Variables Affecting Emissions of PCDD/Fs from Uncontrolled Combustion of Household Waste in Barrels

Transcription

1 TECHNICAL PAPER ISSN J. Air & Waste Manage. Assoc. 53: Copyright 2003 Air & Waste Management Association Variables Affecting Emissions of PCDD/Fs from Uncontrolled Combustion of Household Waste in Barrels Paul M. Lemieux and Brian K. Gullett National Risk Management Research Laboratory, Office of Research and Development, U.S. Environmental Protection Agency, Research Triangle Park, North Carolina Christopher C. Lutes and Chris K. Winterrowd ARCADIS, Inc., Research Triangle Park, North Carolina Dwain L. Winters Office of Pollution Prevention and Toxics, U.S. Environmental Protection Agency, Washington, D.C. ABSTRACT The uncontrolled burning of household waste in barrels has recently been implicated as a major source of airborne emissions of polychlorinated dibenzo-p-dioxins and polychlorinated dibenzofurans (PCDD/Fs). A detailed, systematic study to understand the variables affecting emissions of PCDD/Fs from burn barrels was performed. The waste composition, fullness of the barrel, and the combustion conditions within the barrel all contribute significantly to determining the emissions of PCDD/Fs from burn barrels. The study found no statistically significant effect on emissions from the Cl content of waste except at high levels, which are not representative of typical household waste. At these elevated Cl concentrations, the impact of Cl on PCDD/F emissions was found to be independent of the form of the Cl (inorganic or organic). For typical burn conditions, most of the PCDD/F emissions appear to be associated with the later stages of the burn when the waste is smoldering. Polychlorinated biphenyls (PCBs) were also measured for a subset of the tests. For the nominal waste composition, the average emissions were 76.8 ng IMPLICATIONS The uncontrolled burning of household waste in barrels is an important source of PCDD/Fs. In addition, barrel burning is a common waste management practice in many parts of the developing world. This paper reports on a systematic study to refine the emission factor from barrel burning so that it can be used in the quantitative emissions inventory for air emissions of PCDD/Fs in the United States. Results from this paper, when coupled with estimates of activity, can be used by environmental officials to estimate emissions from barrel burning on a local or national basis. toxic equivalency units (TEQ) WHO98 /kg of waste combusted, which suggests that uncontrolled burning of household waste could be a major source of airborne PCDD/Fs in the United States. INTRODUCTION The uncontrolled burning of household waste in barrels is practiced in many rural areas of the United States when no local waste collection is available and is one of the primary waste management techniques in many parts of the developing world. This activity typically consists of placing household waste in a 208-L (55-gal) drum called a burn barrel and, when a sufficient quantity of household waste has accumulated, igniting the waste and burning it. Three previous studies characterized emissions associated with open burning of residential refuse in a backyard burn barrel. 1 4 Results from these studies indicated that burn barrels could potentially be a major national source of airborne polychlorinated dibenzo-p-dioxins and polychlorinated dibenzofurans (PCDD/Fs), given conservative estimates of the frequency of this practice. However, variability in the PCDD/F emissions of several orders of magnitude was found between seemingly duplicate runs. Although apparent relationships between PCDD/F emissions and airborne HCl and Cu emissions were observed, sufficient data were not available to conduct rigorous statistical analyses and determine whether a causal relationship existed. Many possible parameters could have a significant influence on PCDD/F emissions from burn barrels. Many of these parameters could be caused by variations in practice-related variables that would vary from homeowner to homeowner. Some of these parameters include physical Volume 53 May 2003 Journal of the Air & Waste Management Association 523

2 condition of the waste in the barrel (e.g., fullness of the barrel, degree of compression of the waste, distribution of waste components within the barrel), chemical composition of the waste (e.g., wetness, trace metal content, Cl content, organic vs. inorganic Cl), and combustion conditions resulting from variations in the previously mentioned physical and chemical characteristics. None of these have been examined in detail in the past. The limited amount of data and high degree of variability confounded efforts to incorporate burn barrels into the U.S. PCDD/F inventory, reported in toxic equivalency (TEQ) units. The national emissions from backyard burn barrel sources were estimated to be greater than 1000 g TEQ/yr, although the uncertainty in this estimate was too great for it to be included in the U.S. Environmental Protection Agency s (EPA) quantitative inventory of PCDD/Fs. 5 In an effort to reduce the uncertainty in the emissions estimation, additional testing was performed on burn barrels so that the emissions of PCDD/Fs could be characterized as a function of waste composition, burn conditions, and other physical properties of the waste in the barrels (e.g., degree of compaction and wetness). Initial results from these follow-up tests 6 9 showed that waste composition parameters (e.g., Cl), combustion conditions (e.g., barrel temperature distributions), and resultant emissions (e.g., Cu and CO) can account for a significant portion of the variability of PCDD/F emissions between runs. PCDD/F TEQ values bracketed the 140 ng TEQ/kg used by EPA in its preliminary estimate of burn barrel emissions as a part of its 1998 draft dioxin source inventory. 10 The critical questions that this study attempted to answer are (1) what is the emission factor for PCDD/Fs from barrel burning; and (2) given that various operating conditions and compositions are used in this practice, do they influence emissions? The study attempts to evaluate the representativeness of the testing methodology by performing a preliminary sensitivity analysis on the effect of waste composition. EXPERIMENTAL METHODS In an effort to simulate burn barrel emissions and determine a representative emission factor, experiments varying the composition and burn conditions, coupled with PCDD/F analyses, were conducted to determine which of these variables were important. These experiments resulted in detailed analyses from 25 burn barrel tests and three blank tests that were performed at EPA s Open Burning Test Facility (OBTF). 3 The OBTF consists of an enclosed structure with a measured volumetric influx of ambient air, a weigh scale where the mass of the burning material is continuously weighed, and various sampling devices, including continuous emission monitors (CEMs), thermocouples (TCs), and ambient organic sampling equipment. To represent common practice for residential waste burning, the test container consisted of a seasoned, 208-L (55-gal) steel barrel with 12 2-cm-diameter ventilation holes evenly spaced around the base. The barrel was initially sandblasted 3 to remove residual paint and any remaining contents that might affect emissions. It was placed on an electronic scale platform to allow the mass consumed by combustion to be continuously monitored. An aluminum skirt was placed around the outer circumference of the barrel to minimize the potential for recirculation of combustion gases back through the air inlet holes. High-volume air handlers provided 52.7 m 3 /min (1862 ft 3 /min) of metered dilution air into the burn hut to simulate ambient mixing. Additional fans were set up inside the burn hut to enhance recirculation within the hut. Most tests consisted of 6.8 kg (15 lb) of waste, randomly mixed for a brief time in a portable concrete mixer and dumped en masse into the test container before burning. Six TCs were mounted inside the barrel at various heights. A baseline waste composition was developed containing materials in quantities representative of domestic waste based on a New York State Department of Environmental Conservation survey. The average Cl content of the baseline waste (approximately 0.2%) was based on a family that removes most of their plastic materials from the waste before combustion 3 and was somewhat less than what is typically fed into a municipal waste combustion (MWC) facility (approximately 0.5% Cl). 11 It must be noted that this waste composition does not necessarily reflect rural waste compositions but reflects that of the general public. Eight replicate tests using the baseline (see Table 1) composition were performed so the inherent variability of the burn barrel combustion process could be characterized. Because it was believed that waste Cl content might have a significant impact on PCDD/F emissions, several runs were performed using varying waste Cl levels. Polyvinyl chloride (PVC) plastic was used most of the time to vary the waste Cl content because it is relatively easy to distribute uniformly through the waste. The waste Cl content was perturbed from the baseline Cl content with three PVC levels (0, 1, and 7.5% by weight) using pipe forms. When PVC levels were varied, the bulk waste heating value was adjusted by substituting high-density polyethylene and Fe (both also in pipe form), in an effort to approximate consistent physical and energy properties of the waste across all batches while varying Cl. For two experiments, however, an inorganic Cl source was used instead of PVC. Inorganic Cl levels were derived by 524 Journal of the Air & Waste Management Association Volume 53 May 2003

3 Table 1. Waste composition. Waste Category Waste Description Target wt % Paper Newspaper, books, office paper 32.8 Magazines and junk mail 11 Corrugated cardboard, Kraft paper 7.6 Paperboard, milk cartons, drink boxes 10.3 Plastic resin Polyethylene terephthalate (PET) #1, soda bottles 0.6 High-density polyethylene (HDPE) #2, detergent bottles, pieces 6.6 Polyvinyl chloride (PVC) #3, schedule 40 pipe 0.2 Polystyrene (PS) #6, food trays 0.1 Mixed #7, Poly-Fil polyester 0.1 Food Frozen processed potatoes 5.7 Textile/leather Rubber and leather sneakers 3.7 Wood Chipboard, plywood 1.1 Glass/ceramics Bottles, jars 9.7 Broken ceramics, flower pots 0.4 Metals (ferrous) Iron (cans), dog food cans 7.3 Metals (nonferrous) Aluminum cans, foil, soda cans 1.7 Wire, Cu pipe, batteries 1.1 Total 100 Note: Inorganic Cl tests were conducted with CaCl 2 -saturated newspapers (using Prestone Driveway Heat) such that [Cl] 7.5 wt %. HDPE #2 3.3 wt %, g; PVC 0 wt %, 0 g; iron cans 3.3 wt %, g; the 0 wt % PVC test consisted of HDPE #2 6.7 wt %, g; PVC 0 wt %, iron cans 7.4 wt %, g; the high Cu mix test consisted of bottles/jars 8.7 wt %, g; iron cans 6.4 wt %, g; wire, Cu pipe, batteries 3 wt %, g. soaking the paper portion of the waste in a CaCl 2 -based deicer followed by drying. Tests were also performed to evaluate the combustion characteristics of waste with a higher moisture content, which might occur if the waste was rained on before combustion. For those tests, some of the newspaper was soaked in water before combustion. These test conditions were designated wetted. To simulate other commonly expected practices, tests (designated compacted ) were also performed in which the waste was compacted by allowing a 91-kg (200-lb) person to stand on the bed of waste in the barrel before ignition. Additional test conditions (designated double ) used twice the normal mass of waste, which also required compaction so the larger quantity of waste would fit into the barrel. A set of tests (designated high Cu ) were performed in which additional Cu was added in the form of short pieces of bare Cu wire. (Note that a small amount of Cu wire was contained in the baseline mix as well.) Finally, a single test (designated open pile ) was performed where the baseline composition waste was burned in an open pile as opposed to inside a barrel. The experimental facility and experimental procedures were the same as reported before 3,4 except (1) the ceiling of the hut was lined with aluminum foil to reduce the danger of the hut s catching fire because of the more vigorous combustion that resulted from the double charges; (2) the PCDD/F samples were collected using Method TO9 12 and were analyzed using high-resolution gas chromatography and high-resolution mass spectrometry using EPA Test Method 8290; 13 and (3) several tests were performed where the PCDD/F sampling occurred for three consecutive 30-min intervals to understand the temporal emissions of PCDD/Fs. For those runs, three sampling trains were set up, and the pumps were started and stopped so their samples reflected elapsed times of 0 30, 31 60, and min. The barrel was emptied between runs but was not washed. Assuming that the gases inside the burn hut are perfectly mixed, estimated emissions of PCDD/Fs per unit mass burned can be calculated using eq 1: E C sample Q hut t run m burned (1) where E the estimated emissions in ng/kg waste consumed, C sample the concentration of the pollutant in the sample (ng/m 3 ), Q hut the flow rate of dilution air into the burn hut (52.7 m 3 /min), t run the run time (30 or 90 min), and m burned the mass of waste burned (kg). When analyzing and reporting the results, all nondetects (NDs) and incidences of questionable analytes were set to zero. TEQ emission values caused by PCDD/Fs (TEQ DF ) were calculated using toxic equivalence factors (TEFs) from the World Health Organization (WHO). 14 RESULTS PCDD/F and PCB Data Table 2 lists the total PCDD/F and TEQ DF emissions in terms of ng emitted/kg of waste burned (calculated using eq 1), as well as the total polychlorinated biphenyls (PCBs) and TEQ PCB calculated using the WHO TEFs 14 for those runs where PCB data were acquired. For most runs, PCDD/F isomers were present in quantities above the detection limits, which results in TEQ DF data that are not based on how the nondetects are handled. There were also relatively few instances of estimated maximum possible concentrations (EMPCs) in the data set. EMPCs are cases where a signal is seen in both of the dioxin- or furan-selected ion-monitoring channels of the gas chromatograph/mass spectrometer, but the ratio of areas between the channels for the given analyte does not meet the acceptance ratio. This can indicate a positive interference in the channel potentially overlapping with the signal from the PCDD/F. Nondetects and EMPCs were set to zero for the calculations of total concentrations and TEQs reported in this work, which is a conservative, low-emission assumption. Volume 53 May 2003 Journal of the Air & Waste Management Association 525

4 Table 2. PCDD/F and PCB estimated emissions (ng/kg waste burned and ng TEQ WHO98 /kg waste burned). Run Description TEQ DF Total PCDD PCDF TEQ PCB PCB Total A Baseline , ,663 B Baseline 84 4, ,712 C Baseline 25 1, ,877 K 0% PVC ,411 D Baseline ,869 L 1% PVC , ,354 M 1% PVC , ,452 O 7.5% PVC 3, , ,899 P 7.5% PVC 6, , ,758 E Baseline , ,698 S 0% PVC 28 2,792 NM NM T CaCl ,392 NM NM U CaCl ,549 NM NM Q High Cu 2, ,536 NM NM W Wetted ,679 NM NM G Compressed ,213 NM NM X Baseline 61 4,521 NM NM Z Double 40 1,744 NM NM AA Compressed NM NM Y High Cu 19 1,428 NM NM AC Baseline 50 2,823 NM NM AD Wetted ,714 NM NM AH Open 61 4,760 NM NM AF Double ,217 NM NM AE Double ,504 NM NM Note: NM not measured. The seven baseline tests had emissions ranging from 9 to 148 ng TEQ/kg, a range of more than an order of magnitude. The mean and median emissions were 73 and 61 ng TEQ/kg, respectively. The large variation in baseline emissions, despite careful attention to standardized composition and procedures, suggests that random factors, such as waste orientation and the path of the combustion air through the waste in the barrel, may have a significant impact on PCDD/F emissions. One open-burn (waste-pile) test ( Open ) with baseline waste composition resulted in emissions of 61 ng TEQ/kg. This was within the variability of the baseline burn barrel results and, therefore, no obvious conclusions regarding the potential differences between barrels and piles can be drawn from this test. Figure 1 shows a bar graph of TEQ emissions from all the runs. Note that the y axis of Figure 1 is a logarithmic scale spanning several orders of magnitude. PCDD/Fs from replicate baseline runs were spread over approximately an order of magnitude. The baseline results highlight the inherent variability of emissions from barrel burning that is caused by variables other than those controllable by the investigators. In Figure 1, runs plotted toward the right are varied composition cases. The gray region shows the emission ranges bounded by the baseline tests. Note that most of the tests, including many of those that constituted extremes in possible waste composition, fall in or near that range. The 0% PVC case had one run that was well within the range of variability observed in baseline tests. At high levels of Cl, an effect can be seen, but at levels more closely approximating what is found in practice, other effects dominate. Purposely removing the concentrated sources of chlorinated material from the waste did not eliminate emissions of PCDD/Fs. Time-Resolved Emissions of PCDD/Fs Runs X, Z, AA, AC, AD, and AE were all performed in such a way that PCDD/F samples were acquired at three 30-min intervals during the run. This enabled the measurements of PCDD/Fs with respect to time, and Figure 2 shows a bar graph highlighting these results. In Figure 2, the sum of the three columns for each run totals 1. It is interesting that, for all runs except the wetted run, the majority of the emissions occurred during the later stages of the burn, which represented the smoldering phase. For the wetted run, more PCDD/Fs were emitted during the initial phase as water was driven from the waste. This observation may have implications for landfill fires (burning dumps), where the smoldering stage could constitute the majority of the fire. PCBs versus PCDD/Fs PCBs showed the same general trends in emissions as total PCDD/Fs, which is consistent with the theory that these compounds have either a common or parallel formation mechanism in combustion systems. 15 TEQ PCB also tracked very well with dioxins, although the PCB-based TEQ was strongly influenced by the handling of nondetects because many of the dioxin-like PCBs were present Figure 1. TEQ DF for all runs. 526 Journal of the Air & Waste Management Association Volume 53 May 2003

5 Figure 2. TEQ vs. time. Bars represent normalized estimated emissions. at levels below the detection limits. Total PCBs were approximately a factor of 200 greater than total PCDD/Fs; however, TEQ PCB is approximately 5% of TEQ DF, which is an observation that is consistent with data from MWCs. 16 This suggests that, in burn barrels, the PCBs that are formed are predominantly the nonplanar (ortho-substituted) PCBs that do not exhibit dioxin-like activity in biological systems. Continuous Emission Monitors, Weight, and Thermocouple Data Table 3 lists the weight data (the m burned parameter in eq 1) as well as the CEM and TC average and maximum data. Note that some of the runs include multiple entries. These runs denote experiments where 0 30, 31 60, and min PCDD/F samples were acquired. In general, temperatures and CO rose rapidly after ignition to a peak and then tapered off. Based on qualitative observations, there was an initial phase of the burn where flames were readily visible and occasionally protruded from the top of the barrel, even for those runs where the barrel was not highly filled. This yielded two distinct stages of the burn: a flaming stage lasting approximately min, and a smoldering stage that continued for the remainder of the burn. Generally, the CO rose rapidly and then tapered off as the burn went from the flaming stage to the smoldering stage. The double charge, however, exhibited a generally higher peak temperature; the compressed charge tended to not have as well-defined a flaming stage; and the wetted charge tended to drop into the smoldering stage sooner, with the temperature above the burning bed rapidly returning to near ambient levels much more quickly than for the other conditions. The majority of the weight loss occurs during the flaming stage of the burn, then tapers off as the smoldering stage becomes dominant. Other Data Table 4 lists some of the other measured and calculated data, including the gas-phase HCl and Cu emissions; the maximum burning rate (MAXBURN), defined as the maximum change in charge mass with respect to time (kg/ min) over the entire burn; the time at which the maximum burning rate occurred (MAXTIME); and temperature-based parameters that reflect the relative fraction of the total burn time when the thermocouples showed temperatures within certain key temperature ranges reflective of the optimal PCDD/F formation temperature window. It was desired to develop a dimensionless parameter reflecting the temperature of the barrel over the duration of the transient experiment. Because TC1 and TC2 were positioned inside the burning mass and exhibited wide temporal variations throughout the burn period, parameters were selected that included all thermocouples and that included the set of thermocouples excluding TC1 and TC2. The continuously measured temperature data (1-min averages taken over 90 min) were parsed, and the dimensionless TS1 TS4 parameters were calculated as follows: TS1 1 for each point between 250 and 450 Conall thermocouples; TS2 1 for each point between 250 and 450 C, excluding TC1 and TC2; TS3 1 for each point between 300 and 400 Conall thermocouples; and TS4 1 for each point between 300 and 400 C, excluding TC1 and TC2. The TS1 TS4 parameters represent the fraction of the barrel that was within a given temperature range over the entire duration of the run. Volume 53 May 2003 Journal of the Air & Waste Management Association 527

6 Table 3. Continuously measured data. Test a Description Waste Burned (kg) O 2 (%) O 2 (%) CO 2 (%) CO 2 (%) CO (ppmv) CO (ppmv) TC1 TC1 TC2 TC2 TC3 TC3 TC4 TC4 TC5 TC5 TC6 TC6 Over Barrel TC Over Barrel TC Hut TC Hut TC A Baseline B Baseline C Baseline K 0% PVC D Baseline L 1% PVC M 1% PVC O 7.5% PVC P 7.5% PVC E Baseline S 0% PVC T CaCl U CaCl Q High Cu W Wetted G Compressed N/A N/A X Baseline N/A N/A N/A N/A N/A Z Double AA Compressed Y High Cu N/A AC Baseline AD Wetted N/A N/A N/A N/A AH Open N/A N/A N/A N/A N/A N/A AF Double AE Double a Test results listed chronologically; N/A not available. 528 Journal of the Air & Waste Management Association Volume 53 May 2003

7 Table 4. Other measured and calculated parameters Test Description MAXBURN (kg/min) MAXTIME (min) TS1 TS2 TS3 TS4 A Baseline NM B Baseline NM 2.66 C Baseline NM 2.98 K 0% PVC NM 1.26 D Baseline L 1% PVC M 1% PVC O 7.5% PVC P 7.5% PVC E Baseline S 0% PVC T CaCL U CaCl Q High Cu W Wetted G Compressed X Baseline Z Double AA Compressed NM 1.02 Y High Cu NM 2.74 AC Baseline AD Wetted NM 0.94 AH Open AF Double AE Double Note: NM not measured. Effect of Exhaust Gas Constituents: Statistical Analysis Comparison of the 14 runs in which burn condition factors (Double, Compress, Wetted, Baseline) were changed but the composition was held constant resulted in PCCD/F emissions that ranged from 9 to 992 ng TEQ/kg. Excluding the one high TEQ (and total) value for the Wet run, analysis of variance testing on the mean TEQs and totals for these factors shows no statistically significant differences, likely because of the limited number of runs and the wide variability in emissions. To determine whether this variability could be accounted for by combustion characteristics, the normally distributed log(teq) data were modeled using an SAS STEPWISE regression to choose among various parameters that were suspected of possibly having a statistically significant effect on the emissions of PCDD/Fs, including waste Cl concentration [Cl]; continuously measured parameters of average and maximum TC temperatures (TC1 TC6 AVG and MAX); sampled HCl and Cu (particle-bound) emissions; average CEM values including CO, CO 2, and O 2 ; the time (MAX- TIME) and mass loss rate (MAXBURN) when the waste is at maximum burn rate; and the duration (in minutes) HCl (mg/m 3 ) Cu (mg/m 3 ) that in-barrel TC temperatures were within the common formation window temperature (TS C [excluding TC1 and TC2] and TS C). An optimal model (R ) for log(teq) of these 14 baseline composition runs consisted of three significant ( 0.06) linear predictors: log([hcl]), MAXBURN, and log([cu]). Selection of these predictors suggests that byproduct emissions and burn rate parameters provided the best predictive capability of TEQ emissions. It is interesting to note that the gas-phase [HCl] was a more important predictor than the [Cl] in the waste. Because [HCl] should be closely correlated with [Cl], it suggests that the nonhomogeneity of the waste resulted in an uneven distribution of the [Cl] within the barrel and therefore within the flame zone, leading to [HCl] showing more statistical significance as a predictor than [Cl]. This further suggests the importance of the combustion conditions and possibly the distribution of the waste components in the barrel to PCDD/F emissions. Comparison of the 15 runs in which only Cl levels were changed shows significant ( 0.05) differences in log(teq) values between the 7.5% PVC runs and all other runs, except for CaCl 2. Distinctions in these runs are clearly related to the Cl content of the waste: log(teq) can be modeled with log(cl) alone (R , Q ). This is not surprising because [Cl] was varied over a wide and unrepresentative range. Even with a more rigorous statistical algorithm, no distinction is observed in log(teq) for inorganic (7% Cl in CaCl 2 ) versus organic (7% Cl in PVC) Cl sources. These 15 runs were well modeled for log(teq) (R , Q ) by log([cl]), TC6MAX, and CO. Selection of these parameters indicates the importance of byproduct emissions and temperature trends in predicting PCDD/F emissions, supporting earlier results. Comparison of log(total) suggests significant differences for 7.5% PVC versus all conditions (1% PVC, baseline, and 0% PVC) except for CaCl 2. A model of log(total) for this group results in a single predictor model (R , Q ) using log([cl]). In summary, although Cl in the waste does appear to influence emissions of PCDD/Fs from burn barrels, this effect can be observed only at high levels of Cl, atypical of household trash, and is independent of the source of the Volume 53 May 2003 Journal of the Air & Waste Management Association 529

8 Cl (organic or inorganic). At moderate levels of Cl, a statistically significant effect of waste Cl concentration is not observed, because other more important variables have a much greater influence on the emissions of PCDD/Fs. The results indicate that a high degree of PCDD/F emission variation can be expected because of factors not wholly related to waste composition or burning practice. Random factors, such as waste orientation and its subsequent impact on the path of the combustion air through the barrel, possibly play a significant role in affecting combustion conditions (e.g., as observed by TC variations) and, hence, emissions. Statistical modeling of the results supports this possibility through selection of temperature-related predictors. While the wide variation in PCDD/F emissions and the limited number of runs preclude unambiguous determinations of differences caused by composition and burn condition factors, several trends seem apparent. PCDD/F emissions increased for the runs with very large amounts of Cl, whether organic or inorganic, and higher amounts of Cu catalyst. Test runs at alternative burn conditions (Compress, Wet, Double) resulted in higher mean PCDD/F emissions (203 ng TEQ/kg) and a 6-fold increase in the SD of the TEQ value (260 ng TEQ/kg) from that of the baseline runs. These results suggest widely variant PCDD/F emissions from uncontrolled domestic waste burning. These emissions are partially dependent on practice- and composition-related factors as well as random waste orientation. CONCLUSIONS Experiments were performed to evaluate emissions of PCDD/Fs and other organic pollutants from uncontrolled combustion of household waste in barrels. The main goals of the research described in this paper were to develop a more representative emission factor that could be used to calculate the overall contribution of burn barrels to the national dioxin source inventory and to identify the variables that most affect burn barrel emissions. The experiments were performed nominally at full scale, although there are a host of variables likely implemented in practice that were not addressed at all in this research. Based on the experiments, the following conclusions can be made: Emissions from the replicate baseline runs showed approximately an order of magnitude in variability. The average emissions from the replicate baseline runs was 76.8 ng TEQ DF /kg waste burned. Because of the wide variability in the baseline emissions, many of the attempted perturbations in burn condition and waste composition were not able to achieve a statistically significant effect on PCDD/F emissions. Thus, it is likely that most barrel burning practices would result in emissions that would fall within or near the bounds of the baseline composition runs (i.e., between 9 and 308 ng TEQ/kg waste combusted, which equals the 10 and 90% quantiles of the emissions from tests using the baseline composition). Organic and inorganic Cl sources showed similar propensities to form PCDD/Fs. PCDD/F emissions increased for the runs with very large amounts of Cl, whether organic or inorganic, and higher amounts of Cu catalyst. Test runs at alternative burn conditions (Compress, Wet, Double) resulted in higher mean PCDD/F emissions from that of the baseline runs. In most cases, the majority of the PCDD/F emissions were produced during the smoldering stages of the burn. Total PCBs were approximately a factor of 200 greater than total PCDD/Fs; however, TEQ PCB is approximately 5% of TEQ DF. The single open-pile burning test was within the bounds of the baseline emissions, so it is not possible to generalize whether the estimated emissions from burn barrels could be used to approximate emissions from this practice. The September 2000 draft of the EPA Dioxin Reassessment 17 estimates 1995 PCDD/F emissions from backyard barrel burning to be 628 g TEQ (based on the WHO TEFs 14 ). The reassessment also upgraded the uncertainty status of the estimate from a preliminary estimate to an estimate reliable enough to be included in the quantitative inventory. Barrel burning is now identified as the second largest quantifiable source for 1995, making up 23% of the total release estimate. One implication of this research is that based on the fact that most of the emissions of PCDD/Fs occurred during the later, smoldering stage of the burn, emissions could be high from other potentially important open burning sources such as burning dumps and landfills, where smoldering combustion may constitute a major portion of the total burn time. ACKNOWLEDGMENTS The authors thank Steve Terll and Richie Perry of ARCADIS Geraghty & Miller for their sampling support. REFERENCES 1. Emission Characteristics of Burn Barrels; Report prepared for U.S. Environmental Protection Agency Region 5 by Two Rivers Regional Council of Public Officials and Patrick Engineering Inc.: Chicago, IL, June Burn Barrel Dioxin Test; Western Lake Superior Sanitary District: Duluth, MN, August Lemieux, P.M. Evaluation of Emissions from the Open Burning of Household Waste in Barrels; Volume 1, Technical Report; EPA-600/R a (NTIS PB ); Air Pollution Prevention and Control Division: Research Triangle Park, NC, November Journal of the Air & Waste Management Association Volume 53 May 2003

9 4. Lemieux, P.M.; Lutes, C.C.; Abbott, J.A.; Aldous, K.M. Emissions of Polychlorinated Dibenzo-p-dioxins and Polychlorinated Dibenzofurans from the Open Burning of Household Waste in Barrels; Environ. Sci. Technol. 2000, 34 (3), U.S. Environmental Protection Agency. Exposure and Health Reassessment of 2, 3, 7, 8-Tetrachloro Dibenzo-p-Dioxin (TCDD) and Related Compounds; Public Review Draft; EPA-600/P Bb; National Center for Environmental Assessment: Washington, DC, September Gullett, B.K.; Lemieux, P.M.; Lutes, C.C.; Winterrowd, C.K.; Winters, D.L. PCDD/F Emissions from Uncontrolled Domestic Waste Burning; Organohal. Comp. 1999, 41, Gullett, B.K.; Lemieux, P.M.; Lutes, C.C.; Winterrowd, C.K.; Winters, D.L. PCDD/F Emissions from Uncontrolled Domestic Waste Burning; Chemosphere 2001, 43 (4-7), Lemieux, P.; Gullett, B.; Lutes, C.; Winterrowd, C.; Winters, D. Parameters Influencing Emissions of PCDDs/Fs from Open Burning of Household Waste in Barrels. In Proceedings of AWMA/Environment Canada Specialty Conference Recent Advances in the Science and Management of Air Toxics, Banff, Alberta, Canada, April Lemieux, P.; Gullett, B.; Lutes, C.; Winterrowd, C.; Winters, D. Dioxin Formation: The Barrel Burn Study. In Proceedings of the Eighth Annual North American Waste-to-Energy Conference (NAWTEC VIII), Nashville, TN, May U.S. Environmental Protection Agency. Inventory of Sources of Dioxin in the United States; External Review Draft; EPA-600/P-98/002Aa (NTIS PB ); National Center for Environmental Assessment: Washington, DC, April Rigo, G.H.; Chandler, A.J.; Lanier, W.S. The Relationship between Chlorine in Waste Streams and Dioxin Emissions from Waste Combustor Stacks; ASME Research Report CRTD-Vol. 36, Winberry, W.T.; Murphy, N.T.; Riggan, R.M. Compendium Method TO-9: Method for the Determination of Polychlorinated Dibenzo-p- Dioxins (PCDDs) in Ambient Air Using High-Resolution Gas Chromatography/High Resolution Mass Spectrometry (HRGC/HRMS). In Compendium of Methods for the Determination of Toxic Organic Compounds in Ambient Air; EPA/600/ (NTIS PB ); U.S. Environmental Protection Agency, Atmospheric Research and Exposure Assessment Laboratory: Research Triangle Park, NC, June U.S. Environmental Protection Agency. Method 8290, Polychlorinated Dibenzodioxins (PCDDs) and Polychlorinated Dibenzofurans (PCDFs) by High Resolution Gas Chromatography/High Resolution Mass Spectrometry (HRGC/HRMS). In Test Methods for Evaluating Solid Waste. Integrated Manual (SW-846); Volume 1B: Laboratory Manual, Physical/Chemical Methods; Revised Update III; SW-846 (NTIS PB ); Office of Solid Waste: Washington, DC, June Van den Berg, M.; Birnbaum, L.; Bosveld, A.; Brunström, B.; Cook, P.; Feeley, M.; Giesy, J.; Hanberg, A.; Hasegawa, R.; Kennedy, S.; et al. Toxic Equivalency Factors (TEFs) for PCBs, PCDDs, PCDFs for Humans and Wildlife; Environ. Health Perspect. 1998, 106 (12), Wikstrom, E.; Marklund, S. Influence of Variation in Combustion Conditions on the Primary Formation of Chlorinated Organic Micropollutants during Municipal Solid Waste Combustion; Environ. Sci. Technol. 2000, 34 (4), Alcock, R.E.; Behnisch, P.A.; Jones, K.C.; Hagenmaier, H. Dioxin-Like PCBs in the Environment Human Exposure and the Significance of Sources; Chemosphere 1998, 37 (8), U.S. Environmental Protection Agency. Exposure and Health Reassessment of 2, 3, 7, 8-Tetrachloro Dibenzo-p-Dioxin (TCDD) and Related Compounds; Public Review Draft; EPA-600/P Bb; National Center for Environmental Assessment: Washington, DC, September About the Authors Paul M. Lemieux (corresponding author) is a senior research engineer with EPA. He may be reached at Office of Research and Development, National Risk Management Research Laboratory, Research Triangle Park, NC 27711; lemieux.paul@epa.gov. He is working on formation of products of incomplete combustion from stationary combustion sources and is currently assigned to EPA s Center for Homeland Security Research. Brian K. Gullett is a senior research engineer with EPA. He is working on trace air toxics monitoring technologies and source characterization. Dwain L. Winters is a senior scientist working for EPA s Dioxin Policy Project. Christopher C. Lutes is a principal scientist and business practice manager with ARCADIS, Inc. He works in the areas of open-burning process assessment as well as development and implementation of innovative remediation technologies for soil and groundwater. Chris K. Winterrowd has 10 years of experience in the air-emissions monitoring industry. He serves as a staff research engineer for ARCADIS, Inc. Volume 53 May 2003 Journal of the Air & Waste Management Association 531