USEEA/PAFER INDUSTRY COOPERATIVE ;DIQXIK STUDY "THE.Jd04.MILL :STUDY'' • STATISTICAL.-FINDINGSAND ANALYSES; , ,1990 . p ;,Watiepl^egul4t.i.,pris.;?indrSpandards "40L M' S.tree,t,,':S.W.' Washington,, D.c". - 20460 ------- USEPA/PAPER INDUSTRY COOPERATIVE DIOXIN STUDY "THE 104 MILL STUDY- STATISTICAL FINDINGS AND ANALYSES EXECUTIVE SUMMARY This report describes statistical analyses of the data from the "104 Mill Study." This study was the result of a cooperative agreement between EPA and the U.S. paper industry. The purpose of the study was to characterize the 104 U.S. mills that practiced chlorine bleaching of chemically produced pulps in mid to late 1988. The scope of the study was developed by EPA and industry, and the study was managed by the National Council of the Paper Industry for Air and Stream Improvement, Inc. (NCASI), with EPA overview. The data collected included measurements of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) and 2,3,7,8- tetrachlorodibenzofuran (TCDF) concentrations in three export vectors (pulp, sludge, and effluent); and information on wastewater treatment, bleaching, and Biinufacturing processes. More information was available for kraft mills (155 sach lines) than sulfite (18 bleach lines); therefore, some statistical findings are reported for only kraft mills. The statistical findings are: 1. The detected concentration values of TCDD/TCDF were best approximated by lognormal distributions, estimated separately for each of the export vectors: pulp, sludge, and effluent. 2. Analysis of field and laboratory duplicates indicated excellent agreement between duplicate measurements of TCDD/TCDF concentrations. As a consequence, analytical measurement variability is a very small portion of the total variability in the TCDD/TCDF data. 3. The reported detection levels for the non-detected measurements of TCDD/TCDF demonstrate that the target detection level of 10 parts per quadrillion (ppq) for effluent measurements is achievable. Estimates of the daily total mass output rates of TCDD/TCDF at U.S. cached pulp mills were 0.004 Ibs/day for TCDD and 0.032 Ibs/day for TCDF. ------- Output rates for individual mills varied substantially; however, the per averages were 0.00005 Ibs of TCDD and 0.00048 Ibs of TCDF exported dai,. pulp, sludge, and treated effluent. 5. The relative amounts of TCDD/TCDF partitioned to each of the three export vectors (pulp, sludge, and effluent) were highly variable among mills. 6. Significantly more TCDD/TCDF was exported at kraft mills than sulfite mills. 7. Mills using Activated Sludge (ACT) wastewater treatment systems exported somewhat less effluent-based TCDD/TCDF mass on average and significantly more sludge-based TCDD/TCDF mass than mills using Aerated Stabilization Basins (ASB). The difference in sludge exports can be partially attributed to the fact that ACT sludge samples in the 104 Hill Study consisted of combined primary and secondary sludges. Those from ASB systems consisted only of primary sludge. 8. Total Suspended Solids (TSS) concentrations in ACT systems was found be significantly higher than the TSS concentrations of ASB systems at k mills. 9. When ACT and ASB-type kraft mills were combined, a weakly correlated positive trend was observed between effluent TCDD/TCDF and TSS levels, and a weakly correlated negative trend was observed between TSS and sludge TCDD/TCDF. For kraft mills using only ACT treatment, higher TSS levels were associated with higher sludge-based TCDD/TCDF exports but lower effluent-based TCDD/TCDF exports. 10. Linear regressions of the TCDD/TCDF export rates fit co bleaching measures at each mill (including application rates of bleaching and chemical extraction agents) were found to be poor predictors of individual kraft mill outputs. 11. Greater chlorine usage in kraft mills was found to be statistically associated with higher formation rates of TCDD/TCDF. 12. Increased substitution of chlorine dioxide for chlorine in the C-stag< kraft mills was correlated with slight reductions in TCDD/TCDF formation. ------- 13. Higher chlorine multiples during C-stage bleaching were weakly associated with higher TCDD/TCDF mass formation in kraft mills. 14. Kraft mills that used oxygen delignification in the bleaching process exhibited somewhat lower rates of TCDD/TCDF formation than mills that did not use such methods. ------- TABLE OF CONTENTS PAGE 1. INTRODUCTION 1 1.1 STUDY FEATURES 2 1.1.1 Field Sampling Program 3 1.1.2 Analytical Program 4 1.1.3 Data Handling 5 1.2 INDUSTRY PROFILE 5 1.2.1 Pulping and Bleaching 6 1.2.2 Bleach Line Chemical Usage 6 1.2.3 Uastewater Treatment 11 2. SUMMARY OF STATISTICAL FINDINGS 13 2.1 CHARACTERIZING TCDD/TCDF CONCENTRATION DATA 13 2.2 VARIABILITY IN DUPLICATE SAMPLE ANALYSES 14 2.3 DETECTION LEVELS FOR NON-DETECTED MEASUREMENTS 14- 2.4 TOTAL MASS FORMATION ESTIMATES OF TCDD/TCDF 14 2.5 VARIABILITY IN PARTITIONING OF TCDD/TCDF TO 14 DIFFERENT EXPORT MATRICES 2.6 DIFFERENCES DUE TO PULPING AND WASTEUATER TREATMENT 15 2.7 RELATIONSHIPS BETWEEN WASTEWATER TREATMENT AND TOTAL 15 SUSPENDED SOLIDS 2.8 RELATIONSHIPS BETWEEN TCDD/TCDF FORMATION AND MILL 15 OPERATING CHARACTERISTICS 2.9 EFFECTS OF CHLORINE APPLICATION IN PRE-BLEACHING 16 2.10 EFFECT OF THE CHLORINE MULTIPLE 16 2.11 USE OF OXYGEN IN THE BLEACHING PROCESS 16 2.12 DIFFERENCES IN WOOD TYPES 17 3. CHARACTERIZATION OF THE TCDD/TCDF CONCENTRATION DATA 18 3.1 VARIABILITY IN DETECTION LEVELS 18 3.2 FITTING OF DETECTED CONCENTRATIONS 20 4. ANALYSIS OF FIELD AND LAB DUPLICATE SAMPLES 28 4.1 CORRELATIONS BETWEEN DUPLICATE PAIRS 28 4.2 ANALYSIS OF DUPLICATE SAMPLE VARIABILITY 30 5. PARTITIONING OF TCDD/TCDF MASSES INTO EXPORT MATRICES 52 5.1 VARIABILITY ACROSS EXPORT VECTORS 52 5.2 KRAFT VERSUS SULFITE MILLS 53 5.3 ACT VERSUS ASB WASTEWATER TREATMENT 55 5.4 OVERALL.PARTITIONING OF TCDD/TCDF 56 6. ANALYSIS OF TOTAL SUSPENDED SOLIDS 84 ------- TABLE OF CONTENTS (Continued) 7. MODELING TCDD/F FORMATION AS A FUNCTION OF MILL OPERATING PARAMETERS 7.1 REGRESSION ANALYSES 7.1.1 Effects of Chlorine Bleaching 7.1.2 Effect of the Chlorine Multiple 7.1.3 Chlorine Dioxide Substitution 7.1.4 Use of Oxygen in Bleaching 7.1.5 Differences in Wood Types 7.2 SUMMARY APPENDIX A: DATA LISTINGS APPENDIX B: PROBABILITY PLOTS REFERENCES LIST OF TABLES TABLE 1-1 1-2 1-3 1-4 1-5 1-6 3-1 3-2 3-3 4-1 4-2 4-3 5-1 5-2 5-3 5-4 5-5 5-6 5-7 5-8 6-1 6-2 6-3 6-4 6-5 6-6 6-7 7-1 INDUSTRY PROFILE - PULPING INDUSTRY PROFILE - BLEACHING INDUSTRY PROFILE - BLEACH LINE CHEMICAL USAGE STATUS OF U.S. BLEACHERY OPERATIONS: C-STAGE CHLORINATION AND CHLORINE DIOXIDE SUBSTITUTION C-STAGE CHLORINE MULTIPLE (KAPPA FACTOR) INDUSTRY PROFILE - WASTEWATER TREATMENT DETECTION LEVELS FOR NON-DETECT SAMPLES DESCRIPTIVE STATISTICS FOR TCDD CONCENTRATION DESCRIPTIVE STATISTICS FOR TCDF CONCENTRATION PEARSON CORRELATIONS BETWEEN DUPLICATE PAIRS ANOVA TABLE FOR LAB DUPLICATES ANOVA TABLE FOR FIELD DUPLICATES DESCRIPTIVE STATISTICS FOR DIOXIN DESCRIPTIVE STATISTICS FOR DIOXIN (BY WASTEWATER TREATMENT) DESCRIPTIVE STATISTICS FOR FURAN DESCRIPTIVE STATISTICS FOR FURAN (BY WASTEWATER TREATMENT) DIFFERENCES BETWEEN PULPING PROCESSES DIFFERENCES BETWEEN TREATMENT TYPES STATISTICS FOR TCDD/TCDF (BY MILL PROCESS) STATISTICS FOR TCDD/TCDF (BY MILL PROCESS) DESCRIPTIVE STATISTICS FOR TSS TCDD EXPORTS (TREATED KRAFT MILLS ONLY) TCDF EXPORTS (TREATED KRAFT MILLS ONLY) TCDD EXPORTS FOR ACT TREATMENT/KRAFT MILLS ONLY TCDF EXPORTS FOR ACT TREATMENT/KRAFT MILLS ONLY TCDD EXPORTS FOR ASB TREATMENT/KRAFT MILLS ONLY TCDF EXPORTS FOR ASB TREATMENT/KRAFT MILLS ONLY SUMMARY STATISTICS: BREAKDOWN BY CIO, USAGE 104 105 106 107 107 107 108 108 126 144 164 PAGE 7 7 8 9 10 12 19 24 26 31 33 34 58 59 60 61 66 76 82 83 89 94 95 100 101 102 103 110 ------- LIST OF TABLES (Continued) TABLE PAGE 7-2 SUMMARY STATISTICS: BREAKDOWN BY 02 USAGE 111 7-3 REGRESSIONS OF CHLORINE USAGE (KRAFT MILLS ONLY) 121 7-4 REGRESSIONS OF CHLORINE MULTIPLE (KRAFT MILLS ONLY) 122 7-5 REGRESSIONS OF C102 SUBSTITUTION (KRAFT MILLS ONLY) 123 A-l 104 MILL DATA LISTING 127 A-2 TCDD/TCDF CONCENTRATION DATA 129 A-3 TCDD/TCDF FIELD DUPLICATES 139 A-4 TCDD/TCDF LAB DUPLICATES 141 LIST OF FIGURES FIGURE PAGE 3-1 SAMPLE CUMULATIVE DISTRIBUTION GRAPH: EFFLUENT TCDD 21 DETECTION LEVELS 3-2 SAMPLE CUMULATIVE DISTRIBUTION GRAPH: EFFLUENT TCDF 22 DETECTION LEVELS 4-1 PULP FIELD DUPLICATES/TCDD 36- 4-2 PULP FIELD DUPLICATES/TCDF 37 4-3 PULP LAB DUPLICATES/TCDD 38 4-4 PULP LAB DUPLICATES/TCDF 39 4-5 SLUDGE FIELD DUPLICATES/TCDD 4.0 4-6 SLUDGE FIELD DUPLICATES/TCDF 41 4-7 SLUDGE LAB DUPLICATES/TCDD 42 4-8 SLUDGE LAB DUPLICATES/TCDF 43 4-9 EFFLUENT FIELD DUPLICATES/TCDD 44 4-10 EFFLUENT FIELD DUPLICATES/TCDF 45 4-11 EFFLUENT LAB DUPLICATES/TCDD 46 4-12 EFFLUENT LAB DUPLICATES/TCDF 47 4-13 EFFLUENT LAB DUPLICATES: KRAFT MILLS ONLY/TCDD 48 4-14 EFFLUENT LAB DUPLICATES: SULFITE MILLS ONLY/TCDD 49 4-15 EFFLUENT LAB DUPLICATES: KRAFT MILLS ONLY/TCDF 50 4-16 EFFLUENT LAB DUPLICATES: SULFITE MILLS ONLY/TCDF 51 5-1 PERCENT OUTPUT BY MATRIX/TCDD 62 5-2 PERCENT OUTPUT BY MATRIX/TCDF 63 5-3 ADJUSTED TCDD BY MATRIX 64 5-4 ADJUSTED TCDF BY MATRIX 65 5-5 PERCENT OUTPUT BY TREATMENT/EFFLUENT TCDD 68 5-6 PERCENT OUTPUT BY TREATMENT/SLUDGE TCDD 69 5-7 PERCENT OUTPUT BY TREATMENT/EFFLUENT TCDF 70 5-8 PERCENT OUTPUT BY TREATMENT/SLUDGE TCDF 71 5-9 ADJUSTED EFFLUENT TCDD 72 5-10 ADJUSTED SLUDGE TCDD 73 5-11 ADJUSTED EFFLUENT TCDF 74 5-12 ADJUSTED SLUDGE TCDF 75 5-13 TOTAL TCDD EXPORTS (Ibs/day * E+06) 78 5-14 TOTAL OUTPUT: TCDD (KRAFT AND SULFITE MILLS) 79 5-15 TOTAL TCDF EXPORTS (Ibs/day * E+06) 80 5-16 TOTAL OUTPUT: TCDF (KRAFT AND SULFITE MILLS) 81 ------- LIST OF FIGURES (Continued) FIGURE PACE 6-1 TSS BY TREATMENT 88 6-2 EFFLUENT TCDD OUTPUT 90 6-3 SLUDGE TCDD OUTPUT 91 6-4 EFFLUENT TCDF OUTPUT 92 6-5 SLUDGE TCDF OUTPUT 93 6-6 EFFLUENT TCDD OUTPUT BY TREATMENT 96 6-7 SLUDGE TCDD OUTPUT BY TREATMENT 97 6-8 EFFLUENT TCDF OUTPUT BY TREATMENT 98 6-9 SLUDGE TCDF OUTPUT BY TREATMENT 99 7-1 C12 vs. ADJUSTED TOTAL TCDD 112 7-2 C12 vs. ADJUSTED TOTAL TCDF 113 7-3 C12 vs. ADJUSTED TCDD TOXIC EQUIVALENT 114 7-4 Clz MULTIPLE vs. ADJUSTED TOTAL TCDD 115 7-5 Cla MULTIPLE vs. ADJUSTED TOTAL TCDF 116 7-6 C12 MULTIPLE vs. ADJUSTED TCDD TOXIC EQUIVALENT 117 7-7 PERCENT C102 SUBSTITUTION vs. ADJUSTED TOTAL TCDD 118 7-8 PERCENT C102 SUBSTITUTION vs. ADJUSTED TOTAL TCDF 119 7-9 PERCENT C10Z SUBSTITUTION vs. TCDD TOXIC EQUIVALENT 120 7-10 Clj vs. ADJUSTED PULP TCDD 124 7-11 KAPPA # vs. ADJUSTED PULP TCDD 125 B-l PULP TCDD PROBABILITY PLOT: DETECTED VALUES ONLY 145 B-2 PULP TCDF PROBABILITY PLOT: DETECTED VALUES ONLY 146 B-3 SLUDGE TCDD PROBABILITY PLOT: DETECTED VALUES ONLY 147 B-4 SLUDGE TCDF PROBABILITY PLOT: DETECTED VALUES ONLY 148 B-5 EFFLUENT TCDD PROBABILITY PLOT: DETECTED VALUES ONLY 149 B-6 EFFLUENT TCDF PROBABILITY PLOT: DETECTED VALUES ONLY 150 B-7 PULP TCDD PROBABILITY PLOT 151 B-8 PULP TCDF PROBABILITY PLOT 152 B-9 SLUDGE TCDD PROBABILITY PLOT 153 B-10 SLUDGE TCDF PROBABILITY PLOT 154 B-ll EFFLUENT TCDD PROBABILITY PLOT 155 B-12 EFFLUENT TCDF PROBABILITY PLOT 156 B-13 ADJUSTED PULP TCDD PROBABILITY PLOT 157 B-14 ADJUSTED PULP TCDF PROBABILITY PLOT 158 B-15 ADJUSTED SLUDGE TCDD PROBABILITY PLOT 159 B-16 ADJUSTED SLUDGE TCDF PROBABILITY PLOT 160 B-17 ADJUSTED EFFLUENT TCDD PROBABILITY PLOT 161 B-18 ADJUSTED EFFLUENT TCDF PROBABILITY PLOT 162 B-19 TSS PROBABILITY PLOT 163 ------- 1. INTRODUCTION In October 1987, the U.S. Environmental Protection Agency (EPA) and the U.S. Pulp and Paper Industry jointly released preliminary results from a screening study that provided the first comprehensive results on the formation and discharge of chlorinated dibenzo-p-dioxins (CDDs) and dibenzofurans (CDFs) from pulp and paper mills (1). This screening study of five bleached kraft mills ("Five Mill Study") confirmed that the pulp bleaching process was primarily responsible for the formation of CDDs and CDFs. The partitioning of these compounds between the bleached pulp, wastewater treatment sludge, and final wastewater effluent was found to be highly variable among the mills. The study results also indicated that 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) and 2,3,7,8-tetrachlorodibenzofuran (TCDF) were the principal CDDs and CDFs formed. The final Five Mill Study report was published in March 1988 (2). To provide EPA with more complete data on the release of these compounds by the U.S. paper industry, an agreement was reached in April 1988 between EPA and the industry to conduct a second study to characterize the 104 U.S. mills that practiced chlorine bleaching of chemically produced pulps (3). The scope of the study was developed by EPA and industry, and the study was managed by the National Council of the Paper Industry for Air and Stream Improvement, Inc. (NCASI), with EPA overview. The data from this study provided an estimate of the release of TCDD and TCDF in three environmental export vectors (i.e., bleached pulp, sludge, and effluent) from the U.S. Pulp and Paper Industry as of mid- to late 1988. This section presents the major features of the study design, including the field sampling program, the analytical program, and data handling; and a profile of the industry at the time the study was conducted, comprising pulping and bleaching characteristics, bleach line chemical usage during sampling, and wastewater treatment. ------- The remainder of the report provides details of the statistical analyses and study results, and consists of the following sections: • Section 2, summary of the findings • Section 3, characterization of the TCDD/TCDF concentration data • Section 4, analysis of duplicate samples • Section 5, partitioning of TCDD/TCDF mass rates into mill exports • Section 6, analysis of total suspended solids • Section 7, modeling of TCDD/TCDF formation in terras of mill operating parameters A listing of the data used in the analyses is also provided in appendix A. This report and a separate summary document were prepared independently by EPA. The paper industry, through NCASI, has also prepared a report of the 104 Mill Study (4). Preliminary study results were presented by EPA and NCASI in September 1989 (5) and will be published in Chemosphere. This report includes data received by EPA from NCASI as of April 1990 and comprises more than 98 percent of the data required by the study objectives. When reviewing the study results, it is important to keep in mind that the principal objective of the 104 Mill Study was to characterize exports from the 104 mills in terms of TCDD and TCDF. The study was not designed to address mechanisms of formation of these compounds or to determine the best technologies for treating these compounds in wastewaters. Nonetheless, the study results permit some useful observations in these areas as well. 1.1 STUDY FEATURES All U.S. pulp and paper mills where chemically produced wood pulps are bleached with chlorine and chlorine derivatives were included in the Agreement for the 104 Mill Study (3). Although mills included in the Five Mill Study were not resampled for the 104 Mill Study, TCDD/TCDF data and mill operating and wastewater treatment information from the Five Mill Study have been included in ------- this analysis. Consolidated Paper independently conducted a study at its Wisconsin Rapids, Wisconsin mill. Due to differences in sampling and analytical protocols, the data for TCDD/TCDF from this mill were not included. However, mill characteristics and wastewater treatment information for Consolidated Paper are included in the industry profile presented in subsection 1.2. 1.1.1 Field Sampling Program The Agreement for the 104 Mill Study required that each significant export vector (fully bleached pulp, wastewater sludge, and final wastewater effluent) be sampled and that the samples be composited over a S-day period (3). In most cases, the composite samples consisted of up to eight aliquots obtained throughout the sampling day. Nearly all sampling was performed by mill personnel following guidance established by NCASI. In a few cases, NCASI personnel conducted the sampling. The sampling protocols closely followed those established for the Five Mill Study (2). The pulp samples taken were of the highest brightness pulp produced at each bleach line. At mills with two bleach lines where hardwood and softwood pulps are bleached separately, separate hardwood and softwood composite pulp samples were collected. At mills with a single bleach line where both hardwood and softwood pulps are bleached (i.e., a swing line), sampling was conducted intermittently to ensure that the 5-day composite samples were composed only of hardwood or softwood pulp. A few bleach, lines processed mixtures of hardwood and softwood pulps. The composite samples from these lines were classified by the percent of softwood pulp in the mixture. Sludge samples consisted only of those sludges removed from the wastewater treatment system and disposed of in landfills, by incineration, or by other methods. For mills with Activated Sludge Vastewater Treatment (ACT), the sludge samples generally consisted of combined primary and secondary sludge; for mills with Aerated Stabilization Basins (ASB), only primary sludges were sampled. In most cases, the sludges were dewatered prior to offsite disposal; however, several primary sludges were collected in a low consistency slurry form. ------- More than 90 sampled effluents were collected from mills with biological treatment. For eight mills, the samples consisted of partially treated effluents prior to discharge to municipal wastevater treatment plants. Two mills with direct ocean discharges provided samples of untreated effluents. Another untreated effluent was sampled at a mill that used a percolation pond for wastewater disposal. This sampling scheme generated over 400 samples for isomer-specific TCDD and TCDF analyses. About 80 additional samples were collected as part of the quality assurance/quality control (QA/QC) plan. These samples were analyzed as field duplicates and/or included in native spike determinations. The data is listed in Appendix A. In addition, mill operators were required to provide process operating data for bleacheries and wastewater treatment plants. These data were collected to document operation of the processes at the time of sampling. 1.1.2 Analytical Program The Brehm Laboratory at Wright State University (USU), Dayton, Ohio, performed analytical methods development work for isomer-specific determinations of TCDD and TCDF in pulp and paper mill matrices and completed analyses of all samples for the Five Mill Study (2). Analytical work for the present study was conducted by Enseco-California Analytical Laboratories (CAL) in West Sacramento, California, and USU. Enseco-CAL conducted most of the sludge and effluent analyses, while WSU analyzed most of the pulp samples. The analytical methods used in the 104 Mill Study were consistent with the screening study protocols established for the Five Mill Study (2). Analytical objectives for target detection levels for TCDD and TCDF were 1 ng/kg (parts per trillion [ppt]) for sludges and pulps, and 0.01 ng/kg (ppt) for wastewater effluents. The Agreement specified identification and quantitation criteria for TCDD/TCDF and required that NCASI manage QA/QC programs for the study. NCASI staff performed and coordinated sample preparation, submitted samples to the analytical laboratory, and reviewed laboratory data reports. Nearly all analytical results met the QA/QC objectives established for the study. Several ------- samples required re-analysis to obtain valid data; however, the proportion of such samples was less than 6 percent of the total. 1.1.3 Data Handling To ensure consistent reporting of bleach plant and wastewater treatment information, NCASI developed specific forms for mill personnel to report bleach line operating characteristics, bleach line chemical applications, and wastewater treatment operations. Copies of these forms, as well as schematic diagrams of the bleacheries and wastewater treatment facilities, were provided to EPA by NCASI for most mills. For those few mills which requested confidential treatment of certain data, the forms were submitted directly to EPA by mill operators. NCASI submitted final analytical results to EPA as they were developed in conformance with the QA/QC protocols specified in the Agreement (3). EPA and NCASI independently developed data summaries in spreadsheet format to characterize bleach line operating characteristics; mass flow rates of bleached pulp, wastewater sludge, and wastewater effluent; and mass flows of TCDD and TCDF estimated in mill exports. The respective spreadsheet entries were compared several times and corrections made as appropriate. Prior to conducting detailed statistical analyses, EPA had a contractor further compare the spreadsheets against the original report forms. All discrepancies were resolved and the spreadsheets updated. New databases were then created by uploading the data from the spreadsheets to the EPA mainframe computer. 1.2 INDUSTRY PROFILE At the time the 104 Mill Study field program was underway (mid- to late 1988 for most mills), the U.S. Pulp and Paper Industry was characterized by limited application of those pulping and bleaching practices demonstrated to have the potential to reduce formation of TCDD/TCDF. Since that time, many mill operators have initiated programs to institute improved pulping and bleaching technologies and operating practices. This industry profile, however, does not reflect any changes made by U.S. paper mills since the end of 1988. ------- 1.2.1 Pulping and Bleaching Tables 1-1 and 1-2 present the industry profile for pulping and bleaching of those mills included in the study. This segment of the U.S. industry comprises 86 kraft pulping mills, 16 sulfite mills, 1 soda mill, and 1 mill with both kraft and sulfite pulping. More than half of the bleach lines at kraft mills are used for bleaching softwoods exclusively and 40 percent for bleaching hardwoods. The balance of the bleach lines are either swing lines or used to bleach hardwood/softwood pulp mixtures. For sulfite mills, half the bleach lines are used for softwood pulps, nearly 40 percent for hardwood pulps, and the balance for mixed pulps. 1.2.2 Bleach Line Chemical Usage Table 1-3 summarizes the number and percentage of bleach lines with oxygen delignification systems and other chemical usage in pre-bleaching and final bleaching. The data were provided by mill operators during the sampling surveys. During that period, the industry was characterized by low utilization of oxygen delignification, relatively low utilization of oxygen reinforced extraction, low utilization of peroxide reinforced extraction, and relatively high utilization of hypochlorite in both pre-bleaching and final bleaching. The status of bleachery operations in the U.S. industry in mid- to late 1988 with respect to chlorine usage and chlorine dioxide substitution is summarized in Table 1-4. Note that about 35 percent of the kraft mill bleach lines were operated with no chlorine dioxide in the C-Stage, and less than 2 percent of the kraft mill bleach lines had chlorine dioxide substitution rates greater than 50 percent. Table 1-5 presents a summary of chlorine multiples (Kappa factor) determined for kraft and sulfite bleach lines at the time of sampling. The chlorine multiple is the ratio of the amount of active chlorine used in pulp bleaching in the C-Stage to the amount of lignin contained in brownscock or oxygen delignified pulp as characterized by the Kappa number. Eleven percent ------- TABLE 1-1. INDUSTRY PROFILE - PULPING Type Kraft Sulfite Kraft and Sulftte Soda Total Number of Mills 86 16 1 1 104 TABLE 1-2. INDUSTRY PROFILE - BLEACHING Woodtvoe Hardwood Softwood Mixed HW/SW Total Number of Bleach Lines Kraft Sulfite Soda 67 7 1 89 9 9 2 165 18 Note: Kraft hardwood and softwood bleach line data include 14 swing lines counted as both hardwood and softwood lines. ------- TABLE 1-3. INDUSTRY PROFILE - BLEACH LINE CHEMICAL USAGE Chemical Usage Oxygen Delignification Pre-bleaching C-Stage C12 C-Stage C102 E-Stage 02 E-Stage NaOCl E-Stage H202 Final Bleaching C102 NaOCl Number of Bleach Lines (2) Kraft Sulfite Soda 7 (4.2) - ( 0) - ( 0) 165 105 78 47 2 147 90 25 (100) ( 64) ( 47) ( 28) (1.2) ( 89) ( 55) ( 15) 16 1 4 1 1 4 14 1 ( 89) (5.6) ( 22) (5.6) (5.6) ( 22) ( 78) (5.6) 1 1 1 1 (100) (100) (100) ( 0) ( 0) (100) ( 0) ( 0) ------- TABLE 1-4. STATUS OF U.S. BLEACHERY OPERATIONS: C-STAGE CHLORINATION AND CHLORINE DIOXIDE SUBSTITUTION Kraft Mill Bleach Lines Chlorine Application Lbs Cl,/Ton ADBSP Bleach Lines < 40 40-60 60-80 80-100 100-120 120-140 > 140 15 22 32 36 28 16 16 Substitution Percent Bleach Lines 0 < 5 5-10 10-20 20-30 30-40 40-50 50-60 60-70 > 70 59 16 41 33 9 1 3 1 1 1 TOTAL 165 TOTAL 165 Sulfite Mill Bleach Lines < 40 40-60 60-80 80-100 100-120 120-140 > 140 2 1 2 6 3 4 0 0 < 5 > 5 17 1 0 TOTAL 18 TOTAL 18 Notes: Bleachery operations for swing lines were counted twice, separately for hardwood and softwood pulps. ADBSP - Air-dried brownscock pulp. ------- TABLE 1-5. C-STAGE CHLORINE MULTIPLE (KAPPA FACTOR) Number of Bleach Lines Chlorine Multiple Sulfite 0.10 0.15 0.20 0.25 < 0.10 - < 0.15 - < 0.20 - < 0.25 - < 0.30 > 0.30 4 15 51 54 17 14 TOTAL 155 18 Notes: Chlorine multiple was computed from active chlorine (C12 and C102) applied in the C-Stage. Chlorine multiples could not be computed for 10 kraft mill bleach lines because of incomplete data. 10 ------- of che sampled bleach lines were operated with average chlorine multiples less than 0.15. 1.2.3 Wastewater Treatment The status of wastewater treatment provided at the 104 paper mills is summarized in Table 1-6. The industry standard consists of primary treatment followed by secondary biological treatment. Eight mills discharge to publicly owned treatment works (POTUs) after primary treatment, and two have no treatment. Uastewaters from one mill are disposed of in a percolation pond. About 35 percent of kraft mills have ACT and more than half have ASB. For sulfite mills, nearly 70Z have ACT while almost 20Z use ASB. 11 ------- TABLE 1-6. INDUSTRY PROFILE - WASTEWATER TREATMENT Treatment Type ACT ASB Discharge to POTW Discharge to Other Mill UWTP Percolation Pond No Treatment Kraft 32 45 7 1 2 TOTAL 87 Number of Mills Sulfite Soda 11 3 1 1 1 16 1 43 49 8 1 1 2 104 Note: The mill with kraft and sulfite pulping was listed as a kraft mill for purposes of this table. 12 ------- 2. SUMMARY OF STATISTICAL FINDINGS The following discussion summarizes Che statistical findings from the Mill Study of U.S. bleached pulp mills. The conclusions are necessarily limited in scope, due to the design of the study. More information was available for kraft mills than sulfite; therefore, some statistical findings are reported only for kraft mills. The results do provide, though, the basis for several useful observations. 2.1 CHARACTERIZING TCDD/TCDF CONCENTRATION DATA Examination of the laboratory analyses of samples collected at each mill indicated that the detected concentration values of 2,3,7,8-tetrachlorodibenzo- p-dioxin (TCDD) and 2,3, 7 ,8-tetrachlorodibenzofuran (TCDF) were best approximated by lognormal distributions, estimated separately for each of the export matrices -- pulp, sludge, and effluent. A number of non-detected measurements were also reported in the data. Analysis of the mass formation rates of TCDD/TCDF required that values be associated with these non-detects. For the purposes of this study, such measurements were assigned a value equal to half the detection level. This step allowed non-detect samples to be used in a reasonable and consistent manner without distorting the basic findings: (1) the vast majority of all samples had detectable concentrations, with only 15 percent of all TCDD samples and 4 percent of TCDF samples reported as non-detects, (2) the ratio of detectable levels of TCDF to TCDD was fairly consistent from mill to mill, yet less than 4 percent of all the samples were reported as non-detects for both TCDD and TCDF, (3) every mill was found to have detectable levels of TCDD/TCDF in at least one of the export vectors. Setting non-detect values to half the detection level also represented a compromise between underestimation (assigning non-detect values to zero) and overestimation (assigning non-detect values to the detection level) of the unknown actual concentrations. 13 ------- 2.2 VARIABILITY IN DUPLICATE SAMPLE ANALYSES Approximately 30 percent of all the samples were classified as field sample duplicates or lab duplicate splits. Analysis of these duplicate samples for each matrix (effluent, pulp, and sludge) indicated excellent agreement between duplicate measurements of TCDD/TCDF concentrations. Most sample correlations between pairs of duplicate measurements were found to above 0.95. Consequently, the proportion of total variability in TCDD/TCDF levels that could be attributed to field sampling protocol or analytical technique was in all cases small relative to other sources of variation. In the worst case observed, analytical measurement error was still less than 12 percent of the total variability in TCDF concentrations. 2.3 DETECTION LEVELS FOR NON-DETECTED MEASUREMENTS The reported detection levels for non-detected measurements of TCDD/TCDF demonstrate that the laboratories were capable of achieving the target detection levels of 10 parts per quadrillion (ppq) for effluent measurements. 2.4 TOTAL MASS FORMATION ESTIMATES OF TCDD/TCDF By combining Che TCDD/TCDF concentration data with mill production rates of pulp, sludge, and effluent, rates of TCDD/TCDF mass formation were computed for the export matrices at each mill. Estimates of the daily total mass output rates of TCDD/TCDF at U.S. bleached pulp mills were 0.004 Ibs/day for TCDD and 0.032 Ibs/day for TCDF. Output rates for individual mills varied substantially; however, the per mill averages were 0.00005 Ibs of TCDD and 0.00048 Ibs of TCDF exported daily in pulp, sludge, and treated effluent. 2.5 VARIABILITY IN PARTITIONING OF TCDD/TCDF TO DIFFERENT EXPORT MATRICES The relative amounts of TCDD/TCDF partitioned to pulp, sludge, or effluent vectors were not found to be consistent from mill to mill, but highly variable. While some mills partitioned less than 10 percent of their total TCDD/TCDF mass to effluent, effluent-based TCDD/TCDF accounted for more than 80 percent of the 14 ------- exports at other mills. The variability in partitioning of pulp and sludge export vectors was similar. Among the least extreme cases (middle 50 percent of all mills), the relative percentage of TCDD/TCDF exported to specific matrices differed by more than 30 percent from mill to mill. 2.6 DIFFERENCES DUE TO PULPING AND WASTEWATER TREATMENT Comparisons showed that significantly more TCDD/TCDF was exported at krafc mills than sulfite mills for each matrix type. Differences also emerged between wastewater treatment types Aerated Stabilization Basins (ASB) and Activated Sludge Wastewater Treatment (ACT). There was evidence that mills using ACT exported somewhat less effluent-based TCDD/TCDF mass on average and significantly more sludge-based TCDD/TCDF mass than mills using ASB systems. The difference in sludge exports can be partially attributed to the fact that ACT sludge samples in the 104 Mill Study consisted of combined primary and secondary sludges. Those from ASB systems consisted only of primary sludge. 2.7 RELATIONSHIPS BETWEEN WASTEWATER TREATMENT AND TOTAL SUSPENDED SOLIDS Further investigation was made of the relationships between TCDD/TCDF mass exports in sludge and effluent vectors, wastewater treatment types, and levels of total suspended solids (TSS) from kraft mills. When ACT and ASB-type kraft mills were combined, a weakly correlated positive trend was observed between effluent TCDD/TCDF and TSS levels, and a weakly correlated negative trend was observed between TSS and sludge TCDD/TCDF. For kraft mills using only ACT treatment, higher TSS levels were associated with higher sludge-based TCDD/TCDF exports but lower effluent-based TCDD/TCDF exports. 2.8 RELATIONSHIPS BETWEEN TCDD/TCDF FORMATION AND MILL OPERATING CHARACTERISTICS When the effects of mill bleaching procedures upon TCDD/TCDF formation in kraft mills were analyzed, correlations between mass export rates of TCDD/TCDF and a series of mill parameters, including application rates of bleaching and extraction chemical agents, were generally low. Consequently, linear regressions 15 ------- of the TCDD/TCDF export rates fit to bleaching measures at each mill were found to be poor predictors of individual mill outputs. 2.9 EFFECTS OF CHLORINE APPLICATION IN PRE- BLEACHING Significant positive trends were observed between average TCDD/TCDF formation in kraft mills and the rate of application of chlorine (C12) in the C- Stage bleaching process. Greater chlorine usage was thus found to be statistically associated with higher formation rates of TCDD/TCDF. It was also found that increased substitution of chlorine dioxide for chlorine in the C- Stage was correlated with slight reductions in TCDD/TCDF formation. Lack of chlorine dioxide use at high rates of substitution during the study sampling period precluded more detailed analysis of the impact of chlorine dioxide (C102) substitution. 2.10 EFFECT OF THE CHLORINE MULTIPLE Variables measuring the chlorine multiple (also known as the Kappa" factor) during C-stage bleaching were positively associated with TCDD/TCDF mass formation in kraft mills, though the resulting correlations were fairly weak. These results imply that on average, when accounting for lignin content, greater use of chlorine in the C-stage was linked weakly to higher formation of TCDD/TCDF. 2.11 USE OF OXYGEN IN THE BLEACHING PROCESS Kraft mills that used oxygen delignification in the bleaching process exhibited somewhat lower rates of TCDD/TCDF formation than mills that did not use such methods. The sane mills, however, also tended to have high substitution rates of C102 for C12, so it is not clear whether the lower export rates of TCDD/TCDF observed at these mills were attributable to oxygen delignification, chlorine dioxide substitution, or some combination of both. 16 ------- 2.12 DIFFERENCES IN WOOD TYPES Larger amounts of chlorine were generally applied to softwood pulps than to hardwood pulps per ton of pulp processed in kraft mills, and the average Kappa numbers of softwood pulps were significantly higher than those of hardwood pulps. These findings are consistent with known differences in bleaching practices for hardwood versus softwood pulps. 17 ------- 3. CHARACTERIZATION OF THE TCDD/TCDF CONCENTRATION DATA This section characterizes the laboratory data reported to the U.S. Environmental Protection Agency (EPA) concerning the concentration levels of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) and 2,3,7,8-tetrachlorodibenzofuran (TCDF) found in samples of pulp, sludge, and effluent collected as part of the 104 Mill Study. The reported data were examined for distributional properties and skewness and fit to appropriate probability distributions. The sensitivity of subsequent analyses to non-detected measurements was assessed. Attempts were made to handle non-detected samples in a reasonable and consistent manner that would not distort the basic findings. After examining the raw concentrations, the appropriateness of fitting TCDD and TCDF values to separate lognormal distributions was investigated. Only detected concentration values were examined for distributional fit. Approximately 15 percent of all the TCDD analyses and 4 percent of the TCDF analyses were recorded as non-detects. The detection levels for these non- detected measurements are summarized in Table 3-1. 3.1 VARIABILITY IN DETECTION LEVELS The variation in detection levels reported for non-detects (Table 3-1) can be attributed to several sources. Reliable measurement of TCDD/TCDF levels is matrix-dependent, a fact reflected in the analytical detection level targets for effluent samples, which were different from the targets for pulp and sludge. In addition, the presence of other compounds can make identification of TCDD/TCDF difficult without dilution of the sample, leading to detection levels that can be sample-specific. The Enseco-California Analytical Laboratory (CAL) and the Wright State University (WSU) lab each analyzed at least some samples from every matrix. Almost 80 percent of the pulp samples were analyzed at WSU, while 89 percent of the effluent samples and 81 percent of the sludge samples were handled by CAL. Since these laboratories used somewhat different clean-up and routine handling 18 ------- TABLE 3-1. DETECTION LEVELS FOR NON-DETECT SAMPLES Pulp Non-Detects (ppt^ TCDD TCDF N of Cases Minimum Maximum Mean Standard Dev. Median 39 0.100 4.900 0.667 0.805 0.500 11 0.100 6.800 1.218 1.880 0.800 SludEe Non-Detects (ppt) TCDD TCDF N of Cases 4 0 Minimum 0.300 Maximum 3.000 Mean 1.650 Standard Dev. 1.121 Median 1.650 Effluent Non-Detects (PPQ) TCDD TCDF N of Cases 30 11 Minimum 3.000 2.100 Maximum 17.000 10.000 Mean 7.733 5.764 Standard Dev. 2.789 2.458 Median 7.500 5.800 19 ------- procedures, it would be possible Co expect different detection levels for samples of a given matrix, depending on which lab performed the analysis. Overall, the analytical objectives of the 104 Mill Study were generally met. Ninety-two percent of non-detect pulp samples had reported detection levels at or below the 1 part per trillion (ppt) target level established in the Agreement (3). All but four sludge samples had detectable concentrations of TCDD/TCDF. Of these four, one was below the target detection level. For effluent samples, the target level of 10 parts per quadrillion (ppq) was achieved in the analyses of 83 percent of the TCDD non-detects and 100 percent of the TCDF non-detects (Figures 3-1 and 3-2). 3.2 FITTING OF DETECTED CONCENTRATIONS For the detected sample concentrations, graphical goodness of fit was done via lognormal probability plots (base 10 scale), matching the ordered concentration levels against the expected values of a lognormal distribution. When data are well-approximated by a lognormal density, such plots closely resemble a straight line. Examination of the plots showed that the data were adequately fit by lognormal densities estimated separately for each export matrix of pulp, effluent, and sludge samples (plots are located in appendix B). As noted, only detected values were used to characterize the distributions of TCDD/TCDF concentrations within each matrix. Estimates for non-detects measurements, however, were needed for later stages of the analysis. To handle non-detects in a simple, consistent manner, non-detect values were assigned as half the reported detection level. Decision on the treatment of non-detected samples depends upon the purposes of the analysis and the specific nature of the data. In this case, over 96 percent of all the quantitated samples in the 104 Mill Study exhibited detectable levels of either TCDD or TCDF, including at least one matrix export from every mill. Since the ratio of detectable levels of TCDF to TCDD was fairly consistent from mill to mill, there was evidence that non-detected samples contained small positive concentrations of TCDD/TCDF. Setting non-detects to zero would tend 20 ------- FIGl 3-1 SAMPLE CUMULATIVE DISTRIBUTION GRAPH EFFLUENT TCDD DETECTION LEVELS 0 5 10 15 Cone of 2378-TCDD (in PPQ) ------- HICiURE 3-2 SAMPLE CUMULATIVE DISTRIBUTION GRAPH EFFLUENT TCDF DETECTION LEVELS K) G .2 •»-» o V 1.0 08 - 06 o 0.4 o o. o £ 02 E P ° 0.0 8 10 12 Cone of 2378-TCDF (in PPQ) ------- to underestimate the true concentrations of TCDD/TCDF. On the other hand, EPA has frequently assigned non-detects to their detection levels, since the detection levels provide an upper bound on the actual concentrations present in non-detected samples. Setting non-detects to half the detection level is an arbitrary choice, but has been used with environmental data to steer a "middle ground" between over- and underestimation of the unknown concentrations within non-detected samples (6,7). Since the proportion of non-detects among the total sample set was relatively small, the choice to set non-detects at half the detection level was also considered unlikely to seriously affect the final TCDD/TCDF mass loadings computed at each paper mill. To illustrate this last point, Tables 3-2 and 3-3 present summary statistics of the TCDD/TCDF concentrations under different assumptions concerning the values of non-detects; the first section summarizes detected concentration values only, while the others report all TCDD/TCDF concentrations after setting non-detects equal to either half the detection level, zero, or the detection level. Some differences are apparent in the tables, particularly for pulp and effluent TCDD samples at sulfite mills, but overall, the discrepancies were judged to be relatively minor when weighed against the precision of the data as a whole. In summary, the detected concentration values of TCDD/TCDF were found to be best approximated by lognormal distributions, which were estimated separately for each of the export matrices: pulp, sludge, and effluent. Non-detects were consistently assigned to half the detection level in all subsequent analyses. 23 ------- TABLE 3-2. DESCRIPTIVE STATISTICS FOR TCDO OCHCEHTRATIOHS Mattl« All Sample* Pulp (ppt) HH SW Sludge (ppt) Effluent (ppq) Kreft Saeaplea Pulp (ppt) HH SW Sludg* (ppt) Effluent (ppq) Sulflt* Saeaple* Pulp (ppt) HH SH Sludg* (ppt) ECflu*nt (ppq) Matrix All Sa^lee Pulp (ppt} HH SH Sludge (ppt) Effluent (ppq) Kraft Saejplea Pulp (ppt) HH SH Sludg* (ppt) Effluent (ppq) Sulflt* Pulp (ppt) HH SH Jga (ppt) Luent (ppq) 217 84 114 118 133 194 74 104 97 107 18 8 8 19 25 DETECTED SAMPLES OMLT Lower He an Std Minimum Maximum Quart! la Median 3.50 2.80 4. 12 10.63 15.00 3.55 2.80 4.17 14.00 16 00 2.38 2.00 3.50 3.42 9.72 ROM-DETECTS - 1/2 DETECTION LEVEL 179 65 100 114 103 173 62 98 94 90 4 3 1 18 12 10.44 7.48 12.02 86.32 68.22 10.46 7.50 12.11 100.86 7S.es 6.22 7.13 3.50 13.22 13.33 12.85 9.53 14.73 169.43 100.80 13.00 9.68 14.86 183.08 105.67 5.93 6.92 16.61 5.71 8.66 5.84 10.59 83.42 53.70 9.36 6.32 11.43 97.77 64.47 1.63 2.81 0.62 12.53 8.16 0. 0 0 0 3. 0. 0. o 0. 3 2 2 3 0 4 400 400 500 .400 100 .400 400 500 900 100 000 .000 .500 .400 .500 116.00 55.70 116.00 1390.00 640.00 116.00 55.70 116.00 1390.00 640.00 15.00 15.00 3.50 58.00 23.00 Std 12.29 8.91 14.32 167.23 92.63 12.68 9.25 14.68 181.03 100.34 3.56 5. 15 1. 14 16.42 6.41 Minimum 0.050 0.050 0.100 0.150 1.500 0.050 0.050 0.250 0.700 1.500 0.100 0. 100 0. 150 0 150 2.100 Lower Ba«limmi Quartl le 116.00 55.70 116.00 1390.00 640.00 116.00 55.70 116.00 1390.00 640.00 15.00 15.00 3.50 .00 .00 1.90 0.70 3.20 8.77 6.15 2.40 1.57 3.92 13.50 9.20 0. 15 0. It 0. 19 3.20 3.27 Median 6.00 4. 10 7.60 34.00 30.00 6.00 4.00 7.60 39.00 35.00 3 95 4.40 3.50 4. 75 12.00 Median 4.70 3.30 6.30 32.00 19.00 5.15 3. 50 6.50 37 40 24.00 0.30 0.32 0.32 4.70 4. 50 Upper Quart! la 14.00 7.70 14.75 96.50 82.00 13.50 7.70 15.05 105.25 95.07 12.35 15.00 3. 50 15.25 16.00 Upper Quartl le 11.00 6.00 13.25 95.25 63.00 12.00 6.25 14.00 104. 50 ei.oo 1.47 3.80 1. 10 14.00 12.00 90ii Percentlle 23.00 17.00 26.90 188.00 172.00 24.20 17.00 27.00 203.00 189.00 15 00 15.00 3.50 48.10 22.70 9012 PercenLlle 21.00 16.00 25.50 165.60 138.00 22.00 16.50 26.50 197.00 164.00 5.46 15. OD 3.50 47.00 20.20 ------- TABLE 3-2. DESCRIPTIVE STATISTICS FOB TCDD UUBUJUHATIOBS (OOmillUED) Matrix All SaBBj>L«B Pulp (ppt) BW SH Sludge (ppt) Effluent (ppq) Kraft 8aBf>lejs Pulp (ppt) UN SH Sludge (ppt) Effluent (ppq) Suit It* 8aa*>L«a Pulp (ppt) BW SH Sludge (ppt) Effluent (ppq) Matrt» All S«f>U* Pulp (ppt) HW SH Sludge (ppt) Effluent (ppq) Kraft S«-pl«e Pulp (ppt) HW SH Sludge (ppt) Effluent (ppq) Sulflte Saaiplaa Pulp (ppt) HW SH Sludge (ppt) HOB-DETECTS - 0 Lower He an Std Minimum Maximum Quarttle Median 217 at 114 118 133 194 74 104 97 107 IB 19 25 217 64 114 118 133 194 74 104 97 107 IB 8 a 19 8.61 i.79 10.55 83.39 52.83 9.33 6.28 11.41 97.74 63.80 1.38 2.67 0.41 12.53 6.40 12.33 8.94 14.35 167.25 93.12 12.70 9.28 14.70 181.05 100.76 3.65 5.23 1.24 16.42 7.82 Mean Std 8.71 5.89 10.64 83.45 54.58 9.39 6.35 11.45 97.81 65.15 1.88 2.95 1 20 12.5* 12.26 8.88 14.28 167.22 92. 18 12.66 9.23 14.67 181.01 99.95 3.49 5.07 1.19 16.4} 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 I-DETECTS Minimum 0.100 0.100 0.200 0.300 3.000 0.100 0.100 0.500 0.900 3.000 0.200 0.200 0.300 0.300 116.00 55.70 116.00 1390.00 640.00 116.00 35.70 116.00 1390.00 640.00 15.00 15.00 3.50 38.00 23.00 - DBTECTIOH Ha»lmun 116.00 55.70 116.00 1390.00 640.00 L16.00 55.70 116.00 1390.00 640.00 15.00 15.00 3.50 SB. 00 1.90 0.70 3.20 8.77 5.75 2.40 1.57 3.92 13.50 9.20 0.00 0.00 0.00 3.20 0.00 LEVEL Lonor Quartlle 1.95 1.00 3.20 8.77 8.75 2.40 1.57 3.92 13.50 11.00 0.30 0.32 0.37 3.20 Median 4. 70 3.30 6.30 32.00 19.00 5. IS 3. SO 6. SO 37.40 24.00 0.00 0.00 0.00 4.70 0.00 Median 4.70 3.30 6.30 32.00 19.00 5.15 3.50 6. 50 37.40 24.00 0.60 0.65 0.65 4.70 Upper Quattile 11.00 6.00 13.25 95.25 63.00 12.00 6.25 14.00 104.50 81.00 0.50 3.80 0.00 14.00 12.00 Upper Quartlle 11.00 6.00 13.25 95.25 63.00 12.00 6.25 14.00 104.50 Bl .00 2. IS 3.80 2.20 14.00 90" Percentile 21.00 16.00 25.50 185.60 138.00 22.00 16.50 26.50 197.00 164.00 5.46 IS. 00 3.50 47.00 20.20 90!i Pvrcentlle 21.00 16.00 25.50 185.60 139.00 22.00 16.50 26.50 197.00 164.00 5.46 15.00 3.50 47 00 ------- Pulp (ppt) HW . SW Sludge (ppt) Effluent (ppq) Kraft Sample! Pulp (ppt) HW SW Sludge (ppt) Effluent (ppq) Sulfite See^lee Pulp (ppt) HW SW Sludge (ppt) Effluent (ppq) plaa Pulp (ppt) HW SW Sludge (ppt) Effluent (ppq) Kraft Sanplaa Pulp (ppt) KM SW Sludge (ppt) Effluent (ppq) Sulfite Saa^lea Pulp (ppt) HW SW ~'udge (ppt) luent (ppq) TABLE 3-3. DESCRIPTIVE STATISTICS FOB TCDP OOaCBTTRATIGNS DETECTED SAMPLES OHLT Lower N 206 79 108 115 127 167 72 99 97 104 14 5 7 16 21 Mean 89. S3 55. 83 117.69 697.73 412.30 89. SB 56 08 117.98 796.45 476.19 89.36 73.42 125.43 98.63 112.26 Std 251.14 123.24 326.52 2012.20 1108.94 259.27 124.43 337.06 2174.35 1214.02 166.95 139.82 207.71 143.34 194.37 Minimum 0 0 0 0 2 0 0 0 2 4 1 1 1 0 2 .600 800 .600 .700 .600 .600 800 700 .400 200 .100 100 .400 .700 .800 BOM-DETECTS - N 216 84 113 115 138 192 74 102 97 111 19 8 9 16 25 Mean 85.40 52.52 112.50 697.73 379.66 87.26 54.58 114.52 796.45 446.39 65.90 45.99 97.58 98.63 94.55 Std 245.95 120.20 320.07 2012.20 1069.30 256.25 123.05 332.62 2174.35 1180.41 147.50 112.27 188. 16 143.34 182.20 Minimum 0 0 0 0 1 0 0 o 2 2 0 0 0 0 1 .050 150 .050 .700 050 350 350 400 .400 750 050 150 050 700 .050 Mailn; 2620 661 2620 17100 8400 2620 661 2620 17100 8400 449 323 449 584 840 urn .00 .00 .00 .00 .00 .00 .00 .00 .00 .00 .00 .00 .00 .00 .00 Quartlle 5.67 4.10 6.32 34.50 36 00 6.80 5 32 7.30 35.10 42.25 2.70 4.10 2. 10 26.75 16.00 Median 19.50 15.00 22.50 107.00 82.00 21.00 17.50 26.00 161.00 98.00 6.35 9.90 6.30 63.00 35.00 Upper Quartlle 60. 49. 64. 624. 320. 59. 49 63. 675. 359. 100. 174. 409. 85. 120. 22 00 27 00 00 00 75 90 50 75 25 50 00 75 00 90SS Percentlle 164.20 108.00 230.60 1582.00 864.00 148.20 107.10 185.00 1726.00 1150.00 429.00 323.00 449.00 350.20 376.00 1/2 DETECTION LEVEL Minimum 2620 661 2620 17100 8400 2620 661 2620 17100 8400 449 323 449 "V ) .00 .00 .00 .00 .00 .00 .00 .00 .00 .00 .00 .00 .00 .00 .00 Lower Quarttle 4.22 3 13 5.55 34.50 26.00 5 70 3 97 6.52 35.10 37.00 •0.45 0.30 0 77 26.75 6.00 Median 18.00 14 50 19.00 107.00 69.50 20.00 15.50 22.50 161.00 82.00 3.10 4. 10 3.80 63.00 29.00 Upper Quartlle 58. 46. 61. 624. 312. 59 49. 60. 675. 340. 9. 21. 207. 85. 91 50 50 45 00 50 00 25 22 50 00 90 97 70 75 00 90JS Percentlle 154.20 106.50 207.20 1582.00 841.00 144.90 106.50 176.60 1728.00 1064.00 409.00 323.00 449.00 350.20 328.00 ------- TABLE 3-3. DESCRIPTIVE STATISTICS FCB TCDP OJHUJMHATIOBS (CCMTimjED) Mean •OH-DETBCTS - 0 Lower Uppar Std Mlnlnmro Maximum Quartlla Median Quartlla Percent 1 la All Sables Pulp (ppt) HW SW Sludg* (ppt) Effluent (ppq) Kraft Sanies Pulp (ppt) HH SW Sludg. (ppt) Effluent (ppq) Sulflta Saaftla Pulp (ppt) HW SW Sludg* (ppt) Effluent (ppq) HetrU All Samples Pulp (ppt) HW SW Sludg* (ppt) Effluent (ppq) Kraft SanpLem Pulp (ppt) HW SW Sludg« (ppt) Effluent (ppq) Sulflte Samples Pulp (ppU HW SM Sludge (ppL) Effluent (ppq) 216 84 113 113 138 192 74 102 97 111 19 8 9 16 25 85.38 52.30 112.48 697.73 379.43 87.25 34.37 114.51 796.43 446.16 65.83 43.89 97.36 98.63 94.30 245.96 120.21 320.08 2012.20 1069.38 236.25 123.03 332.62 2174.33 1180.30 147.33 112.32 188.19 143.34 182.34 216 84 113 113 138 192 74 102 97 111 19 8 9 16 25 Ha an 85.41 32.54 112.31 697.73 379.89 87.27 54.59 114.54 796.45 446.62 65.96 46. 10 97.60 98.63 94.Bl Std 245.95 120.19 320.06 2012.20 1069.22 256.25 123.04 332.61 2174.35 1180.32 147.48 112.22 188.17 143.14 182.06 0.000 0.000 0.000 0.700 0.000 0.000 0.000 0.000 2.400 0.000 0.000 0.000 0.000 0.700 0.000 •-DETECTS Minimum 0.100 0.300 0.100 0.700 2.100 0.600 0.700 0.700 2.400 4.200 0. 100 0.300 0. 100 0. 700 2. 100 2620.00 661.00 2620.00 17100.00 8400.00 2620.00 661.00 2620.00 17100.00 8400.00 449.00 323.00 449.00 564.00 840.00 - DETBCTIOB Ha» 1 muro 2620.00 661.00 2620.00 17100.00 8400.00 2620.00 661 00 2620 In) 17100.00 8400.00 449.00 323.00 449 00 584.00 840.00 4.22 3.13 5.55 34. 50 26.00 5.70 3.97 6.52 35.10 37.00 0.00 0.00 0.70 26.75 6.00 LEVEL Lower Quartlla 4.22 3 13 5.55 34.50 26.00 5.70 3.97 6.52 35.10 37.00 0.90 0.60 0.8i 26. 7S 6.25 18.00 14.50 19.00 107.00 69.50 20.00 15.50 22.50 161.00 82.00 3.10 4.10 3.80 63.00 29.00 Median 18.00 14.50 19.00 107 .00 69.50 20 .00 15.50 22.50 161 .00 82.00 3. 10 4 . 10 3.60 63.00 29.00 58.50 46.50 61.45 624.00 312.50 59.00 49.25 60.22 673.50 340.00 9.90 21.97 207.70 85.75 91.00 Upper Quartlla 58.50 46.50 61.45 624.00 312.50 59.00 49.25 60.22 675.50 340.00 1 90 2197 207 70 85 75 91 .00 154.20 106.30 207.20 1582.00 841.00 144.90 106 . 50 176.60 1728.00 1064.00 409.00 323.00 449.00 350.20 328.00 902 Parcentlla 154.20 106.50 207.20 1582.00 84 1 . 00 144.90 106. 50 176 60 1728.00 1064 .00 409.00 323.00 449 .00 350.20 328.00 ------- 4. ANALYSIS OF FIELD AND LAB DUPLICATE SAMPLES Section 4 examines the variability in measurements of 2,3,7,8- tetrachlorodibenzo-p-dioxin (TCDD) and 2,3,7,8-tetrachlorodibenzofuran (TCDF) reported for sets of duplicate samples. Concentration values for duplicate measurements were plotted against each other to assess the degree of agreement, and the total variability in duplicate samples was analyzed to determine what fraction could be attributed to measurement error or differences in sampling and analytical protocols. The fact that the distributions of TCDD/TCDF concentration values could be analyzed as approximately lognormal was important in two ways: to concretely characterize the data from the 104 Mill Study and to analyze the variability in TCDD/TCDF concentrations attributable to duplicate field sampling or repeated laboratory tests. Of the 500 samples of pulp, sludge, and effluent from this study, close to ISO (30 percent) were classified as field sample duplicates or lab duplicate splits. The variation in TCDD/TCDF measurements among duplicate samples was evaluated since a single value representing the TCDD/TCDF concentration of each composite sample was needed to compute the TCDD/TCDF mass exports linked to the bleach lines at each pulp mill. Since the variability among duplicates was found to be relatively small, the TCDD/TCDF concentration values from duplicate analyses were averaged, first setting any non-detected values to half of the reported detection level. 4.1 CORRELATIONS BETWEEN DUPLICATE PAIRS Figures 4-1 through 4-12 (located at the end of this section) plot the concentration values of TCDD/TCDF for all pairs of field and lab duplicate samples, subdivided by matrix into pulp, sludge, and effluent. The dashed line on each plot represents the region of perfect agreement between duplicate measurements. Non-detected samples were assigned a concentration value of half the reported detection level. 28 ------- For purposes of estimating the approximate variability in each scatterplot, particularly the variability orthogonal to the dashed 45-degree line, a 95 percent confidence ellipsoid is also shown. For data that are approximately bivariate normal in distribution, only 5 percent of the data pairs would be expected to fall outside the ellipsoid (since the data are plotted on a log scale, the assumption of bivariate normality is not unreasonable given the goodness of fit results described in section 3.2). The widths of the confidence ellipsoids for lab versus field duplicates or between different export matrices roughly indicate the relative agreement between duplicate pairs in each case. In general, both types of duplicate pairs (lab and field) show very close agreement. Few points indicate any significant discrepancy between the measured TCDD/TCDF concentration levels, although three of the plots involving lab duplicate pairs deserve special notice. In Figure 4-4, two pairs of TCDF pulp samples are more discrepant than the rest, both pairs came from the Champion International mill at Cantonment, Florida. In Figure 4-7, three pairs of ICDD sludge samples stand out; all three were collected from sulfite mills. The laboratories that conducted the analyses noted that producing reliable results was much more difficult for samples from sulfite mills than those from krafc mills. In addition, the three sample pairs of TCDF effluent duplicates in Figure 4-12 show less agreement than the others. Two of the pairs came from the Champion International kraft mill in Houston, Texas; the other pair was collected at the Wausau sulfite mill in Brokaw, Wisconsin. The relative agreement between lab duplicates is of particular interest, since repeated laboratory measurements on the same samples provide an estimate of the variability in concentration levels due to analytical measurement error. Though the variability in field duplicates necessarily contains components due to field sampling protocol and to analytical measurement difference, very few samples were labeled as both field duplicates and lab splits, so the variability of lab duplicates in this study cannot be assumed to be "contained" within the variability of field duplicates. 29 ------- To support Che visual impressions provided by the plots of duplicate pairs, Table 4-1 provides the Pearson correlation coefficients between the various types of field and lab duplicates, subdivided by matrix (pulp, sludge, and effluent) and pulping process (kraft and sulfite). The correlations were computed on the logged data to correspond with the above plots. Except for TCDD measurements computed for sulfite mill lab duplicates, this measure indicated very strong agreement between either field duplicate or lab duplicate pairs. Figures 4-13 to 4-16 (located at the end of this section) illustrate the differences between TCOD/TCDF effluent pairs taken from kraft versus sulfite mills. While almost 90 percent of the kraft sample pairs (22 of 25) show very good agreement, at least 40 percent of the sulfite pairs (4 of 10) indicate significant discrepancy between the duplicate analyses. These findings suggest that samples collected from sulfite mills were more difficult to analyze than counterparts collected from kraft mills. 4.2 ANALYSIS OF DUPLICATE SAMPLE VARIABILITY A formal analysis of variance (ANOVA) was also performed to determine the proportion of variability in TCDD/TCDF concentrations attributable directly to field sampling technique or analytical protocol. The objective of an ANOVA is to examine the total variation in a set of measurements and then partition the overall variability into smaller components representing different sources of error. Since the overall variation is known, the partitioning allows one to weigh each particular source of error relative to the total and hence, to rank the sources of error in degree of importance. Although many sources of variation can be attributed to the TCDD/TCDF concentration data, components resulting from field sampling and analytical error were of primary concern. One source of variability that could not be measured was the potential difference between the two laboratories performing the analytical work. In only a couple cases were duplicate samples "split across labs" before analysis; hence, all members of a duplicate set were generally analyzed by the same lab. Consequently, variability attributed to repeated lab measurement comprises "within lab" differences only. 30 ------- TABLE 4-1. PEARSON CORRELATIONS BETWEEN DUPLICATE PAIRS TCDD TCDF Field Duplicates N_ Correlation JJ Correlation Pulp 20 .952 21 .982 Sludge 9 .988 10 .987 Effluent 12 .985 13 .982 Kraft 11 .989 12 .982 Sulfite 1 --- 1 Lab Duplicates Pulp Sludge Effluent Kraft Sulfite 19 21 17 12 5 .994 .945 .967 .983 .735 16 19 18 13 5 .950 .989 .874 .886 .897 Note: Correlations were computed between pairs of logged concentration values. 31 ------- Tables 4-2 and 4-3 provide a breakdown of the components of total variation in TCDD/TCDF concentration values for field and lab duplicates within each matrix. For each matrix, the total sum of squared deviations (SS) from the overall mean was divided mathematically into two smaller sums of squares. The first sum of squares (SSI) was formed by calculating the average concentration value of each set of duplicate samples and then computing the squared deviations of the duplicate set means from the overall matrix mean. Conceptually, SSI represents the variation due to differences between average TCDD/TCDF values of various duplicate sets. The second sun of squares (SS2) was formed by computing the deviations of individual samples from the average concentration level within each duplicate set and then summing across all duplicate sets within the specific matrix. The second sum of squares is of particular interest since it represents an estimate of the variability due to differences between samples within duplicate sets and hence, is a measure of the analytical measurement error (Table 4-2) or field sampling error (Table 4-3) encountered during the 104 Hill Study. It is important to realize that the two component sums of squares add up to the total variation, so that SS - SSI + SS2. In this context, one can judge whether the percentage of the total variation due to field sampling or analytical measurement error (SS2 percent) is large compared with all other sources of variation, which are lumped together in SSI percent. For the cases in Tables 4-2 and 4-3, if one considers the variability resulting from "within duplicate set differences", with the exception of one case, less than six percent of the total variation can be attributed to differences in either field sampling or laboratory analysis. Consistent with the previous analyses, it can be fairly concluded that a minor portion of the variance in TCDD/TCDF concentrations is attributable to field sampling protocol or analytical measurement. Averaging the concentration values within duplicate sets to form a single value for subsequent analysis appears to be justified. The exceptional case involves effluent lab duplicates for TCDF where 12 percent of the total variation can be attributed -to differences between 32 ------- TABLE 4-2. ANOVA TABLE FOR LAB DUPLICATES Macrix fi SSI SS1X SS2 SS2X Pulp Loglo(TCDD) Log10(TCDF) Sludge Log10(TCDD) Log10(TCDF) Effluent Log10(TCDD) Loglo(TCDF) 32 29 31 27 25 27 11.528 20.572 21.083 19.089 10.001 13.886 99.5 96.8 94.2 99.1 97.5 88.3 0.055 0.678 1.300 0.167 0.256 1.845 0.5 3.2 5.8 0.9 2.5 11.7 SSI- Between Duplicate Set Sum of Squares - Within each matrix, the deviations of duplicate set means from the overall matrix mean SS2- Within Duplicate Set Sum of Squares - Deviations of individual samples from their respective duplicate set means SS- Total Sum of Squares - Equal to SSI + SS2 SS1Z - (SS1/SS)*100 SS2Z - (SS2/SS)*100 33 ------- TABLE 4-3. ANOVA TABLE FOR FIELD DUPLICATES Matrix fi SSI SSU SS2 Pulp Log10(TCDD) Log,0(TCDF) Sludge Log10(TCDD) Loglo(TCDF) Effluent Loglo(TCDD) Log10(TCDF) 37 39 15 17 21 23 9.562 17.971 5.027 8.791 5.016 6.686 97.7 98.9 99.0 99.3 99.1 98.8 0.224 0.207 0.050 0.062 0.043 0.078 2.3 1.1 1.0 0.7 0.9 1.2 SSI- Between Duplicate Set Sun of Squares - Within each matrix, the deviations of duplicate set means from the overall matrix mean SS2- Within Duplicate Set Sum of Squares • Deviations of individual samples from their respective duplicate set means SS- Total Sum of Squares - Equal to SSI + SS2 SS1Z - (SS1/SS)*100 SS2I - (SS2/SS)*100 34 ------- analytical measurements within duplicate sets. While this fraction does not appear to be unreasonably large, it is twice as high as any of the other cases, including the corresponding SS2 percentage for effluent TCDD lab samples. As was noted in Figure 4-12, this finding can be attributed to measurement differences from only 3 of 18 pairs of effluent samples; the remaining duplicates appear to be in very close agreement. 35 ------- FIGURE 4-1 PULP FIELD DUPLICATES TCDD 1000 000 H CL, O, 100000 - 10000 a .2 5 1000 8 a o U a a 0100 0010 0001 o\°° TCDD Concentration in PPT ------- FIGURE 4-2 PULP FIELD DUPLICATES TCDF H Cu a 10000.000 1000000 100.000 a S 10000 cd G *-• a 1000 o U g 0100 0010 0001 ,-J ...J TCDF i uncentration in PPT ------- FIGURE 4-3 PULP LAB DUPLICATES TCDD OO 1000000 100000 H h cu 2 10.000 o o a o o a o U Q Q U H 1 000 0100 0010 0001 TCDD Concentration in PPT ------- FIGURE 4-4 PULP LAB DUPLICATES TCDF 10000 000 1000.000 H PD (X TCDF Concentration in PPT ------- FIGURE 4-5 SLUDGE FIELD DUPLICATES TCDD H OU a 1000.000 100 000 ~ 10000 a o a t> o a o U Q 0 1.000 0100 0010 0001 TCDD Concentration in PPT ------- FIGURE 4-6 SLUDGE FIELD DUPLICATES TCDF H 04 10000000 1000 000 100.000 a •3 10000 ct o u u H 1000 0100 0.010 0001 11mil—i i iinul—i i iiniJ—I i 11mil TCDF Concentration in PPT ------- FIGURE 4-7 SLUDGE LAB DUPLICATES TCDD 1000 000 H PH a- 100000 •s 10000 a 2 *•» 2 £ 1 000 O o a o U 0100 Q a u 0010 0001 • ••••I '••! • • ••••••! 1 i i imJ 1 i inmJ 1 i i mil TCDD Concentration in PPT ------- FIGURE 4-8 SLUDGE LAB DUPLICATES TCDF H OH 10000.000 1000000 100.000 a •3 10000 «H «-• a 8 i.ooo a o U g 0100 0.010 0001 limn i i 111in) 1 I IIIITIJ i i iinii| lllllj . ..••••J • • lll.lJ • lllllJ TCDF Concentration in PPT ------- FIGURE 49 EFFLUENT FIELD DUPLICATES TCDD 1000.000 100000 cm 2 10000 a o 1.000 8 a o U Q Q U H 0100 0.010 0001 ..i.l • • ......I . i ......I i i • mill 1 I MlllJ 1 I Mill TCDD Concentration in PPQ ------- FIGURE 4-10 EFFLUENT FIELD DUPLICATES TCDF 10000000 1000.000 100.000 £ a mH a 5 10000 18 H a § 1000 a o U g 0.100 0010 0.001 / / • / uiiL imL uuL TCDF Concentration in PPQ ------- FIGURE 411 EFFLUENT LAB DUPLICATES TCDD 1000.000 a eu a 100.000 = 10.000 a o a o o a o U a a 1 000 0.100 0.010 0001 1.1..I i • i mill • • i .....I 1 i i mill 1 i i iniJ 1 i nun TCDD Concentration in PPQ ------- FIGURE 4-12 EFFLUENT LAB DUPLICATES TCDF 10000 000 1000000 100.000 2 Ou a ••- a •S 10000 s a o 1.000 u g 0100 0.010 0001 . l.lJ iinul—i i inn TCDF Concentration in PPQ ------- FIGURE 4-13 OD EFFLUENT LAB DUPLICATES KRAFT MILLS ONLY TCDD 1000 000 a PI PL, 100 000 •H 10000 a o 1000 o a o U 0100 Q O 0010 0001 • • •••••! ' '""I - 1 i i ninl - 1 i i mi TCDD Concentration in PPQ ------- FIGURE 4-14 EFFLUENT LAB DUPLICATES SULFITE MILLS ONLY TCDD 1000000 100000 •~ 10000 a .2 a o o 5 a a E 1 000 0100 0.010 0001 rrrnrj 1 i i niii| 1 i 111111] 1 i i niii[ 1 i i iinij .....I i i i. i.iJ 1 i 11 mil 1 i i nil TCDD Concentration in PPQ ------- FIGURE 4-15 l-n O EFFLUENT LAB DUPLICATES KRAFT MILLS ONLY TCDF 10000.000 a cu a -••* a a 8 Q O U U H 1000000 100000 10000 1.000 0100 0010 0001 ,.J iii i.nJ i • i niiJ—• • i.inJ—i i i mid—i i i mi TCDF Concentration in PPQ ------- FIGURE 4-16 EFFLUENT LAB DUPLICATES SULFITE MILLS ONLY TCDF 10000000 1000000 a (L, Ou 100 000 g •3 10000 o o u 1.000 0100 0.010 0001 -J • I HIM! — I I inu TCDF Concentration in PPQ c ^ O r— • ^ —i rn f: Go 3> 8 ^ C -; *. ^ ca O g «o 5- ' C-) Oo ^il ------- 5. PARTITIONING OF TCDD/TCDF MASSES INTO EXPORT MATRICES After analyzing the duplicate lab and field samples, average 2,3,7,8- tetrachlorodibenzo-p-dioxin (TCDD) and 2,3,7,8-tetrachlorodibenzofuran (TCDF) concentration values were computed for each set of duplicates. These average values were then grouped with non-duplicate samples to produce a modified data set consisting of a single pulp concentration value for each bleach line and single sludge and effluent concentrations at any given mill (non-detects being set to half the reported detection level). The goal in this section was to use the modified concentration data to compute estimates of the actual mass formation rates of TCDD/TCDF for each paper mill and then to characterize how the TCDD/TCDF masses were partitioned into the exported vectors of pulp, sludge, and effluent Mass output rates were produced because an estimate of the total amount of TCDD/TCDF generated at each mill could not be made using concentration data alone, since the output flow rates of pulp, sludge, and effluent products varied greatly from mill to mill. The calculations involved multiplication of the concentration level of each pulp, sludge, or effluent sample by the corresponding mass output rate reported for that export vector. Since the pulp, sludge, and effluent outflow rates were reported in different units, appropriate conversion factors were used as necessary to standardize each mass rate. Total mass export rates of TCDD/TCDF are reported in either Ibs/day or Ibs/ton Air-dried Brownstock Pulp (ADBSP). The latter rate represents the total output per day divided by the pulp production rate and hence, provides a mass output that is standardized for the size of the mill. (All tables and figures for section 5 are located after the text.) 5.1 VARIABILITY ACROSS EXPORT VECTORS Tables 5-1 through 5-4 provide relevant descriptive statistics of the mass export rates for TCDD and TCDF, including the number of mills, the mean and standard deviation, the minimum and maximum, the median and upper and lower quartiles, and the 90th percentile of the mass rate distributions. For each 52 ------- matrix and analyte, probability plots (appendix B) indicated that the TCOO/TCDF mass distributions could be approximated as lognormal. The tables provide corresponding statistics for the percentage of the total output at each mill attributable to each export matrix (pulp, sludge, and effluent). The same statistics were also recomputed after the mills were subdivided by pulping process (kraft and sulfite) and waste water treatment (Activated Sludge Wastewater Treatment [ACT] and Aerated Stabilization Basins [ASB]). One of the most apparent findings of these tables is the tremendous variability exhibited from mill to mill within each matrix. Figures 5-1 through 5-4 provide boxplots illustrating the range of variability from different perspectives. The first two figures represent the percentage of total TCDD/TCDF output partitioned to each matrix. Each boxplot was constructed so that the top and bottom edges of the box represent the lower and upper quartiles of the distribution of percentages taken across all mills, while the line dividing the box in two is the median. The two "whiskers" extending from the edges of the box mark a range covering the middle 95 percent of all the data points. Figures 5-3 and 5-4 represent the distributions of TCDD/TCDF mass formation adjusted for the pulp production rate at each mill (Ibs/ton ADBSP). In either case, it is clear that some mills partition much more of their TCDD/TCDF mass to one matrix than the others and that the pattern is not consistent from mill to mill. 5.2 KRAFT VERSUS SULFITE HILLS To test the significance of the differences between kraft and sulfite mills suggested in Tables 5-1 and 5-3, two-sample t-tests were run on the logged observations of TCDD/TCDF exports: one set for the unadjusted mass rates (Ibs/day) and one for the mass rates adjusted by the mill-specific pulp production rate (Ibs/ton ADBSP). The results are summarized in Table 5-5. Since the TCDD/TCDF mass export rates followed approximate lognormal distributions, comparison of these variables was made on the log scale in order to make inferences concerning the t-test as valid as possible. Such inferences are generally valid when the tested data have been sampled from a normal 53 ------- distribution, but not necessarily in other cases. An important consequence of using the logged data is that comparing arithmetic means on the log scale is, equivalent to comparing the geometric means of the mass export rates on the original scale. When data follow an exact lognormal distribution the geometric mean is equivalent to the median. Therefore, the comparison presented here Ls approximately one between the medians of the original data, which have been listed beside the corresponding means of the logged data in Table 5-5. For highly skewed data, such as that encountered in the 104 Mill Study, medians actually provide a better impression of the bulk of the sample since the effect of outlying points on the median is minimal. Several points should be kept in mind when interpreting the results of these significance tests. T-tests are designed to indicate how likely it is that an observed mean difference between two groups of sample data reflects an actual difference between the overall means of the populations from which the samples were taken, the p-value is one measure of this likelihood and represents the probability that if the study were repeated from scratch and a new set of measurements procured, one would observe a difference between the samples at least as great as the difference already observed, assuming that no real difference was expected. Low p-valu.es suggest that real differences between the two groups probably exist (i.e., that the observed differences are statistically significant). When comparing the mass rates that are unadjusted for mill-specific pulp production rates (Ibs/day), the p-values of Table 5-5 indicate that significantly more TCDD/TCDF was exported at kraft mills than sulfite mills when considered on a total basis and for each export matrix separately. When the adjusted mass rates (Ibs/ton ADBSP) were compared, the results changed only slightly: significantly more TCDD/TCDF mass was exported at kraft mills than sulfite mills for pulp and effluent vectors and for all exports combined. However, the difference between kraft and sulfite mills with respect to TCDD/TCDF in sludge was not found to be statistically significant. Nevertheless, in the sample data, kraft mills tended to export more sludge- 54 ------- based TCDD/TCDF on average Chan cheir sulfice counterparts. 5.3 ACT VERSUS ASB WASTEWATER TREATMENTS To interpret the main findings of Tables 5-2 and 5-4 with regard to wastewater treatment differences, Figures 5-5 through 5-8 provide boxpLots of the TCDD/TCDF output rates showing the percentage of total output attributable to sludge or effluent vectors, classified by wastewater treatment type. The boxplots illustrate that the percentages of total TCDD/TCDF output to sludge and effluent vectors were highly variable from mill to mill; however, there was a consistent tendency for the median percentage of TCDD/TCDF outflow to sludge to be much higher for ACT than ASB, and the corresponding percentage of outflow to effluent to be lower. The same differences between treatment types were exhibited by kraft mills considered separately; among sulfite mills, only one with usable data employed ASB-type waste treatment, so a similar comparison was not feasible. In part, the pattern exhibited in Figures 5-5 through 5-8 with kraft and sulfite mills combined is probably attributable to the limitations of the data. Sludge samples taken from ACT treatment systems consisted of both primary and secondary sludges, while those collected from ASB facilities only comprised primary sludge. Had representative secondary sludges from ASB-type treatment systems been obtainable, the estimated sludge-based TCDD/TCDF mass exports for ASB mills would have probably been higher than observed. Since the overall TCDD/TCDF mass rates would also be higher, this would have simultaneously raised the percent of total TCDD/TCDF output typically attributable to sludge and lowered the percent of total TCDD/TCDF output attributable to effluent, making the observed differences between ACT and ASB treatments less dramatic. Figures 5-9 through 5-12 provide boxplots of the effluent and sludge TCDD/TCDF mass export rates (in Ibs/ton ADBSP) on a logarithmic scale, subdivided by type of waste treatment. When considered on a mass rate basis instead of a percentage of total output, sludge-based TCDD/TCDF again appears to be significantly higher on average at ACT mills than ASB mills. How much of this 55 ------- difference is due to the different nature of the sampled ACT sludges versus ASB sludges can not be estimated. Sampled effluents from the 104 Mill Study should be more directly comparable, and in this case, the export rates of effluent-based TCDD/TCDF tended to be somewhat higher at ASB mills than ACT mills, though not in every comparison. Median effluent TCDD exports were slightly higher for ASB mills than ACT mills, but the reverse was true for effluent TCDF exports. In both cases, however, the lower and upper quartiles were larger for the set of ASB mills, suggesting that the middle 50 percent of ASB mills tended to export more effluent TCDD/TCDF than the middle 50 percent of ACT mills. T-tests calculated on the logged TCDD/TCDF mass export rates partially confirmed the visual impressions of Figures 5-9 to 5-12 (Table 5-6). Considered on the basis of production-adjusted mass export rates (Ibs/ton ADBSP), no significant differences at the 5 percent level were found between the median effluent export rates of ACT versus ASB mills. However, mills with ACT-type waste treatment exported significantly more TCDD/TCDF in sludge vectors than mills with ASB-type treatment. The same results were echoed by kraft mills considered separately. It should also be noted that the results were somewhat different when considering unadjusted TCDD/TCDF mass output rates (Ibs/day). In that case, significantly more effluent TCDD was exported by ASB-type waste treatments than ACT-type treatments; the same was not true for effluent TCDF or for kraft mills considered separately. 5.4 OVERALL PARTITIONING OF TCDD/TCDF Pie charts representing the overall partitioning of TCDD/TCDF into pulp, sludge, and effluent are presented in Figures 5-13 to 5-16. To construct each pie chart, total TCDD/TCDF mass exports (Ibs/day) were summed across all mills for each matrix, and the percentage of the total exported to pulp, sludge, or effluent is shown on the chart. Similar pie charts were also constructed for kraft and sulfite mills considered separately. These pie charts indicate 56 ------- the estimated total daily outputs of TCDD/TCDF poundage for all U.S. bleached pulp mills that had usable data. To accompany the pie charts, Tables 5-7 and 5-8 present the total mass outputs of TCDD/TCDF summed across all kraft or sulfite mills, the corresponding average output per mill, and the percentage of the total summed output exported to pulp, sludge, or effluent vectors. The two tables differ in that the first provides total outputs without adjustment for the pulp production rate at each mill, while the second sums the output of each mill after dividing first by the pulp production rate, to normalize for mill size. TCDD/TCDF outputs for kraft mills were considerably larger on any basis than the outputs for sulfite mills. However, kraft and sulfite mills exhibited similar patterns of the percentages of total output partitioned to different matrices. With one exception (TCDD output at sulfite mills), the largest fraction of TCDD/TCDF mass output was partitioned to pulp, being more than 50 percent for TCDF exports from sulfite mills. Considering the total estimated mass outputs of TCDD/TCDF for all matrices combined, these data suggest combined production totals of close to 0.004 Ibs/day of TCDD and 0.032 Ibs/day of TCDF at U.S. bleached pulp mills. Estimates of the per mill averages were close to 0.00005 Ibs/day for TCDD and 0.00048 Ibs/day for TCDF; however, substantial variation in the TCDD/TCDF mass exports was exhibited from mill to mill. 57 ------- TABU 5-1. DESCRIPTIVE STATISTICS KB TCDD TCDD Exports All in Oo TCDD in Pulp (lba/day)MO* TCDD in Sludge ( Ilia/day )MO» TCDD In Effluent (lba/day)*10* Total TCDD (lb»/day>*10* TCDD in Pulp (U»/ton ADBSF)MO* TCDD in Sludge (Ibs/ton ADBSP)*10* TCDD in Effluant (Ibs/ton ADBSP)MO* Total TCDD (Iba/ton ADBSF)*10* X TCDD OUTPUT to Pulp X TCDD OUTPUT to Sludge X TCDD OUTPUT to Effluent Kraft Saatilei TCDD In Pulp (lba/day)*10* TCDD In Sludsa (lba/day)*10* TCDD In Effluent (lba/dey)*10* Total TCDD (Lba/day)MO* TCDD in Pulp (Ibs/ton ADBSP)MO* TCDD in Sludge (Iba/ton ADBSP)*10* TCDD in Effluent < Ibs/ton ADBSP)*10* Total TCDD (Iba/ton ADBSP)*10* t TCDD OUTPUT to Pulp X TCDD OUTPUT to Sludga X TCDD OUTPUT to Effluant Sulflta Saavlaa TCDD In Pulp (lbs/day)*10* TCDD in Sludsa (lbs/day)MO* TCDD in Effluant (lbs/day)«10* Total TCDD (lbs/day)*10* TCDD in Pulp (Iba/ton ADBSP)MO' TCDD in Sludge (Iba/ton ADBSP)MO' TCDD in Effluant (Iba/ton ADBSP)*10* Total TCDD (Ibs/ton ADBSP)*10* X TCDD OUTPUT to Pulp X TCDD OUTPUT to Sludge X TCDD OUTPUT to Effluant N 101 99 97 93 101 99 91 93 93 93 93 84 83 81 80 84 83 81 80 80 80 80 IS 14 13 14 13 14 IS 14 14 14 14 Mean 13.73 13.38 12.07 42.18 1 71 1 28 1.22 4.31 39 92 23 79 34.30 18.33 IS 48 14 09 48.84 1 95 1 44 1.38 4.86 43.03 23.91 33 OS 0.93 1.34 1.31 3.80 0 35 0 37 0 33 1 03 21 99 35 70 42 32 Std 22 08 34 34 20 93 61 33 2 27 2 60 1 90 S.31 22 48 24 39 23 47 23 23 37 34 22 35 64 S3 2.39 2.80 2.03 5.57 20 53 24.34 22 71 1.43 2 31 1 33 3.61 0.77 0 44 0 37 1 19 26 13 23 72 27 57 Minimum 0 072 0 000 0 094 0.307 0 010 0 000 0.011 0 066 2 835 0 000 1 536 0 084 0.000 0.161 0.692 0.010 0 000 0.011 0.066 4 046 0.000 1.S36 0 072 0.026 0.094 0.507 0 020 0 008 0.031 0.206 2 835 1 935 6 981 Maximum 140.80 240.30 123.40 374.00 13.31 15.90 10 88 30.36 91 08 85.79 86.53 140.80 240.30 123.40 374 00 13.31 13 90 10 88 30.56 88 40 85.79 86.08 4 93 8.22 4 30 12.70 3 00 1.37 1 28 4.53 91 08 77 20 86 53 Lower Quartila 1 36 0 45 0 99 5.92 0.30 0.05 0 17 0.96 21 98 4 31 14.63 3 20 0 46 1 43 11 43 0 SO 0.05 0 23 1.21 24 78 3 51 14.66 0 13 0 26 0 24 1 34 0 03 0 04 0 11 0 27 6 20 12 23 12 37 Median -8 86 8 86 4 30 18.60 0 98 0 25 0 57 2 13 40 19 16 67 32 10 10 85 10 85 5 82 24 37 1 16 0 25 0 61 2 80 41 90 15 79 26 84 0 20 0 20 0 85 2 43 0 06 0 16 0 IS 0 46 10 48 38 77 39 54 Upper Quart lie 19 20 7.01 14.13 49 47 2 26 1.30 1 30 5.95 39 03 45 IB 49 30 23 35 7.73 18.04 68 21 2 38 1 46 1 70 6.33 60 39 43 SO 46.43 1 22 1 34 1 78 5 59 0 40 0 69 0 42 1.55 26 87 55 80 65 30 90ii Percent! le 45 02 34 OS 30 11 US 24 4 38 3 88 2 79 11 02 70 08 62 60 72 35 48.58 50.49 31 51 136 78 4 55 4 29 3.01 12 14 70 29 60 62 69 20 04 63 19 1 01 .73 24 11 3 32 78 65 70 98 86 21 ------- in TCDD TCDD In Pulp ------- TABLE 5-3. DESCRIPTIVE STATISTICS TOR TCDF TCDF Emaorta All TCDF in Pulp ( Iba/day )MO* TCDF in Sludge ( Iba/day )MO* TCDF In Effluent ( Iba/day )*10* Total TCDF (lba/day)*10* TCDF In Pulp (Iba/ton ADBSF)MO* TCDF In Sludge (Iba/ton ADBSP)MO* TCDF In Effluent (Iba/ton ADBSP)*10* Total TCDF (Iba/ton ADBSP)*10* X TCDF Output to Pulp X TCDF Output to Sludge X TCDF Output to Effluent Kraft TCDF in Pulp (Iba/day) MO* TCDF in Sludge ( Iba/day )MO' TCDF in Effluent ( Iba/day >MO* Total TCDF (Iba/day) MO* TCDF In Pulp (Iba/ton ADBSP)MO* TCDF in Sludge (Iba/ton ADBSPjMO* TCDF in Effluant (Iba/ton ADBSP)MO* Total TCDF (Iba/ton ADBSPJMO* X TCDF Output to Pulp X TCDF Output to Sludge X TCDF Output to Effluant Sulfite TCDF in Pulp (Iba/day)*!!)* TCDF in Sludge (lba/day)MO* TCDF in Effluant (lba/day)MO* Total TCDF (lba/day)MO* TCDF in Pulp (Iba/ton ADBSP)MO* TCDF in Sludge (Iba/ton ADBSP)MO* TCDF in Effluent (Iba/ton ADBSF)MO* Total TCDF (Iba/ton ADBSPJMO' X TCDF Output to Pulp X TCDF Output to Sludge X TCDF Output to Effluent Lower fi 102 102 99 96 102 102 99 96 96 96 96 85 as 62 80 65 85 82 80 80 80 80 15 15 IS 14 15 15 IS 14 14 14 14 Mean 147 80 82 92 94.14 334.30 20 96 8.75 12.67 43.29 43 96 25 83 30.22 162 67 94.41 106 85 374 93 22.67 9 67 14 37 48.33 46.67 23 32 30.02 52 08 14 26 26 17 89 12 10 79 2 71 3 96 13 81 26 47 40 83 32 70 Upper Std Minimum Maximum Quertlle Median Quartlle 339.14 273.27 229.62 711.90 62 53 23.77 41.60 116.62 23.37 24.98 22.19 363 34 297.17 248.81 764.28 67 S3 25 72 45 38 126 . 16 21 34 23.46 21.41 159 47 39 09 70 82 275 66 26 49 5.41 9 67 38 04 28 18 29 73 28 08 0.053 0 000 0 OS4 0.743 0.010 0.000 0.01B 0.147 0.590 0 000 0.323 0.459 0.000 0.417 2.128 0.090 0.000 0 048 0.147 4.383 0.000 0.323 0.053 0.000 0.054 0.743 0 010 0.000 0.018 0.243 0 590 2 002 3 624 2523.00 2394.00 1542.00 4511.00 524.01 195 59 365.71 953.88 92.18 93.81 86.84 2523.00 2394.00 1542.00 4511.00 524 01 195.59 365 71 953.88 92 18 91 35 86 84 615 70 154 90 273 40 1044 00 85.80 21 59 38 10 145 48 90 70 93 81 86 56 5.26 1 73 4 33 22 SO 0 93 0 17 0 64 3 46 23.33 3.94 11 04 10 93 1.S9 S.07 29.30 1 65 0 12 0 78 4 66 26 04 3.76 11 22 0 18 1 77 0.59 4.31 0 05 0 18 0 19 0 94 6 47 13 81 8 08 31 63 31.63 15.35 74 64 3.94 1 36 2 08 8 62 45.23 18.98 26 23 35 75 35.75 21.96 98 79 4 30 1.32 2 51 10 45 45 49 IS 59 26 99 2 03 2 03 1 61 9 19 0 42 1 40 0 73 3 47 12 10 39 25 25 28 127 &2 41 93 71.96 328.92 13 89 5.26 7 22 30 42 61 64 44.90 44 47 132.20 57 59 77 66 370 95 14.09 6.26 8 11 33 19 64 27 43 01 44.47 8 54 7 46 8 18 22.29 1 98 2 87 4 00 8 45 53 80 62 01 54 09 9012 Percent tie 356 47 189 82 273.40 735.14 45 SB 23 30 29 99 120 54 76 84 62 02 64.62 399 20 203 36 282.64 795.62 44 17 26.32 30 39 122 87 77.09 60 36 64 26 325 42 69 09 153 42 564 41 73 08 11.56 19 57 78 03 74 87 88 62 81 92 ------- TABLE 5-4. DESCRIPTIVE STATISTICS FOB TCDP (BY HASTEHaTKK TREATMOrr) TCDF Erporta TCDF in Pulp (lba/day)*10* TCDF in Sludge (Iba/day )*10* TCDF in Effluent (Iba/day )MO* Total TCDF (lbe/day)MO* TCDF in Pulp (Iba/ton ADBSP>*10* TCDF in Sludge (Iba/ton ADBSP)«10* TCDF in Effluent (Iba/ton ADBSP)MO* Total TCDF (Ibi/ton ADBSP)MO* I TCDF Output to Pulp I TCDF Output to Sludge 1 TCDF Output to Effluent HASTEHATER TREATMENT-ACT 1! 41 41 41 39 41 41 0* 41 39 39 39 39 Mean 111 81 72.40 49 60 233 07 17.75 8.93 7 62 33.33 40.66 37 67 21.68 Std ! 186 86 147.58 86.40 348.74 34.61 15.03 16 36 57 24 23 27 23 67 18.74 Loiter Upper llnlmum Haaimum Quart! la Median Quart lie 0.053 0.000 O.OS4 0.743 0.010 0.000 0.01B 0.243 0.590 0.613 2.264 964.40 846 00 422 00 1484.00 193.81 68 OS 90 95 299.61 90 70 93 81 77.28 5.01 4 72 1.83 20.64 1.06 1 27 0 43 3.76 22 34 19 79 7 76 28 45 28 45 12 00 79.23 4 34 2.87 2 08 11.13 38 90 36 92 IS 25 129 OS 91 09 67 90 361 80 20 23 9 09 6 45 27 85 59.15 54 39 26 64 90S Percent lie 300 96 205 8* 142 08 678.70 56 59 28.43 27 03 119 37 73 96 71 93 52 38 TCDF E«port» TCDF in Pulp (lba/day)*10* TCDF In Sludge (Iba/day)«10* TCDF in Effluent (Lbs/day )*10* Total TCDF (Iba/day)*10* TCDF in Pulp (Iba/ton ADBSP)MO* TCDF in Sludge (Ibs/ton ADBSP)*10* TCDF in Effluent (Iba/ton ADBSP)*10* Total TCDF (Iba/ton ADBSP)*10* t TCDF Output to Pulp X TCDF Output to Sludge X TCDF Output to Effluent HASTEHATER TREATMEHT-ASB Lower Upper Mean Std Minimum Maximum QuartUe Median Quart! le Percentll« 48 48 45 43 48 48 45 45 45 45 45 20S.26 111.53 154 44 486 64 27 68 10 55 19 87 60.21 45 77 19 50 34 74 4S6.76 373 28 321 38 967.80 85 11 31.64 S9.20 160.73 22 76 23 96 20.53 0 319 0 000 0.417 2.128 0 OSO 0 000 0 048 0.147 4.383 0.000 0.323 2523 00 2394 00 1542.00 4511 00 524 01 195 59 365 71 953 88 92 IB 91 35 74.99 7 04 1 &7 5 02 26. 6B 0.72 0 12 0 70 3 06 24 66 2.87 IS 17 38 97 38 97 31 79 96.39 3 94 0 70 1 99 8 44 45 54 6 73 32 83 1S9 57 37 66 124 67 428 10 13 35 3 99 11 32 34 85 63 07 26 75 52 38 631 57 259 60 490 88 1940 00 75.45 36 60 41 48 158.67 77 97 62 27 66 01 ------- FIGURE 5-1 a- K) Q Q O (ft 100 80 60 40 20 OUTPUT BY MATRIX TCDD Matrix ------- FIGURE 5-2 UJ 100 80 60 40 20 OUTPUT BY MATRIX TCDF Matrix ------- FIGURE 5-3 ADJUSTED TCDD BY MATRIX 100.00 oo o W 10.00 » £ Q a o *-> Q Q O H 1.00 0.10 0.01 ^ Matrix ------- FIGURE 5-4 ADJUSTED TCDF BY MATRIX 1000.00 o 100.00 U4 • 10.00 1.00 0.10 0.01 a o Matrix ------- TABLE TCDD Exports Qbs/dav) * 10s Total TCDD Kraft Sulfite Pulp TCDD Kraft Sulfite Sludge TCDD Kraft Sulfite Effluent TCDD Kraft Sulfite TCDF Exports flbs/dav> * 106 Total TCDF Kraft Sulfite Pulp TCDF Kraft Sulfite Sludge TCDF Kraft Sulfite Effluent TCDF Kraft Sulfite 5-5. H 79 14 84 15 76 14 80 15 fi 79 14 85 15 76 14 81 15 DIFFERENCES BETWEEN PULPING KRAFT Median 24.4 2.4 10.8 0.2 10.8 0.2 5.8 0.8 Median 98.8 9.2 35.8 2.0 35.8 2.0 22.0 1.6 vs SULFITE Logged Mean t-stat 1.355 7.371 0.411 0.892 7.804 -0.426 0.474 3.324 -0.191 0.714 5.365 -0.122 Logged Mean t-stat 2.021 4.363 1.050 1.588 4.259 0.302 1.120 2.405 0.466 1.340 3.434 0.416 PROCESSES p- value .000 .000 .003 .000 p- value .000 .001 .027 .003 Note: Two-sample t-tests for difference between logged means 66 ------- TABLE 5-5. DIFFERENCES BETWEEN PULPING PROCESSES (CONTINUED) KRAFT vs SULFITE TCDD Exports fibs/ton ADBSP) * 10* Total TCDD Kraft Sulfite Pulp TCDD Kraft Sulfite Sludge TCDD Kraft Sulfite Effluent TCDD Kraft Sulfite TCDF Exports (Ibs/ton ADBSP) * 10* Total TCDF Kraft Sulfite Pulp TCDF Kraft Sulfite Sludge TCDF Kraft Sulfite Effluent TCDF Kraft Sulfite N 79 14 84 15 76 14 80 15 N. 79 14 85 15 76 14 81 15 Median 2.8 0.5 1.2 0.1 0.25 0.16 0.6 0.2 Median 10.4 3.5 4.3 0.4 1.3 1.4 2.5 0.7 Logged Mean t-stat 0.420 4.792 -0.192 -0.028 5.530 -1.010 -0.478 1.527 -0.794 -0.212 3.677 -0.705 Logged Mean t-stat 1.087 3.026 0.447 0.664 3.044 -0.281 0.169 1.097 -0.137 0.414 2.389 -0.167 p- value .000 .000 .140 .001 D- value .007 .008 .286 .028 Note: Two-sample t-tests for difference between logged means 67 ------- FIGURE S-S 00 Q Q U H o «-« a OUTPUT BY TREATMENT EFFLUENT TCDD 100 80 60 40 20 ACT ASB Treatment ------- FIGURE 56 Q Q a »* 9 O •> M 100 80 60 40 20 OUTPUT BY TREATMENT SLUDGE TCDD ACT ASB Treatment ------- FIGURE 5-7 H, P O H a o W 100 80 60 40 20 0 OUTPUT BY TREATMENT EFFLUENT TCDF ACT ASB Treatment ------- FIGURE 5-8 3 o, tt •O ^ CO 100 80 60 40 20 0 OUTPUT BY TREATMENT SLUDGE TCDF ACT ASB Treatment ------- FIGURE 5-9 ADJUSTED EFFLUENT TCDD N> 100.00 oo o w a, % Q a o Q Q o <*H 10.00 1.00 0.10 0.01 ACT ASB Treatment ------- FIGURE 5-10 ADJUSTED SLUDGE TCDD 100.0000 oo o W 10.0000 II £ g 1.0000 a < 0.1000 § 0.0100 I* 0.0010 p 0.0001 I ACT ASB Treatment ------- FIGURE 5-11 ADJUSTED EFFLUENT TCDF oo O i 8 Q a o w 1000.00 100.00 10.00 100 0.10 0.01 ACT ASB Treatment ------- FIGURE 5-12 ADJUSTED SLUDGE TCDF tjl 1 UUU.UUUU 00 o + 100.0000 W {§ 10.0000 PQ H 1.0000 o ua <*> /\ i f\f\f\ ^ 0. 1 000 UH Q U 0.0100 H 60 *§ 0.0010 55 00001 i ' r ; 1 : " • [ 1 i * O ' ACT • -. ~. . i • ~ * : i ASB Treatment ------- TABLE 5-6. DIFFERENCES BETWEEN TREATMENT TYPES ACT vs ASB All Mills dbs/dav} * 106 Effluent TCDD ACT ASB Sludge TCDD ACT ASB Effluent TCDF ACT ASB Sludge TCDF ACT ASB Kraft Mills (lbs/dav) * 106 Effluent TCDD ACT ASB Sludge TCDD ACT ASB Effluent TCDF ACT ASB Sludge TCDF ACT ASB H 40 43 39 45 42 41 39 45 fi 28 41 28 42 29 41 28 42 Median 2.9 9.4 7.3 11.4 12.0 31.8 28.4 39.0 Median 4.5 10.3 5.8 2.0 22.8 31.8 33.7 6.6 Logged Mean t-stat 0.409 -2.583 0.820 0.566 1.245 0.324 1.111 -1.456 1.403 1.230 1.262 0.954 Logged Mean t-stat 0.625 -1.438 0.862 0.829 2.459 0.341 1.337 -0.489 1.434 1.525 2.745 0.938 p- value .012 .217 .149 .211 D- value .156 .016 .627 .008 Note: Two-sample t-tests for difference between logged means 76 ------- TABLE AIL Mills Qbs/ton ADBSP) Effluent TCDD ACT ASB Sludge TCDD ACT ASB Effluent TCDF ACT ASB Sludge TCDF ACT ASB Kraft Mills Qbs/ton ADBSP) Effluent TCDD ACT ASB Sludge TCDD ACT ASB Effluent TCDF ACT ASB Sludge TCDF ACT ASB 5-6. DIFFERENCES * 108 N 40 43 39 45 41 44 39 45 * 108 N 28 41 28 42 29 41 28 42 BETWEEN TREATMENT TYPES ACT VS Median 0.5 0.7 0.6 0.2 2.1 2.0 2.9 0.7 Median 0.6 0.9 1.0 0.2 3.1 2.0 5.0 0.8 . ASB Logged Mean t-stat -0.351 -1.201 -0.191 -0.205 2.672 -0.699 0.238 -1.074 0.436 0.458 2.462 -0.069 Logged Mean t-stat -0.219 -0.430 -0.158 -0.015 3.518 -0.687 0.489 0.388 0.415 0.681 3.612 -0.090 (CONTINUED) p- value .233 .009 .286 .016 p- value .668 .001 .699 .001 Note: Two-sample t-tests for difference between logged means 77 ------- FIGURE 5-13 TOTAL TCDD EXPORTS (Ibs/day) * E+06 ALL MILLS INCLUDED PULP oo EFFLUENT SLUDGE MATRIX SUM PULP SLUDGE EFFLUENT TOTAL 1.517 1.319 1.170 4.006 ------- FIGURE 5-14 TOTAL OUTPUT: TCDD KRAFT MILLS SULFITE MILLS PULP VO EFFLUENT PULP EFFLUENT SLUDGE SLUDGE ------- FIGURE S-1S TOTAL TCDF EXPORTS (Ibs/day) * E+06 ALL MILLS INCLUDED PULP oo o EFFLUENT MATRIX SLUDGE SUM PULP SLUDGE EFFLUENT TOTAL 14.642 8.429 9.024 32.095 ------- FIGURE 5-16 TOTAL OUTPUT: TCDF KRAFT MILLS SULFITE MILLS PULP PULP EFFLUENT SLUDGE SLUDGt EFFLUENT Note: Percentages may not add to 100% due to rounding error. ------- TABLE 5-7. STATISTICS FOR TCDD/TCDF (BY MILL PROCESS) Mill Process-Kraft TCDD Exports TCDD in Pulp (lbs/day)*106 TCDD in Sludge (lbs/day)*10s TCDD in Effluent (lbs/day)*106 Total TCDD (lbs/day)*10* Mill TCDD Exports TCDD in Pulp (lbs/day)*106 TCDD in Sludge (lbs/day)*10« TCDD in Effluent (lbs/day)*106 Total TCDD (lbs/day)*106 N 80 80 80 80 Sum 1,486 1,280 1,141 3,907 Mean 18.6 16.0 14.3 48.8 XfTotal) 38.0 32.8 29.2 100.0 Process-Sulfite H 14 14 14 14 Sum 12 22 19 53 Mean 0.9 1.6 1.4 3.8 X(Total) 23.0 40.5 36.5 100.0 Mill Process-Kraft TCDF Exports TCDF in Pulp (lbs/day)*106 TCDF in Sludge (lbs/day)*106 TCDF in Effluent (lbs/day)*106 Total TCDF (lbs/day)*10e Mill TCDF Exports TCDF in Pulp (lbs/day)*106 TCDF in Sludge (lbs/day)*10* TCDF in Effluent (lbs/day)*106 Total TCDF (lbs/day)*10s 12 80 80 80 80 Sum 13,525 7,996 8,475 29,996 Mean 169.1 100.0 105.9 374.9 JUTotaD 45.1 26.7 28.2 100.0 Process-Sulfite g 14 14 14 14 Sum 649 214 384 1,248 Mean 46.4 15.3 27.5 89.1 Xf Total) 52.0 17.1 30.8 100.0 Note: Discrepancies may result due to rounding errors. 82 ------- TABLE 5-8. STATISTICS FOR TCDD/TCDF (BY MILL PROCESS) Mill Process-Kraft TCDD Exports TCDD in Pulp (Ibs/ton ADBSP)*10a TCDD in Sludge (Ibs/ton ADBSP)*108 TCDD in Effluent (Ibs/ton ADBSP)*10B Total TCDD (Ibs/ton ADBSP)*108 H 80 80 80 80 Sum 158 119 111 388 Mean 2.0 1.5 1.4 4.9 Mill Process-Sulfite TCDD Exports TCDD in Pulp (Ibs/ton ADBSP)*108 TCDD in Sludge (Ibs/ton ADBSP)*10" TCDD in Effluent (Ibs/ton ADBSP)*10a Total TCDD (Ibs/ton ADBSP)*10e N 14 14 14 14 Sum 4 5 5 14 Mean 0.3 0.4 0.3 1.0 Mill Process-Kraft TCDF Exports TCDF in Pulp (Ibs/ton ADBSP)*10a TCDF in Sludge (Ibs/ton ADBSP)*10e TCDF in Effluent (Ibs/ton ADBSP)*10e Total TCDF (Ibs/ton ADBSP)*10e K 80 1 80 80 1 80 3 Sum ,902 819 ,145 ,866 Mean 23.8 10.2 14.3 48.3 Mill Process-Sulfite TCDF Exports TCDF in Pulp (Ibs/ton ADBSP)*108 TCDF in Sludge (Ibs/ton ADBSP)*108 TCDF in Effluent (Ibs/ton ADBSP)*108 Total TCDF (Ibs/ton ADBSP)*108 fi 14 14 14 14 Sum 97 41 55 193 Mean 6.9 2.9 4.0 13.8 40.7 30.7 28.6 100.0 30.6 36.0 33.4 100.0 49.2 21.2 29.6 100.0 Til Total) 50.3 21.1 28.7 100.0 Note: Discrepancies may result due to rounding errors. 83 ------- 6. ANALYSIS OF TOTAL SUSPENDED SOLIDS Since the preceding analysis uncovered differences between treatment types Activated Sludge Uastevater Treatment (ACT) and Aerated Stabilization Basins (ASB) with regard to the rates at which 2,3.7,8-tetrachlorodibenzo-p-dioxin (TCDD) and 2,3,7,8-tetrachlorodibenzofuran (TCDF) were exported to sludge and effluent vectors, a more extensive analysis was made on a measured variable suspected to affect wastewater treatment performance: total suspended solids (TSS). It has been suggested that ACT and ASB treatments differ significantly with regard to average TSS levels, so the goal of the analysis in section 6 was to assess any potential relationship between TCDD/TCDF formation in sludge and effluent and total suspended solids levels at the waste treatment facilities. Since important characteristics of kraft and sulfite mills were quite different, any potential relationship between TCDD/TCDF formation and TSS might be masked if both mill types were analyzed together. As it was, the number of sulfite mills was small, and only one sulfite mill with usable data employed an ASB-type waste treatment, so the analysis was confined to ACT-treated or ASB- treated kraft mills. (Please note that all figures and tables are located at the end of the text.) Preliminary examination of the TSS data indicated that the distribution of values could be approximated by a lognormal density (appendix B). A subsequent two-sample t-test on the logged TSS values indicated that the average total suspended solids content of ACT systems was significantly higher than that for ASB systems at the 5 percent level. Variation in the TSS data by treatment type is shown in the boxplot of Figure 6-1; descriptive statistics for the TSS levels are provided in Table 6-1, classified by pulping process and wastewater treatment. Given the observed difference in treatment types with respect to average TSS levels, the next step was to determine to what degree TSS levels could explain differences due to wastewater treatment in TCDD/TCDF mass outputs to sludge and effluent. Relationships between TSS and TCDD/TCDF mass exports to 84 ------- sludge and effluent were explored and tested for statistical significance. Using TSS as the independent variable, the dependent variables included TCDD/TCDF mass exports to sludge and effluent in both Ibs/day and Ibs/ton Air-Dried Brownstock Pulp (ADBSP). Examination of the dependent variables and their distributional characteristics via probability plots indicated that the TCDD/TCDF mass output variables might reasonably be characterized by lognormal distributions (appendix B). Plots were then made of TSS versus each of the dependent variables on a log- log scale, which enabled estimation of regression equations from data that resembled bivariate normal scatterclouds, a prerequisite for using normal theory estimates of the stability of the regression lines. Each of the scatterplots was overlaid with a best fitting linear regression and 90 percent confidence bands. The 90 percent confidence bands provide an approximate confidence interval for the estimated regression mean within the range of the data at each value along the independent axis. Computation of each confidence band was based upon the t-statistic for the estimated linear slope and the estimated standard error in the dependent variable at any given point Xo along the independent axis. Visual inspection of Figures 6-2 through 6-5 indicates that for any fixed TSS level, the variability from mill to mill in effluent and sludge TCDD/TCDF mass exports was substantial. The regression lines overlaying the plots estimated the average behavior of the TCDD/TCDF exports as TSS levels varied; however, none of the correlations between TSS and TCDD/TCDF exports was very strong. Clearly, TSS is not the only factor that affects amounts of TCDD/TCDF found in sludge and effluent, and it may not be a dominant factor. The estimated regression equations are presented in Tables 6-2 and 6-3. Note that since the regressions were performed on the logged data, the relationships suggested are not linear in the original units. Rather, the model implies that when the slope coefficient is significantly different from zero, the TCDD/TCDF mass output is proportional to a power of the TSS level. 85 ------- Tables 6-2 and 6-3 confirm chat the correlations between TSS and the corresponding TCOO/TCDF mass outputs were rather weak. The largest fraction of explained variance (as indicated by the Rz statistic) for any of the variables was less than 5 percent. The linear regressions suggest that TCOD/TCDF effluent mass rates increased somewhat with larger TSS levels, while TCDD/TCDF sludge mass rates decreased slightly as TSS increased. However, none of the estimated regression slopes were significantly different from zero at the 5 percent level. Very similar results were found for each matrix and analyte when considering either the unadjusted or adjusted mass export rates. Since ASB and ACT-type treatments were combined in the previous plots, the last step in this section was to subdivide mills by waste treatment and recompute possible linear relationships between TSS and the TCDD/TCDF mass exports. This was considered important primarily because the sludge samples taken at ASB facilities consisted of primary sludge only, while those at ACT facilities consisted of composites samples of primary and secondary sludges. Figures 6-6 to 6-9 are redrawings of Figures 6-2 to 6-5 that indicate the type of waste treatment used at each scatterpoint (ACT or ASB), and a regression overlay corresponding to each wastewater subgroup. The separate regression equations for each type of waste treatment are presented in Tables 6-4 through 6-7. For both wastewater treatment types, large TSS levels were somewhat associated with higher TCDD/TCDF exports to effluent and lower TCDD/TCDF exports to sludge. In each case, however, the data from ACT-type treatment facilities were more sharply sloped.than data from ASB systems. These visual results were supported by the regression statistics listed in Tables 6-4 through 6-7. None of the estimated slopes for the ASB mills were significant at the 5 percent level; however, several of the relationships between TSS and TCDD/TCDF exports to sludge and effluent were significant for ACT mills. Again, the estimated correlations were weak, but in some cases total suspended solids accounted for close to 20 percent of the total variability in TCDD/TCDF mass sludge and effluent exports at mills using ACT treatment. Based on this analysis, it is difficult to determine whether TSS influences the proportions of TCDD/TCDF mass exported to sludge and effluent vectors. The 86 ------- proportion of total variation in the TCDD/TCDF data explained by the TSS level (through the R* statistic) did not exceed 20 percent for any of the regressions calculated. It is also possible that other variables were present in these data that might have masked relationships between TSS and TCDD/TCDF exports. The study design did not permit a more complete analysis. However, there did appear at least a weak link between the TSS level and the TCDD/TCDF sludge and effluent export rates for kraft mills using ACT-type wastewater facilities. If such a link exists, the level of TSS may help to explain the observed differences between ASB and ACT waste treatments with respect to TCDD/TCDF found in sludge and effluent. 87 ------- FIGURE 6-1 1000 r TSS BY TREATMENT TREATED KRAFT MILLS ONLY oo oo 100 to a 10 I T I ACT ASB Waste Treatment ------- TABLE 6-1. DESCRIPTIVE STATISTICS FOR TSS All Hills Kraft Mill* APT ASB H 81 67 25 42 Mean 61. SO 52.61 60.02 48.20 Std 50.48 36.19 34.40 36.91 Minimum 5.800 5.800 14.400 5.800 (laxlmun 273.00 144.60 144.60 143.80 Lower Quartile 25.63 22.40 41.90 18.95 Median 46.30 45.80 47.20 35.70 Upper Quartile 81.15 70.00 78.25 69.88 90UI Percent lie 126.72 115.40 119.80 112.26 Sulfite Mllla 12 111.85 85.69 26.800 273.00 32.44 87.05 182.20 264.18 oo ------- FIGURE 6-2 EFFLUENT TCDD OUTPUT TREATED KRAFT MILLS ONLY vD O 1000.0 ! 1000 «J £ Q Q 10.0 1.0 w 0.1 I i I I I I i 10 100 1000 TSS (mg/l) ------- FIGURE 6-3 SLUDGE TCDD OUTPUT TREATED KRAFT MILLS ONLY 1000.000 E 100.000 10000 rt •o 1000 Q Q H 0100 o 60 •O 0010 0001 • • • 1111 10 100 1000 TSS (mg/l) ------- FIGURE 6-4 EFFLUENT TCDF OUTPUT TREATED KRAFT MILLS ONLY o + U) cd "O Q U H «_• a U 100000 E 10000 1000 100 1.0 0.1 10 100 1000 TSS (mg/1) ------- FIGURE 6-5 SLUDGE TCDF OUTPUT TREATED KRAFT MILLS ONLY 1000000 100000 o + u 10000 ^% ed 5 u» § 10.00 UH Q 1.00 o t>o •o p 0.10 001 1 ' '"I 10 100 • • • • I ll 1000 TSS (mg/1) ------- TABLE 6-2: TCDD EXPORTS (TREATED KRAFT MILLS ONLY) TSS rag/1) vs Sludge TCDD dbs/dav)*10* Equation: Log10(Sludge TCDD) - 1.227 - 0.431 * Log10(TSS) Rz - .022 Adjusted R2 - .006 S.E. of Regression - 0.933 Standard Error t Statistic p-Value Constant 0.596 2.059 0.044 Independent 0.363 -1.187 0.240 TSS (mg/n vs Effluent TCDD (Ibs/dav^lO* Equation: Log10(Effluent TCDD) - 0.315 + 0.268 * Log10(TSS) R2 - .014 Adjusted R2 - .000 S.E. of Regression - 0.687 Standard Error t Statistic p-Value Constant 0.461 0.684 0.497 Independent 0.281 0.953 0.344 TSS (me/1) vs Adlusted Sludge TCDD fibs/ton ADBSP)*10" Equation: Log10(Adjusted Sludge TCDD) - 0.157 - 0.373 * Log10(TSS) R2 - .016 Adjusted R2 - .000 S.E. of Regression - 0.961 Standard Error t Statistic p-Value Constant 0.614 0.256 0.798 Independent 0.374 -0.998 0.322 TSS (mf/1) vs Adlusted Effluent TCDD (Ibs/ton ADBSP)*10a Equation: Loglo(Adjusted Effluent TCDD) - -0.713 + 0.311 * Log10(TSS) Rz - .026 Adjusted R2 - .010 S.E. of Regression - 0.589 Standard Error t Statistic p-Value Constant 0.396 -1.802 0.076 Independent 0.241 1.290 0.202 94 ------- TABLE 6-3. TCDF EXPORTS (TREATED KRAFT MILLS ONLY) TSS fmg/1) vs Sludge TCDF abs/dav)*106 Equation: Logw(Sludge TCDF) - L.599 - 0.277 * Logw(TSS) R2 - .008 Adjusted R2 - .000 S.E. of Regression - 1.010 Standard Error t Statistic p-Value Constant 0.645 2.480 0.016 Independent 0.393 -0.704 0.484 TSS (mq/n vs Effluent TCDF flbs/dav)*106 Equation: Log10(Effluent TCDF) - 0.538 + 0.499 * Log10(TSS) Rz - .037 Adjusted R2 - .022 S.E. of Regression - 0.787 Standard Error t Statistic p-Value Constant 0.528 1.018 0.313 Independent 0.322 1.553 0.126 TSS fmg/n vs Adjusted Sludge TCDF fibs/ton ADBSP)*10a Equation: Log10(Adjusted Sludge TCDF) - 0.530 - 0.2L9 * Log10(TSS) R2 - .004 Adjusted R2 - .000 S.E. of Regression - 1.066 Standard Error t Statistic p-Value Constant 0.681 0.778 0.440 Independent 0.415 -0.527 0.600 TSS (mtt/1) va Adjusted Effluent TCDF (Ibs/ton ADBSP1*108 Equation: Log10(Adjusted Effluent TCDF) - -0.491 + 0.542 * Loglo(TSS) R2 - .048 Adjusted R2 - .032 S.E. of Regression - 0.751 Standard Error t Statistic p-Value Constant 0.505 -0.972 0.335 Independent 0.307 1.765 0.082 95 ------- FIGURE 6-6 EFFLUENT TCDD OUTPUT BY TREATMENT TREATED KRAFT MILLS ONLY V£> + w l» X) Q Q U H W 1000.00 100.00 1000 1 00 o.io 001 9 ASB o ACT 10 100 1000 TSS (mg/1) ------- FIGURE 6-7 SLUDGE TCDD OUTPUT BY TREATMENT TREATED KRAFT MILLS ONLY o i 03 •o Q Q U H 100000 10000 1000 1 00 0.10 001 ASB ACT 10 100 1000 TSS (mg/1) ------- FIGURE 6-8 EFFLUENT TCDF OUTPUT BY TREATMENT TREATED KRAFT MILLS ONLY 1000000 PL) 1000.00 10000 01 X) 1000 u, Q U 1.00 a 0 0.10 001 0 ASB o ACT 10 100 1000 TSS (mg/1) ------- FIGURE 6-9 SLUDGE TCDF OUTPUT BY TREATMENT TREATED KRAFT MILLS ONLY 1000000 1000.00 o w ~ 100 00 n> •a £ 1000 UH Q U o oo •a 1.00 010 001 10 100 0 ASB o ACT 1000 TSS (mg/1) ------- TABLE 6-4. TCDD EXPORTS FOR ACT TREATMENT KRAFT MILLS ONLY TSS fmg/1) vs Sludae TCDD fIbs/dav^*10* Equation: Log 10(Sludge TCDD) - 2.388 - 0.922 * Log,0(TSS) R2 - .113 Adjusted R2 - .073 S.E. of Regression - 0.661 Standard Error t Statistic p-Value Constant 0.939 2.542 0.019 Independent 0.551 -1.675 0.108 TSS (me/1) vs Effluent TCDD (lbs/day)*106 Equation: Log10(Effluent TCDD) - -0.969 + 0.925 * Loglo(TSS) Rz - .140 Adjusted R2 - .101 S.E. of Regression - 0.587 Standard Error t Statistic p-Value Constant 0.834 -1.162 0.258 Independent 0.489 1.892 0.072 TSS fmg/1) vs Adlusted Sludge TCDD fibs/ton ADBSP'>*10a Equation: Log,0(Adjusted Sludge TCDD) - 1.605 - 0.966 * Log10(TSS) R2 - .165 Adjusted R2 - .127 S.E. of Regression - 0.556 Standard Error t Statistic p-Value Constant 0.790 2.031 0.054 Independent 0.463 -2.085 0.049 TSS fmg/n vs Adlusted Effluent TCDD aba/ton ADBSP1*10e Equation: Logu(Adjusted Effluent TCDD) • -1.752 + 0.882 * Log 10(TSS) R2 - .201 Adjusted R2 - .164 S.E. of Regression - 0.451 Standard Error t Statistic o-Value Constant 0.640 -2.736 0.012 Independent 0.375 2.350 0.028 100 ------- TABLE 6-5. TCDF EXPORTS FOR ACT TREATMENT KRAFT KILLS ONLY TSS fmg/11) vs Sludge TCPF Equation: Logw ------- TABLE 6-6. TCDD EXPORTS FOR ASB TREATMENT KRAFT MILLS ONLY TSS fmy/1) vs Sludge TCDD f lbs/dav)*106 Equation: Log10(Sludge TCDD) - 1.128 - 0.495 * Logw(TSS) R2 - .029 Adjusted R2 - .004 S.E. of Regression - 1.023 Standard Error t Statistic p- Value Constant 0.738 1.527 0.135 Independent 0.462 -1.073 0.290 TSS (me/I) vs Effluent TCDD abs/dav)*10s Equation: Log ,0( Effluent TCDD) • 0.582 + 0.164 * Log10(TSS) Rz - .006 Adjusted R2 - .000 S.E. of Regression - 0.723 Standard Error t Statistic p-Value Constant 0.557 1.045 0.303 Independent 0.348 0.472 0.639 TSS (ma/1) vs Adlusted Sludge TCDD (1 PS /ton ADBSP)*10e Equation: Log10( Ad justed Sludge TCDD) - 0.056 - 0.481 * Logu(TSS) R2 - .026 Adjusted R2 - .001 S.E. of Regression - 1.053 Standard Error t Statistic p- Value Constant 0.760 0.074 0.941 Independent 0.475 -1.012 0.318 TSS (rny/n va Adlusted Effluent TCDD fibs/ton ADBSP)*10" Equation: Log10(Adjusted Effluent TCDD) • -0.447 + 0.169 * Log 10 (TSS) R2 - .008 Adjusted R2 - .000 S.E. of Regression - 0.654 Standard Error t Statistic D- Value Constant 0.504 -0.886 0.381 Independent 0.315 0.538 0.594 102 ------- TABLE 6-7. TCDF EXPORTS FOR ASB TREATMENT KRAFT MILLS ONLY TSS (mg/n vs Sludge TCDF ( lbs/dav)*106 Equation: Log,0(Sludge TCDF) - 1.425 - 0.312 * Log10(TSS> Rz - .010 Adjusted R2 - .000 S.E. of Regression - 1.106 Standard Error t Statistic D- Value Constant 0.798 1.785 0.082 Independent 0.499 -0.625 0.536 TSS fmg/n vs Effluent TCDF flbs/dav^*10* Equation: Log10(Ef fluent TCDF) - 0.778 + 0.393 * Log,0(TSS) R2 - .022 Adjusted R2 - .000 S.E. of Regression - 0.879 Standard Error t Statistic p- Value Constant 0.677 1.148 0.258 Independent 0.423 0.929 0.359 TSS fmq/n vs Adjusted Sludge TCDF dbs/ton ADBSP)*10» Equation: Log10(Adjusted Sludge TCDF) • 0.353 - 0.298 * Log10(TSS) R2 - .008 Adjusted R2 - .000 S.E. of Regression - 1.162 Standard Error t Statistic p -Value Constant 0.839 0.421 0.676 Independent 0.525 -0.567 0.574 TSS (mq/n vs Adjusted Effluent TCDF dbs/ton ADBSP)*10e Equation: Log10(AdJusted Effluent TCDF) - -0.251 + 0.398 * Loglo(TSS) R2 - .024 Adjusted R2 - .000 S.E. of Regression - 0.857 Standard Error t Statistic p- Value Constant 0.661 -0.380 0.706 Independent 0.412 0.965 0.341 103 ------- 7. MODELING TCDD/TCDF FORMATION AS A FUNCTION OF MILL OPERATING PARAMETERS Several steps were taken to investigate the effect of mill bleaching procedures upon 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) and 2,3,7,8- tetrachlorodibenzofuran (TCDF) formation. The goal of this section was to determine the strength of relationships between mass export rates of TCDD/TCDF and key chemical bleaching and extraction agents used at U.S. bleached pulp mills. Three dependent measures were used, including the total mass export rates of TCDD and TCDF generated by the combined vectors of pulp, sludge, and effluent (in Ibs/ton Air-Dried Brownstock Pulp [ADBSP]); and the TCDD toxic equivalent export rate, which combines the TCDD total mass rate with one-tenth of the TCDF total mass rate. Though the mass formation rates of TCDD/TCDF varied from bleach line to bleach line, as gauged by pulp sample analyses, effluents and sludges were not sampled at each line but rather at the "downstream" treatment facilities. Consequently, the chemical bleaching application rates for each bleach line were combined to form a mill average, the rates being weighted over different lines depending on the volume of pulp produced. As in the previous section, kraft and sulfite mills were treated separately in the analyses. Since the number of sulfite mills with usable data was quite small, only the analyses of kraft mills were included in this section. The independent variables for which there were enough data to be of utility included the following: chemicals added during C-stage bleaching -- Chlorine (C12), Chlorine Dioxide (C102), C12 Equivalent in C-Stage, and Percentage C10Z Substitution for C12; chemicals added during other stages of bleaching or caustic removal -- Other stage C102, Sodium Hypochlorite, Sodium Hydroxide, and Oxygen (02); and characterizing features of bleach line operation -- Kappa number, Final brightness, C12 Line Equivalent, C12 Multiple (Kappa Factor) in C-stage, C12 Equivalent Multiple in C-stage, and C12 Line Equivalent Multiple. Other variables had for the most part zero values and were not included in these 104 ------- analyses. They included Calcium Hypochlorice, Hydrogen Peroxide, Other Scage C12, and ocher chemical agents which did not contain chlorine derivatives. As was done in the analysis of total suspended solids, exploratory plocs and regression analyses were performed only after the variables of interest were examined for distributional properties and skewness. If warranted, variables were transformed so that their distributions approximated normality as much as possible. (All figures and tables are located at the end of the text.) Two of the independent variables -• 02 and C102 -- contained significant fractions of zero values (almost half of all kraft mills in the case of 02). The analyses assumed an inherent difference between mills vhich, for instance, did not use any C102 in bleaching and those mills which did. Two different distributions of the TCOD/TCDF mass export rates are presented for each of these variables, one for all cases of zero values in 02 and C102 and the other for cases when the two variables were positive (Tables 7-1 and 7-2). 7.1 REGRESSION ANALYSES After analyzing and transforming variables where necessary, plots were made of each dependent measure versus each independent variable and then analyzed for trends. Figures 7-1 to 7-9 are representative of the most significant results. Each plot contains two important interpretive features: a least squares linear regression overlay, drawn over the actual range of data, and a 90 percent confidence band about the estimated regression line. The confidence band provides a visual indication of the degree to which, at any given point x, along the independent axis, the estimated mean of the dependent variate might be in error. Hills in which the calculation of either TCDD or TCDF mass export rates was problematic (such as in cases of seasonal or no waste treatment) were not used in the scatterplots or regression analyses and were considered unreliable data for purposes of the report. Two mills discharged untreated effluents to the ocean, and another five mills had average wastewater retention spans of 105 ------- several months. Ac six mills, Che reported concentration or flow data was incomplete, so TCDD/TCDF mass formation rates could not be calculated. Corresponding co the above plots, equations of che regression lines and relevant summary statistics (including standard errors and Rz values) are given in Tables 7-3 to 7-5. Since the regressions were performed on che transformed variables and not in the original units, the estimated relationships are not linear in the original variables. On the log-log scale, for example, a non- zero linear slope implies that che dependent variable tends to be proportional to a power of the independent variace. The most immediate finding from che analysis is that each of che dependent variables exhibited significant variation at essentially every level of the various chemical application races. Consequencly, Che proporcion of variance explained by any of che regression equations was generally low (as given by R2), indicating that che linear regressions were noc very useful as predictive equations. In face, specific predictions regarding output of TCOD/TCDF at mill Y when a certain level of chemical X was applied would probably have little meaning. The scaccerplocs were useful, however, Co detect che presence or absence of non-zero trends in che estimated regression lines. 7.1.1 Effects of Chlorine Bleaching Variables measuring che applicacion of chlorine Co brownscock pulps (C12, Clj Equivalent in C-SCage, C12 Line Equivalent) were positively associated with che formation of TCDD/TCDF (Table 7*3). Hence, greater use of chlorine in bleaching was associated with higher formation races of TCDD/TCDF. This result was consistent with previous evidence concerning che effect of chlorine bleaching on TCDD/TCDF formation in pulp mills (2); however, none of che estimated regression models involving these variables accounted for more than about 30 percent of the total variance in TCDD/TCDF mass export rates. 106 ------- 7.1.2 Effect of che Chlorine Multiple Since more chlorine cends co be applied when che ILgnin concent of che pulp is high, regressions were also estimated for variables involving ratios between the amount of chlorine applied and the Kappa number (as measured by the ratios Cl, Multiple, C12 Equivalent Multiple, and Cl, Line Equivalent Multiple), the Kappa number being a useful index of lignin content in brownscock pulps. Table 7-4 provides che results for regressions on che C12 Multiple, and again documents a generally significant positive relationship between formation of TCDD/TCDF in mass exports and the Cl, Multiple. Such a result implies chat, on the average, even when lignin content was accounted for or "held constant," greater application of chlorine was mildly associated with higher formation of TCDD/TCDF In this case, the association must be considered mild because che percencage of total variation accounted for by the estimated regression models never exceeded 18 percent. 7.1.3 Chlorine Dioxide Substitution The substitution of C10Z for Clz in Che C-Stage of bleaching produced slight reductions in average TCDD/TCDF formation (Table 7-5), the regression trends being statistically significant at below the 2 percent level. However. the regression models accounted for at most 16 percent of the total variation in TCDD/TCDF mass exports, and since very few mills substituted CIO, for more than 30 percent of their chlorine usage, the regression trends sajapj; be reliably extrapolated to predict reductions of TCDD/TCDF formation at higher C102 substitution races. It was also seen in Table 7-1 that mills that did not use any CIO, exhibited tremendous variation in TCDD/TCDF mass exports. Hence, substitution of CIO, for Cl, was not by itself an adequate predictor of TCDD/TCDF reduction. Use of CIO, may help, however, to reduce TCDD/TCDF formation when considered in conjunction with other reduction strategies. 7.1.4 Use of Oxygen in Bleaching Mills that use oxygen in the bleaching process exhibited a slight but statistically significant trend toward reduction of TCDD/TCDF with increased 107 ------- oxygen application. However, this trend was wholly attributable to those four kraft uLlls that used oxygen dellgnificatlon tðods at the cine of the 104 Mill Study (Table 7-2). Furthermore, the same four mills also tended to have higher substitution rates of C10Z for Cl,, so it cannot be determined whether the lower export rates of TCDD/TCDF observed at these mills were attributable to oxygen delignification, chlorine dioxide substitution, or some combination of both. Use of oxygen in other applications was not statistically correlated with TCDD/TCDF mass formation. 7.1.5 Differences in Wood Types Due to limitations of the study design, softwood and hardwood bleach lines could not be systematically analyzed for differences in TCDD/TCDF mass formation. However, it was observed that greater amounts of chlorine were generally applied to softwood pulps than hardwood pulps per ton of pulp processed, and that the average Kappa numbers of softwood pulps were typically much higher than the Kappa numbers of hardwood pulps (Figures 7-10 and 7-11). Both of these observations were consistent with known differences in the.bleaching practices of softwood versus hardwood pulps. 7.2 SUMMARY To summarize, the most consistently significant independent variables were those involving chlorine application in the C-stage of bleaching: Cl, and Cl, Equivalent. Variables measuring the chlorine multiple (also known as the Kappa factor) were also positively associated with TCDD/TCDF formation, though the correlations were weaker. Substitution of chlorine dioxide for Clz was associated with slight reductions in TCDD/TCDF formation. However, since very few mills reported CIO, substitution rates of more than 30 percent at the time of the study, the effect of higher chlorine dioxide substitution rates could not be gauged with any precision. Barring more detailed information on chemical usage patterns and mill process characteristics, the data at hand preclude the fitting of very precise predictive models. While other variables might significantly impact the 108 ------- formation of 2378-TCDD/TCDF, in che 104 Hill Study only chose measuring chlorine applicacion races were consistently linked to TCDD/TCDF formation at pulp mills 109 ------- TABLE 7-1. SUMMARY STATISTICS: BREAKDOWN BY CIO, USAGE KRAFT MILLS ONLY C10T - 0 N Minimum Maximum Mean Standard Dev. Median Adjusted Adjusted Total TCDD Total TCDF 27 0.186 16.337 .110 .260 4. 4. 2.433 27 0.748 299.613 27.940 61.417 8.228 Adjusted TCDD Toxic Equivaleac 27 0.260 43.026 6.904 9.433 3.256 C10 N Adjusted Total TCDD 52 Adjusted Total TCDF 52 Adjusted TCDD Toxic Equivalent 52 Minimum Maximum Mean Standard Dev. Median 0.066 30.556 5.331 6.152 3.437 0.147 953.875 59.818 149.441 16.088 0.081 118.722 11.313 19.996 4.963 Adjusted Total - Ibs/ton ADBSP * 10s Adjusted TCDD Toxic Equivalent • Ibs/ton ADBSF * 108 110 ------- TABLE 7-2. SUMMARY STATISTICS: BREAKDOWN BY Oa USAGE KRAFT MILLS ONLY N Minimum Maximum Mean Standard Dev. Median Adjusted Adjusted Total TCDD Total TCDF 34 0.117 13.065 .764 .603 3. 3. 2.068 34 0.363 299.613 27.054 55.415 7.946 Adjusted TCDD Toxic Equivalent 34 0.153 43.026 6.469 8.492 2.807 02 > 0 Extraction N Minimum Maximum Mean Standard Dev. Median 02 > 0 Delienification N Minimum Maximum Mean Standard Dev. Median Adjusted Total TCDD 43 0.124 30.556 6.028 6.659 3.589 Adjusted Total TCDD 2 0.066 0.960 0.513 0.632 0.513 Adjusted Total TCDF 43 0.450 953.875 68.447 163.044 15.778 Adjusted Total TCDF 2 0.147 1.747 0.947 1.131 0.947 Adjusted TCDD Toxic Equivalent 43 0.283 118.722 12.872 21.668 5.153 Adjusted TCDD Toxic Equivalent 0.081 1.135 0.608 0.745 0.608 Adjusted Total - Ibs/ton ADBSP * 108 Adjusted TCDD Toxic Equivalent - Ibs/ton ADBSP * 108 111 ------- FIGURE 7-1 C12 vs. ADJUSTED TOTAL TCDD KRAFT MILLS ONLY o W (X, £ p a o 100 00 1000 = 100 P P •a u 010 001 50 100 150 200 C12 (Ibs/ton ADBSP) ------- FIGURE 7-2 C12 vs. ADJUSTED TOTAL TCDF KRAFT MILLS ONLY u> oo O + UJ PH CO PQ Q § u- Q rt u •>- UJ 1000.0 1000 100 10 01 0 50 100 150 200 C12 (Ibs/ton ADBSP) ------- FIGURE 7-3 C12 vs ADJUSTED TCDD TOXIC EQUIVALENT KRAFT MILLS ONLY w 1000.00 100.00 10.00 a ^w p 0 100 0.10 001 0 50 100 150 200 C12 (Ibs/ton ADBSP) ------- FIGURE 7-4 C12 MULTIPLE vs. ADJUSTED TOTAL TCDD KRAFT MILLS ONLY 00 O + w PH 00 03 0 a o «-• CO O Q U H •o o «-> (A 10000 1000 1 00 010 001 00 0.1 02 03 04 C12 Multiple ------- FIGURE 7-5 C12 MULTIPLE vs. ADJUSTED TOTAL TCDF KRAFT MILLS ONLY oo o 1000.0 Q a o W) S3 1000 3 100 Q § 1.0 0.1 00 0.1 02 03 04 C12 Multiple ------- FIGURE 7-6 C12 MULTIPLE vs ADJUSTED TCDD TOXIC EQUIVALENT KRAFT MILLS ONLY <§ 1000.00 |