POINT SOURCES AND ENVIRONMENTAL LEVELS OF 2378-TCDD ON THE MIDLAND PLANT SITE OF THE DON CHEMICAL COMPANY AND IN THE CITY OF MIDLAND, MICHIGAN NOVEMBER 5, I984 THE DOH CHEMICAL COMPANY MIDLAND, MICHIGAN SECTIONS A INTRODUCTION) SUMMARY, RESULTS AND DISCUSSION Agin Atiemo?Obeng Crummett Krume] Lamparski Nestrick Park Rio a Robbins I. . I Tobey ?h 3 Townsend Nestover I I I I SCIENTIFIC REVIEW PANEL Prof. Henry Freiser, Chairman University of Arizona Prof. R. Graham Cooks Purdue University Prof. Dr. Otto Hutzi' University of Federal Repub Prof. Peter C. POINT SOURCES AND ENVIRONMENTAL LEVELS OF 2378-TCDD ON THE MIDLAND PLANT SITE OF THE DON CHEMICAL COMPANY AND IN THE CITY OF MIDLAND, MICHIGAN 11/5/84 EXECUTIVE SUMMARY Background This report summarizes The Dow_Chemical Company investigation of point sources and environmental levels of on the Dow Midland Plant Site and in the City of Midland, Michigan. It is part of the research initiatives program announced by Dow in June 1983 to address public concerns about the trace presence of dioxins in the environment, and the potential health impact of TCDD levels in the Midland community. This study was a comprehensive search for all critical point sources of TCDD to the air, soil and water in the Midland area. Nearly 240 environmental samples were collected and tested by Dow analytical scientists during this research. About 6,000 data points were gathered on the samples. Using the most advanced analytical technology available, the scientists searched for sources by measuring the concentration of TCDD in each sample and the amount of the source effluent from which the samples were taken. A unique ?fingerprinting" technique helped the researchers to identify the most signifi- cant sources of TCDD. Finally, the research team studied how TCDD is trans- ported in the environment and how it can be removed from circulation. Dr. Henry Freiser, professor of Analytical Chemistry at the University of Arizona, served as an independent auditor and monitor for this study. Dr. Freiser is a respected researcher in the field of environmental analytical methodology and trace analysis. The reliability of the analytical procedures used in gathering the data and the validity of the conclusions have also been reviewed by an independent panel of world-recognized authorities on various aspects of data collection and interpretation. This panel, chaired by Professor Freiser, consisted of Professor R. Graham Cooks, Department of Chemistry, Purdue University, Professor Dr. Otto Hutzinger, Chair of Ecological Chemistry and Geochemistry, University of Bayreuth, Federal Republic of Germany, and Professor Peter C. Jurs, Department of Chemistry, State University. Point source identification provides the basis on which Dow will continue to pursue research, and develOp site-Specific controls to further reduce the movement of trace quantities of TCDD into the Midland-environment. The results of this study have been submitted to local, state and federll government regulatory agencies. *The Specific isomer 2378-TCDD will be referred to as TCDD or dioxin. Results Summary 1. Surface soil TCDD levels in the City of Midland are significantly below the part per billion (ppb) concern level for residential areas estab- lished by the U.S. Public Health Service Centers for Disease Control (CDC). Dioxin levels in more than 99 percent of the surface soils examined on the Midland Plant Site also are below that level. Two well-defined areas covering less than one?half percent of the Midland Plant Site have been identified where TCDD levels in soil These two areas are being contained. These two areas are directly associated with the historical manufacture and handling of chlorophenolic compounds. These findings are consistent with available scientific data about the formation and distribution of TCDD: It is known to be a trace contaminant in certain chlorophenolic manufacturing process and a low-level byproduct from the combustion of organic materials in the presence of chlorine- containing compounds. Following a comprehensive study of Midland Plant Site power generating, chemical manufacturing and waste management operations, the laqgest point sourcehgf dioxin emissions_to the air environment was ?gund to be the Waste Incinerator.? Other identifi?d point?sourceS'were shown to be small in comparison to this source and, therefore, were not considered dominant sources. As currently operated, the Waste Incinerator releases approxi- of the atmosphere per year. The concentra- tion of TCDD in the vent from the Haste Incinerator is comparable to, if not lower than, that emitted by several municipal incinerators investi- gated by the U.S. Environmental Protection Agency. This low?level emission rate cannot account for the TCDD levels known to be accumulated in the surface soils of the Midland Plant Site and in the City of Midland. Detailed analyses of past incineration practices, along with studies on the soils and airborne dust particles in the Midland area, show that historical dispersion of ashes and yent stack particulates from historical incineration operations are the probable source of the trace TCDD levels now found in the local environment. The Haste Incinerator is also the most probable source of the 0.6 gram per year of TCDD which enters the Tittabawassee River from the Midland Plant Site. Exhaust gas wash waters from the Waste Incinerator appear to be the major source of the TCDD leaving the Dow Haste Water Treatment Plant. A comprehensive survey of all other sources of TCDD to Midland Plant Site wastewater (described in more detail later in this summary) has identified three historical sources that release TCDD intermittently to the Midland Plant Site Waste Water Treatment Plant. The current outfall to the Tittabawasee River from the entire plant contains about 25 parts per quad? rillion TCDD carried by fine suspended solids. -3- DIOXIN POINT SOURCE RESEARCH STUDY RESULTS: SOIL SAMPLING I. CITY OF MIDLAND SOIL SAMPLING Analyses for TCDD in soil samples taken in the City of Midland and in samples taken just outside the Plant indicate that TCDD soil levels are significantly below the 1 part per billion (ppb) concern level established by the U.S. Public Health Service Centers for Disease Control (CDC) for exposure in residential areas. Levels between 0.0006-0.45 were observed; the levels decrease with increasing distance from the Midland Plant Site suggesting that trace levels in area soils may be associated with current or former Midland Plant Site operations. II. NON-MIDLAND SOIL SAMPLING In order to assess the above findings, soil samples were also taken in indus- trialized areas of 16 other Midwestern and Midatlantic cities. Samples were taken near major steel, automotive or chemical manufacturing facilities, and near municipal solid waste incinerators. According to the data collected, soil levels of TCDD were below 0.0l ppb. Comparison of these data with those on the City of Midland soil samples suggests that: 1. While soil levels close to the Midland Plant Site are below 1 ppb, the levels are higher than those found in other industrialized urban areas. This suggests that thg_?ow Miglg?ghPlant is_a_primary source_of the trace environment. 2. Soil levels found in the City of Midland further from the Plant Site are within the range measured in industrialized areas of other U.S. cities. 3. The widespread presence of TCDD in U.S. urban soils up to 0.0l suggests that local combustion sources are probable sources of trace dioxin at those locations, including both industrial and municipal waste incinerators. DON MIDLAND PLANT SITE SOIL SAMPLING The Midland Plant Site has been involved in manufacturing chlorophenolic compounds since the l930's. 0f 24 commercially significant compounds produced at one time or another, only are shill manufactured at the Site. Chloro- phenolic production site studies were important in investigating potential -4- TCDD sources since TCDD is a known trace contaminant in certain chlorophenolic production processes. For this segment of the study, repeated soil samples were gathered from locations where relatively high TCDD levels were expected or found. Samples were also taken where low levels were predicted, to verify the predictions. Data were categorized in three groups: Group 1--samples from three areas known to be directly associated with current or hiStoric chlorophenolic production and handling; Group 2-?samples from locations known to be associated with incineration of chemical and conventional wastes, and ash storage; Group 3?-samples from a variety of on-site locations away from potential TCDD sources, believed to represent a general TCDD surface soil background level within the Midland Plant Site fenceline. Group 1 TCDD soil levels generally were below 1 with two exceptions. Two areas with localized elevated levels of TCDD up to 50 were identified. These two locations total less than six acres and comprise less than one-half percent of the total lSOO-acre Midland Plant Site. These areas are located where chlorophenols were manufactured historically. Group 2 soils were also below except for two samples with levels of 2 and 4 ppb, values typical of those seen in incinerator ash and particulates. Group 3 soils were all well-below 1 ppb. In summary, with the exception of two localized areas, surface soil TCDD levels in the majority of the Midland Plant Site were below ppb. In areas away from known potential TCDD sources, surface soil levels generally were less than one-half ppb. IV. DIOXIN POINT SOURCES TO MIDLAND SOILS The purpose of this portion of the study was to identify primary TCDD sources. The mere presence of TCDD in a stream or sample does not confirm this as a source. Before an effluent or stream can be considered a primary TCDD source it must meet the following criteria: 1. The concentration must be high enough to significantly contribute to the levels observed in the environment. 2. The quantity, or mass of TCDD being emitted must be sufficiently large to account for the observed quantities of TCDD in the environment. 3. There must be a plausible dispersal mechanism. 4. The source must have a "fingerprint? corresponding to that observed in the environment. -5- Using these criteria several potential TCDD sources were investigated: 1. 3. Midland Plant Site Powerhouse?~This 60-megawatt, two-million-pound- per-hour steam cogeneration plant was evaluated as a potential TCDD source; samples were taken of powerhouse cinders, flyash, stack dust and exhaust gases. Although minute tracg?_gj_ICQD are present in the ash_and stack particulates, none were concentrated enough to be significant sources of TCDD to the area environment. Combustion Unit and Process Vent Stack Effluents--The Midland Plant Site operates combustion units which incinerate chlorinated wastes. Several process vents with?potential TCDD emissions were also analyzed. The largest of these units is the rotary kiln Haste Incinerator. Trace levels of TCDD were found in exhaust streams of each of these units. Not including the Waste Incinerator, the combined TCDD output of all these other units was only 0.001 gram per year. The Waste Incinerator releases about 0.3 gram per year. Therefore, attention was focused on the Midland Plant Site Haste Incinerator as the most likely TCDD source into the environment. Midland Plant Site Haste Incinerator--This unit typically burns more than 300 tons of solid minimize particulate emissions, kiln ashes are quenched in a water slurry, and combustion gases are scrubbed with water in an extensive emission control system. The ash is sent to a licensed Class I landfill. The water from the ash quench and emission control system is routed to the Midland Plant Site Haste Water Treatment Plant. It should be noted that the Waste Incinerator exhaust TCDD concentrations were significantly lower than those from municipal waste incinerators measured and reported by the U.S. Environmental Protection Agency. Municipal incinerators were found to be emitting exhaust gases with TCDD concentrations up to 70 times greater than the Midland Plant Site Incinerator. In samples taken for this part of the study, the quantity of TCDD cur- rently being emitted to the atmosphere was not enough to account for levels being found in the environment. The impact of the Waste Incinerator on the area environment will be discussed later in this summary. Transgort--TCDD surface soil levels inside the Midland Plant Site were higher than in City of Midland residential soils. [his suggests that an airborne tranSport mechanisnLis dispersing the dioxin. To confirm the point source for this airborne TCDD, researchers analyzed dust particulates from various sources. It was noted that the concentra- tions generally were within the range that would be expected for stack particulates dispersed from the Haste Incinerator. -5- V. DIOXIN POINT SOURCE RESEARCH STUDY RESULTS: WASTE HATER STREAM SAMPLIN 0L This study segment identified possible TCDD sources flowing into the waste- water system, distinguishing current from historical sources. Hater, sediment and sludge samples were collected and analyzed. Hater flow rates were approxi- mated for the streams that were sampled. The Midland Plant Site wastewater system handles about l9 million gallons of water daily. All sources that flow into this network were sampled: sewers, the riverbank revetment system, incinerator scrubber waste and landfill dewatering systems. The findings indicate: 1. The Waste Incinerator is a current point source of TCDD to the wastewater system. TCDD was found in all exhaust scrubber water streams from the Incinerator. 2. Samples taken from dewatering wells located on a clgsed gn-site Landfill showed TCDD levels between l.8 and 3.4 ppb. These wells are an important historical source of TCDD. These wells have been deactivated. The land- fill is clay-capped and is surrounded by a clay wall extending to the natural clay bottom to prevent leakage and rainwater infiltration. 3. A shallow_?ump near former chlorophenol production sites produced samples containing about 1 TCDD. ?Ihjsasump fggmerly f19wed_igto the sewer 4. 69 62-111 94-111, Haste Treatment Plant 340 (50) ND (50) >87 95-111 Primary Inlet 92-111, Haste Treatment Plant 480 (50) ND (40) >92 93-111 Primary Inlet 90-111, Waste Water Treatment 4400 ND (90) >83 91?111 Plant Secondary Sludge 1Water samples filtered through 1.2 micron glass-fiber filter paper. 2For a detailed discussion of uncertainty limits, see Section 111, Chapter 1. -57.. CHAPTER 4. MIDLAND PLANT SITE NASTE WATER TREATMENT PLANT Figure 0 11.4 shows a simplified block diagram of the Midland Plant Site Waste Water Treatment Plant. This unit processes 13,000 gallons of water per minute which enters the facility as streams 1, 2 and 10. This flow corre- sponds to 19 million gallons per day or 6.8 billion gallons per year. Approxi- mately 0.6 gm/yr 2378-TCDD is discharged into the Tittabawassee River at a nominal level in the final outfall waters of 26 ppq. Since the waters enter? ing the Haste Hater Treatment Plant contain 2378-TCDD at 1400 ppq, this facility reduces 2378-TCDD levels by a factor of about 50. This represents a 98% efficiency in removing the low parts per quadrillion level. - Table 0 11.8 shows the 2378-TCDD content of Midland Plant Site waste- waters at various stages in the waste treatment process. Despite large variability in the incoming wastewater composition, a steady reduction in 2378-TCDD content is achieved as the uasteuaters pass through primary, . secondary and tertiary treatment over the course of the four to seven days required for processing.. . The data show that the Waste Water Treatment Plant removes 2378-TCDD by continuing the simple mechanism of removing the suspended silts and particu? lates that carry the 2378-TCDD in the waters passing through it. The efficiency of 2378-TCDD removal is simply proportional to the efficiency of suspended solids removal. As the total suspended solids content of the waste- water goes down, so does the 2378-TCDD content. In primary treatment, which involves the settling out of large-size particulate solids, the 2378-TCDD content of the treated water declines greatly as shown in Table 0 11.8. 2378-TCDD levels probably drop by a factor of 10, from the 1500 range to the 150 range. During secondary treatment, whichlinvolves bio-growth, 2378-TCDD levels drop by a factor of 2 or 3 into the 50-75 range. In the trickling filters FIGURE II .4 BLOCK DIAGRAM OF MIDLAND PLANT SITE SEWERS, WASTE INCINERATORS AND WASTE WATER TREATMENT PLANT, THE DOW CHEMICAL CO. 50 Sewer 318 Sewer 318 Treatment Plant . 76 Sewer A . 0 Stack 100 Sewer Vent 200 Sewer Waste . 300 Sewer . Incinerator Exhaust Scrubber 8 C) 400 Sewer Water 9 500 Sewer I Ash. Quench, 601 Sewer And ESP 610 Sewer General Primary Aeration Secondary Tertiary Sewer l\ . . . 3 Clarrfler @p Basm Clarifier Pond is Reuetment Ash Primary Secondary Outfall To Solids Solids Solids Tittabawassee River I -59- TABLE ANALYSES 0F HATER STREAMS IN THE DON CHEMICAL COMPANY MIDLAND PLANT SITE NASTE HATER TREATMENT PLANT1 STREAM 10; GENERAL SENER INLET TD PRIMARY CLARIFIER Access No. 2378-TCDD (PPQI 88-111 8402 93-111 430 (50) 95-111 . 340 (50) 106-111 1,600 115-111 430 (70) 175-111 3.800 177-111 2,400 AVERAGE: 1,400 STREAMS 1. 2 AND 9: 318 TREATMENT 2378-TCDD . on Particulates Access No. 2378-TCDD (ppq) (ppb) 128-11, 50 Sewer 2,500 22 132-11 . 129-11, 76 Sewer 260 (190) 9 (6.6) 133-II 100-111 InIet to Aeration Basin 1,000 (400) Not Measured AVERAGE: 1,300 1Samples were taken at various unreIated times. 2For a detaiIed discussion of uncertai?ty limits, see Section 111, Chapter 1. -50- TABLE 0 11.8 Continued STREAM 11: PRIMARY CLARIFIER OUTLET Access No. 2378-TCDD (ppq) 34 (6) STREAM 12: SECONDARY CLARIFIER TREATMENT (INLET T0 TERTIARY TREATMENT PONDS) 2378-TCDD in Total Sample 2378-TCDD on Total Suspended Suspended Solids Access No. Solids (mg/L) (p99) 96-111 81 Not Measured 127-111 74 23.2 3.2 128-111 81 23.2 3.5 137-111, 138-111 16 (5) 10.0 1.6 (0.5) 145-111, 146-111 80 33.0 2.4 154-111 240 114.0 2.1 171-111, 19 (3) 8.0 2.4 (0.4) 173-111, 174-111 14 (7) 7.6 1.8 (0.9) AVERAGE: 76 2.4 STREAM 13: AFTER TERTIARY HOLDING POND TREATMENT (DUTFALL T0 TITTABANASSEE RIVER) 2378-TCDD in 2378-TCDD on Total Sample Total Suspended Suspended Solids Access No. Solids (mg/L) (999)_ 374111 6.5 (2) 48.8 11 (2) Not Measured 104-111 40 Not Measured 129-111 15 7.6 2.0 140-111 17 (2) 18.0 0.9 (.1) 147-111, 76 38.5 2.0 10.8 1:2 AVERAGE: 26 1.5 -51.. and aerator basins, the biomass grows rapidly, feeding on the other organic components in the waste stream. The biomass need not metabolize or destroy the 2378-TCDD, it need only engulf or agglomerate the particles?carrying the 2378-TCDD. When the mature biomass is filtered off from the clarifiers, the particles carrying the 2378-TCDD it has entrapped and scavenged are also filtered off. The suspended particulates which escape engulfment in the biomass tend to be much smaller in siie, and therefore more difficult to trap or agglomerate. Tertiary treatment relies on a combination of longer times for sediment settling and bio?degradation along with exposure to sunlight for removal of conventional organic contaminants. Two to three fold reductions in 2378-TCDD content are also achieved by this treatment as shown in Table II.8, leaving only 25 or so of 2378-TCDD in the final discharge. The suspended particles which escape tertiary treatment are small, on the order of 10 microns in size, and are increasingly resistant to settling, conventional filtration, and scavenging by micro-organisms. Proof that the 2378- TCDD is carried on suspended particulates can be seen from the results of a pilot plant scale sand- filtration experiment carried out during the course of this investigation. This experiment was conducted using water currently being discharged from the tertiary pond on the Midland Plant Site. The experiment is described in Appendix VI. Table II.9 shows the results. The data show conclusively that 2378-TCDD in actual effluents from the Midland Plant Site is carried on the suspended particulates, and that . 2378-TCDD levels can be reduced by an additional factor of 2 to 5 by efficient sand filtration. -62.. TABLE REMOVAL OF 2378-TCDD FROM DON II.9 TERTIARY POND EFFLUENTS VIA SAND FILTRATION Total Suspended For a detaiied discussion of uncertainty 2378-TCDD Access No. Sampie Desorption - {Egg} (mg/L1 Set 1 139?111 Tertiary Pond Outlet 171 (2.0) 13.0 141-111 aAfter Sand Fiitration 5.0 (2.0) 9.2 Set 2 147-111 Tertiary Pond Outlet 76 38.5 149F111 .After Sand Fiitration 3.4 (1.0) 5.2 Set 3 155?111 Tertiary Pond Outlet 13 10.8 157-111 After Sand Fiitration 2.0 (1.4) 5.6 1 Iimits, see Section Chapter I. -53- SECTION 8. PART . CHAPTER 1. EVALUATION OF UNCERTAINTIES IN 2378-TCDD MASS FLUH CALCULATIONS 1 . . A. Introduction To understand the meaning of the 2378-TCDD mass flow data to be presented here in Part it is necessary to understand the uncertainties associated with the various flow measurements, environmental sampling techniques, and analytical procedures used in gathering the data. The objectives of the study, and the time resources involved in complex TCDD analyses, precluded the use of formal_statistjcal design_jn the sampling procedures. There was little replication or duplication of samples,.so that a complete variability analysis of the data cannot be done. From the beginning of the study the objective was to locate the dominant point sources of This objective requires determining the "order-of?magnitude? size of the possible sources, rather than exactly quantifying the levels of TCDD from these sources. It is possible however to provide:some estimates of uncertainty for portions of the data in this study, recognizing that the sampling locations, and therefore the results, were not randomly selected. These uncertainty estimates should therefore be used only for putting the data in context, rather than being used for further statistical calculations. The estimates are most relevant, and are best defined here in this mass flow section, where possible point sources are being compared and rank ordered. B. Sources of Sample Variation The appropriate uncertainty estimates for the various analyses reported in this study are highly dependent on the type of sample being discussed. Sources of sample variation include! 1. Analytical precision. 2. Homogeneity of the matrix sampled (soil, air, water, sludge, etc. 3. Sampling variation over time. -54- 4. Sampling variation over locations. 5. Variability in flow rates of moving streams. In some cases estimates of these sources of variation are possible. In other cases they are not. It is expected, a priori, that analytical variation and replicate sampling variation will be normally distributed, whereas data which reflects variation in flow rates or nonhomogeneous samples will be better represented by a log-normal distribution. Therefore in the following discussion geometric standard deviations will generally be reported, except where the variation reercts only analytical or replicate sampling variation.. For either distribu- tion, the arithmetic mean is the most appropriate measure for calculating 2378-TCDD yearly flows. The remainder of this chapter reports uncertainty estimates derived from subsets of data from this study for which it is felt that sufficient data exist to calculate a valid estimate. In general, homogeneous sampling matrices which are not affected by variation in flow rates demonstrate good repeatability, approaching the limit of analytical precision. Samples which are not homogeneous or reflect changing flows are more variable, as would be expected. C. Analytical Precision Analytical precision has been checked by analyzing each of two samples four times. Typical results are shown in Table The pooled standard deviation for 2378-TCDD analytical precision in this table is A further discussion of analytical reproducibility can be found in analytical Section of this report. D. Soil Samples; Matrix Homogeneity There were no replicate soil samples taken during the course of this study from which to obtain an exact estimate of soil sampling variability. ComEound 2378-TCDD Total Total Total H7CDD 0CDD Total Total HECDD Total H7CDD 0CDD -55- TABLE SUMMARY OF ANALYTICAL PRECISION TEST Sample A: I Selected Analyses Access Nos. 3-11, 24-11, 30-11, 34-11 Sample B: Selected Analyses Access Nos. 2-11, 8?11, 12-11, 16-11 Samgle Mean Concentration .0169 '.475 1.85 10 39.5 .518 5.5 31.0 298.5 2020 Standard Deviation .0055 .092 .274 2.49 10.5 .028 1.26 2.45 50 234 Relative Error -55- However an upper bound on variation in soil samples can be obtained from the data in Tables I.1 and 1.4. Variation among the soil samples classified as coming from non-LEL areas and background areas in Table 1.4, plus the City of Midland and Dow plant perimeter soil sample groups from Table I.1, can be used as an estimate of an upper limit. This uncertainty estimate contains variation due to different sampling methods and locations, as well as any gradient differences in the data. The pooled geometric standard deviation for the four data sets is 0.57, representing upper and lower uncertainty limits of E. Air Samples; Matrix Homogeneity and Variation Over Time Table 1.9 shows the results of_air?analy?g?_from four locations with samples taken approximately one month apart. For each location a geometric standard deviation was calculated, then pooled across the fbur locations. The pooled standard deviation is .37, representing upper and lower uncertainty limits of F. Sources of Sample Variation in Nastewater Streams Uncertainty estimates are clearly most relevant to this study when estimating 2378-TCDD movement in wastewater streams, and in attempting to locate point sources of 2378-TCDD into these streams. Annualized gram per year movements of 2378-TCDD in flowing streams are calculated by multiplying the average measured concentration by the independently measured average flow rate. Both these measurements show variation with time and location and are affected by sample homogeneity and sampling reproducibility. To quantify the errors limits on both component factors, replicate samples have sometimes been taken. If the two variables were independent, the relative (percent) variance in the product would be approximately additive. 1Uncertainty is reported as (minus and plus one geometric standard deviation, expressed as a percentage). -67.. Although the two variables are probably not independent, the degree of correlation is not quantifiable. Therefore additivity of the variances will simply be assumed as a first approximation. G. Variability in Hater Elow Rates Midland Plant Site water flow rates in the sewer system are strongly affected both by the weather and periodic changes in individual production facility operations. A brief study of the water flow rates through the 76 sewer and the combined 50 and 76 sewer flows into the 318 treatment plant illustrates the magnitude of this flow variability. The results are summarized in Table 111.2. The uncertainty in water flow rates based on these two data sets is for the combined streams and for stream 76 alone. For comparison, the water flow in the general sewer is about 18 million gallons per day i 1 million gallons per day, except during very heavy rainfall periods. This demonstrates that when many streams are mixed the variability in total flow is much lower. The flow from the Tertiary Treatment Pond to the Tittabawassee River is controlled at a fairly constant 19 million gallons per day. H. Sampling Reproducibility The uncertainty associated with taking duplicate environmental samples from a non-homogeneous stream at approximately the same time and location was tested by taking two samples of the sludge layer from the bottom of the general sewer at essentially the same time and place. The results are shown in Table 111.3. These very limited data suggest an uncertainty due to sampling reproducibility of about i 30%. There is no reason to assume that these data are not normally distributed, therefore a symmetric standard deviation is reported. I. Variability in Sample Composition'and Concentration as a Function of Time Throughout the mass flow study a number of samples have been taken from non-homogeneous, continuously-flowing streams. Recognizing that the concen- trations of trace components in these streams vary with time, due both to -53- TABLE 111.2 . VARIABILITY 0F FLOH RATES INTO THE 318 TREATMENT PLANT. MIDLAND PLANT SITE WASTE HATER TREATMENT PLANT, THE DON CHEMICAL COMPANY Arithmetric Flow Rate Mean Geometric gaIIons per gallons per Standard Date Measured minute minute Deviation 76 Sewer 25 April 84 90 26 ApriI 84 250 145 .21 27 Apri1 84 150 28 Apri1 84 90 Combined 50 and 76 Sewers 25 April 34 220 26 Apri1 84 330 240 - .12 27 April 84 230 28 Apri] 84 170 1Uncertainty is reported as (minds and p1us one geometric standard deviation, expressed as a percentage). -59- TABLE 111.3 DUPLICATE SAMPLING REPRODUCIBILITY AT A NON-HOMOGENEOUS SAMPLE POINT Access Numbers (ppb) Mean Standard Re1ative Compound Concentration Deviation Error 2378-TCDD 1.22 .26 22 Total 51 19 39 -70- variations in water flow rate, but also due to sporadic changes in manu- facturing and waste combustion process outputs, several streams in the sewer_system and Haste Water Treatment Plant have been sampled several times at nominally the same sample point. These studies are summarized in Table 111.4. The 2378-TCDD concentration in Stream 10, the primary inlet stream to the Waste Hater Treatment Plant, shows a geometric standard deviation of .40, representing upper and lower limits of +1501). These limits reflect sampling variation over time and over different flow rates entering the sewer. . The geometric standard deviation for the 2378-TCDD concentration in Stream 12, after secondary treatment in the Waste Water Treatment Plant, is .44 For samples taken from the Tertiary Treatment Pond outfall, the corresponding geometric standard deviation seen, the variation in 2378-TCDD concentrations values remains fairly constant across the five streams defined in Table 0 111.4, while the average 2378-TCDD concentration decreases by a factor of almost 500. The pooled geometric standard deviation for all five streams is .39 +150%) or approximately a factor of 2.5. 0. Summary The uncertainty in the numerical value assigned to any 2378-TCDD mass flow is affected by several factors, each of which contributes an uncertainty. The total uncertainty in any given mass flow calculation for 2378-TCDD will depend upon the variability of the component factors, which may be arithmetic means or single numerical values. The uncertainty analyses described here are based upon a small number of data points, but they indicate an overall variation on the order of +150%) for one standard deviation. Since f. 2378-TCDD annualized flows in wastewater streams are calculated by multiplying I the average flow rate times the average concentration, the uncertainty limits on the result can be approximated by the additive variance of the two component factors. SAMPLE REPEATABILITY AS A FUNCTION OF TIME. -71- TABLE 0 111.4 ANKEYTICALIRESULTS FUR 2373-icoo Stream Identification Access No. 50 Sewer 76 Sewer 101-111 Primary Inlet 106-111. 116-111, Tertiary Pond Inlet Tertiary Pond Outlet 104-111, Average ConcentratiOn_ 12,400 4.400 1.400 76 26 Pooled Geometric Standard Deviation Mean plus and minus one Geometric Standard Deviation Geometric Standard Deviation .27 .20 .41 .36 .39? +130%) -72.. For eXample, the estimated 0.6 gm of 2378-TCDD discharged per year from the tetiary pond has lower and upper uncertainty limits of +40%)1 based on a constant volumetric flow and seven representative 2378-TCDD concentration estimates in Table 11.8. Where fewer concentration measurements are avail- able, the uncertainty limits will increase to the maximum of +150%) for an individual measurement. Since the objective of this investigation was to identify critical point sources of 2378-TCDD in a complex system. quantitation beyond this level was not necessary, and was beyond the scope of this study. 1Calculated by representing plus and minus one standard error. -73- CHAPTER 2. THE FATE OF 2378-TCDD IN THE MIDLAND PLANT SITE HASTE INCINCERATDR As presently operated the Midland Plant Site Haste Incinerator consists of the rotary kiln, afterburner and complex gas scrubbing system shown schematically in Figure The unit burns paper and wood trash, solid chemical waste, chemically contaminated waste equipment and a variety of liquid wastes. The unit uses 2040 gallons per minute of water for quenching and cooling the exhaust gases, and to aid in removing particulates from the exhaust. The water sources are 215 of river water and 1825 of water recycled from the Waste Treatment Plant. As shown in Figure 111.1, about half of the water effluent goes to the general portion of the Waste Water Treatment Plant and half goes to the 318 treatment section of the plant to maintain optimal operating temperatures there. Solid trash is fed to the rotary kiln which operates from about with a 30-45 minute bulk solid residence time. Liguid wastes and tars are sprayed into either the _the afterburner. The afterburner_operate?_from about 1000-1100fE?with a residence time of a few seconds. In general, higher BTU go to the afterburner and higher ash-containing feeds go to the kiln; however, this is adjusted depending on the nature of total feeds at any particular time. An attempt has been made to study the fermation and fate of 2378-TCDD in the Haste Incinerator. Three sample sets have been collected, the results of which are summarized in Table Each set was taken over a two hour period during which time the feeds were documented. The samples were con- tinuously taken from the exhaust stack and the four exhaust scrubbing opera- tions. From the analyses, the amount of 2378-TCDD either captured in the exhaust purification system or vented was calculated on an annualized basis. Since the feeds to the burner vary widely during a year, the annualized results projected here must be considered approximate. Figure shows a block diagram of the Waste Incinerator, and the annualized 2378-TCDD inflows and outflows via the feedstock, air, ash and water streams for Sample Set No. 1. Th% total outflow of 2378-TCDD is calcu- lated to be about 13 gms/yr. About one half of the 2378-TCDD is in the FIGURE .1 A SCHEMATIC DRAWING OF THE DIVISION WASTE INCINERATOR FROM Stack Vent Water Stack Li uid Tars Liquid Tars 1 Fuel Quench I: 08- i mister Afterburner Rubbish IL ll ll Water To WWTP Solid Ash To Landfill -75.. TABLE RESULTS FROM THREE SAMPLE SETS OF THE DOH CHEMICAL COMPANY MIDLAND PLANT SITE HASTE INCINERATOR Concentration-of Sample Sample 2378-TCDD Access No. Type Location Flow Rate in Haste Stream Set No. 1 August 1983 51-111 Air Stack Vent 925 Ms/min 71o pg/IV3 Hater ,Ash Trough 50 520 Hater Quench Tower 825 2,100 Hater Venturi Demister 1,000 660 Hater Het ESP 165 4,600 Solid Dry Incin. Ash 30,000 lb/day 1.3 Kiln Feeds: Rubbish/Fiberpaks/Acti?ated Sludge: approx. 200 lb/min Non-Chlorinated Haste: approx. 50 lb/min Afterburner Feeds: 70% Chlorophenolic Haste: approx. 30 lb/min 30% Phenolic Haste Set No. 2, December 1983 . 159-111 Air Stack Vent 925 M3/min 640 pg/M3 165-111 Hater Quench Venturi 1,825 350 Demister 167?111 Hater Het ESP 165 810 Kiln Feed: Hood: 17 yd3/hr- Aromatic Haste: 15.5 lb/min Afterburner Feeds: Chlorophenolic Haste: 39 lb/min Set No. 6, December 1983 130-111 Air Stack Vent 995 MB/min 58 pg/M3 133-111 Hater Quench Venturi 1,825 280 Demister 134-111 Hater Het ESP -165 280 Kiln and Afterburner Feeds: Natural gas only, to maintain temperature. 2378-TCDD Discharged gm/gr 0.028 1.0 0.092 . 1 i For a detailed discussion of uncertainty limits, see Section 111, Chapter 1. FIGURE .2 BLOCK DIAGRAM OF ANNUALIZED MASS FLOWS INTO AND OUT OF THE DOW CHEMICAL CO. MIDLAND PLANT SITE WASTE INCINERATOR BASED ON SAMPLE SET 1. Stack Vent 0.3 gm Exhaust Scrubber Water 6.2 Chlorophenolic Waste Tars 13 Elms- Scrubber Feed Water aste ncmerator 0.06 Ash Quench Water 0.05 Ashes 6.5 gm -77- incinerator ash, this 2378-TCDD is formed during the rotary kiln com- bustion of orga?iEJmatgrials in the presence of available chlorine, in ?with what has been learned about the creation of 2378- TCDD by combustion. The other half of the 2378- TCDD is 95% captured by the exhaust scrubber equipment, allowing only 0. 3 gms/yr to escape tolthe atmosphere. I Although formation of 2378- TCDD in the rotary kiln section is inevitable in the presence of chlorinated feedstocks, the afterburner section is a net destroyer of 2378-TCDD. One of the important feeds to the Waste Incinerator afterburner is the dichlorophenol distillation wastes. These wastes contain 20 of 2378-TCDD. Recent plant records show that about 1.4 million per year of dichlorophenol wastes arerproduced. Under these operating condi- dichlorophenol wastes will carry about 13 gms/yr of 2378-TCDD into the Waste Incinerator. Since these dichlorophenol distillation wastes are fed to the afterburner. any chlorinated dibenzo- -dioxins remaining as a result of the inc_ineration process will either be 232339, or carried in the particulates washed down from the exhaust gas scrubbing operation. They will not be in the incinerator ash. From Figure the Z378-TCDD vented or washed out of the vented gases totals approximately 5.6 gms/yr, so that of the approximate 13 gms/yr entering the afterburner, approximately 6 gms/yr net is destroyed. This control is achieved while 15 million pounds per year of wastes and solvents are destroyed in the afterbunner, and 135 million pounds per year of trash, rubbish and hazardous liquid and solid wastes are rotary kiln. I Futhermore, the estimated 0.3 gm/yr 2378-TCDD which escapes exhaust purification in the'waste Incinerator and is released to the atmosphere on suspended particulates is likely an upper limit. Sample Sets No. 1 and 2 shown in Table 111.5 were both made with heavy chlorophenolic loadings to the afterburner. These wastes were being fed at 10-20 times the average rate required to consume 1.4 million pounds per year. Sample Set No. 6 shows that when all waste feeds to the Waste Incinerator are stopped the 2378-TCDD measured drops sharply. Therefore it is likely that during periods when; chlorophenolics are not being incinerated, the 2378-TCDD measured will drop below 0.3 gm/yr. -73- CHAPTER 3. SUMMARY OF 2378-TCDD MOVEMENT ON THE MIDLAND PLANT SITE .. Figure shows a block diagram of the Midland Plant Site Haste Treatment complex and the important water, exhaust gas, and sludge streams moving through it. The estimated annual 2378-TCDD flows in each of these streams are summarized in Table The 2378-TCDD content of some of these individual streams varies greatly with time. Some have been shown to fluctuate over a tenfold range. The data for the study were collected over the course of at least a year, and often represent only-a snapshot view. Therefore a detailed mathematical ?balancing? of the flows is not justified. However, a summary overview is instructive. From an overall environmental perspective, streams 13 and 14 represent the waterborne and airborne outflows from the Midland Plant Site of most - critical interest. The Midland Plant Site currently discharges approximately 0.6 gm/yr on suspended particulates into the Tittabawassee River r_from the Tertiary Pond: The analyses of this discharge stream are quite constant with time, and are probably meaningful with an uncertainty of less than plus or minus 40% at any given moment.1 The Waste Incinerator currently emits approximately 0.3 gm/yr 2378-TCDD from the stack into the atmosphere and represents the only major airborne on the Midland Plant Site. The stream undoubtedly? fluctuates from day to day, but would not be expected to exceed 0.3 gm/yr for the reasons stated earlier. The Dow waste management practice of incinerating solid and liquid wastes whenever practidal to minimize environmental impact is based on efficient water scrubbing of the Waste Incinerator vent gases to remove toxic compo- nents. This works for As shown in Figure 111.2, of the 13 1See Section Chapter 1. FIGURE BLOCK DIAGRAM 0F MIDLAND PLANT SITE SEWERS, WASTE INCINERATORS AND WASTE WATER TREATMENT PLANT, THE DOW CHEMICAL CO. I 318 Sewer 313 I Treatment Plant 76 Sewer Stack 100 Sewer Vent 200 Sewer Waste Tars Waste . SOOSewer . . Inmnerator Exhaust (9) a. - - - Water 40 Sewer 6 Ash Quench ?Mi, And Esp 610 Sewer I Gmem' Primary Aeration Secondary Tertiary 15 Revetment Ash Primary Secondary Outfall To Solids Solids Solids Tittabamee River Clarifier @p Basin Clarifier @p Pond -30- TABLE ESTIMATED ANNUAL IN THE.MIDLAND PLANT SITE SENERs, NASTE INCINERATOR AND HASTE NATER TREATMENT PLANT, THE DON CHEMICAL COMPANY1 2378- TCDD 1' Flow Rate Average 2378- Stream Table or (million Analysis TCDD Number Access No. Stream Identification (ppq) (gm/yr) re HATER 1 Table 11.2 50 Sewer 0.3 5.1 2 76 Sewer 0.3 1.8 3 Table 11.2 Combined Sewers Feeding 17.1 1.2 the General Sewer 4 Table 11.1 Revetment System 1.1 0.08 ,1 5 43-111, 44-111, Incinerator Scrubber Water 1.5 5.0 1 46-111 6 53?111 Scrubber Hater Feed 1.4 11 (7) 0.02 7 45-111 Incinerator Scrubber Water 1.4 660 1.2 7 8 48-111 Clarified Primary Water 3.0 30 (20) 0.14; 9 100-111 318 Treatment Outlet 5.0 1000 7.0 10 Table 0 11.8 Feed to Primary Clarifier 18.0 1400 34.0 11 105-111 Primary Clarifier Outlet 13.5 34 (6.0) 0.6 to Aeration 12 . Table 0 11.8 Secondary Clarifier Outlet 19.0 75 1.9 13 Table 11.8 T-Pond Effluent to the 19.0 25 0.6 River 1 1 For a detailed discussion of uncertainty limits, see Section 111, Chapter 1. TABLE 0 111.6 Continued 2378- T000 . Average 2378- Stream Table or I Flow Rate Analysis TCDD Number Access No. Stream Identification (MM god) (qu1 jgm/xr} /min) (Picograms/M_l 14 51-111, Incinerator Stack Vent 925 675 0.33 SOLIDS {lb/dag} iEEbi 15 211-111 Incinerator Ash 30,000 1.3 6.5 16 134-11 Filtered Primary 50,000 11 1 2 Sludge plus Lime 56 214-111 Primary Solids 37,500 3.6 17 25-11 Filtered Secondary 19,000 1.5 ?4.7 Sludge plus Lime 18 170-111 Chlorophenolic 3,300 20 13 Haste Tars 1 2Average of two values. For a detailed discussion of uncertainty limits, see Section 111, Chapter 1. -32- of leaving the Haste Incinerator, 98% is removed as incinerator ash or waterborne particulates to the Waste Water Treatment Plant. leaving only 0.3 gm/yr vented from the stack. Having scrubbed the particles carrying 2378-TCDD from the Haste Incin- 'erator exhaust vent into a wastewater stream, the next step in minimizing 23% environmental impact is removal of the particulates from the wastewaters. As summarized in Table 0 111.6, the Waste Hater Treatment Plant is very effective in accomplishing this. The water inflow to the primary clarifier contains 2378-TCDD at 1400 and carries something on the order of 30 gm 2378-TCDD per year from all waterborne sources. The final outfall waters from the Tertiary Treatment Pond into the Tittabawassee River contain 0.6 gm 2378-TCDD per year at a concentration of 25 ppq. This 50?fold reduction factor in 2378-TCDD content, giving a removal efficiency of 98%, relies on the fact that the 2378-TCDD is.carried by the suspended silts and particulates in the water streams, and the fact that the Waste Water Treatment Plant is very effective in removing these waterborne particulates through sedimentation and biomass entrapment. ng637 The final phase of waste treatment on the Midland Plant Site is containment. Fifty tons per day of ash solids and primary and secondary treatment 15, 10. 17) are contained in hazardous waste landfill on Dow property. This material does not enter the general environment. The 2378-TCDD is immobilized on the landfilled particulates, and dispersal requires transport of the particulates on which it is carried. Turning attention to critical point sources of 2378-TCDD to the Midland Plant Site wastewater system, four important sources have been identified. 'Three of them are historical deposits that intermittently release 2378-TCDD into the sewer system. The fourth and apparently largest single source is the Waste Incinerator. It is the only major currently generating source of 2378-TCDD to the wastewater system identified on the Midland Plant Site. -83- In Part IV. "fingerprint" pattern recognition techniques will be de? scribed which help further define the relative importance of these four sources. In accord with the overall objectives of this investigation, cor- rective actions to further reduce or eliminate the flow of 2378-TCDD from all four of these point sources into the Midland environment are being undertaken. -84- CHAPTER 4. HISTORIC INCINERATION OPERATIONS 0N THE MIDLAND PLANT SITE A. Incineration of Liquid Haste Tars Historical records show that as early as 1930 the company was disposing of organic liquid tars by incineration on the Midland Plant Site. Two basic types of burners have been used; a liquid tar burner of several different configurations, and a rotary kiln solid trash incinerator also capable of handling liquid wastes. In the mid 19305 two tar burners were installed just northwest of the present Midland Plant Site Haste Incinerator. These were vertical brick lined towers in which liquid tars were burned and with the combustion gases vented directly to the atmosphere. Fuel oil was used to start up the unit and to maintain the flame when necessary. In 1951 a new vertical tar burner was installed, replacing the two i earlier units. The unit was a brick-lined burner 15 feet in diameter and 50 feet tall containing four tangential feed nozzles using fuel oil as supplemen- tary erl. The unit discharged combustion gases directly to the atmosphere. It was last used in 1974 and demolished in the late 19705. In about 1957 the 707 Building tar burner was constructed just east of the present Midland Plant Haste Incinerator. This unit provided scrubbing capabilities to reduce hydrogen chloride emmisions when burning chlorinated tars. As shown in Figure 111.4 the vent gases could be diverted directly to a 125 foot stack or to a water quench chamber prior to discharge to the atmosphere. This unit was shut down in 1975. I In 1968 the 830 Building liquid tar burner was commissioned to provide higher temperature control in the burner and improved air pollution control. 1 The unit is shown schematically in Figure 111.5. The burner operated at a temperature of a tar feed rate of about 10 gallons per minute and I discharged about 30,000 cubic feet per minute (cfm) of combustion gases from FIGURE .4 SCHEMATIC DIAGRAM OF THE 707 BUILDING TAR BURNER Stack Water Vent Water Spray Spray Butterfly Burner Butter?y Valve Valve Quench Chamber Water To WWTP Fuel Oil Or Liquid Tars FIGURE .5 SCHEMATIC DIAGRAM OF THE 830 BUILDING LIQUID TAR BURNER Water Stack De- miSter Quench Waste Burner Tar Feed Water To WWTP -87- the stack. Since 1975, chlorinated waste tars have primarily been disposed of in the afterburner of the rotary kiln Haste Incinerator. In 1981 the 830 Building tar burner was put into a standby mode to be used only for tar inventory control. Although still technically in service, the unit has not operated since December 1982. In summary, low temperature burning of waste tars on the Midland Plant Site dates from at least the 19305. High temperature combustion was started in 1968 but not fully implemented until 1978 at which time advances in trace analysis showed the need for higher temperatures to minimize the formation of dibenzo-p-dioxins. B. Incineration of Combustible Solid Mastes Prior to 1948 solid burnable rubbish was either landfilled on the Midland Plant Site, or stockpiled for periodic burning in the open air. In 1948 the first Midland Plant Site trash incinerator was installed. This unit was a rotary kiln general purpose unit that handled rubbish, solids, packs, and liquid tars. It is shown schematically in Figure 111.6. Solids were manually shoveled into the feed chute and various liquids were sprayed into the front of the kiln. Combustion gases were vented directly out the stack with no quenching or treatment. The ash was quenched, dewatered and landfilled in the immediate area. I 1 In 1958 the original rotary kiln was replaced with the dual rotary kiln unit shown schematically in Figure 0 From 1958 to 1975 only rotary Kiln No. 1 was used. It depended on natural draft, discharging about 100,000 out the stack. The unit provided more capacity and better burner control. By installing a water-spray quenching system, a positive step forward in pollution control was achieved. In 19?0 secondary combustion was installed in? the transition zone between the kiln and the quench chamber. Three nozzles which fed natural gas into the secondany combustion zone were installed to FIGURE 0 .6 SCHEMATIC DIAGRAM OF THE ORIGINAL ROTARY KILN INCINERATOR Stack Liquid Tar Quench - Water To WWTP Incinerator Ash To Landfill - Trash - Handling FIGURE .7 A TOP SCHEMATIC VIEW OF THE DUAL ROTARY KILN WASTE INCINERATOR Natural Gas After 1970 Water Spray 2 _l 6 Rolary Rotary 5 i Kiln #2 I Transition Quench Zone Stack -90- help reduce stack particulate emissions. From 1970 to 1975 all liquid tars burned in this unit were fed to the front of the kiln, and only natural gas was fed to the secondary burner. In 1975, the No. 1 Rotary Kiln was shut down. At that time No. 2 Rotary Kiln was operated for the first time with the improved afterburner and sophisticated air pollution control system as shown in Figure In 1978 the natural gas supply was increased to the afterburner and the temperature control point was set at about This increase was implemented as a result of advanced studies indicating that this temperature was needed to minimize formation and assure efficient destruction of chlorinated dibenzo-p-dioxins. By 1981 the unit was operating with an electrostatic precipitator installed between the fan and the stack. -A further reduction in particulate emissions was obtained. Since then. the unit has operated essentially the same except for installation of process computer control to help maintain consistent operation. In summary, low temperature combustion of solid wastes on the Midland Plant Site was carried out with no emission controls prior to 1958 and from 1958 to 1978 with pragressively improving emission controls. Modern inciner- ation. with high-temperature secondary combustion to minimize particulate emissions and chlorinated dibenzo-p-dioxin formation began in 1978. Net electrostatic precipitation of small stack particulates was in operation by 1981. C. Historical Discharge of 2378-TCDD to the Atmosphere from the Midland Plant Site As shown in Table and Table current operation of the Midland Plant Site Haste Incinerator emits an estimated 0.3 gm/yr of 2378-TCDD to the atmosphere. This quantity is not sufficient to account for the surface . soil levels of 2378-TCDD now accumulated on the Midland Plant Site and in the City of Midland. The current waste Incinerator stack emission rate of 0.3 gm FIGURE .8 A SCHEMATIC DRAWING OF THE DIVISION WASTE INCINERATOH FROM 1975?1980 Water Liquid Tars L'qu'd Ta" Fuel Afterburner Solid Ash To Landfill De? mister Quench i Water To WWTP Stack Vent Stack -92- 2378-TCDD per year, even if continued over 20 years and concentrated onto just one square mile of soil, would raise the 2378-TCDD level in the top l/Z inch of soil to just 0.1 ppb. or the top 3 inches of soil to only 0.015 ppb. As shown in the preceding sections, incineration has been practiced on the Midland Plant Site since about 1930 with annual stack emissions substantially greater in the past than at present. This fact, coupled with the fact that incineration was carried out at low temperatures prior to 1978 in the rotary kiln, and prior to about 1968 to 1974 in the various burners, leads to the conclusion that formation and emission of chlorinated dibenzo-p-dioxins has historically been substantially higher than at present. Additionally, combustion of the wastes from the now-inactive chlorophenolic production processes listed in Table 1.3 probably caused higher formation rates of chlorinated dibenzo-p-dioxins. From these data and assumptions it is realistic to conclude that historic . manufacturing and incineration operations on the Midland Plant Site have formed and emitted quantities of 2378-TCDD sufficient to account for the surface soil levels now observed. -93- SECTION C. PART IV. CHAPTER 1. USE OF PATTERN-RECOGNITION TECHNIQUES IN THE IDENTIFICATIONIOF 2378-TCDD POINT SOURCES A. Introduction . Parts I, II and of this study show that operations on the Midland Plant Site are associated with the trace levels of 2378-TCDD found in City of Midland soils. Dow plant soils, and in'the Dow Tertiary Treatment Pond water discharged to the Tittabawassee River. The critical point source of the 2378-TCDD to each of these environments has been identified by showing that in .each case: 1. The proposed source has or historically had an estimated annual output great enough and concentrated enough to account for the level accumulated over time in the soil, or to sustain the concentration found in a flowing water or air stream. 2. No alternative potential point source of a larger or similar size can be found. I 3. A workable transport mechanism can be proposed which accounts for the movement of the 2378-TCDD from the proposed point source to where it is found. Since this source identification method depends on uncertain flow, concen- tration and sample reproducibility measurements, an independent verification method was sought. Part IV describes the development and application of such a direct method for identifying point sources of 2378-TCDD. B. Fingerprint Concept The point source identification method relies on the concept that each point source and environmental sample has a unique ?fingerprint?. -94- or pattern of isomer distributions, and that samples can be linked back to their probable source by the simple fact that they show the same fingerprint. As shown by the soil sample analysis in Table IV.1, 2378-TCDD is not usually found by itself in the environment, but accompanied by all the other tetrachlorodibenzo-p-dioxins. the pentachloro-isomer group, the hexachloro- and heptachloro- groups, and by octachlorodibenzo-p-dioxin. The relative amounts of these dibenzo-p-dioxin congener groups in this sample constitutes a "handprint" of the sample. Within the tetrachlorodibenzo-p-dioxin isomer group there are 22 possible isomers. As Table IV.1 shows. the Dow analyti- cal procedure described in this Section of this report permits separation and determination of 20 isomers. The relative amounts of these 20 isomers within the TCDD isomer group comprise the unique detailed "fingerprint" of each sample used for source identification. The sample "fingerprints? described here may be strong or faint, blurred or only partially readable depending on the particular sample, but the pattern always provides useful information. During the data collection phase of this project it became evident that in related samples, such as City of Midland soil samples, 2378-TCDD absolute concentration levels might differ greatly, but the relative levels of the TCDD isomers were remarkably constant. Table IV.2 shows the TCDD isomer distributions in two such City of Midland soil samples. Even though the level of 2378-TCDD differs by two to threefold between the two samples, the relative concentrations of the TCDD isomers are strikingly similar. The fingerprinting method relies upon the assumption that environmental samples with similar patterns of TCDD isomers have a common source, or have been created or equilibrated by a similar chemistry. Once created, the pattern of TCDD isomers in an environmental sample should remain similar to that in its source until further equilibration or decomposition processes in the environment cause a shift in the pattern. -95- TABLE IV.I CHLORINATED DISTRIBUTION IN AN URBAN INDUSTRIAL SOIL SAMPLE '50. SAMPLE NC: 3'11} EUS BARBERIUN UHIU: IBTH ST. SEIL: 103183?7 20.000 . 2239'149?1 SPECIES MONITURED RELATIVE (XI 005 I 1.3 1269.TC00 I 0.6I 1267.TC00 I 0.4I 1289.TC00 0.5 I 0.4) 1.3 1369.TC00 1.4 I 0.5) 3.5 3.0 I -0.5I 7.5 1278.TC00 1.2 I 0.5I 3.0 1268.TC00 1.3 I 0.5I 3.3 2.7 I 0.5I 6.8 1279.TC00 1.3 I 0.5I 3.3 1246.7C00 0.6 I 0.4) 1.5 1473.TCDD 3.5 I 0.5I 1.3 1236.TCDD I 0.4) 1239.TC00 0.5 I 0.9I 1.3 1249.TC00 0.5 I 0.3I 1.3 1368.TC00 9.5 I I 23.9 1379.TC00 5.3 I I 13.3 1378.TC00 3.9 I 0.6I 9.8 1234.TC00 1.5 I 0.6) 3.3 34.2 85.9 2379.TC00 5.6 I I 14.1 13C.2378.TCDD RECOVERY 82% 2373.ICDF 12.0 I I 13C.2378.TCDF RECOVERY 71% 105.0 I I 26.0 123463.HC00 24.0 I I 5.9 164.0 I I 40.6 123469.HC00 I 1.1I 123473.HC00 63.0 I I 15.6 123678.HC00 I I ?9.0 I I 11.9 404.0 100.0 13C.123473.HC00 RECUVEPY 99% 1234679.HC700 950.0 I I 44.2 1234678.1C700 1200.0 I I 55.6 TCTAL 2150.0 00.0 H7CCD CURRECIIUN FACTDE ABSOLUTE 3CDC 6500.0 I I 13C.CC00 RECOVERY ABSOLUTE DEIECTED AT LDC I.E. 2.5x PEAK TC VALLEY NOISE SIGNIFIES )3 25X PEAK T0 VALLEY N015 -95- TABLE IV.2 COMPARISON OF THE TCDD ISOMER CONCENTRATIONS AND RELATIVE TCDD CONCENTRATIONS IN THO CITY OF MIDLAND SOIL SAMPLES SampIe 23?111 TCDD Isomers Concentration Monitored Egt Re1ative 1469-TCDD ND ND 1269-TCDD 4.0 0.2 12677TCDD 2.5 0.1 1289-TCDD 2.0 0.1 1369-TCDD 23 1.2 1247 1248-TCDD 64 3.3 1278-TCDD 18 0.9 1268-TCDD 17 0.9 1237 95 4.9 1279-TCDD 20 1.0 1246-TCDD 4.2 I 0.2 1478-TCDD 10 0.5 5.1 0.3 1239-TCDD 3.3 0.2 1249-TCDD 7.3 0.4 1368-TCDD 1020 52.3 1379-TCDD 360 18.5 1378-TCDD 120 6.2 1234-TCDD 5.6 0.3 2378-TCDD 170 TOTAL 100% Sample Concentration get Relative 11.9 1.0 0.1 4.4 0.5 3.9 0.5 2.3 0.3 3.0 0.4 480 57100% -97- Conversely, samples taken from different geographical areas, or samples bearing traces of 2378-TCDD from different apparent sources seem to show different TCDD isomer distribution patterns. For example, Table IV.3 shows two samples with similar overall TCDD concentrations (600 and 830 ppt) but with very different TCDD isomer patterns. Note especially the great disparity in relative percent levels of 1379-TCDD and 2378-TCDD. Since one sample is from the Dow Powerhouse baghouse ash, and the other is of City of Midland soil collected close downwind from the.Powerhouse in an area_long exposed to fly? ash, it seems reasonable to independehtly conclude, in corroboration of the conclusion drawn in Part I, that the Dow Powerhouse is not a critical point source of 2378-TCDD into the soils of the City of Midland. However, when the TCDD "fingerprint" data from a group of soil samples collected from the City of Midland are compared with the TCDD "fingerprint" of a group of soil samples collected from the Dow Plant Site, a striking similarity is observed, as shown in the bar chart Figure IV.1. In this chart, the set of bars to the left in each pair represents the mean relative concentration of each TCDD isomer found in one group of samples. The bar on the right represents the mean relative concentration of that same TCDD isomer found in the other group of samples. The length of each bar repre- sents plus and minus one standard deviation from the mean in the measured values for each TCDD isomer in its own data set. 'The degree of overlap be- 'tween the two sets of bars measures the degree of similarity between the two groups of samples. Since all 20 barsgoverlap within the confines of the error bar for these two groups of environmeptal samples, the groups are clearly very The conclusion which pattern comparison Figure IV.1 supports is the same as that drawn earlier in Part I,'namely that Midland Plant Site soils and City of Midland soils have a common source of 2378-TCDD and other TCDD isomers.~ I i The bar chart pattern comparisoniillustrated in Figure IV.1 can be mathematically formalized to quantify the degree of similarity between the TCDD distribution patterns in various environmental samples and potential sources. -93- TABLE IV.3 COMPARISON OF THE TCDD ISOMER CONCENTRATIONS AND RELATIVE CONCENTRATIONS OF IN SAMPLES FROM TNO DIFFERENT ENVIRONMENTS Powerhouse Ash Sample Midland Soil Samples Sample TCDD Isomers Concentration Concentration Monitored _ppt_ Relative ._223_ 'Relative 1469-TC00 ND ND 1.1 0.1 1269-TCDD 0.8 0.1 3.0 0.4 1267-TCDD 0.6 0.1 1.3 0.2 1289-TCDD 0.4 0.1 2.7 0.3 13696TC00 4.9 0.8 8.3 1.0 1247 23 3.3 42 3.0 1278-TCDD 7.2 1.2 11 1.3 1268-TC00 9.1 1.5 12 1.4 1237 1238-TC00 1.9 12464000} 2.31 0.4 1 1.0 0.1 1478-TCDD 4.4 0.5 1236-TCDD 2.3 0.4 3.9 0.5 1239-TC00 1.6 0.3 2.3 0.3 1249-TCDD 2.1 0.4 3.0 0.4 1368-TC00 280 47 480 58 1379-TCDD 200 33 73 8.8 1378-TCDD 31 5.2 48 5.8 1234-TCDD 4.0 0.7 4.3 0.5 2378-TCDD 2.4 ii 53 49 TOTAL 100% 100% 1 value reported is the total of both isomers. -1h this sample separation of 1246-TCDD and was not achieved and the FIGURE (3 IV 1 A COMPARISON BETWEEN DOW AND MIDLAND AREA SOIL SAMPLES a 1 237.8 4 -MIDLAND AREA SOIL -DOW PLANT SITE SOIL 10.0?- 10- 100.0 -100- C. Pattern Recognition The field of mathematical pattern recognition includes a number of multi- variate methods for comparing groups of samples by making measurements along a number of different dimensions for each group, and then comparing these meas- urements. In this case the measures are the relative TCDD isomer concentra- tions in two groups of samples, in which the samples are grouped according to their common location or by the environment from which they were collected. The method used to quantify the similarity between the two groups of samples is a generalized n-dimensional distance measure. This measure was first intro- duced by Mahalanobis and is often referred to as Mahalanobis distance; it is a measure of the n-dimensional distance between the average relative TCDD isomer concentrations in the two groups. Details of the calculational procedure, including the method used for estimating TCDD isomer values below the experi? mental detection limit, are described in Chapter 3. Table IV.4 shows the numerical.results of Mahalanobis comparison of the TCDD patterns in three groups of environmental samples; City of Midland soils, Midland Plant Site soils, and airborne dust particulates collected at the Dow fenceline. The numbers in the table represent the multidimensional distance between the average TCDD isomer concentrations for each pair of sample groups. Small numbers in Table IV.4 indicate similar TCDD isomer patterns. Large numbers imply that the TCDD isomer patterns are different for the two groups being compared. These comparisons can be used for descriptive purposes to rank order the similarity between one group of samples and other groups. Formal statistical hypothesis tests have not been adhered to because of the non-randomness of the data collection, Calculation of distances between samples that are known to be similar and consideration of statistical values for descriptive purposes indicate that differences on the order of 3.0 repre- sent only sampling variation. Distances greater than 9.0 are generally indicative of real differences whereas in-between values may or may not be meaningful depending upon, among other things, the number of samples in the groups being compared. -101- TABLE IV.4 MAHALANOBIS DISTANCES BETHEEN SAMPLE GROUPS1 Group 1 Group 2 Group 3 Group 1. City of Midland Soils - 2.8 4.6 Group 2. Midland Plant Site Soils 2.8 - 5.1 Group 3. Fenceline Airborne Dust Particulates 4.6 5.1 - 1Even though 2378-TCDD is the TCDD isomer of primary concern in the overall point-source investigation, special weighting of this isomer was specifically avoided in the Mahalanobis calculations. Any arbitrary weighting would have implied a preconceived notion as to its relative importance during pattern comparisons. .. LJ -102- Table IV.4 shows that City of Midland soils and Midland Plant Site soils are the most similar pair of the three groups (Mahalanobis distance of 2.8), and are not significantly different. This_conclusion is analogous to the one drawn by noting the high degree of overlap in all relative TCDD concentration bars shown in Figure HV.1. Figure IV.2 shows the bar graphicomparison between relative TCDD concen- trations in City of Midland soils and bow fenceline airborne dust particulates. The patterns are not totally superimposable but are strikingly similar in accord with the calculated Mahalanobisidistance of 4.6 in Table IV.4. This result along with the equivalent 5.1 Mahalanobis distance calculated for the similarity between Dow fenceline airborne dust particulates and Midland Plant Site surface soils, indepedently supports the proposal made in Part I that windborne transfer of surface dusts is the probable mechanism by which 2378-TCDD has been historically transferred off the Midland Plant Site into the City of Midland. - Since the Mahalanobis comparisons indicate no significant difference between the TCDD patterns in City of Midland soils, Midland Plant Site soils and Dow fenceline airborne dust particulates, therefore these three sample groups can be combined to generate a single midland soils and dust parti- culates group with a large data base. Table IV.5 shows Mahalanobis comparisons between this combined environ- mental soils group (group 1) and the five possible TCDD point sources to Midland area soils identified in Part I of the Dow study. Bar chart com- parisons for group 1 with groups Table IV.5 are included as Figure The value of pattern recognition techniques in helping to identify the dominant point sources of 2378-TCDD into the surface soils of the Midland Plant Site and City of Midland is illustrated by the following conclusions which can be drawn from these data when used in conjunction with the facts from Part I of this study: -103- TABLE IV.5 MAHALANOBIS DISTANCES BETWEEN SAMPLE GROUPS REPRESENTING MIDLAND PLANT SITE SOURCES AND MIDLAND AREA SOILS Group Group Group Group Group Group Midland Soils, and Dust Particulates Group 2. Haste Incinerator Kiln 5.8 - 5.7 11 15 22 Ash Group 3. Haste Incinerator Stack 6Particulates a 'n Group 4. 11th and StsGroup 5. Powerhouse Baghouse Ash 15 15 14 16 32 - Group 6. 47? Building LEL Note: For each group, samples were included in the calculations. Group 1, 1 Group 2, Group 3, Group 4. Group 5, Group 6. 3:33:33 l?lNI??wwm The following isomers were excluded from the calculations due to the constraints explained in Chapter 3: 1469, 1269. 1267, 1289. 1478, and 1236 TCDD. IV A COMPARISON BETWEEN MIDLAND SOIL AND FENCELINE AIRBORNE PAHTICULATES AREA SOIL AIRBORNE PARTICULATES RELATIVE AMOUNT H10 - BI 5 EIEI TCDD ISOMER FIGURE IV .3 A COMPARISON BETWEEN INCINERATOH ASH AND MIDLAND AREA SOIL 1001) MIDLAND AREA SOIL ASH RELATIVE AMOUNT "TCDDISOMER FIGURE IV 4 A COMPARISON OF MIDLAND AREA SOIL SAMPLES WITH PARTICULATES FROM THE STACK OF THE WASTE INCINERATOR -MIDLAND AREA SOIL IOOI) WASTE INCIN. STACK PARTICULATES RELATIVE AMOUNT H10 TCDD ISOMER FIGURE Iv 5 A COMPARISON BETWEEN MIDLAND SOIL AND SOIL FROM THE 11m 8! AREA LEL 1001) EMIDLAND AREA SOIL 011th a. AREA LEL SOIL .TCDD ISOMER I I I) A COMPARISON OF POWERHOUSE PARTICULATES WITH MIDLAND AREA SOIL 100.0 IZZJMIOLAND AREA SOIL FROM POWERHOUSE .8 RELATIVE AMOUNT 100? 16 BB ?01[521 1:TCDDISOMER FIGURE IV 7 A COMPARISON BETWEEN MIDLAND SOIL AND SOIL FROM THE 477 AREA LEL 1000? IEIMIDLAND AREA SOIL - SOIL FROM 477 LEL RELATIVE 100? AMOUNT TCDDIROMFR -107? 1. The Mahalanobis and bar graph comparisons both show that the TCDD isomer pattern found in Midland soils and st particulates is different from those of the two LELs, and different from that of Dow Power- house ash. It is therefore unlikely that the LELs or the Dow Powerhouse have Plant Site, Dow fenceline or City of Midland surface soil 2378-TCDD level4. This conclusion is the same as that drawn in Part I considering the flows and concentration gradient contributed significantly to Midland surrounding these point sources. I 2. Within the limits of the availaJle data base, including two stack samples and three kiln-ash samples, Mahalanodis comparison shows that the TCDD isomer patterns in the particulate the Midland Plant Site Haste Incinerator closely resemble the TCDD pattern found in Midland soils and dust particulates.- Bar graph comparison suggests that Waste Incinerator kiln ash more closely resembles Midland soils and dust particulates, especially with respect to the relative amount of the isomer of primary interest. 3. Historically, both Haste Incinerator kiln ash and stack particulates are known to have been accessible to the environment in large quantity, and an airborne dispersal method has long existed. Therefore, although the possi- bility of contributions from additional 2378-TCDD sources cannot be excluded, the Midland Plant Site Haste Incinerator is indicated as the most probable dominant point source of the currently found in the Midland Plant Site and City of Midland surface soils. This conclusion is in accord with that reached earlier considering current and historical Haste Incinerator operating conditions and particulate emission controls. 4. Based on a single set of data points for each LEL, the TCDD isomer patterns in the two Midland Plant Site LELs appear to be greatly different from each other, and greatly different the TCDD isomer patterns found elsewhere in Midland soils. These two LELs are evidently the result of two very different historic chlorophenolic processes. The observed pattern differences are in accord with the earlier finding that LELs are localized areas with elevated TCDD surface soil concentrations which have not spread. -108- CHAPTER 2. USE OF PATTERN RECOGNITION TECHNIQUES IN THE IDENTIFICATION AND RANK-ORDERING OF 2378-TCDD POINT SOURCES INTO THE NASIENATEP ON THE DON CHEMICAL COMPANY MIDLAND PLANT SITE Chapter 1 of Part IV has shown the utility of TCDD isomer pattern com- parisons in helping to identify the Midland Plant Site Waste Incinerator as the dominant point source of the trace levels of 2378-TCDD now found accumu- lated in Midland area soils. Historical windborne dispersion of kiln ashes and vent stack particulates from this facility is shown to be the probable mechanism by which this occurred. In Chapter 2, these same TCDD pattern recognition concepts are used to help identify the Waste Incinerator as the most probable source of the 0.6 gram per year of 2378-TCDD which enters the Tittabawassee River from the Midland Plant Site. The data presented also help further establish that sedimentation is the most probable mechanism by which the Dow Waste Water Treatment Plant removes 2378-TCDD from the outfall waters flowing into the Tittabawassee River. The concentrations of TCDD isomers in Midland Plant Site wastewater streams are low enough that only four of the isomer (isomer pair) levels are consistently above the limits of analytical detection and of use for pattern comparison. Despite this severe restriction, which weakens the use of the comparison techniques, it turns out that the differences observed in these four remaining isomer levels are so striking that meaningful differentiation of potential sources still appears to be possible. Table 1V.6 shows the relative TCDD isomer pattern found on the sus- pended solids remaining in the outfall waters from the Tertiary Treatment Pond. This is stream 13 in Figure 111.3. The samples are composites taken over several-day-periods at intervals of several months. Over 95% of the TCDD comprises just five of the 22 isomers; 1368, 1379, 1237+8 (which are grouped), and 2378-TCDD. This permits the required data analyses and comparisons to be summarized in simple tabular form. The interesting point is that the four- isomer TCDD fingerprint of the suspended solids in the Tertiary Treatment Pond -109- TABLE IV.6 RELATIVE TCDD ISOMER CONCENTRATIONS IN TERTIARY TREATMENT POND SUSPENDED SOLIDS FROM THE NASTE HATER TREATMENT PLANT ON THE non CHEMICAL COMPANY MIDLAND PLANT SITE STREAM 13: OUTFALL NATERS TO THE TITTABANASSEE RIVER 1368 1379 1237+8 Access No. Composite Sampie Date TCDD TCDD TCDD - 1 31-111 5/13/83 - 53.3 16.4 15.9 129-111 2/06/84 45.1 28.2 25.3 139-111 3/5?3/9/84 . 39.4 25.6 30.5 148-111 3/12-3/16/84 46.6 28.1 22.1 156-111 3/19-3/23/84 44.5 25.8 25.8 STREAM 12: -INLET NATERS TO THE TERTIARY TREATMENT POND 5 Relative Concentration 2378 TCDD 1.9 Reiative Concentration 1368 1379 1237+8 Access No. Composite Sampie Date. . TCDD TCDD TCDD I 124-111 2/6/84 3 39.4 27.3 30.3 137-111 375-3/9/84 39.5- 28.1 29.9 145-111 3/12-3/16/84 48.6 27.4 21.2 154-111 3/19-3/23/84 43.6 24.9 28.2 171-111 4/6/84 40.0 27.8 28.8 173-111 4/18/84 39.5 29.9 27.2 2378 TCDD -110- outfall remains quite constant over month-long periods. This behavior could be ascribed either to rapid biochemical equilibration of TCDD isomers in the Tertiary Pond; to the characteristic-behavior of a large stirred reservoir whose composition cannot be changed rapidly; or to the simple fact that the feed coming into the Tertiary Pond has a relatively constant TCDD composition and the pond does nothing to alter it. This last explanation appears to be correct. The data in the lower part of Table IV.6 show that the four-isomer TCDD pattern on the suspended solids in the inlet waters to the Tertiary Pond from the Waste Water Treatment Plant also remains very constant with time, and is essentially the same as the TCDD isomer pattern characteristic of the suspended solids in the outfall waters. This suggests that biochemical destruction or equilibration of TCDD's is not a dominant factor in the 2-3 fold Towering of levels which occurs in the Dow Tertiary Treatment Pond. Simple sedimentation of the silts carrying the seems a most reasonable explanation. The data in Table IV.7 extend this conclusion to the entire Dow Waste Water Treatment Plant. From where wastewaters first enter the primary clarifier in stream 10, to where the wastewaters finally flow into the Tittabawassee River in stream 13, the four-isomer TCDD distribution pattern of the suspended solids remains essentially the same. The TCDD pattern of the sediments removed is likewise the same, and all the patterns are strikingly similar. This finding is in accord with the conclusion reached in Part II that the Dow Midland Waste Water Treatment Plant accomplishes its 50 fold reduction in levels primarily by removing the suspended silts and particulates that carry the 2378-TCDD in the waters passing through it. Since the outfall waters to the Tittabawassee River show a constant four-isomer TCDD pattern over month-long periods, the observed pattern most plausibly results either from the weighted average of the various patterns of all the important point sources to the outfall being constant with time, or from one source dominating the TCDD pattern. ?111- TABLE IV.7 RELATIVE TCDD ISOMER CONCENTRATIONS IN MASTEHATERS AND COLLECTED SEDIMENTS FROM THE WASTE HATER TREATMENT PLANT ON THE DON CHEMICAL COMPANY MIDLAND PLANT SITE STREAM IO: INLET NATERS TO THE PRIMARY CLARIFIER Relative Concentration 1368 1379 1237+8 2378 Access No. Composite Samgle Date TCDD TCDD TCDD TCDD 38-111 5/05/33 I '63.0 11.3 13.1 0.9 1 . 176-111 4/06/34 I 39.7 26.5 30.9 _0.05 178-111 4/13/34 43.3 23.6 24.2 0.1 204-111 Dec. 33 Set 1 42.9 23.3 31.1 0.1 205-111 Dec. 83 Set 1 43.3. 24.5 29.5 0.1 STREAM 16: SOLIDS FROM PRIMARY CLARIFIER Relative Concentration 1368 1379 1237+8 2378 Access No. Comgosite Samp1e Date TCDD TCDD TCDD TCDD 214-111_ 7/23/34 3 49.6 22.6 22.6 0.5 -112- Table IV.8 compares the foureisomer TCDD patterns for all the potential major historical point sources of into the Dow Midland Plant Site wastewater system identified in Part II with the TCDD pattern found in the outfall waters to the Tittabawassee River. The TCDD patterns of those histori- cal sources are all strikingly different from the pattern of whatever point source is dominating the TCDD outfall from the Dow Waste Water Treatment Plant. Furthermore, no combination of these patterns at all closely matches the observed four-isomer TCDD pattern in the outfall. The 2378-TCDD and 1368-TCDD isomer contents are too high, and the 1237+8 and 1379 TCDD contents are always much too low. Therefore, even recognizing the limitations of the four-isomer pattern comparison technique, it seems possible to infer that any and all historical sources, taken alone or together, including rainwater wash- off of surface soils, are probably relatively minor contributors to the water- borne flow of 2378-TCDD leaving the Midland Plant Site into the Tittabawassee River. 0n the other hand, the four-isomer TCDD pattern comparison does suggest that the Midland Plant Site Waste Incinerator is the most probable source of the 2378-TCDD passing through the Dow Haste Water Treatment Plant and entering the Tittabawassee River. Table IV. 9 shows the TCDD "fingerprint" data for the kiln ashes and suspended soils removed from the aqueous exhaust-vent scrubber streams of the Midland Plant Site Haste Incinerator. Table IV.10 compares the ranges on these pattern data with the pattern ranges found on the suspended solids in the outfall waters to the river. The TCDD isomer patterns for the waterborne particulates from the Waste Inciner- ator exhaust scrubber train are strikingly similar to the patterns for the suspended solids found in the Tertiary Pond outfall waters. The TCDD pattern in kiln ash, which contacts Haste Incinerator process waters for quenching and cooling, is also very similar. This inference, that the Waste Incinerator is the most probable major source of the 2378-TCDD found in Tertiary Pond outfall waters, is in accord with the finding in Part II using mass flow measurements that the Haste {?ne Tertiary Pon Stream 13. Data from High Table IV.6 Access No. Data Set for Fig. IV.1 -1135 TABLE IV.8 TCDD ISONER IN OUTFALL NATERS TITTABANASSEE RIVER VS. TCDD ISQHER IN HISTORICRL POINT SOURCES TO THE MIDLAND PLANT SITE NASTENATER SYSTEM NATERS TO THE TITTABANASSEE RIVER Outfall Range Low 1368 TCDD 58.3 39.4 Relative Concentration 1379 1237+8 2378 TCDD TCDD 28.2 30.5 1.9 16.4 15.9 0.1 HISTORICAL POINT SOURCES 2378-TCDD Point Source to Nastewater Systeml Closed Landfill Dewatering Hells i Organic Material Entering the General Sewer Shallow Sump in 11th and Sts. Area Midland Plant Soils, Mean Values 1368 TCDD 76.2 86.1 88.8 48.1 .Relative Concentration 1379 - 1237+3 2373 Tcoo Tcoo TCDD 25.0 1.0 11.9 7.8 16.7 -114- TABLE IV.9 RELATIVE TCDD ISOMER CONCENTRATIONS 0N KILN ASH PARTICULATES FROM THE EXHAUST VENT SCRUBBER NATERS OF THE DON CHEMICAL COMPANY MIDLAND PLANT SITE NASTE INCINERATOR. SEE FIGURE 11.3 ReTative Concentration 1368 1379 1237+8 2378 Access No. Comgosite SamQTe Data .1229 .1992 TCDD QUENCH TONER 44-111 I 8/16/83 50.7 26.0 19.6 0.2 5/19/83 53.8 16.4 19.0 3.1 VENTURI 77-111 5/19/83 55.2 11.9 21.4 2.5 COMPOSITE Dec. 83, Set No. 6 38.2 21.2 37.6 .025 165-111 Dec. 83. Set N0. 2 30.7 17.9 48.6 .020 NET ELECTROSTATIC PRECIPITATOR 8/16/83 44.1 25.2 23.7 0.3 80-111 5/19/83 41.6 22.7 27.6 0.7 Dec. 83, Set No. 6 36.2 21.8 39.2 .021 168-111 Dec. 83. Set No. 2 31.8 18.1 47.4 .021 INCINERATOR KILN ASH 211-111 5/11/84 61.4 12.3 11.1 2.0 Dec. 83, Set Nos. 1-6 38.7 23.5 30.2 1.0 8/16/83 46.0 24.7 9.2 3.7 8/16/83 Ash Trough 43.3 26.5 6.6 5.5 Water -115- TABLE Iv.1o TCDD ISOMER CONEENFRATIONS IN OUTFALL NATERS TO THE TITTABANASSEE RIVER VS. TCDD ISOMER CONCENTRATIONS IN DON NIQLAND ELANT SITE NATERBORNEJPARTICULATES OUTFALL NATERS TO THE TITTABANASSEE RIVER 'Re1ative Concentration 1368 1379 1237+8 2378 Tertiarx Pond 0utfa11 Range TCDD TCDD TCDD Stream 13, Data from High 58.3 28.2 30.5 1.9 Table IV.6 Low 39.4 16.4 15.9 . 0.1 I i i NASTE INCINERATOR NATERBORNE PARTICULES Relative Concentration . 1368 1379 1237+8 2378 Waste Incinerator Range TCDD TCDD TCDD TCDD Scrubber Haters, Data High 55.2 26.0 48.6 3.1 . from Table IV.9 Low 30.7 11.9 19.0 0.02 K11n Ash, Data from High i 61.4- -26.5 30.2 5.5 Tab1e IV.9 Low. 38.7 12.3 6.6 1.0 -116- Incinerator is an important point source of into Midland Plant Site wastewaters. It should be re-emphasized that corrective actions to reduce or eliminate the 2378-TCDD flow to the Midland Plant Site wastewater system from all four of the point sources identified here are being undertaken. -117- CHAPTER 3. MAHALANOBIS CALCULATION PROCEDURES The method used in this report to quantify similarity or dissimilarity between groups of samples is a generalized distance measure, Hahalanobis 1 2 distance. Mabalanobis distance is a generalization of geometric or Euclidean distances. 'The axes of the standard Euclidean distance are first normalized so that each variable has equal weighting in the distance measure. The distances are then corrected for Known correlations between the variables, i.e. the n-dimensional space is stretched along directions of correlations. In matrix notation the Euclidean distance between two points and can be calculated as 1 distance2 . . . . where and are Single numbers or vectors. Mahalanobis distance normalizes the-above calculation-by the variance - covariance matrix distance 2 If and are one dimensional number?, then Euclidean distance 2 (x-y)2 and Mahalanobis distance2 i where 52 is the sample variance of replicate-samples. 1SAS User's Guide: Statistics SAS Institute, 1982. 2Morrison, D. F., Multivariate Statistical Methods, McGraw?Hill, 1967. In this application, sets of relative isomer distributions were compared by calculating the distance (Mahalanobis distance) between the group centroids (multidimensional centroids). Calculation of the distance was carried out using the Statistical Analysis System (SAS) procedure CANDISC (1) with the Mahalanobis print option. -Before this could be done a procedure was needed to handle samples with undetectable isomer concentrations. This was necessary since, for each sample, if 131 isomer was not detected (?missing?) that whole sample would be eliminated from-the calculations. Because of the low concentration of some of the TCDD isomers this missing value problem would eliminate the majority of the samples in the database. A-number of methods for handling this problem were evaluated before any calculations were done. The compromise which must be made is that too much estimation of missinglvalues results in a regression towards the mean, and less discriminatory ability, whereas too little estimation reduces sensitivity through a reduction in sample size. The following a priori criteria were decided upon: Step 1: Define the database list all samples to be included). Step 2: Use only the TCDD isomers which are detected in more than half of all the samples in the defined database. The following isomers were excluded as a consequence of the constraints of this step: 1469, 1269, 1267, 1289, 1478, 1236. These isomers were felt to contribute little information to the fingerprint. They were missing in more than half of the samples, and when present did not vary much in relative concentration. Step 3: Exclude all samples for which there are more than ten undetected tetra isomers. The above procedure can be visualized by listing sample numbers down the side of a page and isomers across the top. First, each column (isomer) is not detected. -119- I eliminated if it is not detected in at least half of the rows (samples). 'Then each row (sample) is eliminated if more than half of its column (isomer) are Step 4: For each of the remaining 14 Fsomers, calculate the mean relative concentration across the samples in the database. I . I Step 5: Replace not detected (ND) relative isomer concentrations by the mean calculated in Step 4. i I Step For each replacement in Step 5, check Hhether the mean value is above the limit of detection (LOD) for the relative isomer concentration in that sample. If the substituted value is above the Lou, then the Loo is used rather than the mean for the undetecte? value. I Step 7: For specific subsets of samples calculate Mahalanobis distance using SAS procedure and print the table of pair-wise distances. -- RIVERBANK REVETMENT SYSTEM DOW EACILITIES MIDLAND, MICHIGAN DOW CHEMICAL COMPANY MICHIGAN DIVISION DIVISION ENGINEERING 57E BUILDING. MIDLAND, MICHIGAN 48640 OCTOBER 10, 1984 BY MCDOWELL ASSOCIATES MCDOWELL ASSOCIATES 10659 Galaxie Femdale, Michigan 48220 Phone: 313 - 399-2066 October 10, 1984 Dow Chemical Company Michigan Division Division Engineering 572 Building Midland, Michigan 48640 Job No. 82-77 Attention: Mr. Gary Veurink Subject: Riverbank Revetment System Gentlemen: In accordance with your request, we are providing an outline of the subject system. - BACKGROUND In about mid?1980, discussions were held between Dow Chemical Company and McDowell Associates concerning the potential of groundwater from the plant site reaching the Tittabawassee River, and if so, the potential solutions. It was known that there were areas within the plant site where the groundwater level was at a higher elevation than the river level, and thus, a potential fer movement to the river existed. The river appeared to flow in hardpan, which is a glacial, preloaded clay soil of very low per- meability and is common to the Midland area at about the elevation of the Tittabawassee River. Visually, the banks appeared to be clay, and there appeared to be little, or no, evidence of ground? water exiting the banks. Geotechnical Services - - Materials Testing Inspection 1 I. if i Page . Job No. 82?77 Dow Chemical Company Midland, Michigan 48640 ConCurrently, McDowell Associates was retained to do a study along the north edge of the river for an erosion control and riverbank revetment project. iThis study included drilling borings from a barge along the north edge of the river. These borings generally confirmed the presence of clay hardpan at shallow depths, overlain by what appeared to be river-deposited sediments. Lures: A Collection System design was integrated.into the design of the existing erosion control and riverbank revetment project and installed. This system, completed, involved a semi?impervious barrier (sheet piling) along the north edge of the river, with a drainage system on the plant side of the sheet piling. The sheet piling was driven a few feet into the clay hardpan. and the drain tile was embedded into the hardpan and backfilled with highly porous sand to near the surface. The slope was capped with clay. including the area over the sand backfill. This system provided a positive collection of any groundwater which could have been moving to the river on top of the clay hardpan. There was a short section where deep deposits of granular soils were present. In these areas, the sheeting was driven to refusal,.and the drain tile was installed as deep as practical. This resulted in a partially penetrating cutoff. The partially penetrating cutoff Page Job No. 82?77 Dow Chemical Company Midland, Michigan 48640 was studied by setting piezometers near the top of the bank and near the bottom of the bank. These piezometers were read, and water elevations were compared to the water elevation in the river at different times of the year. The data indicated that the groundwater table on the plant side of the Collection System had been depressed, so that it was generally below the river level. There was one exception, however, where a piezometer indicated a water level higher than the river for-some of its readings. This was in the deepest area of sand, and appeared to be associated with the river level dropping faster than the piezometer could react the elevation of the river varies, and this variation is reflected in the piezometers a higher elevation in a piezometer could indicate a slow-reacting piezometer caused by time lag). It is unlikely that this condition actually resulted in movement to the river. However, because the analysis of the drain tile syStem was not totally conclusive in the deep sand condition, a Well was installed to provide a backup to the drain tile system. DETAILS OF SYSTEM The location of the drain tile is shown on EXhibit 1. This is a six inch perforated drain tile. It extends from the Dow Rail? road Bridge southeast along the north bank of the Tittabawassee River to Saginaw Road, and then returns north to the railroad tracks. Page Job No. 82-77 Dow Chemical Company Midland. Michigan 48640 This involves a length of approximately eleven thousand feet There are six (6) sumps in the system (five of which are operational) where the collected water is pumped to sewers I which transport it to the treatment plant. There are cleanouts on three hundred foot (300') to four hundred foot (400') centers. Exhibit 2 contains Dow Chemical Company Drawings and and shows a typical plan and profile of the sheet piling (Station 1100 to Station 2300) and some typical sections. Note that the sheet pilings were driven several feet into the clay soils. A typical plan and profile oflthe six inch drain tile and some typical sections are contained in Exhibit 3. which contains Dow Chemical Company Drawings and Note that the drain tile is placedlinto the clay strata. The trench which was excavated for the drain tile was backfilled with a well- graded sand which provides good porosity and good filtering action. The surface of the trench was covered with compacted clay which extended from the sheeting, part of the way up the slope (see hatching on Drawing 32?027?810120). The sheet pile wall and drain do not penetrate into clay from about Station 3160 to about Station 3560. This condition is Shown on . . -. . nan 3-.ng 11:5,. - Page Job No. 82?77 Dow Chemical Company Midland, Michigan 48640 EXhibit 4. This is due to the deep sand area previously discussed. The deepest sand condition was found at about Station 3430, where it extended to about Elevation 4. This condition was investigated with piezometers, and the results were not completely conclusive, in that Piezometer 2847 indicated a piezometric head higher than the river during short?term cyclic low water levels in the river. In order to provide additional assurance of flow not reaching the river, a deep Well was installed at about the'location of Soil Boring 2851. This is a twelve inch Well, screened from about Elevation 2 to Elevation 22. A set of data was obtained from the piezometers between August 22, 1983 and August 26, 1983. During this time, there was minimal fluctuation in the river level, and all of the bottom of the bank piezometers recorded water levels lower than the river level. Piezometer 2847, which has been determined to be the critical piezometer, indicated piezometric heads varying from about one foot to three feet below the river level. The pump in the Well was turned on at about 10:00 A.M. on August 26, 1983, and the piezometric head in 2847 dropped to about five and one? half feet below the river level in two (2) hours. A set of readings taken one (1) month later indicated that the pumping was maintaining the piezometric head in 2847, six feet to seven feet below the river level (see EXhibit 5, McDowell Associates' Drawings labeled Group 1, Pages 1?4 and Group 2, Pages We IPage -6- Job Dow Chemical Company - 3 Midland, Michigan 48640 . I concluded from this data that with the Well functioning, are was Hh_pi__ - a significant factor of safety in terms of water flowing from the i river to the Collection System. CLOSURE If we can be of any further service, please feel free to call. very truly yours, MCDOWELL ASSOCIATES W: Robert ?cDowell, P.E. RMcD/dmm 00.1 DO . 0730.021 1.42. 3.3. . .. nImz?rr BEEN whiz m. ?nomarm lozngelq mImmA. . . . 85 @080 .1 xn?_u_Oz NOISIAIU 5. I 55 a 3: :4an vamu Dnmud 02\018- Oto?Z?a H3 MOO . . av 00:: nxmz?br Fm? Damn. nm680 Van ox. A mus?m? Em APPENDIX VII HIGH PERFORMANCE SAND FILTRATION 0F PARTICULATES FROM TERTIARY POND EFFLUENT I INTRODUCTION: This appendix is'a Dow Chemical Company 1984 research report by H. A. Johnson, L. A. Robbins and D. R. Haslam. It describes pilot plant sand- filtration experiments on Midland Plant Site Tertiary Treatment Pond outfall waters. The experiments described in this appendix were carried out to determine how much of the suspended silt now remaining in the Outfall waters to the TittabawasSee River from the Dow Waste Water Tneatment Plant could be rempved by final.filtering the water through specially designed deep sand beds grEated with cationic flocculating polymers. The experiments show that under ideal pilot plant conditions 80-90% of the suspended silts can be removed. This is an important finding because Chapter 11.3 of this report shows that 80?99% of the 2378-TCDD found in Midland Rlant Site wastewaters is' carried along by the suspended silts in it, and that the 2378-TCDD can be removed from the water by removing the'silt. These two findings, taken together, suggeSt that large scale flocculant- polymer-treated deep- -sand?bed filtration could iremove 50- 80% of the 25 parts per quadrillion 2378- TCDD now present in the odtfall waters from the Midland Plant Site. As summarized in Table 11.9 in the main body of this report, actual measurements of the 2378-TCDD content in Tertiary Treatment Pond outfall waters treated by the technology described in this appendix show that a 50-80% . reduction in 2378-TCDD does indeed occur. TABLE g; CONTENTS ListOf Figures ii List Of Tables 11" Introduction Phase Partitioning 0f Chemical Species 2 . Sand 11 tra tion Concep Process EquiPmEnt 1:0 Sample Analys Selection Results Of Sand Filter operation 30 Sand Filter Design Considerations 44 ReferenCes 44 FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE -11- LIST FIGURES Partitioning 0f Chemical Species Onto Solids In A Solids-Hater Stream Sand Filter vessel Effect Of Multiple Backwashs On Solids Removal From Sand Filters Particle Removal Mechanisms Sand Flowsheet For Sand Filter Pilot Plant Sand Filter Pilot Plant Location ?Total Suspended Solids Analytical Method Correlation 0f Total Suspended Solids By The Gravimetric Method With HIAC Particle Volume Measurements Typical Tertiary Pond Particle Size Distribution Particle Counting By HIAC Laboratow Sand Filter Laboratory Sand Filter Polymer-Evaluation Methods Laboratory Performance Of Cationic Polymers Diminishing Filter Performance with Increased Polymer Dosage Environmental Information On Agefloc WT-40 Ef?ect Of Polymer 0n Sand Filter Performance 3.713 FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE LIST Seasonal Variation 0f Solids Concentration . 33 In The Filter Feed Filter Performance At 25 GPM With 1.1 PPM 34 PURIFLOC C-31 Flocculant Filter Performance At 25 GPM Pith l.l PPM 35 PURIFLOC 0-31 Flocculant - Multiple Cycles Filter Performance At 30 GPM with 1.2 PPM 36 PURIFLOC C-31 Flocculant - Multiple Cycles Filter Performance At 35 GPM With 1.4 PPM 37 PURIFLOC 0-31 Flocculant Effect Of Flux Rate 0n Sand Filter .. 39 Performance Using PURIFLOC 0-31 Flocculant Filter Performance At 30 GPM With 1.2 PPM 40 PURIFLOC Flocculant On A Special Single Cycle Filter Performance At 25 GPM With 1.07 PPM 41 Agefloc WT40 - Multiple Cycles Filter Performance At 25 GPM With 0.63 PPM 42 Agefloc WT40 Filter Performance At 30 GPM With 0.79 PPM 43 PURIFLOC C-31 Flocculant TABLE TABLE TABLE TABLE TABLE TABLE TABLE .. 1v- LIST TABLES Linear Partitioning 0f Aroclor 1254 Onto Various Solids Stepper Control Program Filtration Sequence Hith Typical Conditions Polymers Evaluated In The Laboratory Laboratory Polymer Evaluation.Data Filter Performance Data Summary A Sand Filter Haterial Balance Example Calculation 3 15 25 26 32 INTRODUCTION Tertiary pond water discharging to the river carries suspended solids (less than 45 ppm)(l). There are also minute concentrations of various chemical species associated with this total stream. Recent research work (2) has shown that some chemical species (which are not necessarily the same ones as those associated with the tertiary pond stream) are prefer- entially partitioned onto solids suspended in water. Therefore, a reduc- tion of the chemical species present in the total tertiary pond discharge may be possible if a suitable liquidrsolids separation was available. The purpose of the research discussed in'this report is the development.of such a process. Removal of most of the suspended solids from water can be accomplished by a number of solid-liquid separation methods. However, there are several constraints in this particular situation that decrease the alternatives. First, the discharge rate is very high at 20 million gallons per day. Second, the suspended solids are small, with a 1011 average, while the present concentration of 30-45 mg/L of solids must be reduced to 3?4 mg/L. Third, the separated solids that must be disposed need toibe minimized. Fourth, the capital and operating costs must be economically realistic. High performance sand filtration can handle these constraints. Thfse sand filters Operate under.pressure (60 psig) atzrates of 8-12 gpm/ ft versus atmospheric pressure at rates of 2-4 gpm/ft for gravity sand filters. High performance sand filters have demonstrated performance (3) on large volumes of water (to 150,000 gpm) containing low concentra- tions (less 50 mg/L) of small particles. No filter aids are required so no additional solids diaposal is required. These filters are backflushed periodically with a small amount of product water. The capital and oper? ating costs of high performance sand filters are very low compared to other solid-liquid separation processes. High performance sand filtration was therefore selected for removing particulates from the tertiary pond discharge. Successful removal of these particulates would provide two other benefits (1). During the summer months, there are large algae blooms in the tertiary pond that can cause the total suSpended solids (TSS) concentra- tion to exceed, especially during windy days, the 45 mg/L maximum stipu- lated in the water discharge permit. These algae are also included in the biological oxygen demand (BOD) determination. Even though the algae in the discharge to the river would not affect the BOD because they would be alive, the laboratory BOD procedure causes the algae to die. These dead algae are then counted as increased BOD. - PHASE PARIITIONING OF CHEMICAL SPECIES In a recently published research paper (2), Weber, et.al. presented data on the relative importance of various solids found in aquatic systems with respect to the sorption of biphenyls (PCB's). They used a commercial mixture of PCB's, Aroclor 1254, for their study. At low con-' centration levels, they found that "the amount of PCB sorbed by the solid phase versus the amount remaining in solution at equilibrium gave linear phase-partitioning behavior". The partition coefficients that they determined for various sorbents are shown in Table 1. Although the data in Table are not determined from tertiary pond efflu- ent and solids, the values of can be used to project the level of 'chemical species associated wi the suspended solids. The projection should be realistic, because a high percentage of the solids are probably live or dead algae, and because a number of chemical species could have sorption characteristics similar to Aroclor 1254. A number of material balance equations can be written for the tertiary pond water. The overall solid-liquid equation is: - (1) Total stream rate, of total/min Liquid rate, liquid/min Solid rate, solids/min A component balance for all chemical species in the water or associated with the solids has the following form: qu st . (2) x? Concentration of chemical species in total stream, chemical species/gms total stream xL Concentration of chemical species in liquid, chemical species/gms liquid x3 Concentration of chemical species in solids, chemical species/gms solids The definition for total suspended solids, TSS a SIM (3) TSS Total suspended solids, solids/gms total stream I I TABLE 1 LINEAR PARTITIONING 0F AROCLOR 1254 ONTO VARIOUS SOLIDS (REF. Sorbent CLAYS Montmorillonite 1,626 Kaolinite 1,654 Blue Clay, Natural 26,408 Blue Clay, Stripped 3,626 SEDIMENTS Saginaw River 1, Natural 23,760 Saginaw River 1, Stripped - 1,152 Saginaw River 2, Natural 53,543 Saginaw River 2, Extract 150,686 Saginaw River 2, Benzene, Extract. 35,734 Saginaw River 2, Stripped 102,841 Saginaw River 2, 12C 31,859 Saginaw River 2, <75 pm 76,759 Saginaw Bay - 11,143 SUSPENDED SOLIDS Huron River 116,338 Saginaw River 1 88,981 Saginaw RiVer 2 117,693 ORGANISHS Bacteria, Live 464,372 Bacteria, Dead 421,753 Algae, Live 1,289,065 Algae, Dead 578,913 *Grams of Aroclor 1254 per gram of solids divided by grams of Aroclor 1254 per gram of water. and the linear partition coefficient: KP xS/xL (4) KP Linear partition coefficient, of.chemical species per gm of solids divided by of chemical species per gm of liquid and Equation 1 can be substituted into Equation 2 to give Equation 5. xix-rs .LTs_s+Tss] ?2 The fraction of chemical species in the entire stream that are associated with the solids is defined by Equation 6. Sx . (6) F3 Fraction of all chemical species on solids, of chemical species per gm solids divided by of chemical species per gm of total stream Substitution of Equations 3 and 5 into Equation 6 gives Equation 7. F5 TSS (7) T88 '99 Equation 7 was used to calculate, for various_conditions, the fraction of chemical species in the total stream that is associated with the solids. The results of this calculation, presented in Figure 1, show that greater than 902 of the total chemical species for condi ions of TSS 9 and KP 106 are associated with solids. A KP of 10 is probably typical of the tertiary pond environment (algae in Table 1). These numbers indicate that sand filtration could significantly reduce the chemical species concentration of tertiary pond effluent. SAND FILTRATION CONCEPTS In the introduction of this report, the reasons for selection of high per- formance sand filtration for solids separation from tertiary pond effluent are presented. In this section, the mechanics of the separation will be discussed. Fraction of Chemical Species Associated With Solids (F3) 0.3 0.2 0.1 - -..- FIGURE 1 PARTITIONING OF CHEMICAL SPECIES ONTO SOLIDS IN A STREAM '20 Total Suspended Solids 24 28 Solids are removed from a liquid, generally water, by forcing the feed stream either by gravity or pump pressure through multilayers of various types and sizes of granular media. Additives are sometimes added to the feed to improve the filter efficiency. As solids are removed by the bed of granular media, the differential pressure across the bed (head loss) increases. When this pressure increase becomes high enough to signifi- cantly.reduce the flow, to cause concern about plugging, and/or to present difficulties with solids removal during backwash, the feed is stopped, and the backwash sequence is begun. During this sequence, the media are scoured with an upflow of air? The freed solids are then subsequently backwashed upwards through the media by clean liquid. The air scour/back- wash liquid sequence is most efficient with respect to time and liquid usage when kept short and repeated (Figure 3). The single air scour with extensive use of backwash liquid does not remove as many solids per unit volume of liquid as the multisequenee (2 or 3 times) procedure (area under curves). After this solids-remdval procedure, the filter is ready for further production of filtered liquid. The granular media used in this project, anthracite and garnet, have specific gravities of 1.3 and h.2,.respectively. This wide specific gravity difference was selected so that the garnet would settle faster than the anthracite after the bed was backflushed. The garnet would then be below the anthracite. This positioning is shown in Figure 2. The media on the top layer, the anthracite, must be heavy enough to settle rapidly through the water so that the backwash step would not be needlessly prolonged. The anthracite at 12 U.S. mesh is larger than the 30-40 U.S. mesh garnet. This graded media approach is used so that the solids capacity of the filter will be higher than a single layer of media would provide. Solids are separated from the liquid by the 3 mechanisms shown in Figure 4. Surface collection is used for a small amount of particles that are larger than the Openings in the media. If the media were extremely fine, most of the solids would be filtered from the liquid by this mechanism. However, the solids capacity per area of filter would be low, and the pressure drop would be excessive. Particles are also removed from the liquid by voids collection. The particles are physically trapped by vari- ations in size of the void spaces in the media or settle into quiescent openings between them. Higher loading levels are possible with this mechanism compared to surface collection; however, most of the smaller sized particles are not removed. The charge-effect mechanism is the preferred means of particle removal. All particles have surface charges that are due to unequal distribution of ions over the particle and in the surrounding'solution (4). These uneven distributions are dependent on the nature of the particle. Insoluble Upper Drainl Under Drain Layout SAND VESSEL FIGURE .345. op..".955 wt{Mal .- 0.0%9.2 anwl..51.. 5 .. .ia. 7 IN.- ?r I?Ll\.mex91.? ?v5..4. 3.2.0% ymmoo. . 8.9 2 ?run? 3 9 1 .- .J-W.uc.rnv a?as} brvuv: e. . .. Hun-Q?. 99D half-A a. u. . .0. I OWAHA n. Ram-rog?ofv .rmau .. .1. Feed Backwash 0 Out L5 Backwash In Product Air Solids Removal Solids Removal -3- FIGURE 3 EFFECT OF MULTIPLE BACKWASHS 0N SOLIDS REMOVAL FROM SAND FILTERS ingle equence Time Triple 3 equ ence -9- 4 PARTICLE REMOVAL MECHANISMS IN A FILTER Surface Collection 000000 00000 .. - ?Voids Collection 0 0 0?0 0 0 Charge Collection . Media OI ?0 Particle 50 Particle Media 4' And Polymer I -10- oxides can take up or 0H- depending on the pH of the solution. Oxides are positively charged at lo pH and negatively charged at high pH. The relative surface affinity for or OH depends where the pH shift occurs. For example, the zero charge pH for silica is about 2, while- hematite 33203) has a zero charge pH of about 8.5. Organic particles often carry a negative charge due to dissociation of rboxyl group??3 Clay+pinerals attain their charge from exchange-of Si ions by AL or Ca ions without a structure charge. This exchange causes the platelet surfaces to be negatively charged. However, clay particles also form positive charged edges on the platelets and readily adsorb ionic surface active substances So that their net charge becomes environmentally dependent. Natural and substances such as alum and flocculation polymers can be used to enhance the surface effects of all types of particles. Selection of these substances must be based on experimental trials with the filter media and the feed stream being considered. Seventy to eighty percent of the filter capacity is due to charge effects (5). This removal mechanism permits high solids loading, because the solids are distributed throughout the media. Small particles are also efficiently removed. PROCESS EQUTPHENT The sand filter used for this work is part of a complete skid unit con- taining feed pump, air compressor, valving, and control system. This equipment is rented from its manufacturer, Serck Baker, Inc. (3). The fi ter vessel is 24" in diameter and 60" high (cross-sectional area 3.1 ft In Figure 2, the interior view of this vessel is shown. Liquid into the vessel and backwash out passed through a "shower head type" dis- tributor at the top of the sand bed. The bottom discharge grid (underh drain) is a wound, circular screen pipe typically used in ion exchange columns. The openings in these screens are 150 microns. The underdrain layout, also shown in Figure 2, was designed so that the liquid would pass through the sand bed in parallel flow with a minimum of turbulence. Good liquid distribution is necessary for high solids capacity with good fil- trate clarity (no early breakthrough). The upper drain is located 5" below the top of the feed distributor so that the liquid level can be lowered before the air scour is turned on. Lowering the liquid level pre- vents media carryover into the backwash discharge line. The inlet, out- let, and upper drain lines are 1-1/2" pipe. The vent, located on top of the filter, is l" pipe. Media was loaded into the bed after the internals were inspected. Garnet (8-12 U.S. Mesh or 2.38 to 1.68 mm) was used to cover the underdrain by about This material helps in the liquid .-11- collection after filtration has taken place in the finer media above it. The main filter bed was 20" of 30-40 Mesh (0.595 to 0.420 mm) garnet. Then ll" of 12 U.S. Mesh (1.68 mm) anthracite were added. The specific gravity of garnet is 4.2, and of anthracite is 1.3. The flowhseet for the entire process is shown in Figure 5. This process was located near 1069 Building (Figure 6) and enclosed by a temporary building. All the automatic valves on the filter except the vent and air ball valves are the butterfly type. The feed pump is a 1-1/2 2 Aurora with a 7.5 horsepower 3550 motor. Pressure on the top of the sand bed during feeding was 64 psig. The polymer solution pump is a Precision Con- trol Products, Model 12322-11 unit having a flowrate range of 5-70 mil- liters per minute. The feed flowrate is manually controlled by the valve on the effluent line using a MK485 Signet Scientific flowmeter on the feed line. The backwash low is restricted to a preset level less than 55 (typically 16 gpm/ft by a 2" Griswold Model 4075-04.flow controller. The sump pump is a Gorman-Rupp, cast iron, Model 52D3, 115 unit having 34' of head at 60 gpm. The feed tank is 5' in diameter by 4' high. The backwash tank is 6' in diameter by 5' high. Both tanks are made of fiberglass and are open topped. _The filtrate tank is made of 304 stainless steel, is covered with plywood, and is 6' in diameter by 5' high. All the equipment is connected by 1-1/2" diameter galvanized pipe or 2" diameter rubber hose. The polymer solution tank is made of fiberglass and is 2' in diameter by 2-1/2' high. The tank is also covered. The automatic valves and feed pumps are controlled by a combination of timers and a cam-type stepper switch system. The program for the stepper system is shown in Table 2. The filtration sequence, with the 2 backwash cycles used in this work and with typical Operating times, is shown in Table 3. The on-line timer is normally set so that the pressure drop through the filter, due to solids loading, is not excessive (less than 20 psig), before the timer reaches zero. Since unexpected surges in feed solids could plug the filter before the timer went to zero, a differential bed pressure switch is wired to override the timer. The switch setting for this work was 30 psig. SAMPLE ANALYSIS Determination of the separation efficiency of a filter is made by measur- ing the solids in the feed and product. During this project, 3 methods for making this measurement were evaluated. These methods are gravimetric analysis, light transmittance, and particle counting. FIGURE 5 FLOWSHEET FOR SAND FILTER PILOT PLANT Vent . Sand Backwash Filter From Air Compressor Filtrate Feed Automatic Butterfly valves Automatic Ball valves Gate Valves Griswold Controller Upper Drain I FIGURE 6 SAND FILTER PILOT PLANT LOCAIIO lertiary Pond I I . i. 1512 9 Trailer 1 Backwash i i . Filter 1069 Building TABLE 2 -14- STEPPER CONTROL PROGRAM Valves u_ 3 Timers piston-$2 8P8 a: smugaqss??s. 33 saageasgw-ags 54 ?4 a a a Function Switch DeJEnergized 0/0 On-Line Off-Line 2 XIX Drain 3 XIX 0 Air Scour 4 XIX 0 0 Vent 5 XIX XIO Backwash 6 XIX 0 XIO Settle 7 XIX Drain 8 XIX 0 Air Scour 9 0 Vent- 10 XIX 0 XIO Backwash ll XIX 0 XIO 0 Settle 12 x/xo XION Refill 13- 0 0 Rinse To Waste 14 Run-Skip 15 Run-Skip 16 Run-Skip 17 Run-Skip 13 Run-Skip 19 RunqSkip 20 Value Closed. 0 Value Opened Power On Conditions not shown are in de-energized state Step On?Line Off-Line Drain Air Scour Vent Backwash? Settle Drain Air Scour Vent Backwash Settle Refill -15- TABLE 3 FILTRATION SEQUENCE HITH TYPICAL Typical Conditions And Comments 8-12 gpm/ft:2 Feed Equipment?Dependent 2 cfm/ft2 Media Settles l6 gpm/ft2 Feed Equipment Dependent 2 cfm/ft2 Media Settles 16 gpm/ft2 Feed Equipment'Dependent Rinse To Waste Time Depends 0n Product Needs CONDITIONS Trpical Times 6-12 hours 1.0 min 1.5 min 2.0 min 1.0 min 3.0 min (can vary') 1.0 min 1.5 min 2.0 min 1.0 min 3.0 min (can vary) 1.0 min 10 sec 30 sec - - r. - -15- The gravimetric analysis is currently used to measure the total suspended solids (TSS) in tertiary pond.discharge. Filtration of the sample through a fine filter is the basis of this analysis (Figure 7). Although this method gives an accepted TSS value, it takes a long time and is quite dependent on the laboratory technique (experience) of the analyst. The method utilized a Klett-Summerson calorimeter.- This unit was operated by initially zeroing the photoelectric circuit with a blank solution (clean'water) in-a glass cell. The light transmittance through a cell containing the sample was then determined. This method was very quick and-simple, but the results did not correlate well with the gravimetric method (especially at low solids concentrations). Changes in the particle from sample to sample was probably the main cause of this problem. The particle-counting method was easy to use, gave a good correlation with the gravimetric method (Figure 8), and provided a particle size distribu- tion for each sample (Figure 9). This method, which is shown in Figure 10, was developed by Peck (7) for use in a HIAC, Model PC-320, 12 channel, particle size analyzer. The HIAC.0perates on a principle. Particles flow past a window of known area through which a light beam shines on a photodiode. When a particle having a cetain projected area passes by the window, part of the light beam is blocked. This particle is counted as a sphere having the-same projected area. The HIAC particle counting method was used in conjunction with the results shown in Figure 8 to give the T88 values for filter feed and filtrate samples presented in this report. Collecting samples that represented the stream conditions was critical for -the filter evaluation. The cencentration levels of solid in the feed and filtrate streams varied from 1 to 60 ppm. Minute amounts of solids in sampling lines would easily invalidate an analysis. Therefore, the feed samples were taken directly from the feed tank. The continuous flow into the tank and overflow out kept the tank well mixed with fresh feed. The filtrate samples were taken directly from the end of the discharge pipe. POLYMER SELECTION The initial runs with the sand filter pilot plant were made with no polymer addition to the feed. Filtration without polymer, if acceptable operation could be attained, would be simpler and less expensive. How- ever, in runs without polymer, less than 20% of the solids were removed so use of a polymer became mandatory. FIGUEE 7 TOTAL SUSPENDED SOLIDS ANALYTICAL METHOD Total Suspended Solids is defined as the solids retained on a standard glass fiber filter disc (2.4 cm Whatman, effective retention 1.2 m) after filtration of a well mixed sample of water or waste water. They are those solids which are retained and dried to constant weight at Nonhomogeneous particulates such as leaves, sticks, fish, and other such matter should be excluded from the sample. Apparatus 1) Glass fiber filter discs, 2.4 cm Whatman without organic binder. 2) Filter holder, membrane filter funnel or Gooch crucible adapter. 3) Suction flask, 500 or 1000 ml. 4) Gooch crucible, perforated 25 ml. 5) Drying oven 103-105?c. 6) Muffle-furnace at 7) Desiccator. 8) Analytical balance, 200 gm capacity, capable of weighing to 0.1 mg. 9) Tongs or forceps. 10) Graduated cylinders (100 ml, 250 ml, 500 ml) and assorted large orifice pipets. Procedure 1) Preparation of Gooch Crucible: Place a glass fiber filter disc on the bottom of a Gooch crucible. While vacuum is applied, wash the disc with 2 or 3 successive 20 to 25 ml volumes of distilled water. Continue to apply vacuum to remove all traces of water. Remove from vacuum, and dry in an oven of for 5 to 10 minutes. Allow to cool in desiccator. Store in the oven. Immediately prior to use, remove crucibles from the oven, allow to cool in a desiccator, and obtain tare weight. Do not store crucibles at room temperature for an extended period of time before obtaining the tare weight. DOW CONFIDENTIAL 2) -13- FIGURE 7 Treatment Of Sample: Assemble the filtering apparatus and begin suction. The.desired amount of a well shaken sample is'measured with a graduated cylinder or a large orifice pipet and filtered through the combination of Gooch crucible and glass fiber filter disc. The size of the sample depends upon the concentration of suspended material present. Continue to apply suction for about 1 to 2 minutes after filtration is completed in order to remove as much water as possible. Rinse sample in crucible with approximately 25 to 50 ml of distilled water. Shut down suction. Using tongs or forceps, remove crucible from adapter and place in drying oven at for at least 2 hours. Cool in a desiccator at room temperature and weigh. Calculation suspended solids grams of.suspended solids 1000 (mg conversion) 1000 ml of sample used (1) Crucible Dried Sample Weight - Tared Height.0f Crucible HIAC Total Particle Volume, cm3x106 -19- FIGURE 8 CORRELATION OF TOTQL SUSPENDED SOLIDS BY THE GRAVIMETRIO METHOD WITH HEAC PARTICLE VOLUME MEASUREMENTS i I I Slope 1.Gravimetrtc ToFal Suspended Solids, mg/L FEED: FILTRATE: FIGURE 9 TYPICAL TERTIARY POND PARTICLE SIZE DISTRIBUTION ,-63 PARTICLE 29.8 10 cm ?193 463-98-6 ..39 ?43 CUM UDL PERCENT l? I "23 EH -6 PARTICLE VOLUME 2.9M 10 cm; I ?93?s 1aMLx??ggLK I ?153 ?23 ng?f ?NH-ha} 19 i DIRMETER IN BUM V.- -21- . FIGURE 10 PARTICLE COUNTING BY HIAC Apparatus 1) Model PC-320 HIAC with 12 counting channels and a sampling system. 2) Source of deionized water that has been run through a Millipore filter cartridge. 3) Glassware that was cleaned with Dow's bathroom cleaner and well rinsed with the filtered water. Procedure 1) Warm up HIAC for 20 minutes with water in sensor. 2) Stir sample well and then withdraw lO mls, and transfer to beaker containing 200 of water. 3) Put beaker in HIAC sample chamber, and agitate 30-60 seconds before starting to count. 4) Record particle counts in each channel after running HIAC. 5) Repeat counting procedure 3 times. 6) Determine background counts On the water for each series of samples run. 7) Use the Hewlett-Packard desk-tap computer to analyze the data. -22- There are numerous anionic, cationic, and nanionic polymers available from many manufacturers. Since evaluation of a number of these materials at various loading.levels wOuld-require considerable time in the pilot plant, a laboratory-screening method was developed. Ives (7) and others have "empirically-established that if the wall-to-wall distance" in a sand filter test model for the laboratory "is at least 50 times the largest grain size", the effect of the walls would be negligible. The- small garnet sand used in the pilot plant has a size-range of 30-40 U.S. mesh, which yields an average particle size of about 0.5 mm. Thus, a 1" diameter column could be used and would satisfy Ives' criteria. In Figure 11, the laboratory setup is shown. The method used to evaluate the polymers is presented in Figure 12. The polymers that were evaluated are listed in Table 4. The data and the calculated terms are tabulated in.Table The effectiveness factors in Table 5 show that alum and the anionic AP30 flocculant do not perform very well compared to the cati- onic polymers. In Figure 13, the cationic polymers are compared. The Agefloc performed the best in the laboratory experiments, and since it was a commercially available material (CPS Chemical), it was chosen for further pilot plant testing. PURIFLOC 0-31 flocculant was also selected for further pilot plant tests because it gave the best perform- ance of any Dow product. The trade prices (December, 1983 bulk) of these _products were similar -- PURIFLOC flocculant 0.89 $/lb and Agefloc 0.91 $/1b (FOB West Memphis, Ark.). However, the PURIFLOC 0-31 flocculant has a large economic advantage because the internal Dow transfer price may be only 25% of the trade price. "More is better" is not a good philosophy when trying to improve sand filter performance with polymers. Figure 14 shows that filter perform? ance reaches a limit for increased feed concentration of polymer. Also, data on polymer ll58, presented in.Table 5, show that very high polymer concentrations can cause a major catastrophe (plugging) in sand filters. Cationic polymers above certain concentrations are toxic to fish. PURIFLOC 0-31 flocculant has been tested (8) with vertebrate, fathead minnows, and found to give an L050 of 24 mg/L at 95% confidence intervals in water containing 100 mg/L of bentbnite clay. According to Figure 15, the maximum recommended concentration of Agefloc WT40 in potable water applications is 20 mgIL. The use of both polymers at dosage levels less than 5 mgIL for filtering tertiary pond effluent in the pilot plant was reviewed and approved (9). The polymers used in the sand filter are expected to be attached to the solids so that most of them would be associated with the backwash stream. The effluent stream'would then have a significantly lower concentration than the feed stream. Analytical efferts (10) are currently underway to identify the concentration of the polymer in the sand filter streams. *Trademark of The Dow Chemical Company 1 I. p.15 I: an a. IO Cua?hwu?n?urwru? .. a. Vazbnuwcoa . . at.? x: . II v. -23- FIGURE 11 AL Garnet Sand Screen LABORATORY SAND FILTER -24- FIGURE 12 LABORATORY SAND FILTER POLYMER EVALUATION METHODS Procedure Backwash the sand filter test unit shown in Figure 11 for 5 minutes. Measure 1000 of fresh tertiary pond effluent into a beaker (enough efluent should be collected initially so that a series of evaluations can be made on.a uniform feed). Leave water in the bed above) when finished. Treat feed solution in the beaker with the type and amount of polymer desired. Thoroughly mix the polymer with the feed. Pour the feed into the funnel at the top of the bed while trying to maintain a liquid level of about 5" above the bed. The discharge flowrate is maintained at about 15 to 20 mls/min by a screw type hose clamp. Take samples of the effluent after each 1/3 of the feed has been filtered. Analyze samples and average the results. Record data and repeat procedure as necessary. Calculate the-effectiveness factor, E, for each polymer where (Z Solids Weight.Per Feed Solids Weight) or out/TSS in)l(polymer concentration). Flocculant Name PURIFLOC C-31 XE 67629 Ageflox WT40 XD30598.05 M-239 1195 1180 1158 SEPARAN A230 Flocculant Alum XD30484.0 *All liquids are water soluble Form* Liquid Liquid Liquid Emulsion Liquid Liquid Liquid Solid Solid Solid Emulsion -25.. TABLE 4 POLYMER EVALUATEDIIN THE Com?anx Dowell Dow CPS Chemical Dow Dowell Betz Betz Betz Dow Several Dow LABORATORY Other Cationic polyamine Polydadmac ?2 HEN CH3 CH 3 This is the material sold to Dowell and packaged as XE67629 I Cationic emulsion 0-31 plus quat group. Cationic, (10,000 Cationic, Cationic, >5 (1 E'mwt H'mwt Anionic Polyacrylamide H20 Cationic emulsion Polymer (Flocculant) PURIFLOC C-31 PURIFLOC C-31 PURIFLOC C-31 XE67629 XE67629 XE67629 XE67629 XD30598.05 M-239 M-239 1195 1195 1195 1180 1180 1158 1158 1158 SEPARAN AP30 Alum XD 30584.0 (Z Solids Removed)(Feed Load) -25- TABLE 5 LABORATORY POLYMER EVALUATION DATA Polymer Filtrate Feed 2 Solids Effectiveness* Load, TSS, ppm_ TSS, Removed Factor, 1 7 36 80.6 2902 3 3.5 _36 90.3 1084 5 2.2. 23? 92.1 516 0.5 4.7 36 86.9 6257 1 3.1 36 91.4 3290 1 12 60 80.0 4800 3 5.0 60 91.7 1834 12 37 67.6 2501 1 12 37 67.6 2501 5 5.0 37 35.1 260 1 17 60 71.7 4302 3 5.2 60 91.3 1826 5 3.8 60 93.7 1124 34 60 43.3 2598 5 7 60 .88.3 1060 3 Plugged 34 -- -- 1 Plugged ?34 -- -- 0.5 6.6 34 80.6 5481 3 28 59 52.5 1033 5 11 28 60.7 340 3 15 59 74.6 1467 13 LABORATORY PERFORMANCE OF CATIONIC POLYMERS PURIFLOC C-31 Flocculant XE67629 Agefloc WT40 XD30598.05 1195 1180 I I I 2500 3000 3500 4000 Effectiveness Factor For Cationic Polymers At A Dosage 0f 1 4500 5000 Solids Removal In Filter-23- 14 DIMINISHING FILTER PERFORMANCE WITH INCREASED POLYMER DOSAGE Polymer Concentration In Feed, -29- ED 3? . a. UNITED STATES ENVIRONMENTAL PROTECTION AGENCY WASHINGTON. D.C. 20460 2 9 1982 a? ?013? .9 q? (I Dr. David C. Armbruster, Manager EIGm?zls Commercial Development ENVIRONMENTAL INFORMATION ON AGEFLOC WT4O CPS Chemical Company, Inc. P.O. Box 162 Old Bridge, New Jersey 08857 Dear Dr. Armbruster: Based on the information submitted, the coagulant aid listed below is acceptable for potable water application when used within the stated rates. Maximum Recommended Product Concentration Agefloc - 20 mg/l We would not anticipate any adverse health effect resulting from use of this product when used at or below the requested feed rates and assuming the product continues to meet the supplied specifications. We are currently in the process of revising our evaluation procedures as outlined in the Federal Register, Vo1. 44, No. 141, 42775?8, July 20, 1979} When these revisions are completed and the interim procedures are in place, all ex? isting advisories will be periodically reviewed. Our opinion concerning the safety of the product does not con- stitute an endorsement, nor does it relate to its effectiveness for the intended use. If this letter is to be used in any way, we require that it be quoted in its entirety. . Sincere uHanson, P.E., Acting Chief Additives Evaluation Branch Criteria and Standards Division Office of Drinking Water cc: Regional Drinking Water Representatives Hr. Lowell VanDenBerg, Technical Support Division, ODW Holders of the'Water Supply Guidance Series Mr. John Trax, State Programs Division, ODW -.- a -30- RESULTS OF SAND FILTER OPERATION The sand filter pilot plant described in an earlier section of this report was started up on October 18, 1983. After the media had been loaded into the filter, the filter was filled with water, and the media was "aged" for 12 hours so that all the particle surfaces would be well wetted. The media was then backwashed 4 times to remove fines. Fines removal is necessary to prevent plugging of the underdrain. The usual minor difficulties occurred with the equipment and installation but were. quickly remedied. The system was operated without polymer. The operation was poor with less than 20% of the solids being removed, so on November 3, 1983, polymer addition was begun. Figure 16 shows the dramatic effect that the addition of polymer had on filter performance. The-filter was Operated using either PURIFLOC 6-31 flocculant or Agefloc WTAO until being shutdown for several winter months on'December 21, 1983 (the entire pro- cess had been enclosed in a heated temporary building since the middle of November)._ During the weeks of operation, the filter system operated automatically at rates of 20-30 in 6 to 12 hour cycles. Over 2 million gallons of tertiary pond effluent were filtered in these 63 days of operation, which included about 200 backwash sequences (Table 3). Numerous samples were taken and analyzed to evaluate filter performance. A summary of the data is presented in Table 6. A notable trend in this data is the rapid decrease with time in T88 in the tertiary pond effluent (filter feed concentration), see Figure 17. The high solids.concentration in late November was due to high winds stirring up the pond. When the pond froze over, the wind effect was eliminated, and the T53 levels became very low. Performance data for the filter when operating floccu- lant and the high feed solids concentration, Figures 18, 19, 20, and 21, show that there is virtually no interaction between the concentration of suspended solids in the feed and the filtrate. Production plant designers and operators will recognize and appreciate this trait. These data also demonstrate the reason for labeling the ordinate "Total Suspended Solids By HIAC, mg/L" instead of "Removal Efficiency Of Filter, 2 Of Feed Solids In Product". RemOval.efficiencies change as the feed solids concentra- tion changes. The actual solids concentration in the filtrate is the number that represents what the filter is expected to do. A high removal efficiency does not necessarily guarantee low TSS in the filtrate stream.- The data on extended sampling runs of 7 (Figure.l9) and 4 (Figure 20) days show that the sand filter does continually provide product with low TSS. Figure 18 shows the filter performance for the first filtration cycle with polymer. When this-data is compared to the data in Figure 19, an improvement in performance.as measured by the T35 of the product is evi- dent. This improvement is probably due to additional polymer coating of sand grains. 9 7""7 Total Suspended Solids By HIAC, mg/L 48 40 32 24 16 FIGURE 16 EFFECT OF POLYMER 0N SAND FILTER.PERFORMANCE Polymer PURIFLOC Flocculant - At 1.1 In Feed Feed I .. I I I Filtrate Filtration Time, Min . -.-. 21 23 24 25 26 Flux Rate gpm/ft 8.06 8.06 9.68 11.29 9.68 8.06 8.06 9.68 Area-= 3.1 ft AVERAGE STREAM -32- TABLE 6 FIDTER PERFORMANCE DATA SUMMARY 2) PURIFLOC c-31 Flocculant; A Agefloc wrao 1) Feed Stream CONCENTRATION POLYMER Feed Filtrate Cone mg/L _mg/L Type Filter Operation 45.7 5.08 1.1 First time polymer was used on media 31.0 2.99 .l.l Continuation of above for 7 days 26.3 3.33 1.2 System was using polymer (above), was backwashed, and operated at new rate for 4 days 26.7 3.51 1.4 System had been operating for over 12 hours at set conditions before sampling ?15.35 2.37 1.2 System was on-A, was backwashed, and switched to for demonstration run 5.47 -2.37 A 1.07 :System ran with no polymer for a day before starting polymer 6.16 2.31 A 0.63 'System was running with A, the polymer rate was changed, and samples were taken 4.32 2.41 0.79 System was operating on polymer, was backwashed, and operated at new rate Total Suspended Solids By HIAC, mg-331 FIGURE 17 SEASONAL VARIATION OF SOLIDS CONCENTRATION IN THE FILTER FEED 1 I I 1 I 11?22 11-26 11?30 12-4 12-8 Time, Days In 1983 *Span is shown when three or more samples were available 12-12 12?16 Filtration Time, Hours 10 12 In 40 .. Filter Differential Pressure, Lbs/1112 I-I O5 45 8 0 50 Filtrate l? I I I I .34- FIGURE 18 FILTER PERFORMANCE AT 25 GPM WITH 1.1 PPM PURIFLOC 0-31 FLOCCULANT I Filter Differential Pressure, Lbs/1.112 Total Suspended Solids By HIAC, mgFIGURE 19 FILTER PERFORMANCE AT 25 PURIFLOC C-31 FLOCCULANT GPM WITH 1.1 PPM -- MULTIPLE CYCLES Filtration Time, Hours AHA Feed Filtrate Filtration Time, Hours 10 12 - (3 (I) 'Total Suspended Solids By HIAC, mg/L 40 - 50- Filtrate 24 AA -35- FIGURE 20 PURIFLOC C-31 - MULTIPLE CYCLES FILTER PERFORMANCE AT 30 GPM WITH 1.2 PPM Total SuSpended Solids By HIAC, mg/L Filter Differential Pressure, Lbs/in2 DFeed Filtrate 0 COO 0000 000 -37- FIGURE 21 FILTER PERFORMANCE 4T 35 GPM WITH 1.4 PPM PURIFLOC C-?l FLOCCULANT Filtration Time, Hours - - - -33- Figure 22, using data from Figures 19, 20, and 21, shows the effect of the flow-rate on the filter performance. The TSS in the filtrate rises with increasing rate due to the shearing effect of the flow past solids dropped in the sand grains. The initial pressure drop across the bed also goes up as the flow increases. Sand filter vendors (5) say that typical design flux rates are 8-12 gpm/ft2_ Low rates give the lowest pressure-drop and solids concentration in the filtrate but require the largest filters (most capitar). Another-set ofufilter performance data (Figure 23) was collected using PURIFLOC C-3l flocculant in a feed stream containing lower concentrations of solids as part of a special run. The sand filter had been Operating at other conditions before the decision was made to make a run at the best known conditions at that time. The conditions noted on Figure 23 and in Table 6 were.used, samples were taken regularly, and an excellent reproduction of previous operation was attained. This demonstration plus the multiple cycle data (Figures l9 and 20) provided design confidence in the sand filter using PURIFLOC flocculant. The next series of filter performance data were taken using Agefldc WT40. This polymer had provided the best performance in the laboratory screens ing previously discussed in this report. Figure 24 shows the filter per- formance at 25 using this polymer. When these performance data are compared to.the performance data for PURIFLOC polymer at the same conditions, Figure 19, several conclusions are evident. The.initial filter pressure is the same (this result is expected for the same flow rates). The average feed TSS concentration is significantly less due to the seasonal effect (Figure 17). This low loading rate almost eliminates the pressure buildup over the cycle shown. The solids concentration in the product is lower as the laboratory performance data would predict. However, the differences in feed concentration may affect this result. The data presented in Figure 25 show that filter performance can be main- tained when the polymer concentration in the feed is reduced as feeds solids concentration decreases. Polymer costs would be reduced if this operating discipline was followed. In Figure 26, the performance of the filter at the low feed TSS levels is shown with PURIFLOC C-31 flocculant. Comparison of these data with the data in Figure 20 shows that the feed solids concentration does affect the T35 in the filtrate. Thus, future filter operation using Agefloc HEAD with higher solids concentration is necessary to prove the field effectiveness of this polymer. _a 3a Total Snapended Solids Filtrate Initial Pressure Dr0p Across Bed, Lbs/in2 By HIAG, mgEFFECT OF FLUX RATE 9N SAND FILTER PERFORMANCE USING PURIFLOC C-3l FLOCCULANT Filter Flux Rate, gpm/ft2 13 Filter Differential Pressure, Lbs/in2 'Total Suspended Solids By HIAC, mgFIGURE 23 FILTER PERFORMANCE AI 30 GPM WITH 1.2 PPM PUEIELCC C-31 FLOCCULANT ON A SPECIAL SINGLE CYCLE -Feed - Filtrate Filter Time, Hours 'Filter Differential Pressure, Total Suspended Solids By HIAC, mg-41J FIGUREI 24 FILTER PERFORMANCE AT 25 QPM WITH 1.07 PPM AGEFLOC WT40 - CYCLES Feed Filtrate Filtration Time, Hours Filter Differential Pressure, Lbsl?m2 Total Suspended Solids By HIAC, rug-42- FIGURE 25 FILTER PERFORMANCE AT 25 GPM WITH 0. 63 PPM AGEFLOC WT40 Feed 0 Filtrate m-O El cn --C) 10 12 Filtration Time, Hours Filter Differential Pressure, Lbs/in2 Total SuSpended Solids By HIAC, mg-45- FIGURE 2 6 FILTER PERFORMANCE AT 30 GPM WITH 0.79 PPM PURIFLOC cl31 FLOCCULANT I I I Fil trat ion Time, Hours A: AA 13 AS DFeed (3 Filtrate El DD 0 0000 _44 TABLE 7 A.SAND FILTER MATERIAL BALANCE EXAMPLE CALCULATION FEED: 14,000 of water, TSS 35 mg/L FILTRAIE: TSS 3 mg/L Polymer - 1.2 PLURIFLOC 0-31 Flocculant SOLIDS IN: 5884.7 lbs/day SOLIDS OUT: 504.4 lbs/day SOLIDS IN BACKHASH FOR DISPOSAL: 5380.3 lbs/day, Removal 91% POLYMER USE: 201.3 lbs/day FILTER SIZE: (14,000 gpm)l(10 3.73 Filters Four Filters Give Enough For Baokflush Time BACKWASH WATER: (3 min/cycle) (16 gpm/ft )(4)(375 ft a 432,000 gal/day 45 SAND FILTER DESIGN CONSIDERATIONS During Operation of the pilot plant sand filter, observations were made on what was happening or what could happen and.how the design of a pro- duction unit would be affected. Some-of these observations that are dis? cussed in previous sections of this report are 1) the use of the HIAC as a quick analytical method for TSS in the product, 2) the pressure.at the top of the sand bed, 3) the differential pressure across the bed,.4) initial aging and flushing of the media, 5) the need for excellent flow distribu- tion at the top and bottomfof the filter, and 6) the steps of a filter cycle. - - . The shut-off valving for the air must be designed so that.there is absolutely no way for forward air leaks or backward water leaks. A double block and bleed system are necessary. If the water leaks backward, the air supply system will be disrupted. Block valves followed by check valves are a useless piping design for this application. Forward leaks of air during the backwash cycle when water is flowing upward through the sand bed cause high sand carryOVer from the bed. Sand loss reduces filter performance and increases operating cost. The backwash sequence in the pilot plant consisted of two cycles. The first cycle contained over 700 mg/L of TSS, while?the second cycle con- tained less than 100 mg/L (analysis of one cycle). The liquid volumes of the backwash could be reduced if Only the mOre concentrated material was recycled for treatment. The dilute material could possibly be returned to-the filter feed pond. Further sampling will provide more data to confirm this result. An example of material balance calculations for sand filtration of a . defined strEam are presented in Table 7. These calculations show that four 375 ft filters would be needed'for a 14,000 stream contain- ing 35 mg/L of TSS for a 91% removal. The backwash stream for disposal - would have a daily composite of 432,000 gallons of water and 5380.3 lbs. of solids (ppm 1493). Disposal of backwash water and solids from sand filters of tertiary pond effluent is a significant design consideration that was not covered by the scape of the research in this report. I REFERENCES I 1. Anderson, J., Private Communication, Midland, Michigan,-0ctober, 1983. 2- Weber, Halter J., et.al., "Sorption 0f Hydrophobic Compounds By Sedi- ments, Soils, And Suspended Solids - Water Research, Volume 17, No. 10, P33. 1443-1452, 1983. . . me (3., mun. . .I .. w. r'e-ulm, ml.? . .-.. ,.5551 10. -46- Product Literature, Serck Baker Inc., 5352 Research Drive, Huntington Beach, California, 92649. -. . - - Ives, K. J. (Editor), "The Scientific Basis 0f Fiocculation", Sijthoff And Noordhoff, The Netherlands, 1978. Hunter, Garry, Manager 0f Filter Systems, Serck Baker Inc., Private- Communication,_September 8, 1983. - Peck, C. N. and Blanchard, H. R., Particle Size Analysis Of Tertiary Pond Effluent", Dow CRI Report AL-83-30452, November 28,. 1983. Purchas, D. B. (Editor), "Solids/Liquid Separation Equipment Scale- Up", p. 295, Uplands Press LTD, Cryoden, England, 1977. Dill, D. C. and Applegath, S. L., "Evaluation Of The Toxicity 0f PURIFLOC Flocculant To Representative Aquatic Organisms, Dow CRI Report ES-DR-0056-3542-2, October 19, 1983. Anderson, J., Private Communication, November 11 and 21, 1983. Melcher, R. and Break, K., Private Communication, December 16, 1983. I 4