1. Introduction
Epidemiological studies indicate that ozone (O3) and particulate matter smaller than 2.5 μm in diameter (PM2.5) are among a handful of criteria air pollutants that are primarily responsible for adverse impacts on human health (EPA 2005). Inhalation of these pollutants is recognized as a major cause of acute and chronic respiratory and cardiovascular diseases and air pollution–associated premature mortalities (e.g., Fann et al. 2012). The National Air Quality Forecasting Capability (NAQFC) is designed to safeguard the public by providing forecast guidance for surface concentrations of these pollutants at fine enough spatial and temporal resolutions and with sufficient lead times to be useful for official interpretative air quality forecasts issued by state and local air quality forecasters. NAQFC operational O3 forecast guidance for the nation has been issued since September 2007 (Stajner et al. 2012a) and developmental PM2.5 forecast guidance since September 2009 (Mathur et al. 2008).
The NOAA/Air Resources Laboratory (ARL) and the National Centers for Environmental Prediction (NCEP) develop upgrades for the NAQFC forecasting system, and conduct and evaluate pre-implementation testing. The NAQFC comprises an offline coupling of the North American Model (NAM) Nonhydrostatic Multiscale Model with Arakawa B-grid staggering (NMMB) and the Environmental Protection Agency’s (EPA) Community Multiscale Air Quality Model (CMAQ). Surface O3 concentration forecasts are issued out to 48 h twice daily for the 0600 and 1200 UTC cycles (Chai et al. 2013). Predictions for each cycle are available online (http://airquality.weather.gov) by 1300 and 1730 UTC. The accuracy criterion for a successful O3 forecast was determined to be achieving at least a 0.9 value in fraction correct (FC), which is also referred to the as proportion correct (Davidson 2009). FC is calculated as the ratio of the sum of correctly predicted exceedances and correctly predicted nonexceedances defined by the primary National Ambient Air Quality Standard (NAAQS) threshold for the maximum daily 8-h-averaged surface (MDA8) O3 concentration to the total number of measurements. NAAQS for MDA8 was 75 ppbv between March 2008 and October 2015. The monitoring stations used for this accuracy criterion compose the EPA AIRNow surface monitoring network (EPA 2015a). These monitors provide real-time surface measurements of air pollutant concentrations collected using federal reference or equivalent methods operated by local environmental and state agencies. AIRNow collects hourly data from about 1300 O3 reporting stations across the country between June and September and has done so since 2007. In total this network reports about 158 600 MDA8 ozone measurements from June to September. The number of O3 concentration reporting stations in the AIRNow network decreases to about 1000 for the remaining months of the year. During the O3 seasons between 2009 and 2014, the NAQFC O3 forecast guidance exceeded FC of 0.90 or greater for each monthly average. Moreover, the daily FC exceeded 0.90 for all but 2, 5, 6, 8, 2, and 1 days for the 6 yr from 2009 to 2014, respectively (Stajner et al. 2012b, 2014).

Schematic contingency table for an exceedance forecast with N data points.
The NAAQS for MDA8 O3 was set at 75 ppb in March 2008. It coincides with the category “code orange” threshold, beyond which adverse human health effects are expected for sensitive groups. It was strengthened from a previous threshold of 80 ppb that was promulgated in July 1997 and in effect until March 2008. In December 2014, the EPA proposed that this NAAQS was to be further tightened to lie between 65 and 70 ppb (EPA 2015b). On 1 October 2015, the EPA tightened the NAAQS to 70 ppb. This continued trend of tightening NAAQS reflects epidemiological evidence that impacts on human health are found at lower thresholds than those previously established (Tager et al. 2005). This trend poses considerable challenges for NAQFC in terms of its reduced margins for allowable errors. Any proposed upgrade to the NAQFC operational forecast guidance should continue to match and supersede FC for operational ozone predictions and lead to the improvement of PM2.5 predictions for all seasons.
Beginning in January 2015, the same NCEP model that produces operational O3 forecast guidance is also producing developmental PM2.5 forecast guidance. Between 2009 and 2014, the PM2.5 developmental product was generated by a nonoperational version of the NAQFC system with results disseminated to select local and state air quality forecasters participating in the NAQFC Air Quality Forecaster Focus Group (AQFFG). The distinction between the operational and developmental modes of the NAQFC forecast guidance is based on how well they meet requirements for reliable on-time delivery, accuracy, and format compliance, as well as their respective content and scope of recipients for product dissemination. Operational products are guaranteed to be available and to be disseminated on time to the general public in graphical and World Meteorological Organization (WMO) standardized formatted files in links and data portals found online (http://airquality.weather.gov/).
By design, developmental product testing fosters a dialogue between forecast system developers and the AQFFG to accelerate the improvement of developmental forecasting products to satisfy the proposed success criteria. An FC of at least 0.9 is required for the 24-h-averaged forecast surface concentrations for PM2.5 with respect to the primary NAAQS threshold of 35 μg m−3. This FC requirement for the forecast accuracy between 2010 and 2014 was not met. The FC has been evaluated against about 1000 hourly reporting monitoring stations in the AIRNow network. There was a seasonal pattern to the rather dismal performance by the developmental PM2.5 forecast. There was significant underprediction in summer and overprediction in winter. During the five years between 2010 and 2014, the averaged FC for the developmental PM2.5 forecast did not improved much over time. It averaged between 0.67 and 0.74 over the summer months of June–September and between 0.58 and 0.64 over the winter months of November–February. In appendix A (Fig. A1), we illustrate this verification calculation for the ETS criterion across the period for the thresholds between 0.5 and 35 μg m−3. Again, the ETS skill did not improve noticeably over time especially for the high thresholds.
This paper overviews the current NAQFC system and performance for O3 and PM2.5. The following section provides the model description. Section 3 focuses on the CMAQ components of the NAQFC system, and section 4 provides sensitivity analyses and evaluations. Results and discussion are presented in the final section.
2. Model description
Figure 1 shows the schematic of the developmental NAQFC modeling system for the forecasting of surface PM2.5 concentrations. It is a regional system comprising a numerical weather prediction (NWP) model and a chemical transport model (CTM) coupled together in an offline arrangement where the NWP model provides predicted meteorological fields to the CTM at hourly intervals. The NWP model is the NOAA/National Weather Service’s (NWS) NAM. NAM is based on the aforementioned NMMB (Janjić and Gall 2012). NAM provides operational weather predictions for the United States out to 84 h based on a 12-km horizontal grid. The NAM domain covers nearly one-third of the Northern Hemisphere, spanning from near the equator to the North Pole from south to north and with its southwestern corner at around 1000 km southwest of the Hawaiian Islands and its northeastern corner in northern Europe in a rotated latitude–longitudinal map projection (Fig. 2). Although the meteorology is available for all 50 states, the developmental PM2.5 product began with coverage for the CONUS in 2009 and was expanded to cover Alaska and Hawaii by 2010. A modified version of the EPA’s CMAQ model version 4.6, dubbed CMAQv4.6.5, is the CTM in this system (Fig. 1). CMAQ is also run with 12-km horizontal grid spacing, but with a Lambert conformal conic (LCC) map projection. The offline coupling between NAM and CMAQ is achieved by two preprocessors: (i) Prdgen, which handles the horizontal map projection transformation for the meteorological variables from the NMMB grid to the CMAQ LCC grid through the NCEP IPOLATES code, a geometric interpolation package; and (ii) Premaq, which handles grid-staggering transformation, meteorological modulation of emission fluxes, and the collapse of the vertical grid structure from NMMB’s 60 hybrid sigma–pressure layers to CMAQ’s 22 sigma-pressure (σ–p) layers using a pressure-surface-based linear interpolation between the two NAM levels containing a CMAQ layer. These interface processors are described in further detail in sections 2b and 2c, respectively. In addition to the coupled NMMB–CMAQ system, there are other components such as the emission module and the chemical lateral boundary condition builder as well as the product generating postprocessing components. The emission module is described in section 2d. Handling of boundary conditions is described as part of the description for the CTM in section 3.
Schematic of the developmental NAQFC PM2.5 forecasting system: NMMB12 is the NWS mesoscale meteorological model, NAMpost calculates prognostic and diagnostic fields from NMMB12, Prdgen interpolates those meteorological fields from the NAM native grid onto an LCC grid, PREMAQ reads the LCC-grid-projected fields for CMAQ as well as prepares for CMAQ emission and process rate files by using preprocessors SMOKE and MOBILE5b, and CMAQ is the chemical transport model that reads the ICON and BCON respective outputs for initial and chemical boundary conditions. CMAQ with PREMAQ simulates air quality to generate O3 and PM2.5 surface concentration forecasts. The GRIB converter and verification tools postprocess and evaluate the forecast.
Citation: Weather and Forecasting 32, 1; 10.1175/WAF-D-15-0163.1
The NCEP NAM (12 km) domain (outline shown by the dashed line). The NAQFC CONUS domain is also shown by the boldface frame.
Citation: Weather and Forecasting 32, 1; 10.1175/WAF-D-15-0163.1
a. NAM configuration
The NWS regional NAM meteorological model is based on the NMMB (Janjić and Gall 2012), covering all 50 states and territories of the United States since October 2011. The physics packages are similar to those of its predecessors: the NCEP Eta Model that was retired from operation in October 2006, and the NCEP WRF-NMM model, retired in October 2011 (Table 2). These three generations of NAM used a rotated latitude–longitude map projection (Janjić et al. 2001). Regional data assimilation systems are used to assimilate meteorological and land surface observational data and provide initial conditions: NAM Data Assimilation System (NDAS) (Wu et al. 2002), and the Noah land surface model (LSM) based Data Assimilation System (NLDAS) provides land states (Mitchell et al. 2004). The advection time step for the 12-km NMMB is 26.7 s. Turbulence and moist processes are computed every 160 s. Both the horizontal and vertical advection schemes are positive definite and conservative for total kinetic energy and hydrometeors. The vertical grid structure is an application of the general hybrid pressure-sigma and isobaric coordinate system after Eckermann (2009). There are 60 layers spanning the area between the surface and the model top at 2 hPa. The vertical grid spacing is more refined near the surface and around the most probable heights of a fully developed summer planetary boundary layer (PBL) top and of a midlatitude spring tropopause. NAM outputs hourly meteorological and hydrological fields at its native horizontal and vertical coordinates for CMAQ.
Physics options in NMMB.
b. Processing of NMMB output using Prdgen
The NAM output for CMAQ consists of 3D and 2D fields (see Table B1). The Prdgen interface processor handles the variable horizontal map transform from the NAM grid to the CMAQ grid. Horizontal interpolation schemes are used to determine the so-called cross-point values of the resulting grid on the intermediate LCC projection with Arakawa A-grid staggering. Most variables use a bilinear algorithm applied to distance. The remaining variables that exhibit a stronger discrete characteristic use a nearest-neighbor algorithm to adopt the value pertaining to the closest cross point in the NAM grid.
c. Using Premaq to read Prdgen output fields and interpolate to CMAQ vertical levels
CMAQ uses an LCC map projection with Arakawa C-grid staggering. The second interface processor, Premaq, is invoked sequentially after the completion of Prdgen to complete the grid-staggering transformation (Fig. 1 and section 2b). Originally designed and implemented by Otte et al. (2005) for an initial version of the NAQFC that provided surface O3 forecasts for the northeastern United States, Premaq’s basic functionality of grid-staggering transformation and emission flux rate calculations remains valid, but has been substantially improved through a series of upgrades in NAM and CMAQ. A major NAM-related upgrade was the retirement of the Eta step-mountain vertical grid structure and the adoption of a hybrid isobaric and terrain-following vertical structure in NMMB. In addition, Premaq was also modified significantly to accommodate version upgrades in CMAQ that often entailed additional meteorological input. Premaq performs both the redistribution of the Prdgen processed fields in Arakawa A-grid staggering to C-grid staggering and a reduction of vertical levels from 60 to 22 to decrease the wall-clock time required for operational predictions.
Premaq also computes and outputs CMAQ-ready fields for dry deposition velocities, cloud-cover-induced photolytic rate attenuation coefficients, emission altitudes, and flux rates of air pollutants at each hour throughout a 48-h forecast. Premaq produces the same CMAQ-ready input files as the EPA’s Meteorology–Chemistry Interface Processor (MCIP). [See Tables 2–4 in Otte and Pleim (2010).] Premaq adds a few more fields such as snow-cover and clear-sky downward shortwave solar flux to enhance the physics package consistencies between NAM and CMAQ.
The gaseous species dry depositional velocity calculation follows that of the “M3Dry” model in MCIP version 3.4.1 (Otte and Pleim 2010). NAM surface parameters such as canopy water and canopy conductance were provided by NAM to retain consistency between NMMB and CMAQ. Table 3 shows typical values of the depositional velocities used in the NAQFC under ordinary ambient conditions. Satellite-retrieval-based canopy heights (Lefsky 2010) are included in the roughness length estimate for the aerodynamic resistance calculation. A multiplicative factor was employed to scale tree canopy height to roughness length for computing dry deposition velocities (Brutsaert 1982).
Typical dry deposition velocities predicted by the NAQFC.
Photolytic rates are proportional to the ambient actinic flux. Above cloud enhanced photolytic rates due to reflection from clouds are accounted for in the standard releases of CMAQ (Byun and Ching 1999). The in situ photolytic rate attenuation coefficient equals unity at the cloud top. The in situ photolytic rate attenuation coefficient at each height at and below the cloud base is equal to the ratio of shortwave solar radiation reaching that height to the radiation that would reach that height under clear-sky conditions. Photolytic rates within the cloud are interpolated between the cloud-top and cloud-base values using the in situ cloud cover fraction at that height.
To generate CMAQ-ready emission files for forecasting, Premaq provides emissions fluxes for point, area, and nonroad, mobile, and biogenic sources on the CMAQ grid. The following subsections elaborate on the methodology adopted to perform the various emission projections and modulations targeting a given forecasting year.
d. Emission projections
The EPA’s National Emission Inventory (NEI) 2011 version 1 is being incorporated into Premaq’s emission projection schemes. The incorporation is accomplished in two phases, as the second phase requires a coordinated upgrade of the CTM that computationally parallelizes many of the calculations for meteorological-dependent emission processes. Table 4 describes the emission sectors in the first phase of incorporating NEI 2011 into the NAQFC modeling system that became effective on 1 May 2015.
NAQFC emission categories using NEI2011, beginning 1 May 2015.
1) Point sources
The 2005 NEI version 1 (NEI05v1) was used as a first estimate for electric generating unit (EGU) and non-EGU U.S. nitrogen oxides (NOx) and sulfur dioxide (SO2) point source strengths. These point sources were updated with the 2014 Continuous Emission Monitoring (CEM) dataset—a biennial database. EGU projections were computed using the ratios of the emission strengths for 2012–14 and then extrapolating to 2015. In addition, the regionally based Annual Energy Outlook (AEO) SO2 and NOx 2015 emission projection factors from the Department of Energy were used. For offshore large point sources, the EPA’s 2008 offshore emission inventory was used. The Environment Canada (EC) 2011 point source National Emission Inventory (EI) was used for Canada. In Mexico, the 2012 Mexico National EI, version 2.2, was used for the six states in northern Mexico bordering the United States, and the Mexico EI version 1 was used for the interior states. Plume rise calculations account for the effective injection heights of buoyant pollution plumes based on stack height, initial discharge characteristics of the plume, and surrounding meteorological stability conditions (Briggs 1972). The plume-rise equations were then delineated into unstable, neutral, and stable atmospheric conditions (Byun and Ching 1999).
2) Area and nonroad sources
Except for off-road sources, a combination of the EPA’s 2011 version 1 (2011NEIv1) and 2005 NEIv1 was used. The delineation to which inventory should be used for a particular sector depended on whether its processing was compatible with CMAQv4.6.5. Results from 2011 NEI were used for all agricultural NH3; railway and class 1 and class 2 marine emissions primarily representing non-ocean-going activities; vehicular refueling; oil and gas industry related emissions; and residential wood combustion. Also, U.S. off-road emissions in the 2005 NEIv1 were replaced with the 2011 inventories. These inventory data were processed using the EPA’s Sparse Matrix Operator Kernel Emissions (SMOKE) modeling system to represent monthly, weekly, daily, and holiday/non-holiday variations specific for the forecast year. Emissions from wildfires, prescribed agricultural burns, and land-clearing fires based on climatology were removed from the area source emissions and replaced with dynamic fire emission modeling using a version of the U.S. Forest Service BlueSky smoke emission package (O’Neill et al. 2009) and the NOAA/NESDIS Hazard Mapping System (HMS) for fire locations and strength. In terms of procedure, we processed emission inventories by sectors, and for the fire sectors, we did not include prescribed burns and wildfires from the NEIs. The 2006 Environment Canada National Inventory (EC NI) area sources were used for Canada, and the 2012 Mexico NEI nonroad sources were used for Mexico.
3) Mobile source emissions
To reflect recent changes in mobile source emissions, the EPA’s Office of Transportation and Air Quality 2005 on-road emission inventory was adjusted to 2012 basing on the EPA Cross State Air Pollution Rule (CSAPR; available online at http://www.epa.gov/airtransport/CSAPR/). Both 2005 NEIs and the CSAPR-projected emission inventories were based on the MOBILE6 model. The CSAPR projection considered possible emission changes caused by the existing and predicted emission control regulations finalized in early 2009. The methodology generated on-road mobile emissions and the time-activity pattern counts for monthly, weekday/weekend, and diurnal variability for different vehicle types over the CONUS based on the EPA’s emission factor model MOIBLE6 (Tong et al. 2015b). The trends for NOx emissions over large U.S. urban centers between 2005 and 2012 were analyzed using surface measurements based on the EPA Air Quality System (AQS).
The collocated column-integrated retrievals made by the Ozone Monitoring Instrument (OMI) on board the National Aeronautics and Space Administration (NASA) Aura spacecraft were well corroborated with the surface-based AQS data. OMI analyses yielded a 40% decrease over the period between 2005 and 2012 for the urban centers where NOx emissions from vehicular activity were the dominant NOx source sector (Tong et al. 2015b). We used the 2006 EC NI for Canada, and the 2012 Mexico NEI version inventory for Mexico.
4) Biogenic sources
All inputs used in the forecast system were updated to the Biogenic Emission Inventory System (BEIS3), version 3.14. Biogenic emissions were calculated dynamically using the BEIS3 version 3.14, which considered variability in temperature and solar radiation to estimate NO and volatile organic compound (VOC) emissions from forests, grasslands, and croplands.
5) Sea salt emissions
Sea salt can represent a significant mass percentage of the aerosol concentration in coastal areas. Sea salt emissions were parameterized as a function of 10-m wind speed and surf zone category. In open oceans, sea spray is the dominant emission mechanism. In the surf zone, wave breaking contributes a larger amount of sea salt emission than in the open oceans. CMAQ v4.6.5 does not use any heterogeneous chemical or depositional processes for sea salt. Sea salt is treated as an inert species in coarse and accumulation modes. Emissions of trace gases and organic aerosols from the ocean have not been included in the NAQFC system.
6) Suppression of fugitive dust by ice and snow cover
Fugitive dust, a significant component of PM2.5, is predominantly composed of mineral and crustal elements and compounds. CONUS-wide average fugitive dust emission sources from paved and unpaved roads account for approximately 7% and 46% of the total fugitive dust emission, respectively. Both sources have crustal and organic carbon components as their major mass fractions. For paved road sources, 89% and 10% of the emissions in this sector are attributed to the crustal and organic carbon components, respectively. For unpaved roads the attributions are approximately 94% and 6%, respectively. Other significant sources included airstrips, construction sites, and fields undergoing tilling and harvesting. These sources do not apply when the surface is covered by ice or snow. NAM-predicted snow cover fraction is used to scale the emissions fluxes.
7) Intermittent emissions from fire and dust
Windblown dust storms and wildland fires contribute a large amount of fine particulates to the surface. Premaq deals with both of these intermittent emission sources originating in the CONUS domain. Premaq provides CMAQ with a 3D emissions file for these intermittent sources.
The HMS was used to provide fire-point and smoke-plume locations by blending multiple satellite retrievals with human analyst products (Ruminski et al. 2006). The U.S. Forest Service BlueSky smoke emission package (O’Neill et al. 2009; Larkin et al. 2009) was used with HMS to estimate near-real-time smoke emissions. Wildfires that were estimated to last at least 24 h were used as emission sources into the NAQFC. BlueSky provides emissions of PM2.5, particulate matter smaller than 10 μm in diameter (PM10), carbon monoxide (CO), carbon dioxide (CO2), methane (CH4) and VOCs from wildfires. Emission rates of others species contributing to the chemical mechanisms in CMAQ, such as elemental carbon (EC), black carbon (BC), nitrogen oxides (NOx), VOCs, ammonia (NH3), and sulfur dioxide (SO2) were estimated by scaling them to that of CO, a wildfire signature species, using their mass ratios with CO (Hsu and Divita 2011). The BlueSky wildfire heat flux was also estimated and used in the Briggs equation (Briggs 1972), to determine injection heights.
Lifting of crustal and mineral elements by winds constitutes a major intermittent primary particulate matter source over barren land, dry river beds, and large swathes of arid land in the United States. The topography, soil surface water content, and textural characteristics are among the governing factors for crustal particle uplift and suspension. Wind gusts surpassing a certain threshold specific to those characteristics will dislodge the particles. Turbulence and wind shear within the boundary layer keep the particle airborne and may bring it to a higher altitude. Tong et al. (2015a) followed and modified the methodology presented by Owen (1964) to determine the instantaneous lifting rates. The large range of parameters in Owen’s model and the non-erodible and moistened effects of the topsoil result in significant uncertainties for modeled soil particle lifting rates. The developmental NAQFC system also includes the effects of rain and snow on erodible topsoil elements (Tong et al. 2015a). Forecast results and verification from the modeling system for a dust storm active season are shown in section 4.
3. NAQFC CMAQv4.6.5
a. Transport














Advection is implemented using the piecewise parabolic method (PPM) advection scheme (Collela and Woodward 1984; Byun and Schere 2006). For convective mixing, a combined local and nonlocal mixing closure model [Asymmetric Convective Model 2 (ACM2)] is used to vertically distribute trace gases and aerosols. Mass is entrained through a gradual layer-by-layer compensatory subsidence in ACM2 (Pleim 2007a,b). ACM2 is used for small-scale eddies and large-scale turbulent mixing in the boundary layer. In non- or weakly convective conditions the ACM2 scheme simulates the suppressed mixing accurately. Some studies utilizing the NAM output vertical eddy diffusivity and the diffusion equation directly for simulating turbulent mixing in the stable boundary layer have shown promising results comparable with those from ACM2 (Lee et al. 2009).
b. Boundary conditions
At the lateral boundaries, with n denoting the outward normal vector,
Monthly varying lateral boundary conditions for the NAQFC CONUS domain for O3 (ppb) for June (red), July (blue), and August (black) averaged at the (a) south, (b) west (c) north, and (d) east lateral boundaries.
Citation: Weather and Forecasting 32, 1; 10.1175/WAF-D-15-0163.1
c. Model options
Table 5 summarizes the CMAQ model options selected for the various geometric configurations and physical and chemical schemes. The 22 σ–p terrain-following layers represent the atmospheric column between the surface and the model top at 100 hPa, with the first 14 layers covering the lowest 2 km, which is critical for both the meteorological and air quality models (Lee and Ngan 2011). The geometric thickness of layers is the thinnest near the surface and increases unevenly with altitude. The top of the lowest model layer is at 39 m AGL.
CMAQ4.6.5 physics and chemistry options for the PM2.5 developmental forecast.
The interested reader is referred to “The CMAQ Science Algorithm” by Byun and Ching (1999) and subsequent EPA technical reports (e.g., Edney et al. 2007), as well as the MODELS3 website CMAQ Release Notes in Community Modeling and Analysis System (CMAS), for details concerning the parameterization of the physical and chemical processes.
d. Model modifications
Several modifications to CMAQ version 4.6 were made to improve the consistency of the physics package between NAM and CMAQ. In section 2c, we already described the modifications for consistency used in Premaq. Removal of gas constituents by dry deposition in the CMAQ model is calculated by multiplying their respective surface level concentrations with deposition velocities (Pleim et al. 1997; Binkowski and Shankar 1995). The deposition velocities for gaseous species utilize the resistance approach analogous to Ohm’s law in electrical circuits. For aqueous-phase chemistry and in- and below-cloud scavenging, the cloud liquid water content is diagnosed from cloud top and base, and from the predicted rate of convective precipitation.
Additional modifications of the standard CMAQ include the following: 1) faster removal of organic nitrate (NTR) and reduction in its sequestration efficiencies within the Carbon-Bond 2005 (CB05) gas-phase mechanism by increasing the photolysis frequency by a factor of 10 (Saylor and Stein 2012; Canty et al. 2015); this modification typically shortens the predicted lifetime of NTR in CMAQ from about a week to approximately a day (Pan et al. 2014); and 2) a minimum PBL height of 50 m avoids total suppression of vertical diffusive mixing.
e. Aerosol processes
The NAQFC CMAQ v4.6.5 follows largely the EPA’s Aero4 module and the related emission and removal processes found in CMAQ version 4.6. Gas-to-particle conversion, heterogeneous reactions, depositional growth, and coagulation are included (Edney et al. 2007; Carlton et al. 2008; Kelly et al. 2009). The Aero4 module simulates particle formation, condensational and coagulation growth, or evaporative dissipation of existing particles due to ambient chemical, temperature, and humidity conditions. The removal processes are modeled outside Aero4 in the cloud, scavenging, and dry deposition modules. The aerosol processes and their parameterizations pertinent to Aero4 have been described in details by Byun and Schere (2006) and Binkowski and Roselle (2003). Two major modifications from CMAQ 4.6 were adopted for emissions: 1) modulation of fugitive dust emission due to snow and ice cover [see section 2d(6)] and 2) the inclusion of windblown dust [see section 2d(7)].
Aero4 adopts a modal approach to categorize particulate matter by its diameter prescribed as lognormal distributions into two fine modes: the Aitken mode with diameters peaking between 0.01 and 0.1 μm, and an accumulation mode with diameters peaking between 0.1 and 1.0 μm. In addition, a coarse mode is also represented by a lognormal size distribution with diameters peaking between 1.0 and 10.0 μm.
Fine particles in CMAQ often reflect fresh emissions of elemental carbon from incomplete combustion or from new particle formation through binary homogeneous nucleation of sulfuric acid vapor with water vapor. The nighttime heterogeneous conversion of N2O5 to nitric acid can also be a significant pathway for fine particle formation and growth, especially under warm and humid conditions (Bhave et al. 2006). Condensation of semi-volatile carbonaceous compounds from anthropogenic and biogenic sources such as alkanes, aromatics, terpenes, and other VOCs can condense to form secondary aerosols. Coarse mode particles emerge predominantly from emissions related to dislodgements of crustal mineral aerosols by wind, vehicular and agricultural activities, and maritime aerosols such as sea salt from sea spray in the surf zone. Once a fine particle is formed, Aero4’s ISORROPIA (Nenes et al. 1998), a thermodynamic equilibrium model, determines the partition between the gas and the particulate phases of ammonia, nitrate, sulfate, and water species.
Dry deposition velocities for particles are parameterized by size-dependent sedimentation (Binkowski and Shankar 1995). CMAQ treats pollutants differently for in- and- below-cloud scavenging depending on whether a pollutant participates in the aqueous-phase reactions. The accumulation and coarse mode particles are assumed to be completely absorbed by the cloud and rainwater whereas the Aitken mode particles are treated as interstitial aerosols. An Aitken mode particle can coagulate with other particles or be absorbed by a cloud or rain droplet. An absorbed particle can reemerge as a dry particle by resuspension and evaporation.
The aerosol mass wet removal rates depend proportionally on the rate of precipitation. Wet deposition trace species removal consists of in-cloud- and below-cloud scavenging and washout (Binkowski and Shankar 1995). NMMB-modeled hydrometeor fields were not used for wet deposition removal rate calculations in the NAQFC CMAQ. Diagnosed precipitation rates based on a diagnostic cloud volume reconstructed by the relative humidity profile given by NMMB were used instead (Chang et al. 1987).
4. Sensitivity analysis and evaluations
Developmental PM2.5 forecast guidance was upgraded in January 2015. Three changes were made from the 2014 developmental PM2.5 model: 1) intermittent emissions due to windblown dust originating inside the CONUS [section 2d(7)] and wildfires [section 2d(7)], 2) suppression of fugitive dust emissions by snow and ice cover [section 2d(6)], and 3) an accelerated removal of the organic nitrate (NTR) species (section 3).
To quantify the performance improvement from these upgrades (e.g., Fig. 4), we conducted several reforecast sensitivity studies across multiple seasons (Table 6). Experiment 2014-PM2.5 refers to the configuration before the NCEP upgrade in January 2015, which had been the mainstay for our studying the forecast performance for surface PM2.5 concentration. Figure 4 shows that between January 2009 and December 2014 wintertime overestimates had gradually been reduced as a result of a reduction in the overestimation of mobile NOx, resulting in reduced particulate NO3−. Experiment 2015-PM2.5 represents the model configuration since the January 2015 upgrade, together with May 2015 emissions updates described in sections 2d(1)–2d(4) using a partial update from NEI2005 to NEI2011, as explained in Table 4. These emission updates are applied retroactively to all months in 2015 in the experiment 2015-PM2.5 presented here. The No Dust experiment denotes the 2015-PM2.5 configuration without real-time windblown dust emissions [see section 2d(7)]. Similarly, the No Wildfire, No Snow/Ice, and No J*NTR10 cases denote the 2015-PM2.5 configurations without the wildfire emission in section 2d(7), snow and ice cover suppression of fugitive dust in section 2d(6), and the increased photolysis frequency for NTR in section 3, respectively. All experiments in Table 6 were initialized once per day at 1200 UTC, except for the 2014-PM2.5 configuration, where the model was initialized four times per day at 0000, 0600, 1200, and 1800 UTC. Only the prediction results from the 1200 UTC cycle run were used for evaluation. The last four experiments in Table 6 were performed for the season when they were expected to have the largest impact: dust events in late spring in the southern plains and in California, increases in the occurrence of wildfire and photolysis frequencies of carbonyl nitrates in midsummer, and wintertime suppression of emissions of fugitive dust during periods of snow and/or ice cover. Figure 5 shows the diurnal cycle of PM2.5 biases for the 2014-PM2.5 and the 2015-PM2.5 configurations averaged over July 2014. A consistent improvement of 2015-PM2.5 forecasts over that of the 2014-PM2.5 simulation is seen.
Time series of regionally and monthly averaged biases for surface hourly PM2.5 concentrations (μg m−3) for the 2014-PM2.5 forecast guidance between January 2009 and December 2014 (see Table 6) verified against the AIRNow network observations: PC (pink); LM (blue); SE (red); RM (cyan); UM (green); and NE (gray); with an average of 105, 60, 80, 50, 95, and 85 reporting stations over the period, respectively (Insets in Figs. 6 and 7 show the regional definitions.).
Citation: Weather and Forecasting 32, 1; 10.1175/WAF-D-15-0163.1
Sensitivity runs.
Monthly CONUS-wide averaged hourly biases of predicted surface PM2.5 (μg m−3) from the 1200 UTC cycle of the 2015-PM2.5 simulation for July 2014 for black showing 2014-PM2.5 and red showing 2015-PM2.5 forecasts, verified against the AIRNow surface network with about 510 reporting stations.
Citation: Weather and Forecasting 32, 1; 10.1175/WAF-D-15-0163.1
Table 7 shows a few of the commonly used statistical measures for evaluating the various model predictions. The inclusion of intermittent emissions shows little impact. Across the CONUS, the normalized mean error (NME) was reduced from −21% to −20%, and from 29% to 28%, respectively, when the intermittent sources of windblown dust in May 2014 and inside-domain wildfires in July 2014 were included in the forecasting simulation.
Daily averaged surface PM2.5 forecast guidance performance statistics.
Suppression of fugitive dust emission during winter conditions with snow and/or ice cover led to a considerable improvement of the PM2.5 forecasts, as illustrated by the January 2015 results. Suppression of fugitive dust emissions by snow–ice cover during the period reduced the NME from 31% to 13% over CONUS. (Table 7). The ETS skill achieved by this treatment of suppressing the emission of fugitive dust reflected a threefold improvement.
Small impacts of NTR and wildfire treatments were observed for the Pacific coast (PC) region, but windblown dust treatment improves the PM2.5 representation for the Rocky Mountain (RM) region (Figs. 6 and 7). The abbreviations for the other four regions are as follows: UM, Upper Middle; LM, Lower Middle; NE, Northeast; and SE, Southeast of the CONUS.
The PM2.5 observed (gray circles) and predicted concentrations (μg m−3) for the PC region (depicted in the inset) averaged over 117 AIRNow stations during July 2014 for the 2015-PM2.5 developmental NAQFC PM2.5 forecast guidance (blue), and the No NTR and No Wildfire (red) experiments.
Citation: Weather and Forecasting 32, 1; 10.1175/WAF-D-15-0163.1
As in Fig. 6, but for the RM region averaged over 57 AIRNow stations during May 2014 for NAQFC PM2.5 forecast guidance and the No Dust experiment.
Citation: Weather and Forecasting 32, 1; 10.1175/WAF-D-15-0163.1
Table 7 shows forecast guidance performance statistics for hourly averaged surface PM2.5 for May 2014 in RM, for July 2014 in PC, and for January 2015 in UM. During July 2014 there were numerous wildfires, especially in the western states. The northern and central Washington State wildfires beginning on 8 July 2014 caused significant damage to more than 1400 km2. Southern California, the southern parts of Arizona, and New Mexico also experienced more frequent wildfires with larger burnt areas than in an average year. The bias for PM2.5 in PC showed a small reduction from 3.56 to 3.49 μg m−3, in a regional-wide observation of 7.95 μg m−3, and an improved r, from 0.17 to 0.19, when wildfire emissions were included in the model (Fig. 6). The change in the photolysis frequencies for the NTR compounds where there was an abundance of volatile organic compound in relation to that of NOx was also examined for the July 2014 case. The impact was primarily on the gaseous species and secondarily on PM2.5. The largest impact of the NTR change was seen in the SE region where the MDA8 O3 bias was reduced from 8.1 to 7.9 ppbv, a 3% improvement; and the root-mean-square error (RMSE) was reduced slightly from 13.54 to 13.44 ppbv. This was due to the exacerbated overpredictions in rural areas, which nearly offset urban area improvements in the regional-scale evaluation. The shortened life of NTR in urban areas resulted in faster transformation into peroxyacyl nitrates (PAN) and affected photochemical reactions in the downwind rural areas, typically resulting in higher O3 production.
The May 2014 case was chosen to evaluate the impacts of windblown dust emissions due to a high frequency of dust storms in the western states. For instance, on 11 May 2014 multiple windblown dust storms occurred in Southern California, Nevada, Arizona, and Texas. The efficacy in capturing windblown emissions resulted in a considerable improvement in the PM2.5 forecast. The correlation coefficient r for RM increased substantially from 0.12 to 0.27 when the windblown fugitive dust emissions were included (Fig. 7). Table 7 shows correspondingly noticeable improvement in NME, RMSE, and ETS skills by incorporating windblown dust emissions.
The January 2015 experiments evaluated the impact of the suppression of fugitive dust emission due to snow and ice cover. This suppression has the largest positive impact of all the sensitivity cases considered in this study. The bias in UM for the 24-h-averaged daily maximum PM2.5 was reduced from 6.5 to 3.1 μg m−3, which is a 52% improvement (Fig. 8).
The PM2.5 observed (gray circles) and predicted concentrations (μg m−3) for the CONUS averaged over 515 AIRNow stations during January 2015 for the 2015-PM2.5 developmental NAQFC PM2.5 forecast guidance (blue) and the No Snow/Ice experiment (red).
Citation: Weather and Forecasting 32, 1; 10.1175/WAF-D-15-0163.1
5. Conclusions and discussion
In January 2015, NOAA/NWS/NCEP began providing the NAQFC developmental PM2.5 forecast guidance, 2015-PM2.5, using the same modeling system that generates the NAQFC operational O3 forecast guidance. This developmental PM2.5 forecast guidance and its predecessor, 2014-PM2.5, have been disseminated to a select group of air quality forecasters from local and state environmental agencies in the NAQFC Air Quality Forecast Focus Group (AQFFG) for early use and evaluation beginning in 2009. Four major upgrades were included in 2015-PM2.5: 1) incorporation of the base emission inventory from the EPA’s 2011 National Emission Inventory (NEI), except for ocean-going ships and on-road mobile 2) incorporation of intermittent emission sources within the NAQFC domain accounting for wildfire emission projections and windblown dust emissions; 3) suppression of fugitive dust emissions when there is snow or ice covering the ground; and 4) increase of photolysis frequency of alkyl nitrate by one order of magnitude to shorten the lifetime of organic nitrate in the gaseous mechanism. These modifications improved both the PM2.5 and O3 forecasts as demonstrated by verification against observations from the EPA’s AIRNow surface monitoring network. Four sensitivity studies were designed to evaluate PM2.5 forecast impacts specific to regional meteorological and emission characteristics. The No Wildfire case during July 2014 showed that the Pacific coast regional forecast correlation improved slightly, from 0.17 to 0.19 (Fig. 6). The No Dust case for May 2014 showed that Rocky Mountain regional forecasts correlation coefficients improved from 0.12 to 0.27 (Fig. 7). The No J*NTR10 case for July showed a negligible impact for PM2.5 forecasts and a 3% reduction in O3 bias in the southeastern United States. The No Snow/Ice run for January 2015 showed that in the upper Midwest, PM2.5 biases were reduced by 52%, from 6.5 to 3.1 μg m−3 (Fig. 8).
Improvement of PM2.5 forecast guidance continues with efforts focusing on the reduction of seasonal (Fig. 4) and diurnal biases (Figs. 5–7) in model predictions. Current efforts by the NAQFC development team seek to improve the physical processes for modeling atmospheric aerosols. Specific current efforts include the incorporation of the latest aerosol sciences from the newest version of CMAQ, incorporation of aerosol plumes that intrude into the CMAQ domain through derivation of the lateral boundary conditions, and application of bias correction techniques based on historical model performance to provide improved forecast guidance for PM2.5.
Developmental NAQFC forecast guidance for hourly and 24-h-averaged surface PM2.5 concentrations are available online (http://www.emc.ncep.noaa.gov/mmb/aq/) and in Gridded Binary (GRIB2) format upon request.
Acknowledgments
This work was partially funded by NOAA’s National Air Quality Forecast Capability and the U.S. Weather and Research Program. Throughout this work, the authors appreciate numerous valuable discussions with Drs. Rohit Mathur and Jonathan Pleim of the Atmospheric Modeling and Analysis Division, EPA; Dr. Ariel Stein of NOAA/Air Resources Laboratory; and Drs. Bill Lapenta and Geoff DiMego of the National Centers for Environmental Prediction. We are indebted to the advice of Drs. Mark Q. Liu and Sid Boukabara of the NOAA/National Environmental Satellite, Data, and Information Service. The authors thank the insightful comments from the three anonymous reviewers that added value to this paper. Although this work has been reviewed by NOAA and approved for publication, it does not necessarily reflect NOAA policies or views.
APPENDIX
Equitable Threat Score
Figure A1 shows the annual average ETS for the NAQFC developmental forecast for surface PM2.5 averaged over the CONUS with approximately 1000 hourly reporting monitors from the AIRNow surface network between 2010 and 2014. The ETS has been evaluated for the thresholds between 0.5 and 35 μg m−3. The ETS skill did not improve noticeably over time, especially for the high concentration thresholds.
Annually averaged CONUS-wide ETS at thresholds between 0.5 and 35 μg m−3 for the NAQFC developmental surface PM2.5 forecast verified against approximately 1000 hourly stations from the AIRNow monitoring network between 2010 and 2014.
Citation: Weather and Forecasting 32, 1; 10.1175/WAF-D-15-0163.1
NAM-Predicted Fields Made Available to CMAQ
Table B1 provides information on the 3D and 2D fields of the output of the NAM output for CMAQ.
NAM-predicted fields made available to CMAQ. Prdgen uses a bilinear interpolation scheme based on the geometrical distance to calculate cross-point values for scalar quantities at midlayers and wind, TKE, and vertical eddy diffusivity values at full layers, except for those variable names labeled with a superscript octothorpe (#), where a nearest-neighbor scheme was used. The variable names that are tagged with a superscript caret (^) are used repeatedly with qualification on vertical levels in accordance with Table 3 (http://www.nco.ncep.noaa.gov/pmb/docs/on388/table3.html). Variable definitions are available online (http://www.nco.ncep.noaa.gov/pmb/docs/on388/table2.html, http://www.nco.ncep.noaa.gov/pmb/docs/on388/table129.html, and http://www.nco.ncep.noaa.gov/pmb/docs/on388/table130.html).
REFERENCES
Aligo, E., Ferrier B. S. , Rogers E. , Pyle M. , Weiss S. J. , and Jirak I. L. , 2014: Modified microphysics for use in high-resolution NAM forecasts. Proc. 27th Conf. on Severe Local Storms, Madison, WI, Amer. Meteor. Soc., 16A.1. [Available online at https://ams.confex.com/ams/27SLS/webprogram/Paper255732.html.]
Alpert, J., 2004: Sub-grid scale mountain blocking at NCEP. Preprints, 20th Conf. on Weather Analysis and Forecasting/16th Conf. on Numerical Weather Prediction, Seattle, WA, Amer. Meteor. Soc., P2.4. [Available online at https://ams.confex.com/ams/pdfpapers/71011.pdf.]
Bey, I., and Coauthors, 2001: Global modeling of tropospheric chemistry with assimilated meteorology: Model description and evaluation. J. Geophys. Res., 106, 23 073–23 095, doi:10.1029/2001JD000807.
Bhave, P., Sarwar G. , Appel W. , and Dennis R. , 2006: Revised treatment of N2O5 hydrolysis in CMAQ. Preprints, Fifth Annual CMAS Conf., Chapel Hill, NC, Community Modeling and Analysis System. [Available online at https://www.cmascenter.org/conference/2006/agenda.cfm.]
Binkowski, F. S., and Shankar U. , 1995: The regional particulate model 1. Model description and preliminary results. J. Geophys. Res., 100, 26 191–26 109, doi:10.1029/95JD02093.
Binkowski, F. S., and Roselle S. J. , 2003: Models-3 Community Multiscale Air Quality (CMAQ) model aerosol component 1. Model description. J. Geophys. Res., 108, 4183, doi:10.1029/2001JD001409.
Briggs, G. A., 1972: Chimney plumes in neutral and stable surroundings. Atmos. Environ., 6, 507–510, doi:10.1016/0004-6981(72)90120-5.
Brutsaert, W., 1982: Evaporation into the Atmosphere: Theory, History, and Applications. D. Reidel, 300 pp.
Byun, D. W., and Dennis R. L. , 1995: Design artifacts in Eulerian air quality models: Evaluation of the effects of layer thickness and vertical profile correction on surface ozone concentrations. Atmos. Environ., 29, 105–126, doi:10.1016/1352-2310(94)00225-A.
Byun, D. W., and Ching J. K. S. , Eds., 1999: Science algorithms of the EPA Models-3 Community Multiscale Air Quality (CMAQ) Modeling System. Tech. Rep. EPA-600/R-99/030, Office of Research and Development, U.S. Environmental Protection Agency. [NTIS PB2000-100561.]
Byun, D. W., and Schere K. L. , 2006: Description of the Models-3 Community Multiscale Air Quality (CMAQ) model: System overview, governing equations, and science algorithms. Appl. Mech. Rev., 59, 51–77, doi:10.1115/1.2128636.
Canty, T. P., and Coauthors, 2015: Ozone and NOx chemistry in the eastern US: Evaluation of CMAQ/CB05 with satellite (OMI) data. Atmos. Chem. Phys., 15, 10 965–10 982, doi:10.5194/acp-15-10965-2015.
Carlton, A. G., Turpin B. J. , Altieri K. E. , Seitzinger S. P. , Mathur R. , Roselle S. J. , and Weber R. J. , 2008: CMAQ model performance enhanced when in-cloud SOA is included: Comparisons of OC predictions with measurements. Environ. Sci. Technol., 42, 8798–8802, doi:10.1021/es801192n.
Chai, T., and Coauthors, 2013: Evaluation of the United States National Air Quality Forecast Capability experimental real-time predictions in 2010 using Air Quality System ozone and NO2 measurements. Geosci. Model Dev., 6, 1831–1850, doi:10.5194/gmd-6-1831-2013.
Chang, J. S., Brost R. A. , Isaksen I. S. A. , Madronich S. , Middleton P. , Stockwell W. R. , and Walcek C. J. , 1987: A three-dimensional Eulerian acid deposition model: Physical concepts and formation. J. Geophys. Res., 92, 14 681–14 700, doi:10.1029/JD092iD12p14681.
Colella, P., and Woodward P. R. , 1984: The piecewise parabolic method (PPM) for gas dynamical simulations. J. Comput. Phys., 54, 174–201, doi:10.1016/0021-9991(84)90143-8.
Davidson, P., 2009: Research need for air quality forecasting from the NOAA perspective. First Int. Workshop on Air Quality Forecasting Research, Boulder CO, NOAA/ESRL. [Available online at http://www.esrl.noaa.gov/csd/events/iwaqfr/presentations/talks/Davidson_Plenary.pdf.]
Eckermann, S., 2009: Hybrid σ–p coordinate choices for a global model. Mon. Wea. Rev., 137, 224–245, doi:10.1175/2008MWR2537.1.
Edney, E. O., Kleindienst T. E. , Lewandowski M. , and Offenberg J. H. , 2007: Updated SOA chemical mechanism for the Community Multi-Scale Air Quality model. Tech. Rep. EPA-600/X-07/025, U.S. EPA, Research Triangle Park, NC, 50 pp.
Ek, M. B., Mitchell K. E. , Lin Y. , Rogers E. , Grunmann P. , Koren V. , Gayno G. , and Tarpley J. D. , 2003: Implementation of Noah land surface model advances in the National Centers for Environmental Prediction operational mesoscale Eta model. J. Geophys. Res., 108, 8851, doi:10.1029/2002JD003296.
EPA, 2005: National air toxics assessment. [Available online at http://www.epa.gov/ttn/atw/natamain/.]
EPA, 2015a: AIRNow—Air quality monitor maps. [Available online at http://www.airnow.gov/index.cfm?action=airnow.pointmaps.]
EPA, 2015b: National Ambient Air Quality Standards for Ozone—Proposed rule. Fed. Regist., 79, 75 234–75 411. [Available online at http://www.gpo.gov/fdsys/pkg/FR-2014-12017/pdf/2014-28674.pdf.]
Fann, N., Lamson A. D. , Anenberg S. , Wesson K. , Risley D. , and Hubbell B. J. , 2012: Estimating the national public health burden associated with exposure to ambient PM2.5 and ozone. Risk Anal., 32, 81–95, doi:10.1111/j.1539-6924.2011.01630.x.
Gandin, L. S., and Murphy A. H. , 1992: Equitable skill scores for categorical forecasts. Mon. Wea. Rev., 120, 361–370, doi:10.1175/1520-0493(1992)120<0361:ESSFCF>2.0.CO;2.
Hsu, Y., and Divita F. Jr., 2011: SPECIATE 4.3: Addendum to SPECIATE4.2: Speciation data base development documentation. EPA Tech. Rep. EPA/600/R-11/121, 27 pp. [Available online at https://www3.epa.gov/ttn/chief/software/speciate/speciate4/addendum4.2.pdf.]
Janjić, Z. I., 1990: The step-mountain coordinates: Physical package. Mon. Wea. Rev., 118, 1429–1443, doi:10.1175/1520-0493(1990)118<1429:TSMCPP>2.0.CO;2.
Janjić, Z. I., 1994: The step-mountain eta coordinate model: Further developments of the convection, viscous sublayer, and turbulence closure schemes. Mon. Wea. Rev., 122, 927–945, doi:10.1175/1520-0493(1994)122<0927:TSMECM>2.0.CO;2.
Janjić, Z. I., 2000: Comments on “Development and evaluation of a convection scheme for use in climate models.” J. Atmos. Sci., 57, 3686, doi:10.1175/1520-0469(2000)057<3686:CODAEO>2.0.CO;2.
Janjić, Z. I., and Gall R. , 2012: Scientific documentation of the NCEP Nonhydrostatic Multiscale Model on the B grid (NMMB). Part 1 Dynamics. NCAR Tech. Note NCAR/TN-489+STR, 75 pp. [Available online at https://opensky.ucar.edu/islandora/object/technotes%3A502/datastream/PDF/view.]
Janjić, Z. I., Gerrity J. P. Jr., and Nickovic S. , 2001: An alternative approach to nonhydrostatic modeling. Mon. Wea. Rev., 129, 1164–1178, doi:10.1175/1520-0493(2001)129<1164:AAATNM>2.0.CO;2.
Kelly, J. T., Bhave P. V. , Nolte C. G. , Shankar U. , and Foley K. M. , 2009: Simulating emission and chemical evolution of coarse sea-salt particles in the Community Multiscale Air Quality (CMAQ) model. Geosci. Model Dev., 2, 1335–1374, doi:10.5194/gmdd-2-1335-2009.
Larkin, N. K., and Coauthors, 2009: The BlueSky smoke modeling framework. Int. J. Wildland Fire, 18, 906–920, doi:10.1071/WF07086.
Lee, P., and Ngan F. , 2011: Coupling of important physical processes in the planetary boundary layer between meteorological and chemistry models for regional to continental scale air quality forecasting: An overview. Atmosphere, 2, 464–483, doi:10.3390/atmos2030464.
Lee, P., and Coauthors, 2009: Impact of consistent boundary layer mixing approaches between NAM and CMAQ. Environ. Fluid Mech., 9, 23–42, doi:10.1007/s10652-008-9089-0.
Lefsky, M. A., 2010: A global forest canopy height map from the Moderate Resolution Imaging Spectroradiometer and the Geoscience Laser Altimeter System. Geophys. Res. Lett., 37, L15401, doi:10.1029/2010GL043622.
Mathur, R., Yu S. , Kang D. , and Schere K. L. , 2008: Assessment of the wintertime performance of developmental particulate matter forecasts with the Eta-Community Multiscale Air Quality modeling system. J. Geophys. Res., 113, D02303, doi:10.1029/2007JD008580.
Mitchell, K. E., and Coauthors, 2004: The multi-institution North American Land Data Assimilation System (NLDAS): Utilizing multiple GCIP products and partners on a continental distributed hydrological modeling system. J. Geophys. Res., 109, D07S90, doi:10.1029/2003JD003823.
Mlawer, E. J., Taubman S. J. , Brown P. D. , Iacono M. J. , and Clough S. A. , 1997: Radiative transfer for inhomogeneous atmosphere: RTTM, a validated correlated-k model for the longwave. J. Geophys. Res., 102, 16 663–16 682, doi:10.1029/97JD00237.
Nenes, A., Pandis S. N. , and Pilinis C. , 1998: ISORROPIA: A new thermodynamic equilibrium model for multiphase multicomponent inorganic aerosols. Aquat. Geochem., 4, 123–152, doi:10.1023/A:1009604003981.
O’Neill, S. M., and Coauthors, 2009: Regional real-time smoke prediction systems. Wildland Fires and Air Pollution, A. Bytnerowicz et al., Eds., Developments in Environmental Science, Vol. 8, Elsevier, 499–534.
Otte, T. L., and Pleim J. E. , 2010: The Meteorology-Chemistry Interface Processor (MCIP) for the CMAQ modeling system: Updates through MCIPv3.4.1. Geosci. Model Dev., 3, 243–256, doi:10.5194/gmd-3-243-2010.
Otte, T. L., and Coauthors, 2005: Linking the Eta Model with the Community Multiscale Air Quality (CMAQ) modeling system to build a National Air Quality Forecasting System. Wea. Forecasting, 20, 367–384, doi:10.1175/WAF855.1.
Owen, P. R., 1964: Saltation of uniform grains in air. J. Fluid Mech., 20, 225–242, doi:10.1017/S0022112064001173.
Pan, L., Tong D. Q. , Lee P. , Kim H. , and Chai T. , 2014: Assessment of NOx and O3 forecasting performances in the U.S. National Air Quality Forecasting Capability before and after the 2012 major emissions updates. Atmos. Environ., 95, 610–619, doi:10.1016/j.atmosenv.2014.06.020.
Pasquill, F., and Smith F. B. , 1983: Atmospheric Diffusion. 3rd ed. E. Horwood, 437 pp.
Pleim, J. E., 2007a: A combined local and nonlocal closure model for the atmospheric boundary layer. Part I: Model description and testing. J. Appl. Meteor. Climatol., 46, 1383–1395, doi:10.1175/JAM2539.1.
Pleim, J. E., 2007b: A combined local and nonlocal closure model for the atmospheric boundary layer. Part II: Application and evaluation in a mesoscale meteorological model. J. Appl. Meteor. Climatol., 46, 1396–1409, doi:10.1175/JAM2534.1.
Pleim, J. E., Xiu A. , Finkelstein P. L. , and Clarke J. F. , 1997: Evaluation of a coupled land-surface and dry deposition model through comparison to field measurements of surface heat, moisture, and ozone fluxes. Preprints, 12th Symp. on Boundary Layer and Turbulence, Vancouver, BC, Canada, Amer. Meteor. Soc., 478–479.
Ruminski, M., Kondragunta S. , Draxler R. , and Zeng J. , 2006: Recent change to the Hazard Mapping System. Preprints, 15th Int. Emission Inventory Conf.: Reinventing Inventories—New Ideas in New Orleans, New Orleans, LA, NOAA/Office of Air Quality Planning and Standards. [Available online at https://www3.epa.gov/ttn/chief/conference/ei15/finalprogram.pdf.]
Saylor, R. D., and Stein A. F. , 2012: Identifying the causes of differences in ozone production from the CB05 and CBMIV chemical mechanisms. Geosci. Model Dev., 5, 257–268, doi:10.5194/gmd-5-257-2012.
Schwede, D., Pouliot G. , and Pierce T. , 2005: Changes to the Biogenic Emissions Inventory System Version 3 (BEIS3). Preprints, Fourth CMAS Models-3 Users’ Conf., Chapel Hill, NC, Community Modeling and Analysis System. [Available online at https://www.cmascenter.org/conference/2005/abstracts/2_7.pdf.]
Stajner, I., Davidson P. , Byun D. , McQueen J. , Draxler R. , Dickerson P. , and Meagher J. , 2012a: US National Air Quality Forecast Capability: Expanding coverage to include particulate matter. Air Pollution Modeling and Its Application XXI, D. G. Steyn and S. T. Castelli, Eds., Springer, 379–384, doi:10.1007/978-94-007-1359-8_64.
Stajner, I., McQueen J. , Lee P. , Draxler R. , Dickerson P. , and Wedmark K. , 2012b: Recent performance of the National Air Quality Forecast Capability. Preprints, 11th Annual CMAS Conf., Chapel Hill, NC, Community Modeling and Analysis System, 379–384.
Stajner, I., and Coauthors, 2014: National Air Quality Forecast Capability: Recent progress and plans. Preprints, 13th Annual CMAS Conf., Chapel Hill, NC, Community Modeling and Analysis System. [Available online at https://www.cmascenter.org/conference/2014/agenda.cfm.]
Tager, I. B., Balmes J. , Lurmann F. , Ngo L. , Alcorn S. , and Künzli N. , 2005: Chronic exposure to ambient ozone and lung function in young adults. Epidemiology, 16, 751–759, doi:10.1097/01.ede.0000183166.68809.b0.
Tang, Y., and Coauthors, 2009: The impact of lateral boundary conditions on CMAQ predictions over the continental United States. Environ. Fluid Mech., 9, 43–58, doi:10.1007/s10652-008-9092-5.
Tong, D. Q., Bowker G. E. , He S. , Mathur R. , and Gillette D. A. , 2015a: Development of a windblown dust module within the Community Multiscale Air Quality (CMAQ) model: Description and preliminary applications in the continental United States. Atmos. Environ., 107, 70–84, doi:10.1016/j.atmosenv.2015.01.035.
Tong, D. Q., and Coauthors, 2015b: Long-term NOx trends over large cities in the United States during the 2008 recession: Intercomparison of satellite retrievals, ground observations, and emission inventories. Atmos. Environ., 107, 70–84, doi:10.1016/j.atmosenv.2015.01.035.
Wu, W.-S., Purser R. J. , and Parrish D. F. , 2002: Three-dimensional variational analysis with spatially inhomogeneous covariances. Mon. Wea. Rev., 130, 2905–2916, doi:10.1175/1520-0493(2002)130<2905:TDVAWS>2.0.CO;2.
Yarwood, G., Rao S. , Yocke M. , and Whitten G. , 2005: Updates to the carbon bond chemical mechanism: CB05. Final Rep. to the US EPA, RT-0400675. [Available online at http://www.camx.com/publ/pdfs/cb05_final_report_120805.pdf.]
Zobler, L., 1999: Global soil types, 1-degree grid (Zobler). Oak Ridge National Laboratory Distributed Active Archive Center, Oak Ridge, TN, doi:10.3334/ORNLDAAC/418.