Characterization of speciated aerosol direct radiative forcing over California

Authors


Corresponding author: Chun Zhao,Atmospheric Science andGlobalChangeDivision, Pacific Northwest National Laboratory, Richland,WA 99354, USA. (chun.zhao@pnnl.gov)

Abstract

[1] The WRF-Chem model, with the added capability of diagnosing the direct radiative forcing of individual aerosol species, is used to characterize the spatial and seasonal distribution of speciated aerosol direct radiative forcing over California. Overall, the simulation in 2005 is able to reproduce the observed spatial and seasonal distribution of total PM2.5 mass concentration and the relative contribution from individual aerosol species. On statewide average over California, all aerosol species reduce the surface net radiation fluxes, with a total by about 1.5 W m−2 (winter minimum) to 3 W m−2 (summer maximum). Elemental carbon (EC) is the largest contributor in summer (−1.1 W m−2 and ~35%), and sulfate is the largest in winter (−0.45 W m−2 and ~30%). In the atmosphere, total aerosol introduces a warming effect of about 0.5 W m−2 (winter minimum) to 2 W m−2 (summer maximum). EC and dust contribute about 75 − 95% and 1 − 10% of the total warming through the seasons, respectively. At the top of the atmosphere (TOA), the overall total aerosol direct radiative effect is cooling of −1.0 W m−2 through the seasons, with sulfate as the biggest contributor of −0.4 W m−2 (winter minimum) to −0.7 W m−2 (summer maximum). EC produces a TOA warming of up to about 0.7 W m−2, whereas all other aerosol species produce a TOA cooling. The diagnostic method implemented in WRF-Chem can be applied to other regions to understand the roles of different aerosols in the direct radiative forcing and regional climate.

1 Introduction

[2] Atmospheric aerosol particles, from both anthropogenic and natural sources, can interact directly with solar radiation and to a lesser extent with terrestrial radiation and hence substantially alter the energy budget of the atmosphere and the Earth's surface, i.e., the aerosol direct radiative effect. Variations in the radiation budget can lead to perturbations in the hydrological cycle [e.g., Ramanathan et al., 2001, 2005] and the photosynthesis rate of plants [e.g., Chameides et al., 1999]. Because of their relatively short atmospheric lifetime (1 − 2 weeks), aerosols offer a unique possibility for mitigating climate change [Jacobson, 2010]. Unlike greenhouse gasses that warm the Earth's surface and the lower and middle troposphere globally, the aerosol radiative effect is more heterogeneous and has greater impacts at a regional scale, because aerosol amounts and properties are highly variable in space and time. The impact of aerosols on regional climate has increasingly become a topic of intense research over the last decade [e.g., Giorgi et al., 2002; Gu et al., 2006; Novakov et al., 2008; Khan et al., 2010; Kim et al., 2012; Zhao et al., 2011, 2012].

[3] California is one of the most polluted regions in the world, with air quality that is likely affecting the wellbeing of millions of people. Particularly, ambient concentrations of small aerosol particles (i.e., PM2.5 and PM10, particulate matter with aerodynamic diameter less than 2.5 and 10 µm, respectively) have been found to correlate positively with human health [e.g., Pope and Dockery, 2006]. The California Air Resources Board (CARB) was established in 1967 to attain and maintain healthy air quality over California, and its air pollution control program has become one of the most effective in the world. In particular, increasingly stringent emission standards for diesel engines and reformulated low-sulfur diesel fuel have resulted in a decrease in black carbon and other particulate emissions. Several studies have shown that the concentration of black carbon over California has decreased by about 50% from 0.46 µg m−3 in 1989 to 0.24 µg m−3 in 2008, although the relative changes of other aerosol species (e.g., OC, SO42−, NO3) are relatively smaller [e.g., Kirchstetter et al., 2008; Novakov et al., 2008; Bahadur et al., 2011]. The decreasing aerosol concentrations can cause an increase in solar radiation intensity [e.g., Pinker et al., 2005; Wild et al., 2005], surface temperature [e.g., Wild et al., 2007; Novakov et al., 2008], and near-surface wind speeds [Jacobson and Kaufman, 2006]. In addition, the decrease in particulate emissions has the potential to weaken the aerosol effect of regional climate change mitigation over California [Novakov et al., 2008].

[4] To understand fully the potential impact of emission controls on aerosol forcing and the resulting effects on regional climate of California and to provide further information as guidance for future emission control strategies, understanding the seasonal variation and speciation of aerosols over California and their role in air quality and climate change is needed. Furthermore, accurate calculation of the aerosol direct radiative forcing is crucial for simulations of regional hydrological cycle and climate over California. However, quantification of aerosol − radiation interactions has proved difficult and is limited by uncertainties [Forster et al., 2007]. Aerosol direct radiative effects on climate are complicated because of variations in aerosol compositions and concentrations [e.g., Giorgi et al., 2002; Kim et al., 2007]. The main chemical constituents of atmospheric aerosols are inorganic species (such as sulfate, nitrate, ammonium, and sea salt), organic species, elemental carbon (EC), and mineral species (mostly windblown dust). The majority of EC, sulfate, and nitrate come from anthropogenic sources, whereas sea salt and most dust are of natural origin. Aerosol optical properties vary according to aerosol types. In addition, as a result of significant orography and seasonal contrasts in climate, the direct radiative impact of aerosols can vary greatly seasonally and geographically over California.

[5] The regional aerosol forcing in terms of its contribution from individual species, seasonal variation, and spatial distribution over California has not been fully characterized. Previous studies in which global models have more commonly been used do not speciate aerosol radiative forcing and lack the ability to resolve the large spatial variability of both meteorological conditions and emissions that influence aerosol spatial distributions and compositions. This study investigates the aerosol direct radiative forcing over California, including its variations by composition, geographically, and seasonally. A coupled meteorology − chemistry regional model (WRF-Chem) is used, as a first step, to investigate the impact of emission controls on the regional hydrological cycle and climate. This study also represents the first effort to evaluate WRF-Chem for simulating aerosols over California at a relatively high spatial resolution to reflect the region's complex topography and strong seasonal variation. We evaluate the model by comparing the simulations with observations of meteorological conditions from the California Irrigation Management Information System (CIMIS), aerosol mass from the rural Interagency Monitoring for Protected Visual Environments (IMPROVE) program and the urban US EPA Chemical Speciation Network (CSN), and aerosol optical depth from the ground-based Aerosol Robotic Network (AERONET) and the satellite-based Moderate Resolution Imaging Spectroradiometer (MODIS) and Multi-Angle Imaging Spectroradiometer (MISR) instruments. The WRF-Chem model is then used to estimate the seasonality and spatial distribution of major aerosol species in California using a new method for diagnosing the individual aerosol optical depth and direct radiative forcing associated with each species. This article is organized as follows. Sections 2 and 3 detail the WRF-Chem model and the observations used in this study. The seasonal variation and spatial distribution of speciated aerosols and their direct radiative forcing are characterized in section 4. The findings are summarized and discussed in section 5.

2 Model Description

2.1 WRF-Chem

[6] The WRF-Chem model [Grell et al., 2005] a version of the Weather Research and Forecasting (WRF) model [Skamarock et al., 2008] coupled with chemistry, is used. In this study, the WRF-Chem model is based on v3.2.1, but with updates including the GOCART dust emission coupled with MADE/SORGAM and MOSAIC [Zhao et al., 2010] and representation of aerosol direct and indirect radiative feedbacks in the RRTMG radiation and Morrison microphysics schemes [Zhao et al., 2011; Yang et al., 2011], respectively, which were released in v3.3. One of the chemistry options in WRF-Chem, the RADM2 (Regional Acid Deposition Model 2) photochemical mechanism [Stockwell et al., 1990] and the MADE/SORGAM (Modal Aerosol Dynamics Model for Europe [MADE] and Secondary Organic Aerosol Model [SORGAM]) [Ackermann et al., 1998; Schell et al., 2001], was selected. The MADE/SORGAM uses the modal approach with three log-normal modes (Aikten, accumulation and coarse mode) to represent aerosol size distribution. The standard deviation of each mode is prescribed, and the volume mean diameter of each mode is updated from the predicted aerosol mass and number concentrations in the mode during the simulation. Aerosols in each mode are assumed to be internally mixed with the same size distribution after being emitted. All major aerosol components are simulated in the model, including sulfate (SO42−), nitrate (NO3), ammonium (NH4+), EC, organic matter (OM), sea salt, and mineral dust. In this version, the MADE/SORGAM aerosol scheme includes physical and chemical processes of nucleation, condensation, coagulation, aqueous phase chemistry, water uptake by aerosols, and dry and wet deposition. Wet deposition by both resolved large-scale and parameterized cumulus clouds and precipitation is included. Wet deposition by resolved clouds and precipitation includes in-cloud removal (rainout) and below-cloud removal (washout) by impaction and interception, following Easter et al. [2004]. Although aqueous phase chemistry in cumulus clouds is not accounted for in the current version of WRF-Chem, it might not have a significant impact, because our results show that more than 90% of precipitation simulated by WRF-Chem comes from resolved large-scale precipitation at the 12 km resolution used (not shown). Aerosol nucleation follows Kulmala et al. [1998] and dominates the formation of secondary aerosol particles in the sulfuric acid − water system. Aerosol condensation, coagulation, and dry deposition are simulated following the approach of Binkowski and Shankar [1995].

[7] Aerosol optical properties such as extinction, single-scattering albedo, and the asymmetry factor for scattering are computed as a function of wavelength and three-dimensional position. As mentioned above, aerosols in this study are assumed to be internally mixed in each mode; i.e., a complex refractive index is calculated by volume averaging for each mode for each chemical constituent of aerosols. To compute the extinction efficiency (Qe) and the scattering efficiency (Qs) efficiently, WRF-Chem follows a methodology described by Ghan et al. [2001] that performs the full Mie calculations once first to obtain a table of seven sets of Chebyshev expansion coefficients, and later the full Mie calculations are skipped and Qe and Qs are calculated using bilinear interpolation over the Chebyshev coefficients stored in the table. A detailed description of the computation of aerosol optical properties in WRF-Chem has been given by Fast et al. [2006] and Barnard et al. [2010]. Aerosol radiative feedback was coupled with the rapid radiative transfer model (RRTMG) [Mlawer et al., 1997; Iacono et al., 2000] for both shortwave (SW) and longwave (LW) radiation as implemented by Zhao et al. [2011]. Aerosol − cloud interactions were included in the model by Gustafson et al. [2007] for calculating the activation and resuspension between dry aerosols and cloud droplets.

[8] Since aerosols in the model are assumed to be internally mixed, a methodology must be developed to diagnose the optical depth and direct radiative forcing of individual aerosol species. In this study, calculation of aerosol optical properties and radiative transfer is performed multiple times with the mass of one or more aerosol species (i.e., the mass of an individual or a group of aerosol species) and also its associated water aerosol mass removed from the calculation each time. After this diagnostic iteration procedure, the optical properties (e.g., aerosol optical depth [AOD]) and direct radiative forcing for an individual or a group of aerosol species are estimated by subtracting the optical properties and direct radiative forcing from the diagnostic iterations from those estimated following the standard procedure for all the aerosol species. That is,

display math
display math

For example, the optical depth and direct radiative forcing for EC are calculated by subtracting the values from the calculation with aerosol species other than EC from the values calculated with all aerosol species. In this study, we estimate the optical properties and direct radiative forcing for OM, EC, dust, sulfate, and all other aerosol species placed into a single group.

[9] The WRF-Chem simulation was performed on a domain at 12-km horizontal resolution covering California and its surrounding areas with 106 × 140 grid cells (112 − 126°W, 29 − 44°N). The simulation is configured with 35 vertical layers up to 100 hPa. The meteorological initial and lateral boundary conditions are derived from the North American Regional Reanalysis (NARR) data at 32 km horizontal resolution and 6 h temporal interval. The chemical initial and boundary conditions are from a WRF-Chem simulation with a larger domain at 36 km horizontal resolution covering the western United States (98 − 136°W, 24 − 53°N). This larger domain simulation is driven by NARR meteorological boundary conditions and chemical boundary conditions derived from the MOZART global simulation for the same time period. Thus the long-range transport of pollutants from Asia is accounted for in the simulations. The simulation is conducted from 15 December 2004 to 31 December 2005. Only the results for the year of 2005 (referred to as the simulation period hereafter) are analyzed to minimize the impact from initial condition. The Mellor-Yamada-Janjic (MYJ) planetary boundary layer scheme, Noah land surface scheme, Morrison 2-moment microphysics scheme, Grell cumulus scheme, and RRTMG LW and SW radiation scheme are used in this study. Aerosol direct and indirect radiative forcing is accounted for in the RRTMG radiation and Morrison 2-moment microphysics schemes.

2.2 Emission Inventory

[10] Anthropogenic and biogenic emissions over California used in this study are based on the 2008 inventory produced by CARB from the Arctic Research of the Composition of the Troposphere from Aircraft and Satellites—California Air Resources Board (ARCTAS-CARB) campaign. The ARCTAS-CARB inventory covers a roughly 2-week period from 14 to 26 June 2008 and includes day-specific anthropogenic (point, area, mobile, and shipping) and biogenic emissions, but not wildfire emissions. The emissions were available as a gridded inventory with a 4 km horizontal resolution and chemical speciation based on the SAPRC99 chemical mechanism. The SAPRC99 species were mapped to the WRF-Chem species, which include CO, NOx, SO2, NH3, volatile organic compounds (VOCs), sulfate, nitrate, EC, OM, and unspeciated PM2.5 and PM10 mass. Speciation profiles for some unknown PM2.5 and PM10 sources and their emissions are placed in an unspeciated category. To account for all emissions contributing to the total PM, WRF-Chem simulated two artificial tracer aerosol species, unspeciated PM2.5 and PM10, contributed by the unspeciated PM emissions. These two tracer species did not undergo chemical processing, and both were assumed to be nonabsorbing. The day-specific inventory was averaged to create a single daily-average set of hourly emissions, which was applied to the entire simulation period. As a result, daily and seasonal variability of anthropogenic and biogenic emissions is not accounted for in the simulations. The potential impact of this variability is discussed below. The original ARCTAS-CARB emissions at 4 km are interpolated to our 12 km grid. Anthropogenic emissions from geographical areas of the simulation domain not covered by the ARCTAS-CARB inventory are derived from the U.S. EPA NEI-2005 inventory.

[11] Figure S1 (in the Supporting Information) shows the spatial distribution of anthropogenic emissions of NO, CO, SO2, NH3, SO42−, NO3, EC, OM, and unspeciated PM2.5. It's obvious that the emission of unspeciated PM2.5 follows the spatial distribution of other speciated aerosols. Compared with observations (in section 4.2), the simulated EC surface concentration is significantly underestimated (by a factor of ~2), which will significantly affect the aerosol-induced atmospheric warming. Therefore, a sensitivity simulation with anthropogenic EC emission doubled was also conducted. In this simulation, the corresponding increase of EC emission is subtracted from the unspeciated PM2.5 emission to keep the total PM2.5 emission unchanged. The model negative bias of EC surface concentration is discussed in section 4.2.

[12] Biomass burning emissions in 2005 were obtained from the Global Fire Emissions Database, version 3 (GFEDv3), with monthly temporal resolution [Van der Werf et al., 2010]. Dust emission is calculated following the GOCART scheme [Ginoux et al., 2001], coupled with MADE/SORGAM within the WRF-Chem framework implemented by Zhao et al. [2010]. The size distribution of emitted dust and other details of the coupling of the GOCART dust emission scheme with WRF-Chem follows Zhao et al. [2010]. Sea salt emission in the publicly released WRF-Chem model is calculated following Gong et al. [1997]. In this study, the sea salt emission scheme is updated following Gong [2003] to include correction of particles with radius less than 0.2 µm and Jaeglé et al. [2011] to include the sea salt emission dependence on sea surface temperature.

3 Observations

3.1 Meteorology

[13] The California Irrigation Management Information System (CIMIS) is a program of the California Department of Water Resources (DWR) that manages a network of over 120 automated weather stations in California. Various weather data are collected, including solar radiation, air temperature, relative humidity, precipitation, and other surface variables. Further details on the CIMIS weather data, including data quality details and the links to download the data, can be found at http://wwwcimis.water.ca.gov/cimis/data.jsp. Modern-Era Retrospective Analysis for Research and Applications (MERRA) is the most recent reanalysis data produced by NASA's Global Modeling and Assimilation Office (GMAO) using the GEOS Data Assimilation System (DAS) [Rienecker et al., 2011]. The model has a finite-volume dynamic core that is run at a resolution of 1/2° latitude and 2/3° longitude with 72 vertical layers. Observational inputs to MERRA include data from surface land, ship, and buoy observations; rawinsondes; dropsondes; pilot balloons (PIBALs); wind profilers; aircraft; and various satellites [Rienecker et al., 2011]. The data can be downloaded through the MERRA website (http://gmao.gsfc.nasa.gov). Although the MERRA reanalysis is used to evaluate model simulations, we caution that the reliability of some estimates in reanalysis is not very high.

3.2 Aerosols

3.2.1 Surface Mass Concentration

[14] The Interagency Monitoring for Protected Visual Environments (IMPROVE) program was initiated in 1988 to monitor visibility and visibility-reducing particles in mostly remote and rural areas such as national parks and wilderness areas [Malm et al., 1994]. Currently, IMPROVE operates 170 sites in the United States where 24 h samples are collected every third day from midnight to midnight local time for speciated aerosol analyses. Details related to site location, sample collection, and analytical methodology have been given by Hand et al. [2011], and data can be downloaded from (http://views.cira.colostate.edu/fed). The U.S. EPA Chemical Speciation Network (CSN) collects samples with the same frequency and sampling schedule as the IMPROVE network and provides chemical composition data on fine particles (i.e., PM2.5) at hundreds of urban and suburban sites. Some details regarding the network (such as calibration standards, sampling methodology, and frequency) can be found at http://www.epa.gov/ttnamti1/speciepg.html and are given by Hand et al. [2011], and data can be downloaded from http://views.cira.colostate.edu/fed or http://www.epa.gov/ttn/airs/airsaqs/.

[15] In this study, the measured surface concentrations of fine particles (i.e., PM2.5) and their composition from rural IMPROVE sites (~40) and urban CSN sites (~100) in 2005 are used to evaluate model results. Most data from IMPROVE and CSN for comparison were directly downloaded, except for the CSN OM, sea salt, and dust and IMPROVE OM. Because both IMPROVE and CSN report only organic carbon (OC) measurements, in this study we multiply the OC data by 1.4 for converting measured OC to organic carbon mass (OM; to account for hydrogen, oxygen, etc.) [Chow et al., 2006]. The CSN sea salt mass is converted from the CSN measured sodium (Na+) following the formula MassNaCl = MassNa × 58.4/23. The CSN dust is calculated following the same formula that was employed by the IMPROVE [Malm et al., 1994]:

display math

where [Al], [Si], [Ca], [Fe], and [Ti] represent the mass concentration of aluminum, silicon, calcium, iron, and titanium, respectively.

3.2.2 Aerosol Optical Depth

[16] The Aerosol Robotic Network (AERONET) [Holben et al., 1998] has ~100 identical globally distributed sun- and sky-scanning ground-based automated radiometers, which provide measurements of aerosol optical properties throughout the world [Dubovik and King, 2000; Dubovik et al., 2002]. In this study, AERONET-measured AOD at 675 nm and 440 nm from four sites over California or the surrounding areas (Trinidad Head, 124.15°W, 41.05°N; Fresno, 119.77°W, 36.78°N; UCSB, 119.85°W, 34.42°N; Tonopah Airport, 117.09°W, 38.05°N) are used to derive the AOD at 550 nm (using the Angström exponent) for comparison with model results. All the AOD retrievals are quality level 2, and the uncertainty of AOD measurements is about ±0.01 [Holben et al., 2001].

[17] The MODIS instruments with wide spectral range, high spatial resolution, and near-daily global coverage onboard the NASA Aqua platform are designed to observe and monitor the changes in the Earth's surface, land, and atmosphere, including tropospheric aerosols [Kaufman et al., 1997]. The standard MODIS aerosol product does not retrieve aerosol information over bright surfaces (e.g., desert) because of a strong surface spectral contribution in the visible range [Kaufman et al., 1997]. However, a new algorithm, called the “Deep Blue” algorithm [Hsu et al., 2006], has been integrated with the existing MODIS algorithm to retrieve AOD even over bright surfaces. Therefore, the retrieved “Deep Blue” AOD from MODIS (collection 5.1; available only over land so far) is used in this study. The MODIS onboard the Aqua platform passes over the equator at ~13:30 local time (LT) during daytime [Kaufman et al., 1997]. In this study, available 550 nm AOD data for 2005 are used to evaluate the modeling results during the same period.

[18] Since February 2000, the MISR instrument on board the NASA Terra platform observes AOD continuously at nine distinct zenith angles, ranging from 70° backward to 70° forward, and in four narrow spectral bands centered at 446, 558, 672, and 866 nm. MISR's unique blend of directional and spectral data allows use of aerosol retrieval algorithms that do not depend on the explicit radiometric surface properties. Therefore, MISR can retrieve aerosol properties even over highly reflective surfaces (e.g., deserts) [Diner et al., 1998; Martonchik et al., 2004]. The MISR on board the Terra platform passes over the equator at ~10:45 LT during daytime [Diner et al., 2001]. In this study, available 558 nm AOD data for 2005 are used to evaluate the modeling results during the same period.

4 Results and Discussion

4.1 Meteorological Conditions Over California

[19] Meteorological conditions such as winds and precipitation play an important role in regional aerosol concentrations [e.g., Gao et al., 2011]. Before characterizing distributions of aerosols and their direct radiative forcing, it is necessary to evaluate and understand the WRF-Chem simulated meteorological conditions with available observations or reanalysis over California. The spatial distributions of seasonal mean of temperature (at 2 m), relative humidity (RH; at 2 m), and precipitation from CIMIS observations are compared with the corresponding results from the WRF-Chem simulation over California for 2005 (Supporting Information Figures S2, S3, and S4). Winds at 10 m from NARR reanalysis and WRF-Chem simulations are also shown in Supporting Information Figure S2. The NARR data show that the flow during winter is weaker than in other seasons. Over land, the winds are weaker and less homogeneous because of the complex topography. The 2 m temperature from CIMIS shows large seasonal variability, with a maximum in the summer and larger temperature contrast between coastal and inland areas during summer. The CIMIS observed RH shows clear seasonality and spatial heterogeneity. Influenced by the air mass and storms from the Pacific Ocean, RH has the highest value (~90%) during winter, but it reaches very low values (~10%) during summer, when California is under the influence of the subtropical high-pressure system. Northern California is distinctly moister than southern California, particularly during winter, and the Central Valley is drier than the mountainous regions and the nearby coastal areas. Furthermore, the simulated planetary boundary layer height (PBLH) is compared with the MERRA reanalysis (Supporting Information Figure S5), although we caution that the reliability of such estimates in reanalysis is not very high. The comparisons between observations/reanalysis and the WRF-Chem simulations are summarized in Supporting Information Table S1. In general, WRF-Chem reasonably simulates both the seasonal and the spatial distribution of the observed/reanalyzed meteorological fields with mean normalized biases (MNB) within ±20%. The simulation has negative biases for RH and positive biases for precipitation.

[20] A successful simulation of radiation is important to estimate aerosol direct radiative forcing. The CIMIS measurements of downward solar radiation are used for model evaluation. However, the absolute values of downward solar radiation from CIMIS could have significant systematic biases at some sites because of sensor problems (personal communication with the CIMIS engineers). Therefore, we compare the normalized radiation from CIMIS with the corresponding normalized reanalysis and model results. Normalization is performed by dividing the daily downward solar radiative fluxes with the maximum daily value of the year, so our comparison focuses on the seasonal and cloud-induced variation of downward solar radiation. Figure 1 shows the daily-normalized downward solar radiation from the CIMIS measurements, the MERRA reanalysis data, and the WRF-Chem simulation. All are averages over the CIMIS stations. The variation of all-sky downward solar radiation fluxes from the MERRA reanalysis and WRF-Chem simulation are consistent with the CIMIS measurements, with temporal correlation coefficients of 0.96 and 0.98, respectively. Comparing the WRF-Chem and MERRA clear-sky and all-sky downward solar radiation fluxes indicates that clouds are responsible for the high-frequency fluctuations of downward solar radiation fluxes, which are prominent during the cold (wet) season.

Figure 1.

Daily normalized downward solar radiation from CIMIS observations, MERRA reanalysis data, and the WRF-Chem simulations over California in 2005.

4.2 Spatial Distribution and Seasonality of Aerosols Over California

[21] Figure 2 shows the seasonal mean spatial distribution of total PM2.5 mass concentrations at the surface from the IMPROVE and U.S. EPA measurements overlaid on the corresponding WRF-Chem simulations for California in 2005. Overall, the EPA measurements have higher values than the IMPROVE measurements, because the EPA measurement sites are located in urban areas affected directly by anthropogenic emissions, whereas the IMPROVE measurement sites are located in remote areas. The measurements show high mass concentrations of total PM2.5 over the Central Valley and the Los Angeles metropolitan regions. In general, WRF-Chem captures the spatial distribution of observations through the seasons but shows more spatial variability than observations in part because of the limited number of stations and the spatial heterogeneity in the emission data, land surface forcing (e.g., topography), and atmospheric conditions. In addition to the Central Valley and Los Angeles metropolitan regions, the WRF-Chem simulations also show high PM2.5 concentrations over southeastern California, with the former two regions significantly affected by anthropogenic aerosols and the latter dominated by dust aerosols (Figure 3). A distinct seasonality is shown, with maximum surface mass concentrations in winter but lowest concentrations in summer. The seasonality of various aerosol components of PM2.5 is discussed in more detail below.

Figure 2.

Seasonal mean spatial distribution of total PM2.5 mass concentrations at the surface from the IMPROVE (circles) and U.S. EPA (triangles) measurements and the corresponding WRF-Chem simulations over California in 2005. The numbers represent the four AERONET sites: 1, Trinidad Head; 2, Fresno; 3, UCSB; 4, Tonopah Airport.

Figure 3.

Spatial distribution of seasonal mean speciated PM2.5 (sulfate, nitrate, ammonium, organic matter [OM], elemental carbon [EC], dust, sea salt, and unspeciated PM2.5) mass concentrations from the IMPROVE (circles) and U.S. EPA (triangles) measurements and the corresponding WRF-Chem simulations over California in 2005.

[22] Figure 3 shows the spatial distribution of seasonal mean speciated PM2.5 (sulfate, nitrate, ammonium, OM, EC, dust, sea salt, and unspeciated PM2.5) surface mass concentrations from the IMPROVE (Figure 3, circles) and EPA (Figure 3, triangles) measurements overlaid on the corresponding WRF-Chem simulations over California for 2005. In general, the anthropogenic aerosols (i.e., sulfate, nitrate, ammonium, OM, EC, and unspecieated PM2.5) follow a similar spatial distribution. It is evident that the high mass concentrations of total PM2.5 over the Central Valley and the Los Angeles metropolitan regions result from anthropogenic emissions, whereas the high values over southeastern California are due to natural dust. Overall, the model captures well the observed spatial distribution and seasonality of anthropogenic aerosols, with correlation coefficients of 0.7 − 0.9 in most cases. Both observations and simulation show that dust spreads out over large distances from the desert source regions, although the spatial gradient clearly indicates the highest dust concentration over the Sonoran Desert. In contrast, sea salt is confined to the coastal areas because of its short lifetime. It is noteworthy that the magnitudes of OM and EC surface mass concentrations are significantly underestimated. Dust and sea salt are overestimated near deserts and along the coast, respectively, although they are simulated well elsewhere in the domain. The comparisons between observations and WRF-Chem simulations of surface speciated aerosol concentrations are summarized in Supporting Information Table S2. A more quantitative comparison between observations and simulations and an analysis of the seasonality of aerosol concentrations are shown in Figure 4.

Figure 4.

Seasonal mean mass concentrations of speciated PM2.5 such as EC, OM, dust, sulfate, nitrate, ammonium, sea salt, and unspeciated PM2.5 from IMPROVE and U.S. EPA measurements and the corresponding WRF-Chem simulations over California in 2005. The bottom-right panel shows the results from WRF-Chem sensitivity simulation with anthropogenic EC emissions doubled.

[23] Figure 4 depicts the seasonal cycle of statewide averaged surface mass concentrations of speciated PM2.5 (sulfate, nitrate, ammonium, OM, EC, dust, sea salt, and unspeciated PM2.5) from the IMPROVE (rural) and EPA (urban) measurements and the corresponding WRF-Chem simulations over California for 2005. In general, the seasonality of aerosol surface concentrations is determined by seasonality of emissions (anthropogenic emissions are assumed constant throughout the year in this study), meteorological conditions (e.g., vertical turbulent mixing and ventilation), chemical production, and removal processes. Aerosol can either be emitted into the atmosphere or be generated by physical and chemical processes within the atmosphere (primary and secondary aerosols, respectively).

[24] One example of primary aerosols is EC in WRF-Chem. Although total emission of EC is not constant through the year because of biomass burning emissions having a summer maximum and a winter minimum (not shown), both observations and simulation show the seasonality of EC surface concentration in urban areas, with a winter maximum and a spring minimum. This likely is due to stronger vertical turbulent mixing in spring compared with winter, which is reflected in the seasonality of the PBLH, which is the highest in spring and lowest in winter (see Supporting Information Figure S6). The surface winds are also stronger in spring, indicating more efficient ventilation (Supporting Information Figure S2). It is evident that the seasonality of EC surface concentration in urban areas is determined mainly by the vertical turbulent mixing and ventilation. In rural areas, the seasonality of EC is not obvious. Although the boundary layer depth follows the same seasonality in the urban and rural areas (Supporting Information Figure S6), EC in rural areas is mainly transported into the region rather than being produced by local emissions. Therefore, vertical mixing and ventilation should have much less effect on the seasonality of aerosols in rural areas, hence the much smaller seasonal variation. Unspeciated PM2.5, which is another primary aerosol, has a similar seasonal variation in the model.

[25] It is noteworthy that the EC surface concentration is significantly underestimated in the model by a factor of ~2 in both urban and rural areas. The bias may be partially due to the inability of the model to resolve emissions and aerosol/meteorological processes reflected by the measurements at the local scale. Previous studies applying WRF-Chem over central Mexico have reported underestimation of aerosol surface concentrations and attributed the issue in part to the need to account for subgrid variability of aerosol in the simulations [Qian et al., 2010; Gustafson et al., 2011]. This motivated us to conduct a 4-km-resolution simulation to investigate the effect of model resolution on model biases. Our results, however, show that the EC surface concentration of the 4-km-resolution simulation does not change significantly from that of the 12-km-resolution simulation. Uncertainties and interannual variabilities of the anthropogenic EC emission might also contribute to the bias. Since the 2008 emission inventory for a 2 week period is used with meteorological boundary conditions of 2005 in this study, our simulation might not be able to capture the measured EC concentration in 2005. The sensitivity simulation (section 2) with anthropogenic EC emission doubled reproduces the observed EC surface concentration and its seasonal variation in both urban and rural areas very well (bottom right panel in Figure 4).

[26] In this study, OM includes both primary and secondary sources. In urban areas, the seasonality of observed OM surface concentration is similar to that of EC, indicating that anthropogenic primary emissions dominate. In rural areas, the seasonal variability with a summer maximum indicates that secondary production, which peaks in the summer because of more active photochemistry and higher emissions of biogenic VOCs, dominates. The higher biomass burning emission in rural area in summer may also contribute to this summer peak. The surface OM concentration is also significantly underestimated in the model by factors of about 4 and 2 in urban and rural areas, respectively. The underestimation of OM likely is due to uncertainty of emissions and the outdated SOA mechanism used in the current version of WRF-Chem [Shrivastava et al., 2011]. In addition, because seasonal variability is not accounted for in the ARCTAS-CARB emissions inventory, some categories such as residential wood burning, which may contribute to directly emitted OM in winter, are underestimated. As with EC, the result of a simulation with higher resolution (4 km) does not significantly reduce this negative bias. Furthermore, it is noteworthy that, if a factor of 1.8 from a recent study [Simon et al., 2011] instead of 1.4 from Chow et al. [2006] is applied to convert the measured OC to OM, the model negative bias of OM is even greater.

[27] Nitrate (NO3), ammonium (NH4+), and sulfate (SO42−) are important constituents of PM2.5, formed mainly from gas-phase precursors. In general, both observation and simulation show consistent magnitude and seasonality of sulfate, nitrate, and ammonium in urban and rural areas. Sulfate surface concentrations are lowest in winter, when photochemistry is least active, and highest in the summer, when the most active photochemistry occurs [e.g., Dawson et al., 2007; Pye et al., 2009]. This seasonality of sulfate may also be contributed by the seasonality of wet removal (much more precipitation in winter). Nitrate shows a seasonality that is opposite to that of sulfate, with a maximum nitrate surface concentration occurring in winter and a minimum in summer, which can be explained by the combined effects of temperature and vertical turbulent mixing. Increasing temperature (from winter to summer) can lead to a reduction of nitrate aerosol concentration as a result of nitrate − ammonium partitioning to the gas phase [Chow et al., 2006; Dawson et al., 2007]. In addition, the stronger vertical turbulent mixing and ventilation in summer can also result in lower nitrate surface concentration. Ammonium surface concentration shows a maximum in winter and a minimum in spring, reflecting the role of the vertical turbulent mixing and the increased ammonium nitrate production in winter [Chow et al., 2006]. The seasonal variation of ammonium is much smaller than that of nitrate or sulfate, which can be explained by the fact that ammonium is generally associated with nitrate and/or sulfate, which have opposing seasonal variations.

[28] Sea salt and dust are natural aerosols. Compared with anthropogenic aerosols, sea salt and dust are emitted mostly over the rural or remote areas instead of urban areas and their emissions are meteorology dependent. Stronger surface winds, particularly over the ocean, result in larger sea salt emission and hence higher surface sea salt concentration in spring and summer than in winter and fall. Sea salt is a factor of 2 higher in simulations than in observations. It's likely that the model overestimates sea salt emissions, but the measurements may also underestimate sea salt aerosol [Hand et al., 2011]. The observed seasonality of surface dust concentration is more distinct in rural areas, showing a summer maximum and a winter minimum. The simulation shows a similar but weaker seasonality (Figure 3) compared with observations. The dust surface concentration is overestimated by the model. Figure 3 shows that this high bias ccurs mainly for sites close to the source region. However, it is difficult to evaluate the model performance over the source region because of the absence of observations there. The bias near the source region may result from the coarse resolution (1°) prescribed source function used in the GOCART scheme in this study, which might not resolve the emission gradient close to the source region. Finer resolution of dust emission source function and more measurements over the source region may help to constrain the model simulation of dust over southeastern California.

[29] The total PM2.5 surface concentration determined by its components shows a spring minimum and a winter maximum in urban areas but a contrasting seasonality (summer maximum and winter minimum) in rural areas, indicating that the seasonality of total PM2.5 surface concentration is dominated by temperature and photochemistry in rural areas and by vertical turbulent mixing and ventilation in urban areas [e.g., Hand et al., 2012]. The comparison of total PM2.5 surface concentration between observation and simulation is encouraging. WRF-Chem captures well the seasonality and magnitude of total PM2.5 surface concentration, showing that anthropogenic aerosols contribute ~90% and ~70% of the total PM2.5 in terms of mass in urban and rural areas, whereas natural aerosols (dust and sea salt) make up the rest. It is notable that WRF-Chem significantly underestimates the OM and EC surface concentrations but reproduces the total PM2.5 surface concentration. The negative biases of OM and EC are compensated mainly by the unspeciated PM2.5, which is simulated by WRF-Chem to account for the unspeciated PM emission in the emission inventory. The simulated spatial distribution of unspeciated PM2.5 surface concentration correlates strongly with that of OM and EC. Our results suggest that some of the anthropogenic emissions of OM and EC are miscategorized as emission of unspeciated PM2.5 in the emission inventory used in this study (Supporting Information Figure S1) or that the categorization for 2005 could have been different from that applied to the 2008 emission inventory. This motivated our sensitivity experiment in which anthropogenic emission of EC is doubled, but the unspeciated PM2.5 is correspondingly reduced to maintain the same total PM2.5 emission. Again, the mismatch of using the 2008 emission inventory for the 2005 simulation is another possible reason for the negative biases of simulated EC and OM.

4.3 AOD and Absorption AOD Over California

[30] Figure 5 shows the spatial distributions of seasonal mean 550 nm AOD and absorption AOD (AAOD) from the WRF-Chem simulation with anthropogenic EC emissions doubled over California for 2005. In general, the AOD is largely dependent on the aerosol column burden (vertically integrated aerosol mass) and RH. The high AOD simulated over the Central Valley, the Los Angeles metropolitan regions, and the southwestern U.S. deserts is consistent with the spatial distribution of total PM2.5 surface concentration and column mass burden (Figure 2 and Supporting Information Figure S7). Figure 6 shows the hourly 550 nm AOD from four available AERONET sites (Trinidad Head, Fresno, UCSB, and Tonopah Airport, shown as numbers 1 − 4 in Figure 5) and the corresponding MODIS and MISR retrievals and WRF-Chem simulations over California for 2005. The Trinidad Head and UCSB sites are located in coastal areas, the Fresno site is located in central California, and the Tonopah Airport site is located in the remote area of Nevada. The AERONET measurements show the highest annual AOD at Fresno, which is more polluted than other sites. Tonopah has the lowest annual AOD and the smallest temporal variation. In general, the MISR retrieval is consistent with the AERONET measurements, although it occasionally produces unreasonably high values. Compared with the AERONET measurements, the MODIS retrieval exhibits systematic high bias. For the Trinidad Head and UCSB sites, the model generally captures the annual magnitude of the AERONET measurements (0.06 and 0.04 from model vs. 0.08 and 0.07 from AERONET, respectively) and captures some episodes. The observed AOD variations are not fully captured by the model, which likely is due to emission uncertainties and model resolution. For the Fresno site, the model captures the general variation of AOD but has a negative bias (0.04 from model vs. 0.10 from AERONET), which may be due to a negative bias in the emissions. The model better captures the AOD at the Tonopath Airport site, which is the cleanest site among the four AERONET sites (0.03 from model vs. 0.04 from AERONET). The MODIS retrieval at this site is the most problematic, likely because of the bright surface. The model generally captures the seasonality of AERONET-observed AOD, such as that at the Tonopath Airport site; AOD is higher in summer (0.05 from both model and AERONET) than in fall (0.02 from model vs. 0.03 from AERONET). In general, the model underestimates AOD compared with the AERONET observations at all the sites, which may be partially explained by the negative bias of RH (Supporting Information Table S1).

Figure 5.

Spatial distributions of seasonal mean 550 nm AOD and AAOD from the WRF-Chem simulations with anthropogenic EC emissions doubled over California in 2005. The numbers represent the four AERONET sites: 1, Trinidad Head; 2, Fresno; 3, UCSB; 4, Tonopah Airport.

Figure 6.

Hourly 550 nm AOD from four available AERONET sites and the corresponding MODIS and MISR retrievals and WRF-Chem simulations over California in 2005.

[31] The seasonal variation of AOD (Figure 5) is different from that of the total PM2.5 surface concentration (Figure 2). For example, over the Central Valley, aerosol surface mass is similar between spring and fall, but AOD is higher in spring than in fall. One reason is that AOD is correlated with the aerosol column burden rather than the surface concentration, which is largely affected by the vertical turbulent mixing. Although total PM2.5 surface mass concentration shows distinct seasonality, the column burden shows weaker seasonality over California except for the dust source regions (Supporting Information Figure S7). Seasonal variation of RH plays an important role, which can be reflected in the water aerosol mass concentration. RH is higher in winter and spring than in summer and fall (Supporting Information Figure S3), resulting in higher water aerosol column burden in winter and spring (Supporting Information Figure S8), consistent with the seasonality of the AOD spatial distribution (Figure 5). The spatial distribution and seasonality of AAOD follow closely those of AOD. In this study, only EC and dust have prescribed absorbing optical properties. Therefore, the AAOD is determined mainly by the distribution of EC and dust, although other species (e.g., sulfate) can enhance their absorption [Jacobson, 2001a]. From the spatial distribution of seasonal mean AAOD/AOD shown in Figure 5, AAOD is about 5 − 25 times lower than AOD over the Central Valley, where EC is the dominant absorbing aerosol, and 10 − 20 times lower than AOD over the southwestern U.S. deserts, where dust is the dominant absorbing aerosol. The seasonality of AAOD/AOD over the Central Valley is more distinct than that over the southwestern U.S. deserts, because the ratio of EC to other aerosols over the Central Valley has greater seasonal variation than the ratio of dust to other aerosols over the southwestern U.S. deserts [Hand et al., 2011].

[32] Figure 7 shows the seasonal statewide mean AOD and AAOD and their contributions from individual (OM, EC, dust, and sulfate) and lumped (other aerosol including nitrate, ammonium, sea salt, and unspeciated PM2.5) aerosol species from the WRF-Chem simulation with anthropogenic EC emissions doubled over California in 2005. In WRF-Chem, aerosols are assumed to be internally mixed. Therefore, differences between the sum of AOD and AAOD diagnosed for the individual aerosol species and that of the internally mixed aerosol reflect nonlinear interactions among the aerosol species. It is critical to examine the consistency between the diagnosed and the simulated total AOD and AAOD to assess the representativeness of the diagnosed AOD and AAOD for individual aerosol species. It is encouraging that the diagnosed values (sum of AOD and AAOD for individual and lumped aerosol species) are generally comparable to the simulated values for the internally mixed aerosols, which suggests that the diagnostic calculation can provide reasonable estimates of the contribution from individual or lumped aerosol species.

Figure 7.

Seasonal variations of total 550 nm AOD and AAOD and their contributions from sulfate, OM, EC, dust, and other species from the WRF-Chem simulations with anthropogenic EC emission doubled. Other species include nitrate, ammonium, sea salt, and unspeciated PM2.5.

[33] Although the sulfate surface concentration is less than the lumped aerosol species (Supporting Information Figure S9), sulfate AOD is larger than the AOD from the lumped aerosols (sum of nitrate, ammonium, sea salt, and unspeciated PM2.5 in Figure 4) because it has larger column mass burden resulting from aqueous-phase production in the free troposphere (Supporting Information Figure S9). The AOD for EC and OM is small, but the OM AOD may be biased by the model underestimation of OM concentration. The seasonality of AOD and AAOD from individual and lumped aerosol species is different. Dust AOD shows a summer maximum and a winter minimum, whereas sulfate and other aerosols (i.e., anthropogenic aerosols) show a winter/spring maximum and a fall minimum. The seasonality of total AOD, with a maximum (~0.06) in winter and a minimum (~0.04) in fall, is determined by that of anthropogenic aerosols. It is not surprising that EC is the main contributor to the AAOD. Dust follows as the second largest contributor. Although aerosol species other than EC and dust are assumed in the model to be nonabsorbing, they enhance the absorption by EC and dust and account for 15 − 20% of the state-wide averaged total AAOD. The seasonality of AAOD (from 0.0023 to 0.0038) is determined by EC, which shows a maximum in summer and a minimum in winter. This is different from the seasonality of EC surface mass concentration (Figure 4) but is consistent with that of the column mass burden of EC with a summer maximum resulting from biomass burning that peaks in the summer (Supporting Information Figure S9), indicating that biomass burning emission has greater impact on column mass burden than on surface concentrations. In this study, the total AAOD and OM AAOD may be underestimated because of the absorption of OM (i.e., brown carbon), which has been found to contribute up to 20% of warming in the atmosphere [e.g., Jacobson, 1999, 2001b; Chung et al., 2012; Kirchstetter and Thatcher, 2012], but such an effect is not accounted for in the model.

4.4 Aerosol Direct Radiative Forcing Over California

[34] Figure 8 shows the spatial distribution of seasonal mean aerosol direct radiative forcing at the top of atmosphere (TOA), in the atmosphere, and at the surface from the WRF-Chem simulation with anthropogenic EC emission doubled over California for 2005. The seasonality of TOA aerosol direct radiative forcing over California is relatively small. Aerosols result in a TOA cooling (i.e., negative forcing) over all of California, with a statewide annual average of −1.0 W m−2 and a statewide maximum of −3.5 W m−2. In the atmosphere, aerosols introduce a warming effect. The spatial distribution of atmospheric warming follows that of EC and dust and is consistent with that of AAOD. The maximum atmospheric warming reaches 10 W m−2 over the Central Valley and the Los Angeles metropolitan regions. The atmospheric warming has a distinct seasonality with a summer maximum of 2.0 W m−2 and a winter minimum of 0.5 W m−2 on domain average, determined by the seasonality of AAOD and solar radiation fluxes (peak in summer; Figure 1). Aerosols cool the surface over the whole of California by up to −10 W m−2. The aerosol surface cooling shows a winter minimum of −1.5 W m−2 and a summer maximum of −3.0 W m−2 on domain average.

Figure 8.

Spatial distributions of seasonal mean aerosol direct radiative forcing at the top of atmosphere (TOA), in the atmosphere (ATM), and at the surface (BOT) from the WRF-Chem simulations with anthropogenic EC emissions doubled over California in 2005. At TOA and BOT, positive value represents downward radiation; in ATM, positive value represents warming.

[35] Figure 9 shows the seasonal mean aerosol direct radiative forcing and its contribution from individual and lumped aerosol species at the TOA, in the atmosphere, and at the surface from the WRF-Chem simulation with anthropogenic EC emissions doubled over California for 2005. Nonlinear interactions can again be inferred from the comparison between the diagnosed (Figure 9, purple) and simulated (Figure 9, black) total aerosol forcing. Similar to the total AOD and AAOD, it is encouraging that the diagnosed total aerosol forcing is generally comparable to the simulated value. At the TOA, most aerosols introduce a negative radiative forcing, with the largest contribution coming from sulfate (seasonal variation from −0.4 W m−2 in winter to −0.7 W m−2 in summer) followed by dust (seasonal variation from −0.2 W m−2 in winter to −0.3 W m−2 in summer) and other aerosols (seasonal variation from −0.25 W m−2 in spring to −0.3 W m−2 in winter), except that EC leads to a positive forcing (seasonal variation from 0.2 W m−2 in winter to 0.7 W m−2 in summer). In the atmosphere, EC and dust contribute to the majority of total aerosol warming, with a seasonal range of about 75 − 95% and 1 − 10%, respectively. The rest of the aerosol warming is the enhancement of EC and dust warming by all other nonabsorbing aerosols through internal mixing. Again, the atmospheric warming may be underestimated, because OM concentration is underestimated and its absorption is not taken into accounted. At the surface, all aerosols have a cooling effect. In summer, EC is the largest contributor (−1.1 W m−2 and ~35%) to the total surface cooling, followed by sulfate (−0.8 W m−2 and ~25%), dust (−0.6 W m−2 and ~18%), other aerosols (−0.4 W m−2 and ~13%), and OM (−0.3 W m−2 and ~9%), whereas, in winter, sulfate is the largest contributor (−0.45 W m−2 and ~30%), followed by EC (−0.39 W m−2 and ~26%), other aerosols (−0.38 W m−2 and ~25%), dust (−0.2 W m−2 and ~13%), and OM (−0.1 W m−2 and ~6%). Again, the forcing from OM likely is underestimated.

Figure 9.

Seasonal variations of aerosol direct radiative forcing and its contributions from sulfate, OM, EC, dust, and other species at the TOA, in the atmosphere, and at the surface from the WRF-Chem simulations with anthropogenic EC emission doubled. Other species include nitrate, ammonium, sea salt, and unspeciated PM2.5. Black and purple bars represent the total forcing from the simulation and the sum of the diagnosed individual forcings.

5 Discussion and Conclusions

[36] A fully coupled meteorology − chemistry model (WRF-Chem) has been used to study the spatial and seasonal distribution of speciated aerosols and their direct radiative forcings over California for 2005. Model simulations are evaluated with various data sets, including the meteorological fields from CIMIS and MERRA, the surface aerosol mass concentrations from EPA and IMPROVE, and the AOD from AERONET and satellites. In general, the model captures well the observed seasonal and spatial distributions of meteorological fields. The simulations reproduce the observed spatial distributions of aerosol surface mass concentrations, showing high mass concentrations of total PM2.5 over the Central Valley and the Los Angeles metropolitan regions resulting from anthropogenic aerosols. In addition, the simulations also show high PM2.5 concentrations over southeastern California resulting from natural dust aerosol. The spatial gradients of natural aerosol mass concentrations are well simulated, and the impact of natural aerosols is confined to limited regions (coastal areas for sea salt and southeastern California for dust).

[37] The surface concentrations of primary aerosols (emitted directly into the atmosphere) such as EC and unspeciated PM2.5 show a consistent seasonality with a winter maximum and a spring minimum over the urban area, determined by the vertical turbulent mixing and ventilation. As secondary aerosols, sulfate and nitrate show a different seasonality in their surface concentrations. Sulfate has a summer maximum and a winter minimum because of the seasonality of photochemical activity, whereas nitrate shows a contrasting seasonality because of vertical turbulent mixing and the temperature dependence of the nitrate − ammonium gas − aerosol partitioning. OM includes both primary and secondary sources and therefore shows seasonality similar to that of primary aerosols in the urban areas, but the peak surface concentration in rural areas in summer is related to the active photochemistry in summer and the higher biomass burning emissions in summer. Natural aerosols, dust and sea salt, show a summer maximum and a winter minimum surface concentration resulting from meteorological conditions (stronger surface winds) that favor their emission in the summer and lower precipitation. The total PM2.5 surface concentration shows a minimum in spring and a maximum in winter in the urban areas and the contrasting seasonality in the rural areas.

[38] The model captures the general characteristics of 550 nm AOD observed at the AERONET sites. The spatial distribution of 550 nm AOD and AAOD follows that of aerosol mass concentrations. AOD shows a winter maximum and a summer minimum determined by the seasonal variation of RH, whereas AAOD shows a summer maximum and a winter minimum because of the higher column mass burden of EC and dust. The seasonality of AAOD/AOD over the Central Valley is more distinct than in the southwestern U.S. deserts, because the ratio of EC to other nonabsorbing aerosols over the Central Valley has greater seasonal variation than the ratio of dust to other nonabsorbing aerosols over the southwestern U.S. deserts. The seasonality of total AOD is dominated by that of anthropogenic aerosols (mainly composite of sulfate, nitrate, ammonium, and unspeciated PM2.5). Sulfate has the largest contribution to AOD, and EC is the main contributor to AAOD. The diagnosed total AOD and AAOD are consistent with the simulated values, although nonlinear interactions from internal mixing result in a small difference between them. The nonabsorbing aerosol species (e.g., sulfate and nitrate) account for 15 − 20% of the AAOD through their enhancement of absorption by EC and dust.

[39] Aerosols result in a TOA cooling (i.e., negative forcing) over all of California, with a statewide annual average of −1.0 W m−2 and a statewide maximum of −3.5 W m−2, with negligible seasonality. Aerosols introduce atmospheric warming with a maximum reaching 10 W m−2 over the Central Valley and the Los Angeles metropolitan regions, following the spatial distribution of AAOD. The seasonality of atmospheric warming shows, on statewide average, a summer maximum of 2.0 W m−2 and a winter minimum of 0.5 W m−2, determined by the AAOD and the solar radiation fluxes. Aerosols cool the surface over all of California by up to −10 W m−2. The statewide average cooling has a winter minimum of −1.5 W m−2 and a summer maximum of −3.0 W m−2, dominated by the seasonality of solar radiation fluxes. Jacobson et al. [2007] reported a total anthropogenic radiative forcing of −7.7 W m−2 at the surface in February and −7.91 W m−2 in August of 1999 over California. Their larger radiative forcing estimates at the surface compared with this study may be related to the larger anthropogenic emissions in 1999 and/or interannual variability of meteorology. In addition, their radiative forcing, estimated from the baseline and sensitivity simulations, has been defined as a radiative flux perturbation (RFP) [Haywood et al., 2009] that includes the change of radiative fluxes from aerosol − cloud interaction, whereas our diagnostic method strictly estimates only the aerosol direct radiative forcing [Lohmann et al., 2010].

[40] Aerosol direct radiative forcing along with the contribution from each aerosol species over California is presented for the first time. At the TOA, most aerosols introduce a negative radiative forcing, with sulfate as the biggest contributor (minimum −0.4 W m−2 in winter and maximum −0.7 W m−2 in summer), except that EC leads to a positive forcing (minimum 0.2 W m−2 in winter to maximum 0.7 W m−2 in summer). In the atmosphere, EC and dust contribute about 75 − 95% and 1 − 10% of the total aerosol warming (~0.5 − 2.0 W m−2), respectively. The rest of the aerosol warming comes from nonabsorbing aerosols as a result of nonlinear interactions involving internal mixing. All aerosols cool the surface, EC being the largest contributor in summer (−1.1 W m−2 and ~35%) and sulfate the largest in winter (−0.45 W m−2 and ~30%). The TOA radiative forcing of EC over California is comparable to that of total carbonaceous aerosols on global average (0.75 ± 0.25 W m−2, mainly from EC, because the net effect of OM is close to zero if accounting for the OM absorption effect) estimated by Chung et al. [2012], with observational constraints.

[41] The direct radiative forcing of speciated aerosols estimated in this study may be biased by the model uncertainties identified through the comparison with available observations. First, the underestimation of RH may result in negative bias of AOD and AAOD and hence aerosol radiative forcing. The overestimation of precipitation may contribute to the negative bias of aerosol concentrations because of excessive wet removal, which can also lead to negative bias of aerosol radiative forcing. Second, the simulation with doubled anthropogenic EC emission that is used to estimate the aerosol radiative forcing significantly underestimates the OM surface mass concentrations by a factor of 2 − 4 (this underestimation is even larger if a factor of 1.8 instead of 1.4 is applied to convert the measured OC to OM). In addition, absorption by OM (brown carbon) is not treated in this study, so the absorption from OM is underestimated. These OM- related biases might lead to an underestimation of aerosol-induced atmospheric warming. Consequently, the TOA OM radiative forcing (negative in this study) may be overestimated considering that the net effect of OM might be close to zero after its absorption is taken into accounted [Chung et al., 2012]. Third, the radiative forcing of dust and sea salt may have biases near the desert source regions and the coastal areas, respectively. However, evaluation of their biases is difficult because of the absence or uncertainties of observations. Finally, the aerosol radiative forcing estimated in this study is from a 1 year simulation using the meteorological conditions from 2005 and the anthropogenic and biogenic emissions from 2008. Aerosol radiative forcing is influenced by both meteorology and emissions, so uncertainty in our estimates of radiative forcing resulting from interannual variability in meteorological conditions and emissions, in addition to the numerous factors (e.g., model biases) discussed above, should be acknowledged.

[42] Several deficiencies in this study should be underscored. The negative bias of OM may reflect the issue with the outdated SOA mechanism used in the current version of WRF-Chem, the SOA chemistry of which must be improved. Although the model simulation with anthropogenic EC emission doubled significantly reduces the model negative bias of EC, we emphasize that the mismatch of using the 2008 emission inventory for the 2005 simulation complicates the evaluation and interpretation of model performance. Although it is not an uncommon practice in the modeling community to use an emission inventory developed for a very limited time period for simulations of a longer or different period, this study warns of the interpretation of model skill and emission uncertainties based on evaluation of such simulations. Emissions of OM and EC may be misrepresented. Efforts should be made to improve the speciation of emissions (i.e., unspeciated aerosols should be properly categorized and accounted for in the emission of speciated PM), which has important implications for estimating aerosol direct radiative forcing. Neglecting the seasonal variation of anthropogenic and biogenic emissions may also introduce some biases. However, the impact of seasonality of anthropogenic emissions may be small, because the comparison with observations suggests that chemical and physical processes can well explain the seasonality of aerosol surface concentrations, especially in urban locations. The biogenic emissions in seasons other than summer may be overestimated, but their impact is not quantified in this study because OM is already significantly underestimated throughout the seasons.

[43] Although the simulations have been evaluated with extensive meteorological and aerosol mass and optical measurements, more field measurements are needed, especially for dust near the source region, and the vertical profiles from aircrafts would be useful for further evaluating and improving model performance. Measurements regarding the optical properties and size distribution of aerosols are also needed to constrain further the model-estimated aerosol direct radiative forcing. Despite various sources of uncertainty in the model and measurements, the encouraging performance of WRF-Chem, a relatively sophisticated coupled meteorology − chemistry model applicable at high spatial resolution, in simulating aerosols and their direct radiative forcing supports the use of the model for investigating emission control impact on the regional climate over California. The diagnostic method implemented in WRF-Chem can be applied to other regions to understand the roles of different aerosols in direct radiative forcing and the regional climate.

Acknowledgments

[44] This work was supported by the California Air Resources Board (CARB) under contract 08-323 and the DOE Regional and Global Climate Modeling Program. The statements and conclusions in this paper are those of the researcher and not necessarily CARB. We thank Drs. V. Ramanathan and Ranjit Bahadur of the Scripps Institution of Oceanography and Dr. John Seinfeld of the California Institute of Technology for many insightful scientific exchanges on interpretations of our modeling results. The authors thank Dr. Jinho Yoon for his constructive suggestions during the PNNL internal reviewing. This study used computing resources from the National Energy Research Scientific Computing Center, which is supported by the U.S. Department of Energy Office of Science under contract DE-AC02-05CH11231, and PNNL Institutional Computing. Pacific Northwest National Laboratory is operated for the U.S. DOE by Battelle Memorial Institute under contract DE-AC06-76RLO330 1830.

Advertisement