Retrofit of Storm Water Treatment Controls in a Highway Environment

Modification des systèmes existants de traitement des eaux de ruissellement sur un réseau autoroutier

 

Scott Taylor                    Dr. Michael Barrett

RBF Consulting                          Center for Research in Water Resources

P.O. Box 57057                         University of Texas

Irvine, CA USA 92619               Austin, TX USA 78712

staylor@rbf.com                       mbarrett@mail.utexas.edu

RESUME

Un programme pilote, pour le contrôle du traitement des eaux de ruissellement, a été développé pour estimer les coûts et les bénéfices de l’adaptation au réseau routier des BMP (Best Management Practices) structurelles conventionnelles. Le côté unique de ce programme réside dans les  trente neuf installations placées, conçues, construites, mises en œuvre et entretenues utilisant, sur une période de cinq ans, des critères cohérents et des protocoles de test et de suivi. Le budget de ce projet dépassa quinze millions de dollars US. Le concept de cette étude, pour évaluer les coûts (de construction, mise en œuvre, et maintenance)  et les bénéfices (éléments présents et leur diminution au travers de BMP), était basé sur la construction de prototypes d’appareils de mesure sur des sites du réseau de transport de la Californie du sud (Caltrans). La purification estimée par ces appareils a généralement été en accord avec les résultats d’autres études. Néanmoins, certains de ces appareils ont montré que les concentrations d’éléments en sortie étaient indépendantes des concentrations d’éléments à l’entrée. Ceci nécessita le développement d’une méthode alternative pour l’évaluation de l’efficacité de la purification. Les paramètres généralement acceptés des concepts de BMP ont été ajustés, et des stratégies concernant le contrôle de vecteur ont été développées.

ABSTRACT

A storm water treatment control pilot program was developed to assess the costs and benefits of retrofitting conventional structural Best Management Practices (BMPs) into highway infrastructure.  The program was unique in that thirty-nine installations were sited, designed, constructed, operated, and maintained using consistent criteria, monitoring and testing protocols over a five-year period.  The project budget exceeded $15 million US dollars.  Constituent removal of the devices was found to be generally consistent with that reported by other studies; however, some of the devices exhibited constituent effluent concentrations that were independent of the influent concentration, necessitating the development of an alternate method for evaluating removal efficiency.  Refinements in generally accepted BMP design parameters were determined, and strategies for vector control were also developed.

KEYWORDS

best management practices, highway runoff, stormwater treatment, sustainable drainage systems

1.        Introduction

The BMP Retrofit Pilot Program was implemented to determine the costs and benefits of retrofitting highway infrastructure, facility maintenance stations (corporation yards) and park and ride lots with conventional structural Best Management Practices (BMPs).  The program was developed to record costs (construction, operation and maintenance) and constituent removal by constructing field-scale devices at selected locations in the California Department of Transportation (Caltrans) system.  The devices shown in Table 1 were studied as a part of this program:

Table 1. Device Types and Number of Installations

Device

Number of Sites
Device

Number of Sites

Extended Detention Basin

5

Biofilter Strip

3

Infiltration Basin

2

Infiltration Trench

2

Wet Basin

1

Drain Inlet Insert

6

Sand Media Filter

8

CDS**

2

MCTT*

3

Oil/Water Separator

1

Biofilter Swale

6

Perlite/Zeolite Filter

1

* Multi-Chambered Treatment Train

** Continuous Deflective Separation

The study was unique in that thirty-nine BMPs were investigated using identical protocols for constituent removal.  Paired flow-weighted samples were collected from 26 BMP sites and grab samples were collected from an additional 11 sites.  In all, over 13,400 chemical analyses were performed on paired samples collected during three consecutive rainy seasons.  Consequently, this study is thought to be the most comprehensive investigation of conventional structural BMP performance completed to date.

2.       methods

Study sites were selected in two Caltrans southern California Districts using a weighted decision matrix process. Criteria important in the selection of the site for each BMP were listed, and each criterion was then assigned a weighting factor. Further, the order of site selection was also considered to ensure that the devices with the most restrictive criteria were located first, followed by devices with less restrictive criteria in descending order. Infiltration sites generally had the most restrictive siting criteria (space and soil requirements).

The BMPs were monitored to determine their effectiveness at removing a number of conventional constituents commonly observed in highway runoff.  With the exception of the drain inlet insert, oil-water separator and infiltration BMPs, all the sites were outfitted with automatic samplers (Sigma 900 Max Series) and flow meters (Sigma 950 Series) to collect flow weighted composite samples of the influent and effluent of the devices.  Automatic samplers consisted of a peristaltic pump, pump control electronics, a sample distribution system, a power supply, and a housing that contained the composite bottle(s).  Rain gauges (Sigma 2149) were installed at all sites.  A typical monitoring configuration is shown in Figure 1.

 

 

 

 

 

 

 

 

 

 

 


Figure 1 – Typical Monitoring Configuration

 

3.       FINDINGS

3.1.              Technical Design

Each of the BMPs was designed using conventional criteria found in contemporary publications such as Young et al. (1996) and WEF (1998).  Although the study was not specifically designed to evaluate the effects of discrete BMP design parameters, there were a sufficient number of installations for most devices to make some assessments in this regard. 

3.1.1.           Extended Detention Basins

The basic design criteria for detention basins included detention time, length/width ratio, and depth.  Basins were designed for a water quality volume drawdown time of 72 hours, with the objective of providing an average detention time of 24 hours for all storms up to the water quality design storm (a storm with a recurrence interval of about 1 year).  This was accomplished by providing a staged riser-type outlet configuration.  A minimum length to width ratio of 3:1 was used and the depth of maintained between 0.5 and 1.2 meters to prevent resuspension of material.

Five basins were constructed with length to width ratios from 3:1 and exceeding 10:1.  TSS removal performance did not vary significantly with the length to width ratio.  It can be concluded that exceeding a length to width ratio of 3:1 will not enhance removal performance, and it is the author’s opinions from field observation that lower ratios would not significantly impair performance.  It was noted in the field that the basins filled rapidly (on average within 3 hours from the start of rainfall) but had an average residence time of about 18 hours.  This effectively precludes an idealized Type II sedimentation process where the influent moves through the basin in a plug flow fashion in favor of a more classic application of Stokes Law and vertical settling.  Relaxing length to width criteria to less than 2:1 may be acceptable with little performance penalty.

A more critical element was the re-suspension of particulates near the basin inlet structure that was observed at several locations, underscoring the need to provide effective energy dissipation where the inflow jet enters the basin.  Another interesting finding was the discrepancy between calculated and actual draw down times noted at some of the locations.  Orifices in the outlet risers were relatively small (about 3 cm), which, coupled with the inherent error in estimating the orifice coefficient tended to underestimate the actual discharge.  This problem was compounded by flow discharging through boltholes in the outlet structure.  It is recommended that draw down times are field-verified for a variety of basin stages and post-construction modifications to the orifice size made as appropriate.

Two other findings observed during maintenance of the basins are noteworthy.  At one of the pilot locations that exhibited a consistent wind direction, floating trash and debris tended to accumulate at the lee side of the basin.  This observation can be used to advantage for locations with a prevailing wind direction by locating a maintenance ramp where the trash will accumulate and keeping the basin outlet at the extreme upwind location away from floating debris.  During the course of the three-year study, sediment accumulated at a rate of less than 1% by volume per year.  This would indicate a required clean-out frequency in excess of once every 10 years.

3.1.2.           Sand Filters

Sand filters were designed using standards developed by the City of Austin, Texas, specifically for storm water use.  The design sedimentation chamber draw down time was 24 hours, and the assumed permeability of the sand was 1.07 m/d.  The sedimentation chamber was drained through a perforated riser; the rate of discharge from the riser was controlled via an orifice plate in the barrel.

Observation of the filters during rain events and maintenance during the study period afforded several recommendations for future designs.  It was noted that the level spreader used to distribute effluent from the sedimentation chamber over the sand bed was ineffective.  The flow from the sediment chamber was small (on average about 0.01m/s) and with relatively large weirs (greater than 3 meters) resulting in flow only over isolated sections of the weir that varied marginally from the specified grade.  Flow subsequently pooled over small portions of the sand bed, again in areas with the lowest elevation.  Future designs may eliminate the weir in favor of a small bed of riprap at the discharge point to the sand filter chamber.

Maintenance of the filters was hampered by the lack of access to the treatment chambers.  Vertical sidewalls were employed at all sites to preserve space, without an allowance for equipment access.  Future designs should employ maintenance ramps to the sedimentation basin invert and filter bed. 

Performance of the filters was consistent with other published studies such as Glick, et.al. (1998); however, it was interesting to note that the reduction of particulate related constituents was independent of the influent concentration.  This conclusion is attributed to the filtering stage of this device that is highly effective in removing all but the smallest particulate size fractions.  For particle size distributions found in urban storm water, the total suspended solids concentration in the filter effluent is consistently about 8 mg/L.  This finding is of value in cases where a specific TSS effluent concentration is required.  Nutrient removal was poor, with an increase in effluent nitrate concentration but an overall net decrease in total nitrogen.

3.1.3.           Vegetated Treatment

Two types of vegetated controls were reviewed in the study, vegetated strips and swales.  Strips were constructed as stand-alone BMPs as well as a pretreatment device for infiltration trenches.  A single type of vegetation was used, salt grass, which is a plant native to California.  It was selected since it is relatively low growing, forms a dense matt and does not require irrigation.

Strips were designed using guidelines from WEF (1998) with a maximum slope of 12 percent (prototype slopes were all less than 2%), and a length (in the direction of flow) of 8 m.  Similarly, swales were designed following the guidelines from WEF with a target minimum hydraulic residence time of 9 minutes, a maximum longitudinal slope of 5 percent, a maximum velocity of 0.3 m/s and a maximum base width of 2.5 m.  A literature review of these design criteria concludes that they are based on few installations, and virtually no work has been done to refine them. 

One of the primary problems with the current design guidelines is that there is no direct consideration of the relationship between tributary area and treatment area.  The strips constructed in this study had a tributary area to treatment area ratio of from 4 to 43.  Similarly, the swales constructed in the study had a ratio that varied from 13 to 53.  Guidelines specifying a minimum for the tributary area to treatment area ratio are needed to assist the designer to achieve a performance benchmark while using the available space efficiently. Swale design could benefit from more rigorous guidance for depth and maximum flow velocity. Vegetated controls have performance that equals or exceed most other BMPs, yet require little maintenance and have a low capital and maintenance cost.

3.1.4.           Infiltration

Two types of infiltration devices were reviewed as a part of the study, infiltration basins and infiltration trenches.  A review of design criteria in the US showed that recommended in situ soil infiltration rates varied greatly from about 5 mm/hr to about 100 mm/hr.  A minimum in situ infiltration rate of 7 mm/hr and a minimum separation to the season high groundwater table of 0.6 meters was used in the study.  The maximum ponding depth was 0.9 m for basins, and basins and trenches were designed to completely empty the water quality volume within 72 hours.

Studies have show that more than half of all infiltration devices fail within the first five years of operation (Schueler et.al., 1992).  This is generally the result of clogging of the site soils, or due to high groundwater.  Of the four infiltration devices constructed as a part of this study, two failed or did not perform at design level expectations.  One of the infiltration trenches was constructed in fractured sandstone.  The intrinsic permeability of the sandstone was very poor; however, the fractures created sufficient permeability so as to meet the design requirements.  Unfortunately, the fractures in the rock were not homogeneous, and when the trench was constructed, far fewer fractures on a relative basis (as compared to the in drill hole permeability test) were present.   This resulted in trench drain times from a full condition of about 2 weeks.

The second location that failed was an infiltration basin.  The groundwater at this location was measured about 0.6 m from the basin invert during the soil permeability testing.  During construction, the groundwater elevation rose to within 0.3 m of the basin invert.  The design invert elevation was raised about 0.3 m during construction to compensate for the rise in the groundwater table.  However, the groundwater table continued to rise after construction was complete, at times nearly coinciding with the basin invert.  As a result, water often remained in the basin several months following rain events.

Clearly, some of the infiltration devices in this study were sited in marginal conditions.  Since there are readily available alternatives to infiltration, infiltration should only be used where conditions are favorable.  Highly permeable soils will increase the risk of groundwater contamination.  A portion of the study was designed to measure changes in groundwater quality below the infiltration devices using lysimeters to obtain groundwater samples.  This portion of the study was not successful due to the inability to obtain samples.  Further study on the potential of groundwater contamination from storm water is warranted.

3.1.5.           Wet Pond

One wet pond was monitored as a part of the study.  The primary design criteria included a three to one ratio for the permanent pool to water quality volume.  Depth in the open water areas was from 1.1 to 1.9 m with a 0.3 m bench around the pond perimeter to promote vegetation growth and enhance public safety.  Dry weather flow was intercepted from an adjacent channel to sustain the permanent pool.  The pond was sited upstream of a coastal estuary that is impaired for nutrients.

A mass balance for constituents in and out of the pond was developed and it was determined that the dry weather inflow was relatively high in nitrate (about 15 mg/L as N) and phosphorus (about 2.2 mg/L).  Prior to construction of the pond, these nutrients were being contributed to the estuary on a year-round basis.  This contribution was especially critical during the summer months when additional daylight promoted more algal growth in the estuary, and the associated low ambient oxygen levels.  The pond operation reduced each of these nutrients by about half, most likely benefiting the receiving water beyond the treatment provided during storm events.

Two primary research needs for pond design were identified at the conclusion of the study.  First, there is a lack of definitive guidance for the ratio of permanent pond volume to influent volume.  The authors feel that a permanent pool volume of twice the mean annual storm will provide good results. The second research need is the requirement of perennial flow to sustain the pond.  This is a fairly restrictive siting requirement, and the applicability of wet ponds could be greatly expanded if storm water were used to initially fill the dead storage area of the pond at the beginning of the wet season.  Investigation is needed to determine if the water quality in the pond would remain acceptable, if performance is comparable to the traditional design, and if vector control becomes more problematic.

3.2.              Construction Cost

A primary part of the study was concerned with tracking the costs of construction of the devices.  There are sufficient references for the development of BMP construction cost associated with new development or significant redevelopment, however, little documentation is available concerning stand-alone BMP retrofit.  The results of the cost analysis show that stand-alone retrofit construction cost is substantially greater than construction of BMPs with a larger project.  This is due to many factors such as economy of scale, accommodation of existing operations at the site and limited work area.  Overall, the costs of the devices ranged from a high of $ 3,472 (sand filter) to a low of $246 (swale) per cubic meter of water treated.  A premium for retrofit of about 34 percent over the cost of new construction was estimated from the compiled data.  Accordingly, storm water retrofit projects should be combined with other capital improvement projects whenever possible for increased economy.

3.3.              Vectors

Historically, vectors such as mosquitoes, have received comparatively little consideration when designing BMPs.   Many devices retain water for extended periods, or maintain standing water indefinitely (wet pond, Delaware sand filter) creating a potential vector habitat.  The devices in this study were designed to impound water for a maximum of 72 hours, a span less than that required to complete the mosquito breeding cycle.  The wet pond was stocked with mosquito fish and the Delaware filter was treated with synthetic hormones.  The US has recently seen of proliferation of the West Nile virus, which is transmitted through mosquitoes. Future BMP designs must limit the opportunities for standing water and routine inspections should include an assessment of vector breeding.

3.4.              Performance Summary

Analysis of the water quality data collected during the study indicated that in many cases the traditional method of reporting performance as a percent reduction in the influent concentration did not correctly convey the relative performance of the BMPs.  The problem was primarily the result of differences in influent runoff quality among the various sites.  Devices that were installed at park-and-ride lots, where the untreated runoff had relatively low constituent concentrations resulted in a low calculated removal efficiency even though the quality of the effluent was equal to that achieved with the same device at another location.  Consequently, a methodology was developed using linear regression to predict the expected effluent quality for each of the BMPs as if they were subject to identical influent quality.  The study found that a comparison on this basis resulted in a more valid assessment of the relative performance of the technologies. 

Table 1 presents the expected effluent quality for total suspended solids (TSS), total phosphorus, and total zinc that would be achieved if each of the BMPs were subject to runoff with influent concentrations equal to that observed on average for highway and maintenance stations during the study.  Effective concentrations of zero are shown for the infiltration devices, since there is no discharge to surface waters.

 

Table 1  Effluent Expected Concentrations for BMP types

Device

TSS                        (Influent 114 mg/L)

Total Phosphorus (Influent 0.38 mg/L)

Total Zn                   (Influent 355 ug/L)

Austin Sand Filter

7.8

0.16

50

Delaware Sand Filter

16.2

0.34

24

EDB unlined

36.1

0.24

139

Wet Basin

11.8

0.54

37

Infiltration Basin

0

0

0

Infiltration Trench

0

0

0

Vegetated Swale

58.9

0.62

96

Vegetated Strip

27.6

0.86

79

 

A series of figures were created for the devices ranked by life cycle cost from most to least expensive and graphed against constituent concentration and load reduction. The life-cycle costs include the expected maintenance cost estimated from the final maintenance protocol for each device developed at the conclusion of the study.  One could reasonably expect that those devices on the left side of the graph would have lower effluent concentrations and greater load reduction since they have a higher capital cost. As shown in the example for TSS in Figure 2, this is not always the case. Error bars on the graph indicate the reliability of the estimated effluent concentrations and load reductions. This uncertainty indicates the 90 percent confidence interval of the estimate of the mean effluent concentration.

 

Figure 2 Predicted TSS Effluent Concentration

4.       Conclusions

Several of the BMPs in this study were found to be inappropriate for the study sites due to poor constituent removal, excessive complexity and cost, relatively high maintenance requirements or a combination of these factors.  The rejected BMPs included the oil/water separator, drain inlet inserts, the multi-chambered treatment train and the Stormfilterâ.

A family of performance graphs were developed indicating constituent removal efficiency by device ranked by relative life-cycle cost.  The graphs were generated from consistent influent constituent concentrations to allow their application by the designer for any given set of conditions.  These graphs are available from the authors on request.

Acknowledgements: This research effort was funded by the California Department of Transportation under Contract Nos. 43A0004A and 43A0051. Prime contractor for the study was RBF Consulting and monitoring and operation of the tests sites were conducted by KLI and Brown and Caldwell and Mactech.

LIST OF REFERENCES

 

Glick, Roger, Chang, George C., and Barrett, Michael E., 1998. “Monitoring and Evaluation of Stormwater Quality Control Basins,” in Watershed Management: Moving from Theory to Implementation. Denver, CO, May 3-6, 1998, pp 369 ‑ 376.

Schueler, T.R., Kumble, P.A. and Heraty, M.A.,1992. “A Current Assessment of Urban Best Management Practices, Techniques for Reducing Non-point Source Pollution in the Coastal Zone,” Metropolitan Washington Council of Governments, Washington, D.C.

 

Water Environment Federation (WEF) and ASCE, 1998. “Urban Runoff Quality Management,” WEF Manual of Practice, No. 23, ASCE Manual and Report on Engineering Practice No. 87.

Young, G.K. et al.,1996.  Evaluation and Management of Highway Runoff Water Quality. Publication No. FHWA-PD-96-032, U.S. Department of Transportation, Federal Highway Administration, Office of Environmental and Planning.