Environ. Sci. Technol. 2005, 39, 8790-8796
Emissions of Polychlorinated Dibenzodioxins and Dibenzofurans and Polychlorinated Biphenyls from Uncontrolled Burning of Garden and Domestic Waste (Backyard Burning) BJO ¨ R N H E D M A N , * , † M O R G A N N A¨ S L U N D , † CALLE NILSSON,‡ AND STELLAN MARKLUND† Chemistry Department, Environmental Chemistry, Umeå University, SE-901 87, Umeå, Sweden, and NBC Defence, NBC Analysis, The Swedish Defence Research Agency, SE-901 82 Umeå, Sweden
To assess emissions of dioxins (chlorinated dibenzodioxins and dibenzofurans) and PCB from uncontrolled domestic combustion of waste (“backyard burning”), test combustions in barrels and open fires were monitored. The waste fuels used were garden waste, paper, paper and plastic packaging, refuse-derived fuel (RDF), PVC, and electronic scrap. Combustions including PVC and electronic scrap emitted several orders of magnitude more dioxins than the other waste fuels. Emissions from the other fuels had considerable variations, but the levels were difficult to relate to waste composition. Emission factors of PCDD/F and PCB from the backyard burning ranged from 2.2 to 13 000 ng (WHO-TEQ)/kg. The levels found in ash usually were less than 5% of the total. For assessment of total emissions of dioxins and PCB from backyard burning of low and moderately contaminated wastes, an emission factor range of 4-72 ng (WHO-TEQ)/kg is suggested. These figures imply that combusting waste in the backyard could contribute substantially to total emissions, even if the amounts of fuel involved are equivalent to just a few tenths of a percent of the amounts combusted in municipal waste incinerators.
Introduction Combustion-related activities are important sources of polychlorinated dibenzo-p-dioxins and polychlorinated dibenzofurans (PCDD/F) and polychlorinated biphenyls (PCB) released to the ambient air. Municipal waste combustion, which used to be a major contributor, currently is the most intensively regulated and monitored combustion activity. It is generally considered to have a low impact on the total emissions to air, due to improved operational guidelines and improvements in emission cleaning equipment. However, there are large uncertainties regarding the extent of emissions from diffuse combustion sources, such as small-scale biofuel combustion, accidental fires, and small, domestic-scale burning of waste (“backyard burning”). Furthermore, few data have been compiled on the frequency of backyard burning or the volume and composition of the waste involved. * Corresponding author phone: +46-90-786-6664; fax: +46-90128133; e-mail:
[email protected]. † Umeå University. ‡ The Swedish Defence Research Agency. 8790
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 22, 2005
FIGURE 1. Schematic view of the experimental setup. The scale is approximate. T1-T6 are thermocouples. Thus, the Swedish government initiated a national survey of unintentionally produced persistent organic pollutants (POP’s), including the study of “backyard burning” reported here (1). There are probably large variations in the amounts and composition of waste combusted in backyard burning both between countries (due to differences in legislation and waste assessment policies) and within countries (due to variations in the availability of regulated waste handling facilities, lifestyle, etc.). However, the scientific literature on these activities is very scant. Apart from a few studies from Belgium (2) and Japan (3, 4), the majority and the most profound published studies reflect conditions in the United States, where legislation often permits the burning of household waste in backyard barrels, and it is a common practice in some areas (5-8). Because of expected differences in the composition of burned waste, results from these studies may not be directly extended to other countries. A rough estimate based on responses to a questionnaire sent to Swedish local government officials responsible for environmental issues suggests that about 1/10 of the waste fuel in domestic waste burning is ordinary household waste, whereas the proportion of garden waste is suspected to make up to about three-quarters (1). However, there is an apparent lack of knowledge about emissions from open, uncontrolled combustions in which garden waste is a major ingredient. Thus, this study was designed to gather data on various possible backyard burning scenarios, primarily reflecting waste-handling habits that are thought to be common in Sweden.
Experimental Section Experimental Setup. Nineteen test combustions of waste in a 200-L steel barrel and two in open fires were monitored. The experimental setup is shown schematically in Figure 1. The barrel had an inner diameter of 570 mm and an inner height of 850 mm. To supply air, 12 holes, each of 20-mm diameter, were drilled at equal intervals around the barrel about 50 mm from the bottom. A conical fume hood of 750mm diameter was placed approximately 100 mm above the barrel, connected to a 100-mm steel tube. To decrease the flow of flue gas sufficiently to enable isokinetic sampling, this tube was connected to a 160-mm tube. The flow was maintained at a constant level of approximately 410 m3/h with a radial fan at the end of the latter tube. A 3 m long, 100 mm wide tube was connected as a fume stack at the outlet of the fan. To reduce emissions of pollutants to the environment, an activated carbon filter was mounted on the stack. 10.1021/es051117w CCC: $30.25
2005 American Chemical Society Published on Web 10/13/2005
TABLE 1. Composition of Combusted Waste Fuel test no.
date
composition
1 2 3 4
Aug 5 Dec 13 Aug 12 Aug 10
5 6 7 8 9 10 11
Dec 15 Aug 5 Dec 13 Aug 25 Dec 15 Dec 15 Aug 18
12
Aug 12
13 14 15
Aug 10 Aug 12 Aug 25
16 17 18 19 20 21
Aug 26 Aug 18 Aug 18 Aug 25 Sep 3 Sep 3
6 kg of garden waste 6 kg of garden waste 6 kg of garden waste 4 kg of garden waste, 1.7 kg of paper pack., 0.3 of plastic pack. 4 kg of garden waste, 2 kg of plastic pack. 4 kg of garden waste, 2 kg of RDF 4 kg of garden waste, 2 kg of RDF 4 kg of garden waste, 0.5 L of waste motor oil 4 kg of garden waste, 2 kg of silage film (polyethylene) 4 kg of garden waste, 2 kg of PVC 4 kg of papers and magazines, 1.7 kg of paper pack., 0.3 of plastic pack. 4 kg of papers and magazines, 1.7 kg of paper pack., 0.3 of plastic pack. 4.5 kg of RDF, 1.5 kg of paper pack. 4 kg of RDF, 2 of kg paper pack. 3.4 of kg RDF, 2.65 of kg paper pack., 0.95 car tire, 0.5 L of waste motor oil 4 kg of RDF, 2 kg of paper pack., 2.95 kg of computer scrap 6 kg of straw 6 kg of straw 4 kg of straw, 2 kg of silage film (polyethylene) 6 kg of garden waste 4 kg of garden waste, 2 of kg RDF
a
No logging of flue-gases.
b
circumstances
ash sample
a b
yes
a
yes
a
yes
a a
yes yes
Cl, % (estd value) 0.14 0.12 0.09 0.14 0.28 0.36 0.31 0.13 0.41 19 0.08 0.08
c
0.68 0.55 0.38
a
1.2 0.34 0.29 0.53 0.11 0.35
b b open fire open fire
Moistured fuel. c Air-dried RDF.
Temperatures were measured by thermocouples placed as shown in Figure 1 and were continuously logged with a Picolog TC-08 (Pico Technologies Ltd). A probe in the tube was connected to a multigas analyzer (Electra Control) to continuously measure CO2 and CO by means of infrared detectors and O2 by means of an electrochemical cell. A control of displayed levels was performed with calibration gases (CO and CO2) before the tests and zero-calibrations were performed with air before each sampling. Composition and Origins of Waste Fuels. The composition of the waste fuels and other important factors affecting the different tests are summarized in Table 1. The garden waste, originally from private gardens, was collected at a municipal recycling center. In the tests it consisted of a mixture of approximately 50% tree branches and 50% dry leaves and grass (by weight). The refuse-derived fuel (RDF) consisted of municipal waste where the combustible fractions (e.g. paper, textile and soft plastics) had been mechanically sorted out from noncombustible waste and decomposable material at a waste sorting plant. The paper and the paper and plastic packaging were collected from recycling containers. A local farmer supplied straw (barley) and polyethylene stretch film that had been used to bale silage in round bales (a common practice in Sweden). Used motor oil was collected from a do-it-yourself service hall. No analysis of chlorine or other contents was performed on the oil. The computer scrap (a diskette drive unit, motherboard, power pack, and plastic cover) consisted roughly of two-thirds metal and one-third plastic. The PVC (poly(vinyl chloride)) waste consisted mainly of pieces of piping. Combustion Tests. Prior to combustion tests, large pieces of waste (e.g. branches and PVC pipe) were cut in pieces of a few decimeters and mixed. Wastes used in more than one test were divided in advance to ensure a fairly uniform composition. The waste fuels to be combusted were placed in layers in the barrel with coarse material (e.g. branches and paper packaging) preferably in the bottom to allow air to flow through it. The fuel was set on fire with an LPG (liquefied petroleum gas) burner through the holes in the bottom and from the top of the barrel. Sampling of organic
compounds in flue gas and logging of temperatures started immediately, but logging of gases started a couple of minutes later, after the first outburst of thick fuel gas (to avoid temporary instrumental malfunctions due to overloading by hydrocarbons). To protect the fan from heat, the flue gas was kept below 200 °C by cooling the tube with water or snow. This had the additional effect of keeping the sampling temperatures below 200 °C, since the port for sampling organic compounds in the emissions was located about 25 cm before the fan. Condensation in the tube was avoided by interrupting the cooling when the temperature was approaching 150 °C. When a major part of the fuel had been combusted and the temperature in the barrel (T1) had decreased to less than ca. 100 °C, the smoldering remains were stirred once or twice with a stick to stimulate further combustion. Flue gas samplings were ceased when the temperature (T1) decreased below the previously chosen temperature 70 °C. The remaining ash and unburned material were weighed afterward. The majority of the tests were done in August and September with ambient temperatures of 15-25 °C. Test nos. 2, 4, 7, 9, and 10 were done in December with ambient temperatures around 0 °C. In these tests and for test no.16, the gas logging equipment was not available. In the five December tests, ashes were sampled to analyze the PCDD/F and PCB in them. To minimize contamination of equipment from previous combustion tests, they generally were performed in order of expected increasing emission generation, i.e., from garden waste to domestic waste to computer scrap in the autumn tests. There were deviations from this order due different circumstances, and some of the earlier tests were repeated in the December tests. In test nos. 20 and 21, the waste was combusted in open fires. In each of these cases the fire was placed on a 1 × 1 m plate of steel and the hood was mounted about 0.6 m above the plate. Except for the thermocouples in the barrel, the same equipment was used and the same procedures were followed as in the barrel tests. PCDD/F and PCB in flue gas were isokinetically sampled with a cooled probe and sampling train according to VOL. 39, NO. 22, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
8791
TABLE 2. Approximate Levels of Ash and Carbon Contents (given as percentages of dry matter, % DM), Used in the Calculations leaves, grass branches straw RDF papers and magazines paper packaging
ash
C
5 1 6 15 10 8
46 52 46 42 40 40
plastic packaging silage film car tire used motor oil computer scrap PVC
ash
C
5 2 10 1 67 0
75 75 60 75 25 14
European standard methods (EN 1948 1-3) (9), except that the sampling duration (and hence volume) was reduced because supplies of some of the waste fuels were limited. The sampling time was usually about 1 h and the sampling volume about 0.5 m3. Analysis. The dry substance of waste fractions was estimated by drying approximately 10-g pieces overnight in 105 °C. The extraction, cleanup, and analysis of PCDD/F and PCB with dioxin-like toxicity (according to WHO) in the flue gas and ash samples followed the general scheme for combustion-related samples described by Liljelind et al. (10) Multivariate Data Analysis. Multivariate data analysis is a common term for a number of methods used to analyze patterns and to evaluate data matrixes with many variables statistically (11). In principal component analysis (PCA), every observation is represented by a point in a multidimensional space with as many coordinates as the number of variables. These points then are projected on planes of lower dimensions, giving orthogonal principal components (PC’s) describing most of the variation in the dataset, and are visualized in score-plots. The first PC, PC1, describes the largest contribution to the variation in the data and so on. The corresponding loading-plots show the variables that have the strongest influence on the distribution of the observations. Projection to latent structures (PLS) is a regression technique where multidimensional spaces of x-variables and y-variables are fitted to each other in order to predict y-variables from the x-variables. In this work, PCA was used to distinguish between profiles of dioxin homologues emitted in different tests. The homologues and the TEQ-values used in the model were normalized by dividing the absolute values with ∑PCDD/F. The models were mean-centered but not further scaled. PLS was used in an attempt to predict emission levels of dioxins from the combustion temperature and the composition of the combusted waste. x-Variables were the percentage of different waste constituents (listed in Table 2); dry substance; calculated literature derived levels (12-14) of C, ash (Table 2), and Cl (Table 1); combustion temperature (max. temp of T3); and ambient temperature. y-Variables were ∑PCDD/F, PCDD/F homologues, 2,3,7,8-congeners, PCDD/F (WHOTEQ), PCB (WHO-TEQ), CO, and CO2. The absolute values were scaled to unit variance. Calculation of Emission Factors. The emissions of the chlorinated organic compounds per mass unit of fuel consumed by combustion, or “emission factors” (EF) were estimated from
Ccomb ) Cfuel(mdw - mres)/(mdw - mash)
(1)
f ) (Ccomb/Cmeas)(Aflue/Aprobe)
(2)
EF ) fmsample/mfuel
(3)
The mass of carbon in the combusted fuel (Ccomb, in kg) was estimated by summing literature-derived values of carbon contents of the different components of the fuel (12-14) (Table 2). The resulting values were corrected for unburned 8792
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 22, 2005
combustible materials as in eq 1, where Cfuel is the calculated mass in kilograms of carbon in the fuel, mdw is the dry weight of the fuel, mres is the gravimetrically derived mass of the residues after combustion, and mash is the mass of literaturederived ash content (12-14) (Table 2). The factor f in eq 2 and 3 describes the proportion of the emitted flue gas that is collected in the sample, which depends on the proportions both of the flue gas that is collected in the fume hood and of the gas collected in the fume hood that is subsequently sampled in the sampling train. The first proportion can be approximated by dividing the carbon content of the combusted fuel (Ccomb) by the summed carbon content of the measured emissions of CO and CO2 (Cmeas) and the second by dividing the cross section area of the flue at the sampling point (Aflue) by the cross sectional area of the sampling probe (Aprobe) (eq 2). The emission factor, EF, then can be estimated by eq 3, where msample is the mass of analyte in the sample and mfuel is the mass of waste fuel (gross weight) combusted during sampling.
Results and Discussion Choice of Fuels for the Study. The various waste fractions (fuels) used in this study were selected to reflect likely scenarios for backyard burnings in Sweden and can be regarded as representative of the situation in a country with a fairly developed and extensive system for collecting household waste. The largest waste fraction remaining after ordinary waste collection is expected to be garden waste. This is a bulky type of waste that varies widely in abundance in different seasons and burning by the residents can be regarded as a conceivable way of disposal. In such burnings, other bulky, easy combustible types of waste, e.g. paper and paper and plastic packaging, may also be burnt. Thus, the fuels burned in our tests were largely mixtures of these types of waste (including RDF, representing unspecified combustible household waste). Occasionally, less common but potentially more polluting types of waste may also be combusted in backyard burnings. In our tests, these wastes were represented by used motor oil, a rubber tire, PVC, and electronic equipment (computer components). Straw and plastic (polyethylene) film used for wrapping silage were included, since burnings of these wastes on agricultural fields is considered to be one of the most common types of uncontrolled waste combustions in Sweden. Emission Results. The calculation of emission factors is accomplished through a carbon balance that partly is based on theoretical values. There is also assumed a similar degree of combustion of combustible parts of all fuel components. A deviation from estimated values of carbon content in fuel or emitted flue gas (Cfuel and Cmeas in eqs 1 and 2) will give the same relative error in emission factor, whereas a deviation from the estimated ash content gives approximately 1/10 relative error. A deviation from the assumed similarity in combustion degree may in extreme cases (e.g. if all the oil of test no.15 should have remained in the unburned residue) would have given a 5% error in the estimation of combusted carbon (Ccomb in eq 1) and the same error in emission factor. However, in the tests where the contents of carbon differed largely between the waste constituents, the mass of unburned remains was not large, and generally this error can be assumed to be negligible. The nominal accuracy of the instrument used for CO2 measurements were (0.4% in the range in question; however, controls with calibration gases indicated an accuracy of less than (0.1%. This nevertheless introduces an uncertainty of approximately (10% in the estimated emission factors, and the uncertainties of CO, including the nearness to detection limit, introduce another (5-10% uncertainty. In most of the tests, the temperature registered by the thermocouples T1-T5 (Figure 1) peaked within 5 or 10 min
FIGURE 2. Temperatures (T1-T3) and emissions of CO and CO2 in test no. 4. T1 indicates an uneven distribution of fire inside the barrel. The rise in temperatures and emissions after approximately 43 min are due to stirring of the smoldering remains.
TABLE 3. Emissions and Pollutions in Ash from Combustion Tests emission factors in flue gas emissions flue gas emissions % CO % CO2
test no.
flue gas tempa °C
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
684
0.1
2.1
646 716
0.1
