Supplement of Development of PM 2.5 source impact spatial fields using a hybrid source apportionment air quality model

Transcription

1 Supplement of Geosci. Model Dev., 8, , doi: /gmd supplement Author(s) CC Attribution 3.0 License. Supplement of Development of PM 2.5 source impact spatial fields using a hybrid source apportionment air quality model C. E. Ivey et al. Correspondence to: C. E. Ivey (sunni.ivey@gmail.com) The copyright of individual parts of the supplement might differ from the CC-BY 3.0 licence.

2 Note A During the chosen period of study, the CSN network performed thermal optical transmittance analyses to determine carbon species concentrations. As a result, organic and elemental carbon concentrations were converted to IMPROVE or total optical reflectance equivalents using previously derived methods (Malm et al. 2011). Additionally, measurements below detection limit are handled by setting the concentration to ½ of the measurement detection limit (MDL) and setting the measurement uncertainty to ⅔ of the MDL (Marmur et al. 2006).

3 Table S1. Median and mean CTM-RM and SH adjustment factors for withheld CSN observation locations Source Categories R CTM RM R SH R CTM RM R SH RMSE Agricultural Burning Aircraft Emissions Biogenic Emissions Coal CMᶧ Diesel CMᶧ Dust Fuel Oil CMᶧ Livestock Emissions Liquid Petroleum Gas CMᶧ Lawn Waste Burning Metal Processing Meat Cooking Mexican CMᶧ Mineral Processing Natural Gas CMᶧ NRᶳ Diesel CMᶧ NRᶳ Fuel Oil CMᶧ NRᶳ Gasoline CMᶧ NRᶳ Liquid Petroleum Gas CMᶧ NRᶳ Natural Gas CMᶧ Other NR Sources Open Fires Onroad Diesel CMᶧ Onroad Gasoline CMᶧ Other CMᶧ Sources Other PM Sources Prescribed Burning Railroad Emissions Seasalt Solvent Emissions Wildfires Woodfuel Burning Woodstoves ᶳNR = Nonroad, ᶧCM = Combustion

4 Table S2. January 2004 domain-wide PM 2.5 emissions and uncertainties. Source Categories Emissions (metric tons/day) Uncertainty (factor) Rank Agricultural Burning ± 5 12 Aircraft Emissions 25.1 ± Biogenic Emissions 0.0 ± Coal CMᶧ ± Diesel CMᶧ 2.0 ± Dust ± 10 1 Fuel Oil CMᶧ ± Livestock Emissions 1.3 ± Liquid Petroleum Gas CMᶧ 13.7 ± Lawn Waste Burning ± 5 6 Metal Processing ± Meat Cooking ± Mexican CMᶧ 13.2 ± Mineral Processing ± Natural Gas CMᶧ ± NRᶳ Diesel CMᶧ ± NRᶳ Fuel Oil CMᶧ 33.8 ± NRᶳ Gasoline CMᶧ ± NRᶳ Liquid Petroleum Gas CMᶧ 2.9 ± NRᶳ Natural Gas CMᶧ 0.3 ± Other NR Sources 5.8 ± Open Fires ± 5 9 Onroad Diesel CMᶧ ± Onroad Gasoline CMᶧ ± Other CMᶧ Sources ± Other PM Sources ± Prescribed Burning ± 10 7 Railroad Emissions 64.4 ± Seasalt ± Solvent Emissions 61.3 ± Wildfires 45.7 ± Woodfuel Burning ± Woodstoves ± 5 2 ᶳNR = Nonroad, ᶧCM = Combustion

5 Table S3. Mean observations and model simulations for January 2004 for withheld CSN observations (N = 75 observations). OBS CMAQ-DDM HYB SH Avg. Std. Avg. Std. RMSE Avg. Std. RMSE Avg. Std. RMSE PM OC EC NO NH SO Na Mg Al Si P Cl K Ca V E-4 3.1E E-4 5.0E Fe Cu Zn Se E E Pb E-4 3.8E E-4 7.8E Note: All averages and standard deviations are expressed in µg m -3.

6 Table S4. Linear regression and correlation coefficients for model simulations vs. observations for January 2004 for withheld CSN observations (N = 75 observations). Regression equation: Conc model = α + β Conc obs CMAQ-DDM vs. OBS HYB vs. OBS SH vs. OBS α SE α β SE β r α SE α β SE β r α SE α β SE β r PM OC EC NO NH SO Na Mg Al Si P E Cl K Ca V E E E E-4 8.8E E E Fe Cu E E Zn E E E Se E E-4 2.5E E Pb E E-4 9.7E E-4 7.9E

7 Table S5. Mean observations and model simulations for January 2004 for SEARCH monitors (N = 8 monitors). OBS CMAQ-DDM SH Avg Std. Avg Std. RMSE Avg Std. RMSE PM OC EC NO NH SO Al Si K Ca Fe Cu Zn Se 8.00E E E E Pb E E Note: All averages and standard deviations are expressed in µg m -3.

8 Table S6. Mean observations and model simulations for January 2004 for IMPROVE monitors (N = 38 monitors). OBS CMAQ-DDM SH Avg Std. Avg Std. RMSE Avg Std. RMSE PM OC EC NO NH SO Al Si K Ca Fe Cu 7.00E E Zn Se 9.00E E E Pb E E Note: All averages and standard deviations are expressed in µg m -3.

9 Table S7. Linear regression and correlation coefficients for model predictions vs. observations for January 2004 for SEARCH monitors (N = 8 monitors). Regression equation: Conc model = α + β Conc obs CMAQ-DDM vs. OBS SH vs. OBS α SE α β SE β r α SE α β SE β r PM OC EC NO NH SO Al Si K Ca Fe Cu 7.00E E Zn E Se E E Pb E E E

10 Table S8. Linear regression and correlation coefficients for model predictions vs. observations for January 2004 for IMPROVE monitors (N = 38 monitors). Regression equation: Conc model = α + β Conc obs CMAQ-DDM vs. OBS SH vs. OBS α SE α β SE β r α SE α β SE β r PM OC EC NO NH SO Al Si K Ca Fe Cu 3.00E E E E Zn Se 7.00E E E E Pb 8.00E E E E

11 Table S9. January 2004 domain-wide precursor emissions estimates (metric tons/day). Source Categories NOx VOC PM 2.5 SO 2 NH 3 CO EC POA Agricultural Aircraft Emissions Biogenic Emissions Coal CMᶧ Diesel CMᶧ Dust Fuel Oil CMᶧ Livestock LPG CMᶧ Lawn Waste Burn Metal Processing Meat Cooking Mexican CMᶧ Mineral Processing Natural Gas CMᶧ NRᶳ Diesel CMᶧ NRᶳ Fuel Oil CMᶧ NRᶳ Gasoline CMᶧ NRᶳ LPG CMᶧ NRᶳ Nat. Gas CMᶧ Other NR Sources Open Fires Onroad Diesel CMᶧ Onroad Gas CMᶧ Other CMᶧ Sources Other PM Sources Prescribed Burning Railroad Emissions Seasalt Solvent Emissions Wildfires Woodfuel Burning Woodstoves All Sources ᶳNR = Nonroad, ᶧCM = Combustion, LPG = Liquid Petroleum Gasoline

12 Figure S1. Cumulative distributions of original and interpolated hybrid adjustment factors for witheld CSN observation locations.

13 Figure S2. Spatial fields of temporally-averaged, kriged adjustment factors (R j SH ) for dust, onroad diesel combustion, on-road gasoline combustion, and woodstove sources for January Values were averaged over 9 observations days in the month. Note that each figure has a different scale.

14 Figure S3. Distribution of Rj values generated by hybrid analysis of CSN data for January 2004 (n = 26,400; 0.1 < Rj < 10). Note that the y-axis is on a log scale and Rj values are log transformed.

15 Figure S5. Hybrid adjustment of biomass burning impacts on PM 2.5 on January 4 th and 22 nd in Biomass burning fields are produced by aggregating source impacts from agricultural burning, lawn waste burning, open fires, prescribed burning, wildfires, woodfuel and woodstove burning. (a) CMAQ-DDM spatial field for January 4th. (b) SH spatial field for January 4 th. (c) CMAQ-DDM spatial field for January 22 nd. (d) SH spatial field for January 22 nd.

16 Figure S6. Hybrid adjustment of metals processing impacts on PM 2.5 on January 4 th and 22 nd in (a) CMAQ-DDM spatial field for January 4th. (b) SH spatial field for January 4 th. (c) CMAQ-DDM spatial field for January 22 nd. (d) SH spatial field for January 22 nd.

17 Figure S7. Hybrid adjustment of natural gas combustion (point and area sources) source impact fields on January 4 th and 22 nd in (a) CMAQ-DDM spatial field for January 4th. (b) SH spatial field for January 4 th. (c) CMAQ-DDM spatial field for January 22 nd. (d) SH spatial field for January 22 nd.