2015 The 17 th GEIA Conference Influence of

2015 -The 17 th GEIA Conference : Influence of Urbanization on Emission Worldwide Quantitative analysis of uncertainty in sectorbased emission factors Junyu (Allen) ZHENG Zhuangmin Zhong South China University of Technology Beijing, November, 19, 2015

Why Uncertainty Analysis in Emission Inventory? l Uncertainties are in inherent in compiling air pollutant emission inventories (EIs); l Characterizing uncertainty in EIs can help prioritize source categories for future improvement and assess the quality of Eis; l Quantitative uncertainty information in EIs are fundamental input data for analyzing the impact of uncertainty in EIs on air quality modeling results and thereby can help improve air quality model performance.

Key Challenges in Conducting Uncertainty Analysis Statistical Analysis Output Uncertainty Input n Input 1 Monte Carlo Simulation Emission Factor Activity Factor Emission Inventory Models Total Emission • In most cases, a comprehensive quantitative uncertainty analysis is difficult to be conducted, or cannot be done； • Currently available emission factor database has no quantitative information on uncertainty in emission factors； • Lack of information on uncertainties in emission factors or other parameters is the key challenge.

AP -42 Emission Factor Uncertainty Assessment • In 2007, EPA conducted a emission factor uncertainty assessment • Quantitative uncertainty information were given for only 43 Arated and 1 B-rated AP-42 emission factors

Objectives l To develop a sector-based emission factor database with quantitative uncertainty information with the use of available emission testing data and various emission factor database sources such as from AP-42, IIASA, EEA for those sectors where possible； l To help quantify uncertainties or judge possible ranges in urban, regional or even national-scale emission inventories；

Methods and Tools l For those sectors where emission testing data or measurements or investigation are available, bootstrap simulation was used to quantify uncertainty in emission factors with the aid of Auv. Tool. Pro; l For those sectors where emission testing data or measurements are not available, or not enough, a comprehensive review was first made on available emission factor database and literatures (including recently developed Chinese domestic emission factors and internationally well-known database such as USEPA AP-42 database, EEA , IIASA, and others), then probabilistic distributions were developed to represent uncertainties in pollutant-based emission factors with the use of statistical analysis and expert judgment approaches;

Example 1: Industrial Coating Sector Activity Data : raw materials Data from emission testing or Uncertainty range investigation Sector Pollutant Emission factor(kg/kg) Uncertainty Distribution 95% confidence interval Wood furniture coating Automotive industry VOCs 0. 45 0. 42 Weibull(0. 51, 3. 09) Gamma(6. 29, 0. 06) 0. 41 -0. 50 0. 29 -0. 57 -9%~11% -31%~36% Shipbuilding VOCs 0. 36 Normal（1. 15, 0. 36） 0. 20 -0. 54 -44%~50% Metal surface coating VOCs 0. 35 Gamma（6. 33, 0. 05） 0. 27 -0. 44 -23%~26% Plastic surface coating VOCs 0. 34 Normal（0. 45, 0. 36） 0. 17 -0. 50 -50%~47% Fabric coating VOCs 0. 4 Weibull（0. 48, 0. 88） 0. 11 -0. 91 -73%~128% Uncertainty of EF based on raw materials < Uncertainty of EF based on production Activity Data: production 95% confidence interval Uncertainty range 0. 83 -2. 1 -42%~48% Gamma（1. 79, 11. 38） 8. 84 -37. 32 -56%~84% t/ship Weibull（9. 21, 0. 61） 1. 44 -45. 29 -90%~230% 1. 07 kg/metal production Weibull（1. 25, 0. 72） 0. 53 -2. 87 -50%~168% VOCs 0. 59 kg/plastic production Normal（0. 45, 1. 62） 0. 37 -0. 82 -37%~79% VOCs 27. 46 8. 93 -49. 54 -67%~80% Sector Pollutant Emission factor Units Wood furniture coating VOCs 1. 42 Automotive industry VOCs 20. 3 kg/car Shipbuilding VOCs 13. 72 Metal surface coating VOCs Plastic surface coating Fabric coating Uncertainty Distribution kg/piece of furniture Weibull（1. 12，0. 71） kg/m 2 cloth Normal（26. 18, 17. 80）

Example 2: Power Plant Sector Fuel Capacity Coal <125 MW Coal >125 MWbut<300 MW >300 MWbut<600 MW Polluta Emission nt factor(Mean) Unit Uncertainty Distribution NOx 7. 16 kg/t Gamma(9. 13, 0. 79) NOx 6. 48 kg/t NOx 8. 14 Data from emission factor 95% confidence database or Uncertainty range interval literatures 5. 53 -8. 93 -23%-25% Normal(6. 52, 2. 62) 4. 07 -9. 14 -37%-41% kg/t Normal(8. 17, 2. 36) 5. 92 -10. 67 -27%-31% Coal >600 MW NOx 5. 11 kg/t Weibull(5. 80, 3. 13) 3. 33 -7. 08 -35%-39% Heavy oil all NOx 9. 75 kg/t Normal(9. 77, 4. 26) 7. 49 -12. 50 -23%-28% Gas all NOx 12. 89 kg/m 3 Gamma(0. 57, 22. 81) 4. 64 -28. 24 -64%-119% Coal all PM 10 9. 28 kg/t Normal(9. 39, 7. 75) 3. 2 -16. 59 -66%-79% Heavy oil all PM 10 0. 78 kg/t Normal(0. 78, 0. 06)) 0. 72 -0. 85 -58%-59% Gas all PM 10 0. 09 kg/m 3 Gamma(4. 54, 2. 71) 0. 03 -0. 18 -67%-100% Coal all PM 2. 5 4. 5 kg/t Normal(4. 54, 2. 72) 2. 49 -6. 68 -45%-48% Heavy oil all PM 2. 5 0. 58 kg/t Gamma(24. 2, 0. 01) 0. 55 -0. 62 -65%-77% Gas all PM 2. 5 0. 07 kg/m 3 Gamma(1. 34, 0. 05)) 0. 03 -0. 14 -57%-100% Coal all VOC 0. 13 kg/t Weibull(0. 13, 1. 15) 0. 05 -0. 24 -62%-85% Heavy oil all VOC 0. 08 kg/t Weibull(0. 10. 2. 27) 0. 05 -0. 12 -38%-50% Gas all VOC 0. 08 kg/m 3 Gamma(2. 27, 0. 04) 0. 04 -0. 14 -50%-75% Coal all CO 1. 18 kg/t Normal(1. 15, 1. 11) 0. 05 -2. 32 -36%-37% Heavy oil all CO 0. 48 kg/t Weibull(0. 54, 3. 27) 0. 31 -0. 65 -35% Gas all CO 1. 12 kg/m 3 Weibull(1. 22, 1. 44) 0. 47 -2. 02 -58%-80%

Example 3: Residential Combustion Sector Data from emission factor database or literatures Subsector Pollutant Emission factor(Mean) Unit Uncertainty Distribution 95% confidence interval Uncertainty range Honeycomb briquet BC 0. 2 kg/t Weibull（0. 11, 0. 52） 0. 06 -0. 47 -70%-135% lump coal BC 2. 47 kg/t Weibull（0. 83, 0. 41) 0. 50 -6. 70 -80%-171% soft coal BC 1. 55 kg/t Weibull(1. 40, 0. 78) 0. 55 -3. 27 -65%-111% blind coal BC 0. 01 kg/t Gamma(0. 98, 0. 01) 0. 01 -0. 02 -50%-100% Honeycomb briquet OC 1. 58 kg/t Weibull(0. 86, 0. 53) 0. 48 -3. 72 -70%-135% lump coal OC 1. 76 kg/t Weibull(1. 52, 0. 72) 0. 81 -3. 25 -54%-85% soft coal OC 3. 35 kg/t Normal(3. 34, 1. 73) 2. 30 -4. 36 -31%-30% blind coal OC 0. 15 kg/t Gamma(0. 65, 0. 23) 0. 07 -0. 30 -53%-100% Honeycomb briquet PM 10 1. 26 kg/t Normal(1. 30, 1. 46) 0. 03 -2. 61 -98%-107% lump coal PM 10 5. 41 kg/t Normal(5. 58, 4. 86) 1. 33 -9. 36 -75%-73% soft coal PM 2. 5 7. 86 kg/t Normal(7. 75, 3. 10) 4. 70 -11. 58 -40%-47% blind coal PM 2. 5 2. 09 kg/t Gamma(1. 59, 1. 25) 0. 75 -4. 34 -64%-108%

Example 4: Dust Source Sector Type Pollutant Emission factor(Mean) Construction dust PM 10 0. 14 g/(m 2·h) Road dust PM 10 2. 83 Road construction dust PM 10 Construction dust Unit Uncertainty Distribution Data from field emission testing and emission factor database 95% Uncertainty confidence range interval Lognormal(2. 22, 0. 9) 0. 08 -0. 24 -43%-71% g/VKT Gamma(2. 22, 1. 29) 2. 03 -3. 66 -28%-29% 2. 12 g/（m 2. d） Weibull(1. 91, 0. 87) 0. 23 -5. 60 -89%-164% PM 2. 5 0. 04 g/(m 2·h) Weibull(0. 04, 1. 05) 0. 02 -0. 07 -50%-75% Road dust PM 2. 5 0. 45 g/VKT Weibull(0. 50, 1. 44) 0. 33 -0. 62 -27%-38% Road construction dust PM 2. 5 0. 66 g/（m 2. d） Weibull(0. 67, 0. 94) 0. 17 -1. 82 -74%-176%

Example 5: Biomass Burning Sector Data from emission factor database or 95% confidence Uncertainty interval literatures range Types Pollutant Emission factor(Mean) Unit Uncertainty Distribution Firewood BC 0. 83 kg/t Normal(0. 85, 51) 0. 47 -1. 18 -43%-42% Straw-household BC 0. 69 kg/t Weibull(0. 75, 1. 25) 0. 38 -1. 07 -45%-55% Straw-opening burnign BC 0. 44 kg/t Gamma(2. 88, 0. 15) 0. 32 -0. 57 -27%-30% Forest fire BC 0. 42 kg/t Normal(0. 42, 0. 29) 0. 11 -0. 71 -74%-69% Firewood OC 1. 52 kg/t Gamma(2. 51, 0. 58) 0. 92 -2. 32 -39%-53% Straw-household OC 2. 77 kg/t Gamma(1. 29, 2. 13) 1. 54 -4. 79 -44%-73% Straw-opening burnign OC 2. 23 kg/t Normal(2. 26, 1. 43) 1. 47 -3. 08 -34%-38% Forest fire OC 4. 47 kg/t Normal(4. 51, 2. 92) 1. 3 -7. 35 -69%-64% Firewood PM 10 3. 17 kg/t Gamma(1. 42, 2. 23) 1. 91 -4. 98 -40%-57% Straw-household PM 10 8. 57 kg/t Gamma(2. 98, 2. 88) 6. 36 -11. 36 -26%-33% Straw-opening burnign PM 10 8. 21 kg/t Normal(8. 33, 3. 10) 6. 2 -10. 19 -24% Forest fire PM 10 27. 71 kg/t Gamma(0. 91, 28. 61) 5. 9 -69. 595 -79%-151% Grassland fire PM 10 6. 25 kg/t Normal(6. 23, 0. 86) 5. 2 -7. 28 -17%-16% Firewood PM 2. 5 3. 81 kg/t Gamma(2. 86, 1. 34) 2. 69 -5. 14 -29%-35% Straw-household PM 2. 5 8. 49 kg/t Normal(8. 44, 4. 27) 6. 95 -10. 39 -18%-22% Straw-opening burnign PM 2. 5 7. 11 kg/t Normal(7. 17, 3. 33) 5. 64 -8. 73 -21%-23% Forest fire PM 2. 5 25. 09 kg/t Gamma(0. 93, 25. 89) 6. 62 -55. 07 -74%-119% Grassland fire PM 2. 5 5. 84 kg/t Weibull(6. 11, 10. 81) 5. 07 -6. 40 -13%-10%

A Case Study: Uncertainty Analysis of NH 3 Emissions from Agricultural and Husbandry Source in the PRD Region 1. Development of uncertainty analysis model 4. Uncertainty analysis Emission(t) 95% Confidence interval Uncertainty range 235357 (159248，323382) （-30%，42%） dairy-L 888 (478，2024） （-51%，108%） dairy-S 383 (137，660） （-56%，111%） Beef-L 551 (185，1204） （-69%，104%） Beef-S 8215 (2872，18046） （-68%，103%） yellow cattle-L 1000 (548，2013） （-50%，82%） yellow cattle-S 14907 (7966，29379） （-51%，80%） buffalo 11762 (5246，19146） （-50%，81%） Sow-L 13030 (7769，41648） （-59%，118%） Sow-S 13282 (5568，28943） （-60%，110%） Pig-L 64471 (14653，123856） （-72%，134%） Pig-S 60771 (10904，86327） （-71%，126%) Layers 4904 (2968，11379） （-52%，83%） Dorking 26619 (11873，81683） （-68%，122%） Laying duck 3543 (2875，4638） （-23%，24%） Duck 5178 (4165，7081） （-26%，27%） Goose 4485 (1533，23984） （-86%，125%） Pigeon 397 (315，464） （-19%，19%） Rabbit 618 (382，785） （-35%，33%） Goat 352 (204，850） （-50%，110%） N fertilizer application 238593 (81963，541835） （-68%，111%） Total 473950 (287728，781726） （-41%，61%） Source Livestock 2. Distribution types dairy-L Emission factor 27. 63 kg/a Gamma(13. 43, 2. 06) 95% confidence interval 23. 26 -32. 83 dairy-S 28. 51 kg/a Lognormal(3. 32, 0. 23) 24. 49 -33. 73 -14%-15% Beef-L Beef-S yellow cattle-L yellow cattle-S buffalo Sow-L 20. 94 16. 13 14. 93 12. 24 6. 82 kg/a kg/a Normal(21. 01, 4. 39) Gamma(22. 85, 0. 91) Normal(15. 96, 6. 46) Gamma(4. 79, 3. 11) Gamma(11. 33, 1. 08) Normal(6. 79, 4. 07) 17. 79 -24. 03 17. 99 -24. 04 11. 34 -20. 32 11. 06 -19. 75 8. 65 -17. 02 2. 36 -11. 22 -15%-13% -14%-13% -30%-21% -26%-24% -29%-28% -65%-39% Sow-S 9. 2 kg/a Weibull(10. 38, 2. 40) 5. 34 -13. 38 -42%-31% Pig-L 3. 48 kg/a Gamma(9. 53, 0. 36) 2. 32 -4. 85 -33%-28% Pig-S 9. 21 kg/a Weibull(10. 37, 2. 41) 5. 54 -13. 12 -40%-30% Layers Dorking Laying duck Duck Goose Pigeon Rabbit Goat N fertilizer application 0. 34 0. 17 0. 28 0. 16 0. 21 0. 19 0. 36 2. 34 kg/a kg/a Normal(0. 34, 0. 11) Normal(0. 17, 0. 09) Lognormal(-1. 29, 0. 17) Weibull(0. 15, 0. 83) Weibull(0. 24, 2. 12) Weibull(0. 12, 0. 54) Gamma(2. 64, 0. 14) Gamma(1. 12, 2. 11) 0. 26 -0. 43 0. 06 -0. 26 0. 23 -0. 33 0. 03 -0. 41 0. 12 -0. 32 0. 01 -0. 71 0. 15 -0. 68 0. 41 -5. 62 -24%-21% -65%-35% -18%-15% -81%-61% -43%-34% -95%-73% -58%-47% -82%-58% 0. 19 kg/t N Weibull(0. 22, 3. 37) 0. 16 -0. 24 -16%-21% Source 3. Bootstrap simulation Unit Uncertainty Distribution Uncertainty range -16%

Future Work l Continuing to develop probabilistic distributions in more detailed sector-based and pollutant-based emission factors , hopefully to have a sector-based emission factor uncertainty database for most emission source sectors if possible; l Demonstration of this emission factor uncertainty database to quantify uncertainty in the Pearl River Delta regional air pollutant emission inventories;

Thank you for your attention!