Harvard University, Cambridge, MA02138

Size: px

Start display at page:

Download "Harvard University, Cambridge, MA02138"

Evan Stokes
5 years ago
Views:

1 Impact of trans-boundary transport of carbonaceous aerosols on the regional air quality in the United States: A case study of the South American wildland fire of May Hee-Jin In 1+, Daewon W. Byun 1*, Rokjin J. Park 2, Nan-Kyoung Moon 1, Soontae Kim 1, and Sharon Zhong 1 1 Institute for Multidimensional Air Quality Studies, University of Houston, Houston, TX Division of Engineering an Applied Sciences and Department of Earth and Planetary Sciences, Harvard University, Cambridge, MA02138 *Corresponding author: Prof. Daewon W. Byun Science and Research 1 Building, Rm 312 University of Houston 4800 Calhoun Rd. Houston, Texas daewon.byun@mail.uh.edu + Present Affiliation: Department of Environmental Science and Engineering, Ewha Womans University, Seoul, , Korea Short title: transport of carbonaceous aerosols For submission to Journal of Geophysical Research

2 Abstract The present work is an attempt to improve the performance of a regional air quality model by means of linking it with a global chemistry transport model in order to provide initial and lateral boundary conditions. The current Community Multiscale Air Quality (CMAQ) model uses a set of constant lateral background condition profiles of the pollutant species, without reflecting temporal and spatial variations at the boundaries. A modeling study of a severe biomass burning event during May 1998 in Mexico and Central America, attributable to an ENSO-related drought, is presented. In this study, the GEOS-Chem global model output is incorporated as the CMAQ lateral boundary and initial values through an interpolation reconciling the differences in the vertical and horizontal coordinates and the chemical species representations of the two models. Simulated daily and monthly mean aerosol concentrations were evaluated by comparing these predicted concentrations with observational data from the Interagency Monitoring of Protected Visual Environments (IMPROVE) surface network. CMAQ, using the GEOS-Chem output to provide the lateral boundary conditions, improves model simulation of carbonaceous aerosols, such as the elemental carbons (EC) and organic carbons (OC). The correlation coefficients between overall simulated versus observed monthly mean correlation concentrations of EC was 0.60 and OC was 0.85, demonstrating successful simulations of trans-boundary transport of aerosols. Analyses of the model sensitivity simulations to assess carbonaceous emissions using the US EPA s National Emissions Inventory for 1999 (NEI99) and to explore potential sources of carbonaceous aerosols in the United States suggest a low bias of EC if NEI99 was used without the biomass fire emissions. Mass budget estimations of EC and OC show Tg/month (EC) and Tg/month (OC) of emissions and show Tg/month (EC) and Tg/month (OC) of depositions, respectively. The inflow fluxes of carbonacious aerosols through boundaries are well balanced by the outflow fluxes. Visibility degradation due to the carbonacious aerosols from the Mexican fire resulted in

3 an increase of the aerosol extinction coefficient by 40 % over the background level in the Southern United States during the May 1998 episode.

4 1. Introduction The current CMAQ model uses a set of pre-defined fixed lateral background condition profiles which do not properly account for the trans-boundary transport influences in cases where the in-flux mass of pollutants is non-negligible. Significant progress has been made in linking global circulation models to regional atmospheric models [Busuioc et al.,1999; Cubasch et al., 1996; Risbey et al., 1996; von Storch et al., 1993; Wigley et al.,1990], but little effort has been made to address linkage issues between global chemical transport models and regional air quality models. Atmospheric aerosols can be transported over long distances and can impact the environment in significant ways, namely, visibility degradation, effects on the biosphere and human health, and indirect climate forcing [e.g., Novakov and Penner, 1993]. Among the various chemical components of aerosols, elemental carbon (EC) and organic carbon (OC) are especially important. The presence of EC is notable because it results in positive radiative forcing, due to its light absorbing property, and organic carbon (OC) is of interest because of its abundance and detrimental effects on human health in the United States [Malm et al., 2000]. Major sources of carbonaceous aerosols in the atmosphere include combustion of fossil fuel, biofuel and biomass burning [Penner et al., 1992], biogenic emissions [Mazurek et al., 1991], and partial oxidation of hydrocarbon precursors [Hildemann et al., 1993; Turpin and Huntzicker, 1991]. Biomass burning, in particular, is a significant global source of atmospheric aerosols, and therefore plays a major role in the biogeochemical cycling of carbon compounds. In 1998, from April to June, severe biomass burning events occurred in central and South America as a result of enhanced drought conditions [Bell et al., 1999]. A characteristic anomalous northward flow, which was the strongest recorded in the preceding two decades [Rogers and Bowman, 2001], provided favorable conditions for the transport of the fire plumes.

5 The presence of large amount of aerosols, emanating from the biomass burning area, was verified by the Atmospheric Radiation Measurement (ARM) program in Oklahoma and Kansas [Peppler et al., 2000]. A previous study by Park et al. [2003] showed a considerable enhancement in the carbonaceous species concentrations, which has an important implication for the U.S visibility plan under the U.S EPA regional haze rule [EPA, 2003]. This paper presents a case study of the trans-boundary transport of carbonaceous aerosols emitted from biomass burning in Mexico and Central America, for the period of May 1998, using the US EPA s Community Multi-Scale Air Quality Model (CMAQ) [Byun and Ching, 1999; Byun and Schere, 2006]. This study focuses on a quantitative estimation of carbonaceous aerosol transport budgets and on an assessment of the episodic impacts of the Mexican forest fire on the visual environments in the United States. The initial and boundary conditions for the CMAQ simulation are defined by incorporating the global chemical transport model, GEOS-Chem (version 4.23, with CMAQ through a dynamically consistent linking algorithm. 2. Model description The US EPA CMAQ is a community model for multi-scale regional air quality that has been widely used by international communities since first its release in 1998 [Byun and Ching, 1999; Byun and Schere, 2006]. The CMAQ version 4.3 was selected for the current simulation. Currently, CMAQ relies on the Penn State-NCAR fifth generation mesoscale model (MM5) [Grell et al. 1994] to provide the meteorological fields. MM5 was configured with a single domain using 36 km 36 km horizontal grid spacing, which covers the continental United

6 States, including the northern part of Mexico and southern part of Canada. In the vertical direction, 43 unevenly spaced terrain-following sigma levels were employed with grid spacing stretching from approximately 30 m above surface to 4000 m at the model top near 18 km (50 mb). The physical parameterizations used in the simulation include the cloud radiation scheme [Dudhia 1989], the mixed phase cloud microphysics scheme [Reisner et al. 1998], the Kain- Fritsch cumulus parameterization [Kain and Fritsch 1990], the Medium-Range Forecast (MRF) model boundary layer parameterization [Hong and Pan, 1996], and the simple five-layer soil model. The MM5 was assimilated every three hours using the National Centers for Environmental Prediction (NCEP) s operational Eta Data Assimilation System (EDAS). CMAQ was configured with the same 36 km horizontal resolution as that of the MM5 but the domain of CMAQ is slightly smaller than the MM5 domain. Vertical collapsing in the preprocessor (Meteorology Chemistry Interface Program, MCIP) reduced the 43 MM5 layers to 23 CMAQ layers. Among the several physical process options available in CMAQ, the following options were selected for the current study: the carbon bond 4 (CB4) chemical mechanism, the AERO3 aerosol process, the Eulerian backward iterative (EBI) chemical solver, and the parabolic piecewise (PPM) advection scheme. The 1999 US EPA National Emission Inventory (NEI99) was used to provide anthropogenic emission sources for the EC/OC simulation. The emissions inventories were processed using the Sparse Matrix Operator Kernel Emission (SMOKE) [Coats, 1996; Houyoux et al., 2000] version 1.4 for providing emission input compatible with CMAQ [Byun et al., 2003]. However, NEI99 does not include emissions from biomass burning.. The absence of the fire emissions led us to adopt a global scale dry mass burnt amount, as a monthly total compiled at 1º 1º resolution [Duncan, 2002] and re-gridded for CMAQ domain configuration, subject to a mass-conserving constraint after mapping. We approximated the emission factors, of 2 and 14

7 g/kg for EC and OC [Park et al., 2003] respectively, and the monthly total biomass burning emission was distributed for the entire modeling period with a uniform intensity because of the lack of timing information with the assumption of rigorous vertical mixing from ground level up to the boundary layer heights. Monthly total domain integrated EC and OC emissions were 93 Gg/month (NEI99: 34 Gg/month, Biomass burning: 59 Gg/month) and 569 Gg/month (NEI99: 154 Gg/month, Biomass burning: 415 Gg/month) respectively. NEI99 exhibited a general spatial deviation with high emissions rates in the Eastern United States and in California while the biomass burning emissions mostly occurred in Mexico and some areas of Texas, Louisiana, and Florida in the continental US during the May 1998 episode. We carried out three sets of CMAQ simulations as sensitivity tests by switching the boundary conditions and emissions inputs. The first case was a simulation with the NEI99 emissions input with a set of fixed profile boundary conditions, which is the current default CMAQ boundary treatment (NEI99_Profile). The second case used combined biomass burning and NEI99 emissions with the profile boundary conditions (NEI99+Biomass_Profile). The third case used the dynamic boundary conditions provided from the GEOS-Chem linkage (NEI99+Biomass_Linkage) with the same emission inputs as the second run. By comparing the first case (NEI99_Profile) with the second (NEI99+Biomass_Profile), we assessed the contribution of the additional biomass burning emissions to the enhancement of EC/OC concentrations over the case of NEI99 alone. By subtracting the results of the second run (NEI99+Biomass_Profile) from that of the third run (NEI99+Biomass_Linkage), we show the benefits of linking the global chemistry transport model outputs with the regional CMAQ simulations.

8 3. Results and Discussion 3.1. Evaluation of carbonaceous aerosols simulations The simulated monthly mean spatial distributions of EC and OC concentrations in the bottom layer (sigma value=0.998) from the three sets of simulations and the observed monthly mean values from the IMPROVE network are shown in Figure.1. The simulation with NEI99 and fixed boundary values (NEI99_Profile) reproduces an air quality pattern showing high concentration areas located in several eastern states and the west coast states (California, Oregon, and Washington), reflecting the influence of the high anthropogenic emissions inputs. The simulation, however, fails to track carbonaceous aerosol concentrations in several southern states (Arizona, Texas, Arkansas, Georgia, and Tennessee) due to the lack of biomass burning emissions in NEI99. EC/OC concentrations from the simulation with NEI99 plus biomass burning emission but without the GEOS-Chem boundary conditions (NEI99+Biomass_Profile), shows higher concentration areas in Mexico in response to the additional emissions. The biomass burning emissions in the Mexican region (included in modeling domain) also causes enhancement of EC and OC in the southern United States. But the areas demonstrating such effects are not as broad compared to the case generated by the NEI99+Biomass_Linkage run, which shows EC and OC spatial concentration patterns with characteristic high concentration areas in Mexico and the southern and southeastern US, and clearly reflects the trans-boundary transport effects. The added biomass burning emissions in Northern Mexico, inside the CMAQ domain, increases the EC and OC concentrations up to 3 (EC) and 20 µg m -3 (OC) in the source regions and the effect spreads out, with gradually decreasing concentrations, over the southern part of the United States. Although the large-scale features of EC/OC distributions for the transboundary transport and biomass burning emission cases are similar, they display certain

9 differences in the detailed spatial patterns. The core of the trans-boundary plume is located between o W and extends to the eastern United States with strong intensity, while the added biomass burning emission case shows weakened effects after passing through Texas. The western states neighboring Mexico (NM, AZ, CO) show EC/OC enhancements in the transboundary case, but there is little change from the biomass burning emissions in the northern Mexico. Scatter plots of the simulated monthly mean from the best CMAQ(NEI99+Biomass_Linkage) run versus observed EC and OC concentrations at the IMPROVE sites show the correlation coefficients of R 2 =0.37 for EC and R 2 =0.72 for OC, respectively (Figure 2). The slopes of the reduced major axis regression lines, however, reveal a low bias of EC (0.69) and a slight high bias of OC (1.10) concentrations. The low bias of EC suggests an emission uncertainty in the NEI99 data. Figure 2 also shows results from the NEI99_Profile and NEI99+Biomass_Profile runs to demonstrate the benefit of linking CMAQ with GEOS-Chem and the addition of biomass burning emissions. The NEI99_Profile run results show simulated EC and OC concentrations with the slopes of regression lines as 0.22 and 0.32, respectively. The EC concentration shows a more serious low bias than that of OC. The NEI99+Biomass_Profile run shows increased slopes of 0.42 for EC and 0.68 for OC, but they are still underestimating observed concentrations. It shows that combining the biomass burning emission inside the regional modeling domain and GEOS_Chem prescribed boundary conditions helps improve accuracy of the simulated aerosol concentrations considerably. The approach brings the regression line closer to one-to-one line and reduces errors existing in other cases. Statistical performances of the simulated monthly mean EC/OC concentrations are summarized in Table 1. We selected six IMPROVE sites for the comparisons of day-to-day concentrations in the

10 CMAQ simulation and observed EC/OC concentrations. They are: Canyonlands National Park (CANY1,Utah; 38.45ºN, ºW, 1799m), Chassahowitzka National Wildlife (CHAS1, Florida; 28.74ºN, 82.55ºW, 2 m), Chiricahua National Monument (CHIR1, Arizona; 32.00ºN, ºW, 1570m), Big Bend National Park (BIBE1, Texas; 29.30ºN, ºW, 1075m), Upper Buffalo Wilderness (UPBU1, Arkansas; 35.82ºN, 93.20ºW, 723m), and Great Smoky Mountains National Park (GRSM1, Tennessee; ºN, ºW, 815m) as marked in Figure 1. EC/OC observations displayed high concentrations on May 13 and 20 at CANY1, on May 20 at CHIR1, on May and 20 at BIBE1, on May at UPBU1, and on May GRSM1 sites. The CMAQ simulation with the NEI99+Biomass_Linkage condition traces the high concentrations at CHIR1 and CANY1 (May 20) well, demonstrating its ability to represent the long-range transport of smoke originated from Mexico (Figure 3). This could be achieved only when the GEOS-Chem boundary conditions were incorporated, demonstrating aerosols from out-domain southern Mexico emission. The concentration behaviors in CANY1 and CHIR1 from the CMAQ simulation with the NEI99+Biomass_Linkage and the GEOS-Chem model coincide very well even though the two models depict different grid resolutions and coordinate systems. However, the CMAQ with GEOS-Chem boundary values and biomass emissions case shows overestimation of EC/OC concentrations with abrupt increases at BIBE1 and UPBU1 while GEOS-Chem presents gradual temporal changes that are in general agreement with the observations. The large enhancement of EC/OC concentrations suggests possibilities of the overestimation of EC and OC boundary inflow fluxes and uncertainties in the added Mexico biomass burning emission obtained through the mapping of global 1º 1º resolution biomass burning emission inventory into 36 km x 36 km CMAQ grid domain. The emission intensities of EC/OC summed over the grid cells of CMAQ tend to be stronger than that of coarser grid

11 GEOS-Chem (2º 2.5º) although both of them use the same global inventory. CHAS1 and GRSM1 sites show generally low EC concentrations in both the CMAQ and GEOS-Chem simulations. OC concentration is not biased much at the CHAS1 sites but is underestimated at the GRSM1 site in both the CMAQ and GEOS-Chem simulations. Model performances at individual sites vary depending on the dominant factors affecting the carbonaceous aerosol concentrations, such as trans-boundary transport, biomass burning emissions, and local emissions Effects of biomass burning on visual environments of United States A useful index for quantifying visibility impairment by the presence of aerosol particles is either the deciview or the aerosol extinction coefficient developed by Pitchford and Malm [1994]. The deciview (dv) scale is defined as: β ext haziness (dv) = 10 ln 0.01km sp 1 1 1, where [ km ] = β [ km ] β. 1 β is the aerosol extinction coefficient that must be adjusted by 0.01[ ] ext sp km. The CMAQ aerosol module calculates visibility indices (β ext and deciview) using two different methods: an approximation of Mie theory and a mass reconstruction technique. Mass reconstructed extinction method, which is an empirical approach, is explained by Malm [1994] as following. β [1/ km] = f ( RH ) {[ ammonium sulfate] + [ ammonium nitrate]} sp [ organic mass] [ Light Absorbing Carbon] [ fine soil] [ coarse mass] deciv = 10 ln ( βsp / 0.01)

12 Figure 4 compares the spatial distributions of areas with an increased aerosol extinction coefficient (1/Km) due to the biomass burning (difference between NEI99+Biomass_Linkage run and NEI99_Profile run) for May 8-9 and 14-15, with the distribution of the aerosol index observed by Total Ozone Mapping Spectrometer (TOMS) satellite. Visibility degradation, resulting from trans-boundary carbonaceous aerosols, started from Northern Mexico and stretched towards the Gulf Coast states, encompassing eastern Texas. A high aerosol index area is shown over the Gulf of Mexico on May 8 in the TOMS data. However, the model simulation shows a shift of the maximum aerosol extinction coefficient area toward eastern Texas and Mexico. The simulated transport path shows good agreement with that of TOMS on May 9. On May 14-15, significant aerosol influx occurred again and resulted in a very high aerosol index in southern Texas and the Mexican region. The simulated evolution pattern of the aerosols extinction coefficient enhancement, due to the biomass burning event, shows in general reasonable agreement with the satellite observation, but the simulation did not capture the details in the location and timing of the maximum. We conducted statistical evaluation of the wind simulation of MM5 to examine a potential relationship between the discrepancy in the spatial distribution of aerosol patterns and wind errors in the meteorological simulation. The zonal and meridional wind components at 950, 850, 700 and 500 mb pressure levels, representing the lower and middle tropospheric atmosphere, were compared with 70 routinely available radiosonde sounding observations over the continental United States. Simulated winds showed correlation with observation in the range of 0.58 ~ However, zonal winds were underestimated and meridional winds were overestimated at the levels above 850 mb (Table 2). Such errors caused further northward and less eastward shift of the fire plume transport and contributed to the discrepancy of spatial pattern of the simulated aerosols as shown in Figure 4.

13 The TOMS aerosol index and aerosol extinction coefficient, simulated by three CMAQ runs over the southern US (20-35 N and W) for May 8-9, shows an increase of the extinction coefficient due to the Mexican fire impact and a positive correlation tendency between them (Figure 5). Aerosol extinction coefficients are in the range of (1/Km) where TOMS aerosol index is greater than 3.0. The mean aerosol extinction coefficients by the three CMAQ simulations are 0.036, 0.041, and 0.050, respectively. On the same days, ground level carbonaceous aerosol concentrations are 0.43, 1.88, and 6.61 in the southern US. The increase of aerosol extinction coefficient is about 40% (0.050/ ) in the southern States but carbonaceous aerosol concentration is increased by 1440 % (6.61/0.43-1). The visual range affected by the carbonaceous aerosol is less than the concentration enhancement itself because carbonaceous aerosols do not contribute to the visibility change as much as hygroscopic aerosols such as sulfate and nitrate Mass budget of carbonaceous aerosols Table 3-1 shows the carbon aerosol mass fluxes for the NEI+Biomass_Linkage run, which are estimated by summing up the multiplication of the normal wind component and aerosol concentrations over the four boundaries of the domain. The mass in-fluxes are dominant at the southern (EC: Gg/month, OC: Gg/month) and the western (EC: 43.8 Gg/month, OC: Gg/month) boundaries. Although the concentrations at the western boundary are much lower than those of the southern boundary, persistent and strong westerly flows contribute to the large mass in-fluxes at the western boundary. The in-flux through the southern boundary is mostly compensated by the out-flux through the eastern boundary (EC: 92.2 Gg/month, OC: Gg/month) due to the prevailing wind pattern of this event. The sum of the in-fluxes, through the four sides, balances with that of out-fluxes (i.e., zero net flux), demonstrating that

14 the long range mass transport processes are simulated reliably. Analysis of mass budget by emission, deposition, and boundary fluxes over the entire domain is given in Table 3-2. Monthly total emission amounts of EC and OC are 93 Gg/month and 569 Gg/month, respectively. 95 Gg/month of EC and 574 Gg/month of OC are removed by surface deposition, which are comparable to the emissions inputs. Net fluxes through the southern boundary (in fluxes out fluxes) are 50 Gg/month for EC and 322 Gg/month for OC (Table 3-1), which indicate that the net carbonaceous aerosol amounts across the southern boundary are equivalent to 53.7 % (EC) and 56.2 % (OC) of the domain total emissions amounts. The net mass amount in the entire domain (emission + final loading - deposition - initial loading) of EC is about 3% (2.8 Gg/month) of the emission amounts demonstrating a good balance of the simulated mass budget. However, OC shows positive net mass budget of 12 % (71.8 Gg/month) of the emissions, which can be interpreted as the chemical production of the secondary organic aerosols from the gas phase precursors. 4. Summary and Conclusion We simulated carbonaceous aerosols using the US EPA s CMAQ model for long-range transport for a severe biomass burning event in central and south America occurring in May of The present study demonstrates that the air quality in the United States was greatly affected by such an event. The Penn State-NCAR fifth generation mesoscale model (MM5) output, the US EPA National Emission Inventory (NEI) 99, and the 1º 1º global scale grid resolved dry mass burnt EC/OC emissions were used for the simulations. Downscaling of the GEOS-Chem global chemical transport model output provided initial and boundary values for CMAQ. The linkage approach brings the most improved results, demonstrating the benefit of using

15 the dynamic lateral boundary conditions from the global simulation. The correlation between the simulated monthly mean EC/OC values and those observed at the IMPROVE network sites were 0.60 and 0.85 for EC and OC, respectively. The slopes of the reduced major axis regression lines, however, revealed a low bias of EC (0.69) and a slight high bias of OC (1.10) concentrations. The benefit of the linkage approach is most evident in the southern United States. The carbonaceous aerosols simulated by linking CMAQ with the GEOS-Chem output showed a reasonable qualitative agreement of the general spatial distribution patterns with the satellite observations. The mass budget analysis of the carbonaceous aerosols over the whole modeling domain showed that 93 Gg/month of EC and 569 Gg/month of OC were from emissions, while 95 Gg/month, 574 Gg/month were lost through the deposition of EC and OC, respectively. The net mass amount (emission + final loading - deposition - initial loading) of EC was about 3% (2.8 Gg/month) of the emission amounts, showing a reasonably balanced mass budget.. However, OC showed positive net mass budget, in the amount of 12 % (71.8 Gg/month) of the emissions, that can be interpreted as the contribution from the chemical production of the secondary organic aerosols from the gas phase precursors. The mass fluxes through the boundaries were well balanced between the in-fluxes and out-fluxes. Visibility degradation, due to the enhanced carbonaceous aerosols from the Mexican fire, was shown to increase the aerosol extinction coefficient by 40 % as compared to the background level in the southern United States. However, the biomass burning emissions compiled on a coarse resolution grid leads to a general overestimation of the EC and OC concentrations. This highlights the need to compile EC/OC biomass burning emission inventories, resolved at the fine regional scale resolution, for a more realistic description of their behaviors. The present study did not resolve the causes of the large high bias of the OC concentrations and, therefore, further investigations will be

16 necessary to identify factors that induce such large bias and to improve model performance under active fire conditions. Acknowledgements This research is supported by the United States Environmental Protection Agency through Grant R to the University of Houston and the Joint Venture Program of the University of Houston with the United States Department of Agriculture, FS-NC

17 References Bell, G. D., M. S. Halpert, C. F. Ropelewski, V. E. Kousky, A. V. Douglas, R. C. Schnell, and M. E. Gelman, Climate assessment for 1998, Bull. Am. Meteorol. Soc., 80, S1-S48, Busuioc A., H. von Stroch, R. Schnur, Verification of GCM generated regional seasonal precipitation for current climate and of statistical downscaling estimated under changing climate conditions, J. Climate, 12, , Byun, D.W. and J.K.S. Ching, ed., Science Algorithms of the EPA Models-3 Community Multiscale Air Quality (CMAQ) Modeling System, EPA Report, EPA/600/R-99/030, NERL, Research Triangle Park, NC, Byun, D.W., and K. L. Schere, 2006: Review of the Governing Equations, Computational Algorithms, and Other Components of the Models-3 Community Multiscale Air Quality (CMAQ) Modeling System. Applied Mechanics Reviews, Volume 59, Number , Byun, D.W., S.T. Kim, F.Y. Cheng, S. B. Kim, A. Cuclis, N. K. Moon, Information infrastructure for air quality modeling and analysis : Application to the Houson-Galveston Ozone non-attainment area, J. Environ. Informa., 2(2), 38-57, Coats, C.J.J. and Houyoux, M.R., Fast emissions modeling with the Sparse Matrix Operator Kernel Emissions Modeling System, in the Emissions Inventory: Key to Planning, Permits, Compliance, and Reporting, Air and Waste Management Association, New Orleans, LA, USA, Cubasch U., H. von Storch, J. Waskewitz, E. Zorita, Estimate of climate change in Southern Europe using different downscaling techniques, Climate Research, 7, , Duncan, B. N., R. V. Martin, A. C. Staudt, R. Yevich, J.A. Logan, Interannual and seasonal variability of biomass burning emission constrained by satellite observations, J. Geophys. Res., /2001JD000501, Hong, S.-H., and H.-L., Pan, Nonlocal boundary vertical diffusion in a medium-range forecast model. Mon. Wea. Rev., 124, , 1996.

18 Houyoux, M. R., Vukovich, J. M., Coats, C.J.J., Wheeler, N.W. and Kasibhatla, P.S., Emission inventory development and processing for the seasonal model for regional air quality (SMRAQ) project., J. Geophys. Res., 105(D7), , Hsu, N. C., J. R. Herman, P. K. Bhartia, C. J. Seftor, O. Torres, A. M. Thompson, J. F. Gleason, T. F. Eck, and B. N. Holben, Detection of biomass burning smoke from TOMS measurements, Geophys. Res. Lett., 23, , Malm, W. C., J. F. Sisler, D. Huffmans, R. A. Eldred, T.A. Cahill, Spatial and seasonal trends in particle concentration ad optical extinction in the United States. J. Geophys. Res., 99, , Malm, W. C., M. L. Pitchford, M. Scruggs, J. F. Sisler, R. Ames, S. Copeland, K. A. Gebhart, and D. E. Day, Spatial and Seasonal Patterns and Temporal Variability of Haze and Its Constituent in the United States: Report III, Cooperative Institute for Research, Colorado State University, Fort Collins, Colorado, Mebust, M. R., B. K. Eder, F. S. Binkowski, S. J. Roselle, Models3 Community Multiscale Air Quality (CMAQ) model aerosol component. 2. Model evaluation, J. Geophys. Res., 108(D6), doi: /2001jd001410, Penner, J.E., R.E. Dickinson, and C.A. O'Neill, Effects of aerosol from biomass burning on the global radiation budget, Science, 256, 1432, 1992b. Peppler R. A., et al., ARM southern Great Plains site observations of the smoke pall associated with the 1998 Central American fires, Bull. Am. Meteorol. Soc., 81, , Park, R. J., D. J. Jacob, M. Chin and R. V. Martin, Sources of carbonaceous aerosols over the United States and implications for natural visibility, J. Geophys. Res., 108(D12), 4355, doi: /2002JD003190, Risbey, J.S., P.H. Stone, A case study of the adequacy of GCM simulations for input to regional climate change assessments, J. Climate, 9, , Rogers, C. M., K. P. Bowman, Transport of smoke from the Central American fires of 1998, J. Geophys. Res., 106(D22), , 2001.

19 von Storch H., E. Zorita, U. Cubasch, Downscaling of global climate change estimates to regional scales: An application to Iberian rainfall in wintertime, J. Climate, 6, , Wigley T.M.L, P. D. Jones, K. R. Briffa, G. Smith, Obtaining subgrid scale information from coarse-resolution general circulation model output, J. Geophys. Res., 95, , 1990.

20 Table 1. Statistical performances of simulated monthly mean EC and OC concentration Run OBS Mean OBS STD 1) SML Mean SML STD 2) Regression Bias ERR Slope R 2 RMSE 3) SDE 4) NEI99_Profile EC NEI99+Biomass_Profile NEI99+Biomass_Linkage NEI99_Profile OC NEI99+Biomass_Profile NEI99+Biomass_Linkage ) STD : Standard deviation ERR : Error 3) RMSE : Root Mean Square Error 4) SDE : Standard Deviation Error Table 2. MM5 simulated wind error statistics at four pressure level on every 00UTC for May 1998 by comparing radiosonde sounding over the continental United States. Pressure Wind Observed Simulated Regression Mean Std Mean Std Slope R BIAS ERR RMSE SDE 950 mb U V mb U V mb U V mb U V

21 Table 3-1. Carbonaceous aerosol mass fluxes through four boundaries of the domain for the simulation period. (unit: Gg/month) East West South North In Out In Out In Out In Out Net* EC ~0 OC ~0 Net*: Total in fluxes - total out fluxes Table 3-2. Carbonaceous aerosol mass budget for simulation period (May1998) over the whole domain (unit: Gg/month) Emission Deposition Burden Initial Final Net* EC (Dry :5.7, Wet :89.5) OC (Dry :35.2, Wet: 539) Net*: (Emission -deposition)+(final burden-initial burden)

22 Figures Figure 1. Simulated monthly mean spatial distribution of EC (left) and OC (right) concentrations in the lowest layer of model (sigma value=0.998) for three CMAQ run scenarios. The top panels show results of NEI99 emission with profile background boundary conditions (NEI99_Profile). The middle panels are results of NEI99 and biomass burning emission with profile background boundary conditions (NEI99+Biomass_Profile). The bottom ones are results of NEI99 and biomass burning emission with GEOS-Chem linking boundary conditions (NEI99+Biomass_Linkage). Six selected IMPROVE sites (CANY1, CHAS1, CHIR1, BIBE1, UPBU1, and GRSM1) are marked on the upper left panel. Figure 2. Scatter plots of EC (left) and OC (right) monthly mean simulated versus observed values at IMPROVE sites for three CMAQ runs which are set in order same as Figure 1. Thick solid lines show the reduced major axis linear regression. The regression equation and correlation coefficients are also indicated. Figure 3. Comparisons of day-to-day EC concentration at six IMPROVE sites. Thin lines are results of NEI99_Profile run, thick lines are results of NEI99+Biomass_Profile run, and thicker lines are results of NEI99+Biomass_linkage run. Dashed line is result of GEOS- Chem simulation and dots show observed concentration at the IMPROVE sites. Figure 4. The spatial pattern of aerosol extinction coefficient (1/Km) revealed by NEI99+Biomass_Linkage run (left) and aerosol index (right) by Total Ozone Mapping Spectrometer (TOMS) satellite on 8-9 and May Figure 5. The correlation distribution between aerosol extinction coefficients for the three runs of CMAQ and TOMS aerosol index selected on 8-9 May 1998 over the southern United States.