Characterization of Non-methane Hydrocarbons Emitted from Various Cookstoves Used in China

Transcription

1 Environ. Sci. Technol. 2003, 37, Chrcteriztion of Non-methne Hydrocrbons Emitted from Vrious Cookstoves Used in Chin STELLA MANCHUN TSAI, JUNFENG (JIM) ZHANG,*,, KIRK R. SMITH, YUQING MA, R. A. RASMUSSEN, AND M. A. K. KHALIL # University of Medicine nd Dentistry of New Jersey - School of Public Helth, Pisctwy, New Jersey, Environmentl nd Occuptionl Helth Sciences Institute, Pisctwy, New Jersey, Environmentl Helth Sciences, University of Cliforni, Berkeley, Cliforni, Institute for Techno-economics nd Energy System Anlysis, Tsinghu University, Beijing, Chin, Deprtment of Environmentl Sciences nd Engineering, Oregon Grdute Institute, Beverton, Oregon, nd Deprtment of Physics, Portlnd Stte University, Portlnd, Oregon Emission contributions from cookstoves to indoor, regionl, nd globl ir pollution lrgely depend on stove nd fuel types. This pper presents dtbse on emission fctors of specited non-methne hydrocrbons (NMHCs) for 16 fuel/stove combintions burning 2 types of crop residue, wood, 4 types of col, kerosene, nd 3 types of gseous fuels. The emission fctors re presented both on fuel mss bsis (compound mss per fuel mss) nd on cooking tsk bsis (compound mss per unit energy delivered to the pot). These fuel/stove combintions cover lrge spectrum of the cookstoves used in both urbn nd rurl households in Chin. Up to 54 hydrocrbons were identified, some of which re rective precursors of photochemicl smog. Bsed on published mximum incrementl rectivity (MIR) vlues for NMHCs, we estimted stove-specific nd fuel-specific ozone forming potentils (OFPs). The results indicte tht rw col powder, wood, nd crop residues hve higher OFP vlues thn the other types of fuels tested. Strikingly, burning the col briquette nd honeycomb col briquette produced OFP vlues more thn 2 orders of mgnitude lower thn burning unprocessed (rw) col, even in the sme vented metl stove, for every 1 MJ delivered to the pot. Introduction In mny prts of developing countries, household cookstoves re the mjor sources of ir pollution exposure (1-3). Elsewhere we hve presented results from studies of cookstove emissions, including pilot study in the Philippines * Corresponding uthor phone: (732) ; fx: (732) ; e-mil: jjzhng@eohsi.rutgers.edu. Corresponding uthor ddress: EOHSI, Room 358, 170 Frelinghuysen Rod, Pisctwy, NJ University of Medicine nd Dentistry of New Jersey - School of Public Helth. Environmentl nd Occuptionl Helth Sciences Institute. University of Cliforni. Tsinghu University. Oregon Grdute Institute. # Portlnd Stte University. (4), nd two reporting the results of more extensive studies in which emissions from 28 fuel/stove combintions were mesured in Indi nd in Chin (5, 6). In both the pilot nd the more extensive studies, we mesured stove performnce nd emissions of severl importnt greenhouse gses (GHGs) nd other pollutnts of helth concern, including crbon dioxide (CO 2), crbon monoxide (CO), methne (CH 4), totl non-methne hydrocrbon (TNMHC), totl suspended prticles (TSP), oxides of nitrogen (NO x) (for Chinese stoves only), nd sulfur dioxide (for Chinese stoves only). These studies found tht gret frction of fuel crbon ws converted to products of incomplete combustion (PICs), rther thn CO 2, in simple household stoves nd tht burning solid fuels generted substntilly greter emissions of totl PICs thn burning liquid nd gseous fuels to deliver the sme mount of energy to the pot. The studies lso found tht the emission of non-co 2 GHGs (e.g., CH 4, nitrous oxide, nd CO) cn be significnt contributor to totl globl wrming potentils for mny mesured cookstoves (7). In the forementioned extensive studies of cookstove emissions, subset of the fuel/stove combintions were lso mesured for specited NMHCs, the results of which hve not yet been reported. In this pper, we present dtbse of emission fctors of NMHCs for the 16 fuel/stove combintions mesured in Chin. The dtbse for individul NMHCs cn provide better informtion for future study on helth risk ssessment bsed on prticulr toxic compounds emitted from household cookstoves. Respirtory illnesses, including cute respirtory infection, chronic obstructive lung diseses, nd lung cncer, hve been strongly ssocited with exposure to indoor smoke from solid-fuel stoves (8). Other effects, including tuberculosis, dverse pregnncy outcomes, ero-digestive cncer, ctrcts, nd sthm, hve lso been ssocited with such exposures, but the evidence bse is still smll (8-10). Synergistic nd ntgonistic effects of toxic compounds present in the cookstove smoke my contribute to the observed dverse heth effects. Knowing detiled constituents of the smoke will certinly help in understnding its toxicologicl properties. In ddition to some NMHCs tht hve direct helth concerns, some of the compounds identified re rective precursors, in the tmosphere, leding to the formtion of ozone (O 3) nd photochemicl smog. The dtbse on individul NMHCs, thus, cn be useful in modeling locl or regionl O 3. In this pper, we compre stove- nd fuel-specific ozone forming potentils (OFPs) using the mesured emission fctors nd the published rectivity indices for NMHCs. The nlysis of stove/fuel OFPs my provide useful energy policy implictions in Chin from stnd point of O 3 control strtegies. Methods Fuel/Stove Combintion Mesured. The fuels mesured included biomss fuels (crop residues nd wood), severl types of cols, kerosene, liquefied petroleum gs (LPG), col gs, nd nturl gs. These selected fuels covered fuels of common use in rurl nd urbn households in Chin in when the mesurements were mde. The stove selection ws bsed on populr models for ech fuel type. Since 1996, usge of gs fuels my hve been incresed in some mjor Chinese cities (e.g., Beijing) (11, 12). However, no significnt chnges in household fuel types hve occurred in most res of Chin. The stove/fuel combintions tested in this study re still commonly found in Chin s whole, lthough some stove/fuel combintions found in hve been replced with others in some cities or res. Detiled /es CCC: $ Americn Chemicl Society VOL. 37, NO. 13, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY Published on Web 06/04/2003

2 TABLE 1. Description of the Fuel/Stove Combintions Tested in Chin description seril no. symbol fuel/stove fuel stove 1 Honey- honeycomb col briquette metl col stove with flue 2 Honey-Metl honeycomb col briquette metl col stove without flue 3 Honey-Imp. honeycomb col briquette improved metl col stove without flue 4 Briq-Metl col briquette metl col stove without flue 5 Wsh wshed col powder metl col stove with flue 6 unprocessed col (col powder) metl col stove with flue 7 unprocessed col (col powder) brick stove with flue 8 Wood- fuel wood brick stove with flue 9 Wood-Imp.-v fuel wood improved brick stove with flue 10 Whet- whet residue brick stove with flue 11 Mize- mize residue brick stove with flue 12 Mize-Imp.-v mize residue improved brick stove with flue 13 Kero-Wick kerosene kerosene wick stove without flue 14 LPG- liquefied petroleum gs LPG trditionl stove without flue 15 col gs trditionl gs stove without flue 16 Nturl nturl gs trditionl gs stove without flue Flue code: v ) vented, i.e., with flue. The fuel ctegory includes: #1-7 for col; #8-9 for wood; #10-12 for crop residue; #13 for kerosene; nd #14-16 for gseous fuel. powder is rw col with lrge pieces being sieved out. power consists of col prticles (pieces) with dimeters typiclly smller thn 1 cm. powder is the most common unprocessed col used in Chinese households. informtion bout the fuel sources, fuel compositions, nd stove chrcteristics cn be found in previous published pper (6). Briefly, for the stoves using piped gseous fuels, the mesurements were conducted in ctul homes. The emissions of ll the other fuel/stove combintions were mesured in simulted villge kitchen locted in the rurl cmpus of Tsinghu University, Beijing, Chin. Among the 28 fuel/stove combintions mesured primrily for mjor GHGs, 16 were lso successfully mesured for specited NMHCs. These 16 fuel/stove combintions, described in Tble 1, represent the most commonly used ones nd covers t lest one fuel/stove combintion within ech brod fuel type (col, wood, crop residue, kerosene, nd gs). With the exception of Honey- combintion, every fuel/stove combintion ws mesured for NMHCs only once for complete burn cycle, due to the budgetry constrint. However, we were ble to mke three repeted mesurements for the Honey- to get sense bout the vrition from one burn cycle to nother burn cycle. This ws evluted by reporting the coefficient of vrition (CV) determined for this prticulr fuel/stove combintion. Uncertinties ssocited with emission fctors for the other tested fuel/stove combintions were discussed bsed on the CV vlues of TNMHC reported in the previous pper nd n nlysis of uncertinty sources (6). Mesurement Methods. Additionl detils on experimentl design nd smple collection procedures cn be found in Zhng et l. (6). Briefly, flue gs smples were collected using smpling configurtion tht included, from upstrem to downstrem of the smpling trin, stinless steel probe, filter holder, smpling pump (SKC Inc., U.S.A.), nd clen Tedlr bg (SKC Inc., U.S.A.). The filter holder contined 37 mm qurtz fiber filter tht ws used to collect TSP. For stoves with flue, the probe ws inserted in the flue. For those hving no flues, the stoves were plced under hood built for the test purpose nd the probe ws plced inside the hood exhust duct. The flue gs smples hd gone through dilution nd cooling before they were collected in the Tedlr bgs. Thus, the temperture of the flue gs smples ws the sme s the mbient temperture. The flow rte of the smpling pump, rnging from 0.2 to 2 L/min, ws djusted to fill one or two 80-L Tedlr bgs throughout whole burn cycle (i.e., from fire strt to fire extinction). If two Tedlr bgs were used, time-weighted frction of ir from ech of the two bgs ws tken nd then mixed in the third Tedlr bg for finl smple nlysis. Hence, the flue smples collected were time-verged emissions covering complete burn cycle consisting of different combustion stges. Two mbient ir smples were collected nerby the simulted kitchen for bckground corrections. Aliquots of ir smples were tken out of the Tedlr bgs for vrious subsequent chemicl nlyses. These included the determintion of flue gs concentrtion of CO 2, CO, CH 4, nd TNMHC, nd crbon in irborne prticultes, ll necessry for the construction of crbon blnce model, long with mesured crbon mss in the fuel combusted nd in combustion residues (for solid fuel only). The crbon blnce model ws used to determine emission fctors of ech mesured species in the flue gs (6). As described in detil by Zhng et l. (6), the crbon blnce pproch only requires the mesurement of emission rtio, i.e., [species concentrtion]/[co 2 concentrtion] of n emitted irborne species. The emission fctor of given species in the flue gs cn be clculted from the CO 2 emission fctor nd the emission rtio of the species. Therefore, here we report emission rtios of ll specited NMHCs, which were used to derive emission fctors. Emission fctors were reported, by convention, on dry fuel mss bsis (i.e., on moisture free bsis). The moisture content of ech fuel ws mesured through stndrd fuel moisture nlysis, s described in detil in the previous pper (6). Since different mounts of fuels re needed for the sme cooking tsk using different fuel/stove combintions, tskbsed emission fctors (pollutnt mss per cooking tsk) rther thn the fuel mss bsed is better performnce index to compre the ir pollution potentil of different fuel/stove combintions (2, 13). For this reson, the emission fctors were lso reported s mss of compound emitted per unit energy, one meg-joule (MJ), delivered to the pot. We chose 1 MJ s unit energy becuse bout 5 MJ of delivered het is roughly wht typicl fmily mel would require, lthough this obviously vry lrgely by mel size, type, nd cooking method (5). The conversion between the mss-bsed nd tsk-bsed emission fctors ws chieved using the fuel energy content (lso known s clorific vlue) (MJ/kg) nd stove therml efficiency (%), both of which were mesured in the study. Therml efficiencies [energy effectively received by the pot]/[totl energy in the fuel burned] of the stoves were determined using wter boiling tests, s described in detil in the previous pper (6). The mesurement of flue gs concentrtions of specited NMHCs involved filling clen evcuted 850-mL stinless ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 13, 2003

3 steel cnister with the flue gs collected in the lrge Tedlr bgs to tm pressure using bttery-powered pump. The cnisters were preclened nd provided by the lbortory t Oregon Grdute Institute (OGI), Beverton, OR. The cnisters contining flue gs smples were shipped bck by ir crriers to the OGI lbortory for nlysis. Storge time before nlysis vried from one btch to nother but no more thn 30 dys. There were no field blnk or duplicte smples collected due to budget constrint. The bckground concentrtion ws determined from two mbient smples collected nerby the simulted kitchen nd ws used to correct for net flue gs concentrtions. The hydrocrbon specition nlysis ws mde using the procedure estblished s EPA Compendium Method TO-14A (14). A system of gs chromtogrphy (GC) equipped with flme ioniztion detector (FID) nd mss-selective detector ws used for the nlysis. In ech nlyticl run, smll liquot of ir smple ws collected from the cnister into the nlyticl system through dryer (to remove moisture), chromtogrphic vlve, nd then cryogenic trp. The trp ws heted rpidly, nd the smple ws injected onto n OV-1 cpillry column. High purity helium ws used s the crrier gs. A stndrd mixture contining 74 NMHCs (Scott Specilty Gses, Inc., Sn Bernrdino, CA) ws used to generte clibrtion curves for every btch of smples. All clibrtion curves hd liner regression R 2 vlues > The detection limit vlues vried by individul compound but were in sub-ppb rnges. Clcultion of Ozone Forming Potentil. Among the NMHCs identified in the cookstove flue gs, totl of 44 compounds were selected for ozone forming potentil clcultion. These compounds re on trget list, which ws developed by the U.S. EPA Photochemicl Assessment Monitoring Sttions (PAMS) progrm, due to their rectivity in terms of producing tropospheric ozone nd photochemicl smog. We estimted ozone forming potentils (OFPs) of the 16 fuel/stove combintions using the mesured emission fctors nd vlues of the Mximum Incrementl Rectivity (MIR) scle developed by Crter (15) for most of the NMHCs mesured, except o-xylene nd sec-butylbenzne. Since only totl xylenes (o+m+p) were mesured, the published MIR vlue for m,p-xylenes ws pplied to totl xylenes. Similrly, since 1,2,4-trimethylbenzene nd sec-butylbenzene were not seprted for the flue gs smples, we pplied the MIR vlue for 1,2,4-trimethylbenzene to the sum of 1,2,4-trimethylbenzene nd sec-butylbenzene. The MIR of individul NMHC ws defined in terms of grms of ozone formed by photochemicl rections in the tmosphere per grm of crbon (C) in NMHC emitted (g ozone/g C emitted). The totl ozone forming potentil (OFP) for ech fuel/stove combintion is the sum of OFPs of individul NMHCs. The OFP from ech individul NMHC ws clculted s follows. Per kg fuel mss OFP: OFP (g ozone/kilogrm of dry fuel) ) [MIR of given NMHC (g ozone/g C)] [emission fctor of the NMHC (g C/ kilogrm of dry fuel)] Per tsk (1 MJ energy delivered) OFP: OFP (g ozone/mj energy delivered) ) [MIR of given NMHC (g ozone/ g C)] [emission fctor of individul NMHC (g C/MJ energy delivered)] Results Molr Emission Rtios. Non-CO 2 crbon-contining compounds detected in the flue gs re typiclly clled products of incomplete combustion (PICs). The rtio of n individul species to CO 2 in the flue gs, nmely emission rtio, represents reltive bundnce of the species in the flue gs. Tble 2 presents molr emission rtios of 54 specited NMHCs, mong the 74 trgeted species, tht were detected for ny of the 16 fuel/stove combintions tested. The emission rtios were clculted using net concentrtions of individul NMHC nd net concentrtions of CO 2. (The net concentrtion equls flue gs concentrtion minus bckground concentrtion). The vlues for totl specited NMHC were the sums of the molr emission rtios of ll the quntified NMHCs listed in Tble 2. The rnking of reltive bundnce of totl specited NMHC to CO 2 is lso shown in Tble 2. Burning biomss fuels (whet residue, mize residue, nd wood) generlly hd greter emission rtios thn burning gseous fuels. It is striking tht col type hd profound effect on the emission rtio of the sum of ll the quntified NMHCs. Among the 16 tested fuel/stove combintions, the stoves burning unprocessed col powder nd wshed col powder hs highest emission rtios while the vented honeycomb col stove hd the lowest emission rtio. The results from the three repeted mesurements of Honey- showed lrge rnge of CV (8-113%) for the emission fctors of the NMHCs detected in the flue gs, providing rough ide bout run-to-run vritions in NMHC emissions. By nture, the emission cn vry lrgely from one burn cycle to nother cycle. Therefore the uncertinty ssocited with the single mesurement vlues reported in this pper is expected to be lrge, which is discussed lter in the Discussion Section. Emission Fctors. Tble 3 presents emission fctors of ech identified compound, by fuel/stove combintion, on dry-fuel mss bsis. The emission fctors in Tble 3 represent the mount of individul compound in grms (g) generted by burning 1 kg of dry fuel with ech fuel/stove combintion. Tht is, for exmple, burning 1 kg of honeycomb col briquette using metl col stove with flue (Honey-) produced 1.98 mg of benzene. The emissions of totl specited NMHC from burning 1 kg of fuel rnged from 11.9 mg to 6.48 g cross the fuel/stove combintions mesured. Even within fuel type, the rnge of totl specited NMHC emission fctor cn be lrge for different fuel species nd stove types. The col/stove combintions hd rnge from 11.9 mg to 6.48 g; the wood stoves hd rnge from 1.36 nd 2.21 g; the crop residues hd rnge from 0.73 to 2.28 g; nd the stoves burning gseous fuels hd rnge from to 0.42 g. Tble 4 presents emission fctors on delivered energy bsis. The dt on stove efficiency nd fuel energy content (mesured s low-het clorific vlue) re lso included in Tble 4. These vlues were used to convert between fuelmss-bsed nd tsk-bsed emission fctors. Emission fctors in Tble 4 represent the mount of ech compound generted from delivering 1 meg-joule (MJ) het to the pot by given fuel/stove combintion. These tsk-bsed emission fctors of totl specited NMHCs lso hd lrge rnge, from 1.11 mg to 2.41 g per 1 MJ het delivered. Within-fuel type rnges were 3.28 mg to 2.41 g for cols, 0.57 g to 0.82 g for wood, 0.37 g to 1.52 g for crop residues, 0.02 g for kerosene, nd 1.11 mg to 17.5 mg for gseous fuels. Ozone Forming Potentils. The estimted ozone forming potentils (OFPs) on dry-fuel mss bsis rnged from 9 mg to 11 g ozone per kilogrm of dry fuel from 16 fuel-stove combintions tested in this study. The estimted men OFP ws 6.03 g ozone per kilogrm of dry fuel for col (col powder), 0.03 g for col briquettes nd honeycomb col briquettes, 1.68 g for wood, 1.75 g for crop residue, 0.74 g for kerosene, 0.14 g for LPG, 0.03 g for col gs, nd 0.06 g for nturl gs. The estimted OFPs on delivered energy bsis rnged from 1.41 mg to 4.05 g per 1 MJ of het delivered to the pot. Figure 1 shows delivering the sme mount of het to the VOL. 37, NO. 13, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

4 TABLE 2. Molr Emission Rtios to CO 2 (Unit in 10-7 ) Honey- CV fuel/stove b men c (%) Honey- Metl Honey- Imp. Briq- Metl Wsh Wood- Wood- Imp.-v Whet- Mize- Mize- Imp.-v benzene ,3-butdiene nd n nd nd nd nd nd nd nd xylenes (o+m+p) nd nd 3.36 nd nd styrene nd n nd nd nd nd nd nd nd nd nd nd nd nd 243 nd ethne ethylene cetylene propne propene i-butne n 1.88 nd nd nd i-butene nd nd butene n n-butne nd trns-2-butene nd n ,2-dimethylpropne nd n nd nd nd nd nd nd nd nd nd nd nd nd 3.19 nd nd cis-2-butene nd n nd nd methyl-1-butene nd n nd nd nd nd nd i-pentne n 2.52 nd nd nd nd nd pentene nd n nd nd 8.37 nd nd methyl-1-butene nd n nd nd nd nd 4.17 nd nd n-pentne nd nd nd nd trns-2-pentene nd n nd nd nd nd nd 4.15 nd nd nd nd cis-2-pentene nd n nd nd nd nd nd nd 2.71 nd nd nd 2-methyl-2-butene nd n nd nd nd nd nd nd nd nd nd cyclopentene nd n nd nd nd nd nd nd nd nd nd nd 1.48 nd nd nd 4-methyl-1-pentene nd n nd nd nd nd nd nd nd nd nd nd nd nd nd nd cyclopentne nd n nd nd nd nd nd nd nd nd nd nd 2,3-dimethylbutne nd n nd nd nd nd nd nd nd nd nd nd nd 2-methylpentne nd nd nd nd nd nd nd methylpentne nd nd nd nd nd nd nd nd nd nd 1-hexene nd n nd nd nd 7.22 nd nd n-hexne nd nd nd methylcyclopentne nd n nd nd nd nd nd nd nd nd nd nd ,4-dimethylpentne nd n nd nd nd nd nd nd nd nd nd nd nd nd nd cyclohexne nd n nd nd nd nd nd nd nd nd nd nd 13.6 nd nd 2-methylhexne nd nd nd nd nd nd nd nd nd nd nd ,3-dimethylpentne nd n nd nd nd nd nd nd nd nd nd nd nd methylhexne n nd nd nd nd nd nd nd nd ,2,4-trimethylpentne n nd nd nd nd nd nd 1.45 nd nd 3.82 nd nd nd n-heptne nd nd nd nd nd methylcyclohexne n nd nd nd nd nd nd 2.04 nd nd nd 2,3,4-trimethylpentne nd n nd nd nd nd nd nd nd nd nd nd nd nd nd nd toluene methylheptne nd n nd nd nd nd nd nd nd nd nd nd 3-ethylhexne nd n nd nd nd nd nd nd nd nd nd nd nd n-octne nd nd nd 1.50 nd ethylbenzene nd nd nd nd n-nonne nd nd nd nd nd nd 4.43 nd nd i-propylbenzene nd n nd nd nd nd nd nd nd nd nd nd nd nd nd n-propylbenzene nd n nd nd nd nd nd nd nd nd nd nd nd nd nd nd p-ethyltoluene nd n nd nd nd nd nd nd nd nd nd nd nd nd nd nd m-ethyltoluene nd n nd nd nd nd nd nd nd nd nd nd nd nd nd nd 1,2,4-trimethylbenzene nd sec-butylbenzene 2.43 n nd 2.57 nd nd nd 2.93 nd nd nd nd nd nd nd nd n-decne nd nd nd 529 nd nd nd nd nd nd totl specited NMHC rnking (1 ) highest) Key: nd ) not detected or bckground level g flue gs concentrtion; CV (coefficient of vrition) ) (stndrd devition)/men; n ) vrince clcultion cnnot be mde, only one or no mesurement with detected level. b See Tble 1 for fuel/stove codes. c Averge of 3 burning cycles using Honey- combintion. Kero- Wick LPG- Nturl pot by different col/stove combintions would contribute quite differently to the ozone forming potentils. The results indicted tht burning col powder nd biomss fuels would contribute highest ozone forming potentil mong the 16 tested fuel-stove combintions on delivered energy bsis. Strikingly, burning the col briquette nd honeycomb col briquette produced OFP vlues more thn 2 orders of mgnitude lower thn burning unprocessed col, even in the sme vented metl stove. However, the interprettion of these OFP results is very crude becuse the MIR vlues used in our estimtion were from the dtbse developed for the United Sttes photochemicl sitution which my be different from the Chinese sitution due to the possible difference in reltive mkeup of ozone precursors (NO x nd NMHCs). To put these cookstove OFPs into perspective, we compre them with the published vehiculr OFPs reported by Chng et l. for gsoline-powered nd LPG-powered vehicles representtive of the Tipei fleet (16). The OFPs were estimted bsed on emission fctors of some 56 NMHCs nd their MIR vlues: 1.8 g/km-vehicle for (verge) gsoline-powered vehicles nd 0.97 g/km-vehicle for (verge) LPG-powered vehicles. If we use 5 MJ delivered energy s wht is roughly needed for cooking mel (5), cooking one mel would produce OFPs rnging from g (the stove) ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 13, 2003

5 TABLE 3. Emission Fctors (mg of Compound per kg of Dry Fuel, by Fuel/Stove Combintion) b fuel/stove c Honey Honey- Honey- d Metl Imp. Briq- Metl Wsh Wood- Wood- Imp.-v Whet- Mize- Mize- Imp.-v Kero- Wick LPG- Nturl benzene ,3-butdiene nd nd nd nd nd nd nd nd xylenes (o+m+p) nd nd 1.10 nd nd styrene nd nd nd nd nd nd nd nd nd nd nd nd nd 177 nd ethne ethylene cetylene propne propene i-butne nd nd nd i-butene nd nd butene n-butne nd trns-2-butene nd ,2-dimethylpropne nd nd nd nd nd nd nd nd nd nd nd nd nd 1.62 nd nd cis-2-butene nd nd nd methyl-1-butene nd nd nd nd nd nd i-pentne nd nd nd nd nd pentene nd nd nd 1.81 nd nd methyl-1-butene nd nd nd nd nd nd nd n-pentne nd nd nd nd trns-2-pentene nd nd nd nd nd nd nd nd nd nd cis-2-pentene nd nd nd nd nd nd nd nd nd nd 2-methyl-2-butene nd nd nd nd nd nd nd nd nd nd cyclopentene nd nd nd nd nd nd nd nd nd nd nd nd nd nd 4-methyl-1-pentene nd nd nd nd nd nd nd nd nd nd nd nd nd nd nd cyclopentne nd nd nd nd nd nd nd nd nd nd nd 2,3-dimethylbutne nd nd nd nd nd nd nd nd nd nd nd nd 2-methylpentne nd nd nd nd nd nd nd methylpentne nd nd nd nd nd nd nd nd nd nd 1-hexene nd nd nd nd 1.88 nd nd n-hexne nd nd nd methylcyclopentne nd nd nd nd nd nd nd nd nd nd nd ,4-dimethylpentne nd nd nd nd nd nd nd nd nd nd nd nd nd nd cyclohexne nd nd nd nd nd nd nd nd nd nd nd 7.96 nd nd 2-methylhexne nd nd nd nd nd nd nd nd nd nd nd ,3-dimethylpentne nd nd nd nd nd nd nd nd nd nd nd nd methylhexne nd nd nd nd nd nd nd nd ,2,4-trimethylpentne nd nd nd nd nd nd nd nd 3.09 nd nd nd n-heptne nd nd nd nd nd methylcyclohexne nd nd nd nd nd nd 1.42 nd nd nd 2,3,4-trimethylpentne nd nd nd nd nd nd nd nd nd nd nd nd nd nd nd toluene methylheptne nd nd nd nd nd nd nd nd nd nd nd 3-ethylhexne nd nd nd nd nd nd nd nd nd nd nd nd n-octne nd nd nd 1.22 nd ethylbenzene nd nd nd nd n-nonne nd nd nd nd nd nd 3.96 nd nd i-propylbenzene nd nd nd nd nd nd nd nd nd nd nd nd nd nd n-propylbenzene nd nd nd nd nd nd nd nd nd nd nd nd nd nd nd p-ethyltoluene nd nd nd nd nd nd nd nd nd nd nd nd nd nd nd m-ethyltoluene nd nd nd nd nd nd nd nd nd nd nd nd nd nd nd 1,2,4-trimethylbenzene 1.77 nd 1.90 nd nd nd 1.47 nd nd nd nd nd nd nd nd nd sec-butylbenzene n-decne nd nd nd 260 nd nd nd nd nd nd totl specited NMHC Zhng et l. (6) presented detiled informtion on clcultion for emission fctors using the crbon blnce model. b Key: nd ) not detected or bckground level g flue gs concentrtion. c See Tble 1 for fuel/stove codes. d Averge of 3 burning cycles using Honey- combintion. to 20.3 g (the stove). At the lower bound of the cookstove OFP rnge, cooking 257 mels would produce OFP equivlent to driving gsoline-powered vehicle for 1 km, nd cooking 139 mels would produce OFP equivlent to driving LPG-powered vehicle for 1 km. At the upper bound of the cookstove OFP rnge, in contrst, cooking 1 mel would produce OFP equivlent to driving gsoline-powered vehicle for bout 11 km or equivlent to driving LPG-powered vehicle for bout 21 km. Discussion Uncertinties for Emission Fctors. Due to the budgetry constrint, only one fuel/stove combintion (Honey-Metlv), mong the 16 combintions mesured for NMHC emissions, ws repetedly mesured for three independent burn cycles (experimentl runs). These mesurements showed high run-to-run vritions (8-113% CV) in emission fctors estimted using the crbon blnce model for the Honey- combintion. High CV vlues were lso expected for the emission fctors of specited NMHCs for the other 15 fuel/stove combintions bsed on the nlysis of uncertinty sources of the crbon blnce pproch. There were two mjor sources of uncertinty for this pproch. One ws the vribility in emissions from one burn cycle to nother, lrgely ssocited with vritions in fire tending behvior nd vritions in fuel prmeters (e.g., fuel size, fuel lyout in the stove). This type of vribility ws obviously greter for the solid fuel stoves thn for the LPG nd gs stoves. The other source ws the mesurement errors for ll the prmeters nd concentrtion vlues used in the crbon VOL. 37, NO. 13, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

6 TABLE 4. Emission Fctors (mg Compound/MJ Delivered to the Pot, by Fuel/Stove Combintion) fuel/stove Honey Honey- Honey- b Metl Imp. Briq- Metl Wsh Wood- Wood- Whet- Mize- Mize- Imp.-v Imp.-v Kero- Wick LPG- Nturl stove efficiency (%) low-het CV(MJ/kg fuel) benzene ,3-butdiene nd nd nd nd nd nd nd nd xylenes (o+m+p) nd nd nd nd styrene nd nd nd nd nd nd nd nd nd nd nd nd nd 7.56 nd ethne ethylene cetylene propne propene i-butne nd nd nd i-butene nd nd butene n-butne nd trns-2-butene nd ,2-dimethylpropne nd nd nd nd nd nd nd nd nd nd nd nd nd nd nd cis-2-butene nd nd nd methyl-1-butene nd nd nd nd nd nd i-pentne nd nd nd nd nd pentene nd nd nd 1.21 nd nd methyl-1-butene nd nd nd nd nd nd nd n-pentne nd nd nd nd trns-2-pentene nd nd nd nd nd nd nd nd nd nd cis-2-pentene nd nd nd nd nd nd nd nd nd nd 2-methyl-2-butene nd nd nd nd nd nd nd nd nd nd cyclopentene nd nd nd nd nd nd nd nd nd nd nd nd nd nd 4-methyl-1-pentene nd nd nd nd nd nd nd nd nd nd nd nd nd nd nd cyclopentne nd nd nd nd nd nd nd nd nd nd nd 2,3-dimethylbutne nd nd nd nd nd nd nd nd nd nd nd nd 2-methylpentne nd nd nd nd nd nd nd methylpentne nd nd nd nd nd nd nd nd nd nd 1-hexene nd nd nd nd 1.25 nd nd n-hexne nd nd nd methylcyclopentne nd nd nd nd nd nd nd nd nd nd nd ,4-dimethylpentne nd nd nd nd nd nd nd nd nd nd nd nd nd nd cyclohexne nd nd nd nd nd nd nd nd nd nd nd nd nd 2-methylhexne nd nd nd nd nd nd nd nd nd nd nd ,3-dimethylpentne nd nd nd nd nd nd nd nd nd nd nd nd methylhexne nd nd nd nd nd nd nd nd ,2,4-trimethylpentne nd nd nd nd nd nd nd nd nd nd nd n-heptne nd nd nd nd nd methylcyclohexne nd nd nd nd nd nd nd nd nd 2,3,4-trimethylpentne nd nd nd nd nd nd nd nd nd nd nd nd nd nd nd toluene methylheptne nd nd nd nd nd nd nd nd nd nd nd 3-ethylhexne nd nd nd nd nd nd nd nd nd nd nd nd n-octne nd nd nd nd ethylbenzene nd nd nd nd n-nonne nd nd nd nd nd nd nd nd i-propylbenzene nd nd nd nd nd nd nd nd nd nd nd nd nd nd n-propylbenzene nd nd nd nd nd nd nd nd nd nd nd nd nd nd nd p-ethyltoluene nd nd nd nd nd nd nd nd nd nd nd nd nd nd nd m-ethyltoluene nd nd nd nd nd nd nd nd nd nd nd nd nd nd nd 1,2,4-trimethylbenzene nd nd nd nd nd nd nd nd nd nd nd nd nd sec-butylbenzene n-decne nd nd nd 157 nd nd nd nd nd nd totl specited NMHC Key: nd ) not detected or bckground level g flue gs concentrtion. b Averge of 3 burning cycles using Honey- combintion. blnce equtions. The mesurement errors ppered to be greter for those compounds present t level below or close to the method detection limit. The gs stoves emitted more compounds t low concentrtion levels. The estimted CV vlues of emission fctors reflected the overll contribution of these two uncertinty sources. This nlysis is supported by lrge CV vlues reported for the emission fctors of TNMHC for the 28 Chinese fuel/stove combintions including the 16 ones mesured for specited NMHCs (6). The CV vlues for the TNMHC emission fctors rnged from 13% to 173% cross the 28 fuel/stove combintions with the highest vlue being for the Honey-. The CV vlues for the specited NMHCs ppered to be within the CV vlue rnges for the TNMHC for the Honey-, the only fuel/stove combintions tht were repetedly mesured for both specited NMHCs nd TNMHC. [The sum of mesured individul NMHCs did not ccount for ll the HCs present in the smples. It represents, in theory, frction of TNMHC. However, TNMHC nd specited NMHCs were mesured using different methods (see ref 6), nd thus the sums of specited NMHCs cnnot be directly compred with the concentrtions of TNMHC. Mesurement precisions for TNMHC nd specited NMHCs were, however, expected to be similr.] Bsed on these dt, we expect the emission fctors of specited NMHCs reported in this pper my hve lrge CV vlues, but probbly smller thn 173%. Another source of uncertinty nd reson to do more tests in field settings is tht wood fuels seem to hve higher emission rtes of PICs during the smoldering phse of combustion (17). Since smoldering cn continue in the field fter cooking ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 13, 2003

7 FIGURE 1. Ozone forming potentil (mg ozone/mj) for the 16 fuel/stove combintions (log-scle). The stndrd devition br for Honey- ws derived from three repeted mesurements. No repeted mesurements were mde for the other 15 fuel/stove combintions. session is completed, the emissions from the smoldering re difficult to mesure in simulted settings. This is only serious issue with wood fuels, however, s our tests of the Chinese fuel/stove combintions did include smoldering phses for col nd crop residues do not smolder significntly fter fueling stops. Gses nd liquids, of course, hve no such smoldering phses. Specited NMHCs were lso mesured in the pilot study in which 6 fuel/stove combintions in the Philippines were tested using the crbon blnce pproch (2, 4, 18). The 6 fuel/stove combintions tested included the stoves burning LPG, kerosene, chrcol, nd wood fuels. As it ws typicl in Mnil, none of the tested stove hd flues or chimneys. Compred to the extensive study (5, 6), the pilot study mesured fewer prmeters tht needed to construct the complete crbon blnce model nd did not use hood to promote mixing of flue gses. No repeted mesurements were mde in the pilot study. Hence, the results from the pilot study cn be expected to be less ccurte thn those from the extensive study. Nevertheless, both the pilot study nd the present study showed tht burning wood fuels yielded greter emissions of benzene ( g s crbon per kg dry fuel) thn burning other fuels. 1,3-Butdiene emissions mesured in the present study, however, were substntilly lower (by 1 or 2 orders of mgnitude) thn those mesured in the pilot study for the similr fuel/stoves tested in both studies. Reltively lrge differences in other NMHC emissions were lso found when the results from the two studies were compred. A very limited number of other studies of cookstove emissions only mesured certin specific compounds of interest nd usully did not report NMHCs. Therefore, it is necessry to conduct future systemtic studies nd uncertinty nlyses to better understnd the nturl vribility nd to reduce the mesurement errors ssocited with emission fctors for vrious fuel-stove systems. Ozone Concerns. Locl nd regionl pollution of ozone nd photochemicl smog is common worldwide nd occurs in mny urbn nd suburbn res in Chin. Among the specited NMHCs identified in this study, some compounds hd higher ozone forming potentils thn others. Sturted orgnic compounds such s lknes usully hve lower MIR vlues thn lkenes nd romtic species tht re more ctive in terms of forming ozone. When both MIR vlues nd mount of emissions were considered, ethylene, mong the 54 specited NMHCs, ws the mjor contributor to OFP vlues for most fuel-stove combintion tested in this study. In most fuel/stove combintions, emission of ethylene contributes more thn 50% of ozone (rnging from 12% to 85%), followed by cetylene (1-46%), xylenes (0-32%), styrene (0-31%), propene (10-24%), cis-2-butene (0-19%), 1,2,4-trimethylbenzene nd sec-butylbenzene (0-18%), nd toluene (0-10%). The type nd mount of fuels used in Chinese households vries in urbn nd rurl res (11, 12). In urbn res of Chin, the mjor types of fuel used in the residentil sector re col, electricity, LPG, nturl gs, nd col gs. In rurl res, biomss fuels (crop residues nd wood) re the mjor types of fuel used in the residentil sector. Becuse groundlevel ozone ccumultion results from complex chemicl rections, in the tmosphere, between NMHCs nd NO x, the ctul ozone formtion my be limited by NO x concentrtions in rurl res where no significnt sources of NO x re present. Therefore, here we only discuss ozone-forming potentils for urbn res. However, some NMHCs emitted from rurl household stoves cn be trnsported to urbn res, contributing to the urbn ozone pollution. Tble 5 shows the estimted totl ozone forming potentils from domestic cookstoves in the urbn res of Chin bsed on the fuel usge nd the specific OFP vlue for ech type of fuel mesured in this study. The estimtes in Tble 5 were VOL. 37, NO. 13, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

8 TABLE 5. Estimted Yerly Ozone Forming Potentil (OFP) in the Urbn Are of Chin relted to household cookstoves fuel types v OFP (g ozone /kg fuel) mount of consumption in the urbn re of Chin in 1995 (metric ton 10 6 ) estimted totl OFP per yer in the urbn re (ton ozone) 10 2 rw col b 3515 col briquette kerosene LPG col gs nturl gs totl 3525 Reference 11. b Totl mount of rw col nd wshed col consumed. fuel specific rther thn fuel/stove specific, becuse dt on fuel consumption were only fuel specific. We brek down the col consumption dt by col type to seprte rw col nd col briquette (including honeycomb col briquette) consumptions. Our estimtes indicte tht the totl ozone mount ttributble to urbn residentil NMHC emissions in 1995 were pproximtely tons per yer. The results indicted tht col-fired cookstoves would contribute > 99% of totl ozone mount ttributble to NMHC emissions from ll the cookstoves used in urbn Chin. This lrge col contribution ws minly driven by the high OFP vlue for rw col nd the lrge mount of rw col consumption. However, relible dt sources of col briquette consumption or production re uncertin becuse most col briquettes were produced in smll workshops where no effective reporting or trcking systems were in plce to record the production dt. We suspect tht the production of col briquettes might be substntilly underestimted in the Chin ntionl energy sttistics nd, hence, the OFP estimte for col briquette in Tble 5 my be lower thn the ctul vlue. Bsed on the sttisticl informtion from the Chin Energy Yerbook (11, 12), the rw col usge hs decresed grdully yerly nd replced by col briquettes, kerosene, nd gseous fuels. However, since the mount of rw col consumption is much lrger nd rw col hs the highest OFP vlue mong ll the fuel tested, the totl OFP from the urbn residentil sector (for cooking) should hve chnged insignificntly over the pst few yers. Our nlysis shows, for exmple, in 1997 rw col still contributes more thn 95% of totl OFP ttributble to household cookstove emissions of NMHCs. This type of OFP nlysis using the mesured emission fctors of NMHCs my provide useful informtion for energy decision-mking, from stndpoint of ozone control strtegy, in Chin or other countries hving similr suite of fuel/stove combintions. By switching the fuel/stove combintions, the NMHC contribution to ozone formtion cn be significntly reduced (see Figure 1). For exmple, on verge, switch of col powder to nturl gs would led to reduction of OFP by 70 times, nd switch of col powder to col gs would led to 40-fold OFP reduction. Among different types of col fuels, burning col briquettes nd honeycomb col briquettes contribute much less OFPs thn burning col powder. However, our estimtes were subject to lrge uncertinties due to the uncertinty ssocited with the emission fctors mesured, the uncertinty ssocited with fuel consumption dt, nd the uncertinty ssocited with MIR vlues (difference in tmospheric NO x/nmhc conditions between the United Sttes nd Chin urbn res). It will probbly be necessry to conduct mesurements under field conditions to reduce these uncertinties sufficiently to mke definitive policy decisions bout ozone control policy in Chin. Other Potentil Applictions of the Dtbse. The specited NMHC dtbse cn be useful in source pportionment studies. Emissions from numerous cooking stoves my be significnt source to prticulte mtter nd voltile orgnic compounds nd thus significntly influence the locl nd regionl ir qulity including the formtion of photochemicl smog. Sorting out the reltive importnce of sources is lwys ttrctive to the scientific community nd policy mkers. Source pportionment techniques hve been widely used to estimte reltive contributions of vrious sources to the concentrtions of given (trget) pollutnts mesured t the receptor. Using VOCs, similr to wht we mesured in the study, to do source pportionment hs been done in previous studies (19, 20). For exmple, Anderson et l. (20) pplied personl exposure mesurements for VOCs to four receptor-oriented source pportionment models. The previous studies used the VOC dt collected in the Totl Exposure Assessment Methodology (TEAM) studies conducted in New Jersey nd Cliforni s well s the dt collected in the Cliforni Indoor Exposure study to evlute severl receptor models. The models evluted include the Chemicl Mss Blnce model, Principl Component Anlysis/Absolute Principl Component Scores, Positive Mtrix Fctoriztion, nd Grphicl Rtio Anlysis for Composition Estimtes/Source Apportionment by Fctors with Explicit Restriction. The source pportionment results from the four models greed resonbly well for the New Jersey dt but were less consistent for the Cliforni dt. Model performnce vries from severl fctors such s vilbility of informtion on emission composition, number of receptor smples, rectivity of compounds used s trcers, mesurement uncertinty, frction of missing dt, frction of belowdetection-limit dt, number of fctors determined, fctors interprettion for sources, whether ll importnt sources re included, nd whether there is collinerity within/mong the source profiles used in the model. Nevertheless, this indictes tht n ccurte source profile is the key to the successful implementtion of source pportionment models. A previous nlysis by Zhng nd Smith (2) evluted lifetime cncer risks ssocited with exposure to benzene, 1,3-butdiene, styrene, nd xylenes relesed from the 6 fuel/ stove combintions tested in the Mnil pilot study. The nlysis found tht estimted cncer risk of benzene nd tht of styrene from the use of the tested wood stove could exceed published risk estimtes from ll sources of irborne benzene or styrene (excluding ctive tobcco smoking) in the United Sttes. These kind of importnt findings bsed on the limited dt of the pilot study should be further evluted using similr or more sophisticted nlysis bsed on the more extensive dt collected in this study. In the future nlysis, cncer risks for ll the tested fuel/stove combintions cn be compred to mke quntittive recommendtions on fuel or stove switching (2). Additionl nlyses my be preformed to lso ssess the noncncer helth risks. For exmple, this dtbse my be used to conduct chemicl mixture risk ssessment using groups (or clsses) of hydrocrbons bsed on crbon number frctions (e.g., C 5-C 8,C 9-C 16), their fte nd trnsport, nd toxicity chrcteristics. These will be discussed in more detiled helth risk nlysis pper to follow. Acknowledgments We gretly pprecite the effective nd timely peer review process of our pper. We pprecite the comments from Dr. Rufus Edwrds of University of Cliforni t Berkeley nd editoril ssistnce from Dr. Steven Miller of the New Jersey Deprtment of Helth nd Senior Services nd Mr. Robert C. Hrrington of Environmentl nd Occuptionl Helth Sciences Institute, New Jersey. We pprecite the ssistnce from Dr. Jonthn Sinton of Lwrence Berkeley Ntionl Lbortory in providing informtion on col consumption in Chin. The originl dt collection ws funded through ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 13, 2003