hourly data into individual storms and then filtered out all storms with a total of less than 0.1 inch as non-runoff-producing events. The upper limit for sizing the WQCV was 99.5% to exclude very large rainfall events. A total of 2,163 individual runoff-producing storms were identified over the 60-year period, of which only 10 storms exceeded the 99.5% exclusion limit.

We ran the software to find the maximized WQCV for an extended detention basin with a 40-hour drain time. The Horton’s runoff options was used with parameters for clayey soils assumed to have initial infiltration rate of 3.0 inches per hour, which rapidly decays to 0.5 inch per hour, a 0.1 impervious depression loss, and 0.4 pervious depression losses. The maximized WQCV was found to be 0.28 watershed inch based on RVCR and 0.23 watershed inch based on ECR. The RVCR maximized volume fully captured 82.7% of all the runoff volume and 88.9 of all runoff events in total, while the ECR maximized volume fully captured 85.7% of all storms and 77.1% of all runoff volume.

The difference in the maximized WQCV vessel size between the ECR and RVCR is almost 22%, which means that the extended detention has to be 22% larger and will take up that much more land space. It will also cost more to construct and maintain just to capture little more runoff volume from very large flood-producing runoff events. To implement a 95% runoff capture standard based on RVCR would require an additional 55% to 60% larger volume and correspondingly larger surface area for the EDB with proportionate increases in capital, maintenance, and eventual rehabilitation costs. If this standard were applied to an EDB with a 40-hour drain time, the result would be a large number of smaller events passing through the BMP with a significantly reduced detention time and corresponding reduced pollutant removals, because the release rates would be higher to accommodate the 95% RVCR standard.

Conclusions
When designing a stormwater quality control facility system, one needs to balance the runoff capture capability, effectiveness in protecting receiving waters, and lifecycle facility costs, including construction, maintenance, and eventual rehabilitation. A WQCV is an integral part of BMPs such as extended detention basins (dry), retention ponds (wet), wetland basins, media filters, and rain gardens. Using the simple maximization techniques developed by the authors is one method to help designers balance these concerns.

However, the authors recognize that localities may have different standards that need to be followed and have developed continuous simulation software, WQ-COSM, for the Urban Watersheds Research Institute that can generate not only the maximized WQCV but a list of WQCV sizes along with the corresponding percentages of total runoff volume and events captured in total. In keeping with the Institute’s mission, this handy, easy-to-use software costs nothing to download and use. It can be accessed through the UWRI website at www.urbanwatersheds.org or through the UDFCD website at www.udfcd.org under Downloads. Finally, the authors recommend using the Event Capture Ratio rather than the Runoff Volume Capture Ratio as the most appropriate method for the sizing of all BMPs that have a WQCV. This approach recognizes that the more typical population of storm runoff events that occur frequently is the driving force behind most geomorphic and aquatic habitat impacts that can be attributed to urbanization and not the very few events that produce disproportionally large volumes of runoff and cause flooding and serious economic damages.