Long-Term Field Performance of Various Passive, Construction Sediment Control BMPs

Aug. 3, 2012

Historically, long-term performance data for construction stormwater best management practices (BMPs) have been virtually nonexistent. The performance data that do exist are primarily associated with short-term performance test protocols using single pass-through flows through passive sediment control BMPs. A better understanding of long-term performance for construction BMPs is essential, as future construction water-quality regulations will undoubtedly continue to become more stringent. Long-term BMP performance, in the context of this study, is defined relative to the variability of effluent turbidity over time durations consistent with actual field construction associated with typical construction projects. Construction projects can last as short as three months for home building projects to nearly three years for large commercial and roadbuilding projects. Performance data from passive construction BMPs for time durations associated with a typical construction project, using turbidity as a general indicator parameter for water quality, are far more useful and relevant, as engineers struggle to design passive treatment trains in the field to meet more demanding effluent standards.

In September 2009, Fiberweb Inc. of Old Hickory, TN, and Civil & Environmental Consultants Inc. (CEC) began the operation of a full-scale applied research project to determine the long-term sediment control performance capabilities for two conventional, nonproprietary passive BMPs and a proprietary passive BMP produced by Fiberweb. With the cooperation of the Williamson County, TN, government, a test site was identified and constructed at the Williamson County landfill facility in Leipers Fork, TN, southwest of Nashville. The goal from this work was to examine the variability in effluent turbidity from the given sediment control BMPs over an extended time period, where the BMPs are subjected to multiple storm events of varying intensities and runoff volumes. Another goal was to observe changes in effluent turbidity over time as materials degrade, structural BMPs are damaged, and/or BMPs are clogged or blinded by solids. This information can be used as part of a growing body of data to assist in the development of detailed design approaches for stormwater treatment trains, in order to meet regulatory requirements for effluent quality from construction sites.

Materials and Methods
Test Scenario 1: Straw Wattles versus Fiberweb Typar Geocells with Shredded Wood Mulch-The test sites established at the Williamson County landfill consisted of two research plots that were carefully designed by the engineering team at CEC. The twin test plots were constructed by CEC and personnel from the Williamson County landfill. The plots were located on a graded area with overall dimensions of 28 meters long by 9.2 meters wide. The areas were completely denuded and consisted of exposed silty clay soils with chert. Each individual test plot measured 4.6 meters in width. The twin plots were designed and constructed to provide identical watershed characteristics to simultaneously compare water-quality treatment between two different types of BMP applications at one time. The slope of each plot was an average of 8% (Figure 1). The size of the test plots was designed based on known internal hydraulic (porous flow) capabilities of each BMP, relative to each BMP’s chosen length and associated cross-sectional surface area. The internal hydraulic capacities were based on previous flume test results performed by CEC on various proprietary and nonproprietary BMPs in 2008 and 2009. The plots were designed to store the runoff volume from the two-year, 24-hour storm event for Williamson County behind the treatment systems without overtopping, thereby treating all of the volume of runoff from the two-year, 24-hour storm and smaller events. The attenuating mechanism for solids-retention BMPs is primarily via particle settling behind the BMP structure, with a very minor contribution, if any, via filtration.

Figure 1. Plan view layout of test plots
Figure 2. Straw wattle test plot-view of south test plot

Figure 3. View, facing north, of the automatic samplers and flow meters equipped with
solar panel

The two passive sediment control BMPs that were initially tested for comparative performance, designated as Scenario 1, consisted of two 30.5-centimeter straw wattles stacked atop one another versus a single DT1 Typar Geocell unit filled with shredded wood mulch. The Typar Geocells are manufactured by Fiberweb Inc., a specialty industrial materials company based in Tennessee. Typar Geocells are a cellular confinement system composed of a series of multi-chambered geotextile cells forming a web configuration. The open cells are filled with a specified material, or ballast, such as sand, earth, rocks, or a similar substance to produce a stable self-supporting structure. Typar Geocells are manufactured in four different widths (i.e., with designations DT1, DC2, DC3, and DC4). The DT1 unit is the narrowest width, followed by the DC2, DC3, and DC4 with correspondingly increasing widths. The width of the DT1 is approximately 0.6 meter. Each unit is 0.5 meter in height and 4.6 meter in length. Smaller lengths of the Geocell can be easily cut and fitted to the dimensions of a ditch cross-section. In addition, the Geocell units can be readily stacked to achieve a desired height. The units can also be supplied with slits in the fabric walls to allow for additional hydraulic flow-through capacity, if desired, for certain sediment control applications.

The wattles were installed per typical industry requirements, including trenching the wattles into the ground surface 5 to 7 centimeters. The Geocell units were not inset into the ground surface, per manufacturer’s requirements. Sections of each type of BMP measuring 4.6 meters were constructed at the downslope end of the test plots. Automatic samplers and flow meters were dedicated to each test plot. Specifically, four Teledyne ISCO Model 3700 full-size portable samplers, with 24 1-liter sample bottle capacity, coupled to four Teledyne ISCO 4230 bubbler flow meters were employed on site (two meter/sampler units used per test plot). Both the influent runoff flow rates/volumes and associated turbidities and the effluent flow rates/volumes and turbidities were monitored via the ISCO units. For the measurements of runoff flow rates and volumes, the ISCO flow meters were connected to influent and effluent rectangular flume systems (four total). The watershed immediately above the upper influent flume for each test plot measured 18.3 meters long by 4.6 meters wide. The end of each flume was fitted with a plastic 0.6-meter H-flume. A bubbler tube was anchored to the base of each H-flume and was connected to the ISCO 4230 flow meter (Figure 2).

The samplers were coupled to the flow meters and were programmed to collect flow-weighted discrete samples throughout each storm event, based on flow increments measured by the flume/flow-meter system. As a result, the event mean turbidity (EMT) values for each storm event were measured, calculated, and recorded. EMT is the overall average turbidity measured from the runoff from a given storm event, i.e., based on the overall mean turbidity as calculated from the flow-weighted discrete samples taken from a given storm event and their associated turbidities.

The ISCO units were powered by lead-acid rechargeable batteries connected to one 100-watt solar panel (Figure 3). Onsite rainfall was collected via a HOBO data-logging rain gauge manufactured by Onset Computer Corp.

Finally, piezometers with dedicated pressure transducers and data-loggers were installed in front of the wattles and the Geocell unit in order to assess the fluctuations in hydraulic head at the front of each BMP throughout each storm event (Figure 4). Monitoring of backwater head is important for tracking if and when the BMP is overtopped, dramatically reducing its efficacy. The backwater head readings, along with the topographic survey of the test plots, also allowed for monitoring of the backwater surface area behind the BMPs. The ratio of the discharge rate from the BMP to the pooled surface area behind the BMP, i.e., up-flow velocity, directly correlates with particle capture efficiencies via settling.

After each storm event, a team of CEC professionals retrieved the collected samples for turbidity measurements and downloaded data from the flow meters, transducers, and the rain station. Turbidity was measured at the CEC laboratory in Franklin, TN, using a Hach 2100Q turbidimeter. The Hach units have a maximum measurement capability of 1,000 nephelometric turbidity units (NTUs). For samples with readings that initially measured above 1,000 NTUs, the sampling team prepared and measured dilutions of the samples in the laboratory to estimate the actual turbidity readings from samples with excessive solids content. The goal of the sampling teams was to pull and measure turbidity samples within 48 hours per US EPA 180.1 requirements (USEPA 1993).

The monitoring and sampling of effluent quality from the straw wattles versus the single DT1 Typar Geocells with mulch were carried out from September 2009 to May 2010.

Scenario 2: Rock Check Dams versus Fiberweb Typar Geocells with Crushed Stone-In the summer of 2010, the straw wattles and the DT1 Typar Geocell with mulch were dismantled and replaced with two different passive BMPs for comparative purposes relative to turbidity effluent quality. In one test plot, an enhanced rock check dam was installed per Tennessee Department of Transportation (TDOT) standard drawing EC-STR-6A; in the other, DT1 Typar Geocells with clean, crushed stone as fill material were installed. The enhanced rock check dam was constructed from Class A-1 machined riprap with two layers of geotextile fabric placed across the upstream face. In addition, a 30-centimeter layer of No. 57 crushed stone was placed on top of the geotextile along the upstream face. See Figure 5 for dimensions and details for the enhanced rock check dam. A photograph of the TDOT enhanced rock check dam is shown in Figure 6. The DT1 Typar Geocell with crushed stone photograph is shown in Figure 7.

Figure 4. DT1 Typar Geocell filled with mulch
Figure 5. TDOT enhanced rock check dam standard drawing (Source: TDOT EC-STR-6A)

The setup of the test plots and the sampling equipment and instrumentation in Scenario 2 was essentially the same as the setup for the straw wattle and DT1 Geocell with mulch testing in Scenario 1. In contrast to the Scenario 1 testing, the testing during the comparative evaluation of the enhanced check dam and the stone-filled Geocell had an added approach. The Typar Geocells provide a much lower footprint than rock check dams. Consequently, the goal for Fiberweb was to determine what Geocell configuration would result in an effluent quality that was equivalent to that produced by the enhanced rock check dam. As a result, there were several modifications to the Geocell configuration during the Scenario 2 testing period for determining an optimum configuration for the Geocell unit that would be equivalent to the performance of the check dam. At the start of Scenario 2 testing on June 8, 2010, a single DT1 Geocell with 5- to 8-centimeter size stone was set up with slits in the front, interior, and rear of the fabric of the unit. The first modification occurred on August 20, 2010, when the larger stone size was replaced with No. 57 crushed stone. On April 13, 2011, the final modification, and what ultimately became the optimum arrangement, was made when two DT1 Geocell units filled with No. 57 stone were positioned adjacent to each other (see Figure 7).

The monitoring of effluent quality from the TDOT enhanced rock check dam versus the DT1 Typar Geocells with stone, Scenario 2, was carried out from June 2010 to December 2011.

As anticipated, the influent and effluent turbidity values obtained from both test scenarios were highly variable. There were 13 storm events that were sampled for Scenario 1, the straw wattle and Geocell with mulch series of comparative tests from September 2009 to May 2010. The two-year, 24-hour storm event for the test site is 96 millimeters. The only storm event that surpassed the two-year, 24-hour event during the Scenario 1 testing was the May 1 and 2, 2010, event, which was the 1,000-year storm for middle Tennessee. There were 33 storm events that were sampled for Scenario 2, the enhanced rock check dam and Geocell with stone series of comparative tests from June 2010 to December 2011. No storms during the Scenario 2 period of testing exceeded the two-year, 24-hour storm event.

Figure 6. TDOT enhanced rock check dam
Figure 7. Dual DT1 Typar Geocells filled with No. 57 crushed stone

Results

Due to periodic sampling equipment error, there were certain storm events where a number of aqueous samples were not taken. This was especially the case during Scenario 1 testing. As a result, there are occasional gaps in the data when plotted on a time series scatter graph.

A scatter plot of the final data set for EMT values taken from the Scenario 1 study is shown in Figure 8. A plot of event rainfall totals is also shown on the same graph (as a bar graph). Event rainfall totals are often larger than the 24-hour rainfall, including rain totals that may fall over two or more days in a row. A consistently lower effluent turbidity from the Geocell with mulch is evident within the plotted data. Table 1 gives descriptive summary statistics for the Scenario 1 comparisons. Bootstrapping statistical techniques were used to calculate the 95% confidence intervals about the mean for all effluent EMT values. The bootstrap is a resampling statistical method in which random samples with replacement are taken from the original data set and analyzed for some descriptive statistic, such as the mean (Efron and Tibshirani 1993). This method is most applicable when the distribution of the true population data set is unknown. Bootstrapping was chosen as a method to decrease the spread (i.e., variability) in the confidence intervals.

The mean of all EMT effluent values from the Typar Geocell with mulch is an order of magnitude lower than the mean of all EMT effluent values from the straw wattles. The bootstrap 95% confidence limits about the mean of all EMT values, the minimum and maximum values from the data sets, and the median of all EMT values also clearly demonstrates a superior effluent quality from the Geocell with mulch.

A scatter plot of the final data set for EMT values taken from the Scenario 2 study is shown in Figure 9. A bar chart plot of event rainfall totals is also plotted on the same graph. Figure 9 includes all data obtained during the study period for Scenario 2. Observations of the scatter plots reveal that the enhanced rock check dam outperformed the single DT1 with stone for the early period of the study from June 2010 to April 2011. Table 2 gives descriptive summary statistics for the Scenario 2 comparisons where all data from the study period are incorporated. Data for the TDOT enhanced rock check dam show an average for all effluent EMT readings of 281 NTUs. This result from the enhanced rock check dam is a surprisingly low value for turbidity from a rock check BMP.

Table 3 gives summary descriptive statistics for only the data from April 13, 2011, to the end of the study (December 2011). During April 2011, the configuration of the Geocell BMP was changed to a dual DT1 with No. 57 stone. From April 13, 2011, to the end of the study, the Geocell with stone outperformed the enhanced rock check dam in terms of turbidity effluent quality. Table 3 shows descriptive statistics for all effluent EMT readings from the dual DT1 units to be approximately an order of magnitude lower than those same statistics for the enhanced rock check dam.

Statistical analysis of both influent and effluent turbidities using all EMT values measured during Scenario 1 and Scenario 2 are presented in Table 4. Table 4 shows the 95% bootstrap confidence intervals about the mean for both measured influent EMT and effluent EMT data. The maximum percent reduction in turbidity from influent to effluent, based on the confidence interval of the mean of all influent and all effluent EMT values for each BMP, are also given in Table 4. The most significant decreases in turbidity from influent to effluent were produced by the single DT1 Typar Geocell with mulch fill, the TDOT enhanced rock check dam, and the dual DT1 Typar Geocells with No. 57 stone fill (after the April 13, 2011, modification).

The lowest percent reduction shown in Table 4 was measured for the straw wattle at just above 73%. The highest percent reductions were measured for the Typar Geocell with mulch and for the dual Typar Geocell with No. 57 stone at near 99%. The TDOT enhanced rock check dam had a consistent turbidity reduction near 96% when considering all data and also when considering only the data from April 13, 2011, to December 2011.

The maximum influent turbidity measured from the flow-weighted discrete samples during the overall study period from September 2009 to December 2011 was 15,536 NTUs, occurring during the Scenario 2 testing for the enhanced rock check dam influent. Confidence intervals for the mean influent turbidity for this study fall very close to the mean turbidity values typical for construction runoff, i.e., 4,000 NTUs, as reported by Nelson (1996) and Pitt (1998).

For Scenario 1, wattle backwater heads were consistently lower than Geocell heads, most likely due to higher flow-through infiltration rates for the wattle and due to water movement at the interface between the two stacked 30.5-centimeter wattles. This would indicate that the wattle is less effective at holding back water for increased settling time for suspended particles. It is clear that during the May 1 and 2, 2010, flood event, overtopping of the wattle did occur. It is probable that some overtopping of the Geocell with mulch occurred during the May flood event as well. This was the only event where overtopping of either BMP occurred.

For Scenario 2, there is no evidence of any overtopping for either the enhanced rock check dam or the stone-filled Geocell. The backwater head for the enhanced check dam was consistently higher than the head readings for the stone-filled Geocell, until April 13, 2011. After the final modification where the dual DT1s were installed, the head readings behind the Geocells were significantly higher than the heads behind the enhanced rock check dam. This would indicate that the dual DT1s with stone are at least equivalent to the TDOT enhanced rock check dam at holding back water and increasing settling time for suspended particles.

Conclusions
Comparative performance data from this comprehensive study of both nonproprietary and proprietary passive BMPs reveal that straw wattles produced effluent with the higher turbidity measurements, with the 95% confidence interval for the mean for all EMT values calculated at [603 NTUs, 1,538 NTUs]. Conversely, effluents with much lower turbidity readings were observed from the single DT1 Typar Geocell with shredded wood mulch fill, producing a 95% confidence interval for the mean for all EMT values at [30 NTUs, 189 NTUs]. When all data are included from the single DT1 Typar Geocell with stone arrangement, the overall 95% confidence interval for the mean for all effluent EMT values is calculated to be at [436 NTUs, 1,832 NTUs]. However, when the dual DT1 Geocells with stone fill are positioned side by side, the effluent quality dramatically improves with the 95% confidence interval for the mean for all effluent EMT values calculated at [16 NTUs, 50 NTUs]. Surprisingly, the TDOT enhanced rock check dam performed consistently well in turbidity reduction, reducing turbidity in the effluent to a 95% confidence interval for the mean for all EMT values to [186 NTUs, 429 NTUs].

Information relative to the potential effluent quality that can be achieved from these sediment control BMPs is only a first step.  he next step in the analysis of data from this study is to further develop and refine useful design information and guidance for sediment control BMP applications. It is essential to understand how the BMPs are functioning and how they are capable of achieving the observed effluent turbidities and under what conditions. It is unrealistic to expect such BMPs to produce low turbidities if they are subjected to exceptionally large watershed areas with high solids loadings (i.e., significant disturbed surface areas). Therefore, we must seek to better understand the application limitations for each single sediment control unit. The research team for this work is currently analyzing the porous-flow rates from these BMPs along with the associated backwater surface areas contained behind the units during the study period. The discharge flow rate to backwater surface area ratio (also referred to as [ITALIC]upflow velocity or [ITALIC]upflow flux rate) has a direct bearing on the soil particle sizes retained behind these units. The ultimate goal is to provide more informed design approaches for BMP treatment trains, including refined guidance for the maximum watershed acreage allowed to drain to each single BMP type, based on the internal porous-flow restrictions of the BMP types investigated, and incorporating considerations of critical particle settling velocities with associated upflow flux rates. These design refinements will be presented in future publications.

Acknowledgments

The authors would like to thank the Williamson County government for assistance and cooperation during this research effort, with special thanks to the county solid waste director, Nancy Zion.

References
Efron, B., and R. Tibshirani. 1993. An Introduction to the Bootstrap. Chapman and Hall.

Nelson, J. 1996. Characterizing Erosion Processes and Sediment Yields on Construction Sites. MSCE Published Thesis. Department of Civil and Environmental Engineering, University of Alabama at Birmingham.

Pitt, R. 1998. “Sediment Control in Alabama.” 41st Annual Transportation Conference. Montgomery, AL.