PROOF COPY [EE/2007/024823] QEE

Transcription

1 1 2 Comparison of Stormwater Solids Analytical Methods for Performance Evaluation of Manufactured Treatment Devices 3 Shirley E. Clark, Ph.D., P.E., M.ASCE 1 ; and Robert Pitt, Ph.D., P.E., D.WRE., DEE, M.ASCE Abstract: As more manufactured stormwater treatment devices enter the market, stormwater managers are searching for effective and 6 rapid methods for evaluating device performance. Many agencies require vendors to test full-scale versions of their devices under 7 controlled conditions. The most common parameter used to document performance is suspended solids for several reasons: 1 many 8 pollutants attach to solids; 2 a solids simulant is relatively easy to generate; and 3 solids are comparatively easy and inexpensive to 9 quantify. However, a controversy still exists in the profession and some regulatory agencies as to whether total suspended solids TSS or 10 suspended sediment concentration SSC, or both, should be measured. This paper focuses on the comparability of the two methods/ 11 protocols used for sample solids analysis, including lessons learned during recent evaluations of two manufactured treatment devices. 12 Analysis of 215 sample pairs where both TSS and SSC were measured on aliquots of the same sample showed that statistically the TSS 13 measured using the wide-bore pipet method and SSC results were indistinguishable from one another and from the original simulant 14 mixture. 15 DOI: XXXX 16 CE Database subject headings: Suspended solids; Stormwater management; Analytical techniques; Water analysis; Performance 17 characteristics; Evaluation Introduction 21 The marketplace of new manufactured stormwater treatment devices is growing rapidly. Past history of these devices has shown that the performance claims vary widely, based on the vendor s 24 testing. The result has been that, once these devices are installed 25 in the field, the performance may not be at the level claimed by 26 the vendor. Because of these discrepancies and in order to help 27 the stormwater manager select a device based on comparable testing, several states and the federal government have developed and implemented standard testing protocols, which are managed 30 through approved verification agencies. For example, the U.S. 31 Environmental Protection Agency USEPA has NSF International, Inc. s Water Quality Protection Center NSF Inc acting as its verification agency as part of its environmental 34 testing-verification ETV program. The verification agency/ 35 organization typically develops the required testing protocol and 36 then selects an organization to conduct the actual testing. At the 37 end of testing, the verification organization provides a recommendation for the anticipated performance range of a particular treat ment device, or verifies the vendor s performance claims. 1 Assistant Professor, Dept. of Environmental Engineering, Penn State Harrisburg, 777 W. Harrisburg Pike TL-105, Middletown, PA corresponding author. seclark@psu.edu 2 Cudworth Professor of Urban Water Systems, Univ. of Alabama, Box , Tuscaloosa, AL rpitt@eng.ua.edu Note. Discussion open until September 1, Separate discussions must be submitted for individual papers. To extend the closing date by one month, a written request must be filed with the ASCE Managing Editor. The manuscript for this paper was submitted for review and possible publication on February 20, 2007; approved on July 13, This paper is part of the Journal of Environmental Engineering, Vol. 134, No. 4, April 1, ASCE, ISSN /2008/4-1 XXXX/ $ Two possibilities generally exist for performing these fullscale verification activities: field installations versus laboratory testing. Because of the time required and the lack of control on 42 the field test site, laboratory testing is often preferred by vendors 43 and is gaining favor with some agencies for, at a minimum, interim certification. A comparison between two or more devices claiming to address a specific pollutant is much easier with controlled laboratory testing Many of the laboratory-testing protocols focus first on solids 48 removal for two reasons: 1 it is the primary parameter of interest 49 for most regulatory agencies; and 2 it is considered the easiest 50 parameter to simulate. There are no interfering reactions to simulate and no associations between dissolved and particulate pollut ants to investigate. Laboratory testing does have its limitations, 53 however. Many of these concerns were addressed by Guo and include the ability to accurately capture and quantify the particles represented in the prescribed test mixtures. This paper fo cuses on the differences found in the two most common methods 57 of solids measurement total suspended solids TSS and suspended sediment concentration SSC. The large sample sets dis cussed in this paper were generated using simulant solids mixtures used by two teams of researchers at Penn State Harrisburg PSH and the University of Alabama UA as they verified the 62 solids removal performance of two prototype and full-scale devices. In addition, these results are compared with a set of field data generated from ETV testing of one of the two treatment 65 devices. The results of aliquots of same sample analyzed for both 66 TSS and SSC are compared statistically. 67 Methodology Much of the past research on stormwater runoff particulates has 69 used the TSS methodology either by EPA or by Standard Meth JOURNAL OF ENVIRONMENTAL ENGINEERING ASCE / APRIL 2008 / 1

2 Table 1. Comparison of Three TSS/SSC Analytical Methods EPA TSS S.M. TSS 2540D USGS SSC D B Filter nominal Not specified 2.0 m 1.5 m pore size Sample mixing Shake vigorously Stir plate Decant supernatant and flush bottle with DI water Aliquot size Not specified Not specified Entire sample Method of aliquot collection Pour aliquot into graduated cylinder Pipet at middepth in bottle and midway between wall and vortex Pour from original bottle 71 ods to quantify these solids. However, the U.S. Geological Survey USGS in much of their water-quality monitoring efforts has used a different methodology to quantify the amount of particles 74 in a water sample the SSC method. They argue that TSS methods miss many of the larger solids often found in water quality samples collected in streams Gray et al For this reason, 77 currently many of the testing protocols developed by verification 78 agencies require the testing organization to analyze samples for 79 both TSS and SSC. For example, the New Jersey Department of 80 Environmental Protection, through their Technology Assistance 81 Reciprocity Protocol TARP Tier II Protocol, including the New 82 Jersey amendments NJDEP 2003, and EPA/NSF ETV required 83 analysis for both TSS and SSC, in contrast to the City of Portland 84 Bureau of Environmental Services BES guidance City of Portland 2004 which only requires SSC analysis. While both TSS and SSC are intended to quantify the concentration of suspended solids in a sample, many researchers have shown that these two methods often give varying results because the range of 89 particle sizes each method can quantify differs Gray et al. 2000; 90 Guo A lack of understanding of the distinction between the 91 methods also permeates some testing protocols. The protocols 92 require specific, and identical, TSS and SSC influent concentrations, even when the test solids have a substantial fraction in the fine sand or larger size range, which typically are not recovered in 95 the TSS analysis. 96 In addition to the concerns over the ability to analyze across 97 the variety of solids found in stormwater runoff, different 98 organizations/agencies have slightly different methods for analyzing TSS. Two of the most common TSS methods are from the US EPA Method USEPA 1999 and from Standard Methods 101 for the Analysis of Water and Wastewater, Method 2540D APHA These solids characterization methods differ in two important ways: 1 the specification, or lack thereof, of the nominal pore size of the filter or of the efficiency rating ; and 2 the 105 method used to collect the sample aliquot. Both of the TSS methods are documented to have problems capturing the larger par ticles found in simulated stormwater. For example, Standard 108 Methods requires sampling middepth, yet that sampling location 109 is likely to miss the larger particles that cannot be adequately 110 suspended evenly in the water column. The EPA method requires 111 a shake-and-pour aliquot selection, with the larger particles settling in the original sample container before pouring is completed Therefore, several verification agencies have also been requiring 114 SSC measurements. The purpose of requesting both measurements is so that comparisons to historical data can be performed Most historical stormwater solids concentrations were measured 117 using one of the two TSS methodologies. 118 The SSC analysis method was developed by the American 119 Society for Testing and Materials ASTM, and endorsed by the 120 USGS, as a method better able to capture and quantify larger 121 particles that may be collected in natural water, including stormwater, samples. That analysis technique is ASTM D B 122 ASTM 1997 reapproved in A summary of the three 123 methods is shown in Table As part of two verification projects one for a sedimentation 125 device and one for a filtration device, the researchers performed 126 TSS and SSC on both influent and effluent samples, ending up 127 with approximately 215 matched pairs. The simulant for each of 128 these tests consisted of a solids distribution similar to that required in the NJ DEP Laboratory Testing Protocol clay 1 2 m, %; silt 2 8 m, 15%; silt 8 50 m, 25%; very fine sand m, 15%; fine sand m, 30%; medium sand m, 5%; coarse sand 500 1,000 m, 5%. These simulants were made using sieved sand and Sil-Co-Sil 250. For the filtration device, the simulant had a portion of the sand replaced 135 by a ground fertilizer to add phosphorus to the challenge mixture. 136 The mix water source was an artesian well in the case of the 137 sedimentation device and tap water for the filtration device. The 138 simulant particles were silica based to meet the NJ DEP recommendation of a specific gravity of Each 500-mL influent and effluent sample from each device 141 was split using a Dekaport cone splitter into five subsamples. An 142 error analysis was performed for the cone splitter during a prior 143 project, and the results showed that, over the particle size and 144 concentration range of interest for this project, the median coefficient of variation COV calculated as standard deviation/mean between splits of the same sample was 5%, with lower COVs for 147 higher concentration samples. One split was retained for particle 148 size distribution analysis using the Coulter Multisizer 3, and four 149 were used for solids analysis TSS, TSS sieved to 250 m, 150 SSC, and SSC sieved to 250 m. Because the sample bottles 151 were filled to the brim, the actual volume split exceeded 500 ml 152 and were approximately ml. The TSS analyses were 153 performed following the Standard Methods protocol 2540D using 154 a wide-bore pipet. To reduce analytical bias, the pipet tip was not 155 allowed to contact the bottom of the stirred sample bottle, even as 156 the water level dropped in the sample bottle. In addition, the 157 convex nature of the bottom of the bottles ensured that the pipet 158 aliquot was not unduly biased by sampling those solids that did 159 not suspend but traveled in a slurry along the bottle bottom. Additional testing Clark and Siu 2007 by the research team of the wide-bore pipet TSS method has shown that the percentage of 162 sample in the aliquot did not influence the solids recovery. The 163 two subsamples were sieved to remove particles larger than m in order to quantify the mass of large particles in the 165 original samples. These results, along with the results of the TSS/ 166 SSC sieved, are not relevant to and are not included in this paper / JOURNAL OF ENVIRONMENTAL ENGINEERING ASCE / APRIL 2008

3 Fig. 1. SSC versus TSS samples collected during two device verification activities Fig. 2. TSS/SSC versus mass balance MB measurements of influent concentration 168 Results 169 Analysis of Entire Data Set 170 The 215 paired influent and effluent TSS and SSC samples were 171 compared through a scatterplot analysis shown in Fig. 1. A linear 172 regression of these data, assuming that the intercept is zero, 173 showed that the SSC measurements were predicted to be 1.4% 174 less than the TSS measurements from the same sample SSC 175 =0.986 TSS. The slope coefficient was significant with 176 p However, the 95% confidence interval on the slope 177 coefficient was , indicating that the slope coefficient, 178 while statistically significant, could not truly be distinguished 179 from 1.0 which is represented by the line SSC=TSS. These 180 results differ from those of Gray et al. 2000, who found that 181 TSS measurements were typically 25 34% less than the corresponding SSC measurements for paired samples. The data were reanalyzed using only the samples generated with no fouling 184 agents or fertilizer in the influent mixtures. The remaining sample pairs had the following relationship: SSC=0.960 TSS. 186 This analysis states that the SSC concentrations were typically 187 4% lower than the TSS measurements for the same sample. The 188 slope coefficient was significant with p However, the % confidence interval on the slope coefficient was , 190 indicating that the slope coefficient again could not truly be distinguished from 1.0. The direction of this statistically insignificant difference was in line with Gray et al. 2000, but not of the same 193 magnitude. 194 Fifteen runs of the sedimentation device with multiple TSS 195 and SSC measurements collected per run also had the influent 196 concentration calibrated through a mass balance calculation. The 197 solids were dispensed using a dry feeder and the solids flow rate 198 was measured over a 5-min period. The weights of these collected 199 samples were combined with the measured flow rate to generate 200 the hopper influent concentration by mass balance. These results 201 are shown in Fig. 2 and include the regression lines for the SSC 202 versus hopper influent concentration designated MB and for the 203 TSS versus hopper influent concentration. The regression equations for these analyses show that the relationship between the SSC and the MB calculation is SSC =1.02 MB, whereas for TSS, it is TSS=0.72 MB. Both slope 206 coefficients are statistically significant with p An analysis of variance ANOVA between the three groups of samples TSS, SSC, MB indicates that, for the number of samples in each 209 group, none of the groups could be distinguished statistically 210 from the other p 0.01 for all combinations. Therefore, while it 211 appears in Fig. 2 that the SSC and the MB are the best replicates 212 of each other, the TSS data cannot be statistically distinguished 213 from them. From a reporting standpoint, the entire dataset could 214 be combined and used to document performance, as opposed to 215 the separate reporting for TSS and SSC required by the agencies. 216 Given the variability in the data set, substantially more samples 217 would be needed for the differences in the data sets to be confirmed statistically Influence of Subsampling Techniques 220 As noted in the method discussions, subsampling techniques can 221 greatly influence the results Fig. 3 shows the same data as above 222 but broken down by analyst. Analysts 1 3 and Analyst 5 were 223 trained in the methods thoroughly by previous analysts prior to 224 commencing sample work, including on the difficulties of obtaining a representative subsample for TSS analysis. Analyst 4 was given the TSS method using Standard Methods protocol and a 227 set of samples. Analyst 4 therefore followed the Standard Methods recommendation of sampling midway between the vortex and the wall and at middepth, while the others were within 1 cm of 230 the bottom of the bottle and midpoint between the wall and the 231 vortex. For Analyst 4, TSS measurements were approximately 232 one half those of the corresponding SSC measurements. Sampling 233 at the middepth location was insufficient for retrieving a representative fraction of the larger particles which tended to slide along the bottom of the beaker or bottle. Unlike the treated sanitary wastewater that this method was designed to analyze, the JOURNAL OF ENVIRONMENTAL ENGINEERING ASCE / APRIL 2008 / 3

4 Fig. 3. SSC versus TSS for 215 full-scale verification tests Fig. 4. SSC versus TSS for 112 full-scale verification tests using influent samples 238 solids in these simulants were not well mixed in the water column The regression equations and the 95% confidence intervals on 241 the slope coefficient are shown in Table 2. All slope coefficient 242 values were statistically significant at the p level. R 2 values ranged from 0.66 to Most of the confidence intervals contained the slope coefficient of 1.0, indicating that for the various analysts, TSS and SSC measurements for this data set were statistically indistinguishable from one another. Only Analyst 3 s 247 confidence intervals for the slope coefficient did not contain 1.00, 248 thus statistically indicating a slight bias. 249 According to Gray et al. 2000, an analysis of paired TSS and 250 SSC samples showed that TSS results were typically between and 34% less than the paired SSC sample. According to this testing, with the exception of Analyst 4, SSC results were between and +10% of the TSS values. For Analyst 3 the only analyst 254 whose 95% confidence interval on the slope coefficient does not 255 contain 1.00, the SSC bias was approximately +10% of the TSS 256 values. TSS values were not consistently negatively biased, unlike the results found by Gray et al Additionally, the results shown as Analyst 4 demonstrate the importance of subsampling near the bottom of a container when using the TSS method if the goal is to have the TSS results reflect all solids in 261 the water and not just those that are well mixed in the water 262 column. TSS subsampling near the top or middle of sample bottles that have a substantial fraction of larger particles will lead 263 to the biases similar to or larger than those documented by Gray 264 et al Influence of Large Particles The second notable result from Gray et al was that the 267 size of the difference between the SSC and TSS result was greatest in samples where the sand fraction 62 m equaled/ exceeded 25% of the sediment. Their results showed that TSS 270 was relatively similar to the SSC results when most of the particles are finer than 62 m. To determine the effect of a substan tial fraction of particles being sand sized 250 m for these 273 tests on the TSS and SSC results, the data in Fig. 3 were reanalyzed using only the influent samples, since it was shown that the effluent samples had very few, if any, particles 250 m less 276 than 10% of the effluent samples had solids 250 m. The 277 influent-only sample results are shown in Fig. 4 and Table According to the testing, with the exception of Analyst 4 see 279 description above of effect of in-bottle sampling location on results, SSC results were between 13 and +10% of the TSS val ues. TSS values were not consistently negatively biased in the 282 influent-only analysis. The regressions of this data show that the Table 2. Regression Equations for Influent+Effluent and Influent-Only Sample Sets Analyst Influent+effluent equation SS=a TSS Influent-only equation SSC=a TSS 1 SSC=0.904 TSS R 2 =0.953, CI: 0.757, 1.05 SSC=0.896 TSS R 2 =0.927, CI: 0.690, SSC=1.004 TSS R 2 =0.715, CI: 0.857, 1.15 SSC=1.003 TSS R 2 =0.707, CI: 0.797, SSC=1.104 TSS R 2 =0.953, CI: 1.04, 1.17 SSC=1.104 TSS R 2 =0.952, CI: 1.01, SSC=1.902 TSS R 2 =0.661, CI: 0.998, 2.08 SSC=2.015 TSS R 2 =0.669, CI: 0.385, SSC=0.994 TSS R 2 =0.870, CI: 0.828, 1.16 SSC=0.878 TSS R 2 =0.865, CI: 0.647, 1.11 Field SSC=0.935 TSS R 2 =0.954, CI: 0.838, 1.03 SSC=0.898 TSS R 2 =0.943, CI: 0.742, 1.05 All analysts SSC=0.986 TSS R 2 =0.872, CI: 0.935, 1.04 SSC=0.960 TSS R 2 =0.848, CI: 0.884, / JOURNAL OF ENVIRONMENTAL ENGINEERING ASCE / APRIL 2008

5 284 slope constants for all equations were statistically significant at 285 the p 0.05 level the largest p value equaled The R values ranged from 0.67 to Table 2 also provides a comparison of the equations generated from the two data analyses. In general, the slope coefficients are generally farther from 1.0 in the influent-only samples compared to the influent+effluent samples. In addition, the R 2 values are slightly smaller and the confidence intervals are larger. This reinforces the difficulty in adequately subsampling for TSS using the wide-bore pipet and indicates that the direction of the bias for each analyst typically is consistent. The magnitude of the bias simply increases when the samples have more sand particles. The effect of sand particles on the bias between TSS and SSC is similar to that found by Gray et al. 2000, although the magnitude of the bias is not the same for all analysts. These results were compared to the ETV field testing of one of the two manufactured treatment devices at a public-works yard in Harrisburg, Pa. denoted Field in Figs. 3 and 4 and in Table 2. The field samples were collected using automatic samplers and were single composited influent and effluent samples from each storm event. The median percent of mass 250 m in the influent was 15% and in the effluent 2%. The equations for the field data were, for the influent+effluent, SSC=0.935 TSS, and for the influent only, SSC=0.898 TSS. The field data results were similar to those from the laboratory testing with biases less than that seen in Gray et al The bias increased when the effluent samples, which had a much smaller percentage of samples with the large-sized particles, were eliminated from the data analysis. While the testing results document the existence of a small, statistically insignificant bias between SSC and wide-bore TSS analyses, these results show that the bias is not consistent between samples where the maximum particle size is between 1,000 and 1,700 m compared to those whose maximum particle size is 250 m. In addition, the size of the bias is not as great as seen by Gray et al For the laboratory, the average bias between SSC and TSS was 1.4% for the combined sample set and 4% for the influent-only sample set. In addition, statistically the TSS and SSC analyses could not be distinguished from each other for 215 sample pairs. The lack of statistical significance shows that while to the untrained eye, the differences appear to be of concern, they, in fact, are not and analyzing the results by either of the two methods should generate statistically similar results. For the 215 data pairs in this project, and assuming a statistical power of 80% and a COV of 0.5, the minimum bias between the two methods that could be detected at 95% would be 15 20%. To detect the differences seen in this paper, given the variability in each data set, would require at least 500 and likely closer to 1,000 sample pairs. Mass balance calculations, which are available for the influent for these controlled full-scale laboratory testing activities for solids removal performance, should be the most accurate measure of the concentration of solids entering the device. These results show that the mass balance calculations are well replicated by the SSC measurements, but not as well by the TSS measurements. However, the statistical analysis shows that the three groups are not statistically distinguishable from each other. Therefore, the mass balance, which should be a simpler measurement to obtain, should suffice as the measure of influent concentration in these tests. Reliance on the mass balance calculation also removes the difficulty documented in these data sets of adequately capturing the larger particles in a traditional TSS/SSC water sample analysis. Conclusions As more manufactured treatment devices enter the marketplace, 347 stormwater managers need to document that the device performs 348 as anticipated. For this reason, many verification agencies have 349 incorporated a laboratory testing component into their approved 350 protocols for documenting device performance. However, questions remain about setting up a representative laboratory program that can adequately represent field conditions and about how to 353 accurately measure the solids concentration in the device s influent and effluent Stormwater researchers and managers have been debating as to 356 whether to use TSS or SSC to quantify the solids concentration in 357 runoff samples and the ability of the treatment device to remove 358 these solids. Other researchers, including those of the USGS, 359 found that TSS measurements were consistently lower than paired 360 SSC measurements by up to 25 34%. This research has outlined 361 the inconsistencies of the three traditional solids analysis methods 362 such as the filter pore size which, if specified, differs between 363 methods and between researchers and the method of obtaining 364 subsamples for filtering and the differences obtained between the 365 two methods for this laboratory. The results have shown that, for 366 the same analytical filters, the bias between TSS Standard Methods 2540D and SSC is not as great as seen by the USGS for the particle-size distribution of the challenge solution samples, and 369 the differences are not statistically significant. However, substantial differences can exist between analysts, if training methods are not consistent. This is particularly crucial when performing TSS 372 analysis, which uses a small subsample of the original sample. If 373 the aliquots are not collected at a location in the mixing sample 374 where a representative fraction of the larger solids in the mixture 375 can be collected by the pipette, SSC results will be measurably 376 greater than the TSS results for the same sample. 377 This research also showed that the mass balance method of 378 calculating influent solids concentration where we know what 379 we added the simplest and most reliable can be well represented in both the TSS and SSC measurements. Therefore, there is an argument to be made that mass balance calculations are 382 comparable to the more time-consuming laboratory analyses and 383 could be substituted for the analytical efforts on the influent side. 384 This research shows that they are statistically identical. However, 385 if it is desired to have an analytical result that most resembles 386 the entire mass of solids added, then the SSC method is recommended since it provides the closest results to the known solids concentration. If the goal is to measure the solids that are well 389 mixed in the water column, then the TSS method may be most 390 appropriate. 391 This research has shown that many questions still exist in how 392 to document the performance of manufactured treatment devices, 393 especially when developing and implementing full-scale laboratory testing protocols. Therefore, it is incumbent upon stormwater managers who are interpreting the results of verification testing to 396 determine the methods of sample analysis used and match the 397 analytical methodology with the goals of the program. Differences in each of these between vendor verification testing can impact on the results. The purpose of standardized laboratory testing has been to ensure that the comparison is apples to apples but the lack of attention to the details of testing and analysis has 402 left the profession examining devices using a red apples to green 403 apples approach JOURNAL OF ENVIRONMENTAL ENGINEERING ASCE / APRIL 2008 / 5

6 405 Acknowledgments 406 The writers would like to express their sincere appreciation to all 407 the current and former graduate and undergraduate research assistants who contributed to this research. James Elligson, Brad Mikula, Christine Siu, Christopher Roenning, and Julia Hafera of 410 Penn State Harrisburg, and Uday Khambhammettu, Renee 411 Morquecho, Yukio Nara, and Alex Maestre of the University of 412 Alabama are some of the students who have spent countless hours 413 helping to develop and then to evaluate these methods to ensure 414 high quality research results. 415 References 416 American Public Health Association APHA Standard methods 417 for the examination of water and wastewater, 19th Ed., American 418 Water Works Association and the Water Environment Federation, 419 Washington, D.C. 420 ASTM Method D (B): Standard methods for determining 421 sediment concentration in water, West Conshohocken, Pa. 422 City of Portland Appendix B: Vendor submission guidance for 423 evaluating stormwater treatment technology. Bureau of Environmental Services Stormwater management manual, Portland, 424 Ore. Clark, S. E., and Siu, C. Y. S Measuring solids concentration in 425 stormwater runoff: Comparison of analytical methods. Environ. Sci. 426 Technol., in press. 427 Gray, J. R., Glysson, G. D., Turcios, L. M., and Schwarz, G. E Comparability of suspended-sediment concentration and total suspended 429 solids data. U.S. Geological Survey Water-Resources Inves- 430 tigation Rep. No , U.S. Geological Survey, Denver. 431 Guo, Q Development of adjustment and scaling factors for measured 432 suspended solids removal performance of stormwater hydrody- 433 namic treatment devices. Proc., World Water Environmental Resources 434 Congress CD-ROM, Anchorage, Alaska, American Society 435 of Civil Engineers, Reston, Va. 436 New Jersey Department of Environmental Protection NJDEP The technology assistance reciprocity partnership TARP protocol 438 for stormwater best management practice demonstrations. adopted 439 by California, Massachusetts, Maryland, New Jersey, Pennsylvania, 440 and Virginia, NSF International, Inc Environmental technology verification 442 statement and report: Terre Hill Concrete Products Terre 443 Kleen 09. Rep. No. 06/29/WQPC-WWF, Ann Arbor, Mich., Terre_Kleen_Statement%20final.pdf USEPA Method 160.2: Total suspended solids (TSS) (gravimetric, 447 dried at C), Revised Ed., Washington, D.C / JOURNAL OF ENVIRONMENTAL ENGINEERING ASCE / APRIL 2008