Atmospheric Environment (v.40, #26)

A performance evaluation of the 2004 release of Models-3 CMAQ by Brian Eder; Shaocai Yu (4811-4824).
This performance evaluation compares a full annual simulation (2001) of Community Multi-scale Air Quality (CMAQ) (Version 4.4) covering the contiguous United States against monitoring data from four nationwide networks. This effort, which represents one of the most spatially and temporally comprehensive performance evaluations of the model, reveals that CMAQ varies considerably in its ability to simulate ambient air concentrations of critical gas and particulate matter species. Simulations of the peak 1- and 8-h ozone (O3) concentrations during the “O3 season” (April–September, 2001) were relatively good (correlation (r)=0.68, 0.69; normalized mean bias (NMB)=4.0%, 8.1%; and normalized mean error (NME)=18.3%, 19.6%, respectively). The annual simulation of sulfate (SO4 2−) was also good ( 0.77 ⩽ r ⩽ 0.92 , depending upon network) with relatively small error (25.0%⩽NME⩽42.0%), though slightly negatively biased (−2.0%⩽NMB⩽−10.0%). The quality of ammonium (NH4 +) simulations is similar to that of SO4 2− ( 0.56 ⩽ r ⩽ 0.79 ; −4.0%⩽NMB⩽14.0%; 35.0%⩽NME⩽63.0%). Simulations of nitrate (NO3 ), elemental carbon (EC) and organic carbon (OC) are relatively poor, as compared to the simulations of the other species. For NO3 , the simulation resulted in: 0.37 ⩽ r ⩽ 0.62 ; −16.0%⩽NMB⩽4.0%; 80.0%⩽NME⩽94.0%. For the carbon species, the r ranged from 0.35 (OC) to 0.47 (EC), with fairly large amounts of error (NME=68.0% for OC, 58.0% for EC) though small amounts of bias (NMB=−6.0% for EC and 12% for OC). The quality of the PM2.5 simulations, like PM2.5 itself, represented a compilation of the quality of all of the simulated particulate species (0.51⩽r⩽0.70, NMB=−3.0% and 45.0%⩽NME⩽46.0%).
Keywords: Community multiscale air quality (CMAQ) model; Performance evaluation; Ozone; PM2.5; Sulfate; Nitrate; Ammonium; Organic and elemental carbon;

A comprehensive model evaluation has been conducted for MM5-CMAQ for the period of 1–10 July 1999 of the Southern Oxidants Study episode with a coarse-grid horizontal spacing of 32-km and a nested fine-grid spacing of 8-km. Meteorological and chemical predictions from simulations with both grids are compared with observations from both routine monitoring networks (e.g., CASTNet, IMPROVE, and AIRS-AQS) and special studies (e.g., SOS99/SOS99NASH, SEARCH, and ARIES). In this Part I paper, five simulated meteorological variables (i.e., temperature, relative humidity (RH), wind speed, wind direction, and planetary boundary layer (PBL) height) are evaluated. While MM5 reproduces well the diurnal variations for temperature and RH, and the minimum temperatures at all sites, it tends to overpredict maximum temperatures and underpredict both maximum and minimum RHs on most days at most sites. MM5 predictions agree reasonably well for wind speeds but poorly for wind direction and the maximum mixing depths. The significant overpredictions in the PBL heights can be attributed to the positive biases of the maximum 2-m temperatures and to the Medium Range Forecast (MRF) model PBL scheme and the Oregon State University (OSU)-Land Surface Model used in the MM5 simulations. For wind speed/direction and the U- and V-component of the wind speed, the normalized mean bias (NMB) and the normalized mean bias factor (NMBF) are the most robust statistical measures because of the dominance of the extremely small observed values in the normalization for those variables.
Keywords: Model evaluation; CMAQ; MM5; SOS; Statistical measures;

A comprehensive performance evaluation of MM5-CMAQ for the Summer 1999 Southern Oxidants Study episode—Part II: Gas and aerosol predictions by Yang Zhang; Ping Liu; Ashley Queen; Chris Misenis; Betty Pun; Christian Seigneur; Shiang-Yuh Wu (4839-4855).
Gas and aerosol predictions from CMAQ simulations with horizontal grid spacings of 8- and 32-km are evaluated against available observations from CASTNet, IMPROVE, AIRS-AQS, SOS99/SOS99NASH, SEARCH, and ARIES for the southeastern US for the period of 1–10 July 1999. The predictions evaluated in this work include mixing ratios of O3 (hourly, maximum 1-h, and 8-h average), NO x , HNO3, NO y , and mass concentrations of PM10, PM2.5, and PM2.5 components.Our evaluation has shown that CMAQ tends to underpredict maximum 1-h O3 mixing ratios on high O3 days at some sites. It overpredicts the maximum and minimum hourly O3 mixing ratios for most low O3 days, the daytime and nighttime hourly, and the maximum 8-h average O3 mixing ratios on most days at all sites. The model performance for hourly O3 mixing ratios generally meets EPA's criteria but deteriorates for maximum 1- and 8-h average O3 mixing ratios. CMAQ underpredicts the mass concentrations of PM10, PM2.5, and PM2.5 composition and fails to reproduce their temporal variations (except for sulfate). Largest underpredictions occur for organic matter (OM2.5) and nitrate2.5 among all PM components. These underpredictions and overpredictions may be caused by inaccurate meteorological predictions (e.g., the PBL height, wind speed/direction, vertical mixing, temperature, and relative humidity) and boundary conditions for chemical species (e.g., O3), underestimation in emissions (e.g., NO x , NH3, and primary OM), as well as uncertainties in model assumptions and treatments in aerosol chemistry and microphysics.
Keywords: Model evaluation; CMAQ; SOS; PM2.5; Ozone;

As Part III of a comprehensive evaluation of CMAQ for the summer 1999 Southern Oxidants Study episode, the observed number (N), volume (V), surface area (S), and size distributions of accumulation-mode particles during the Aerosol Research Inhalation Epidemiological Study (ARIES) are used to evaluate CMAQ's capability in reproducing PM size distributions. CMAQ underpredicts V, S, and geometric number mean diameter (by a factor of 1.24–1.54), and overpredicts N, geometric standard deviation, and geometric volume mean diameter (by a factor of 1.46–2.2) on most days. In addition to inaccurate meteorology and emissions, insufficient condensational growth of PM and uncertainty in the initial size distribution may contribute to the underpredictions in V and S. An overestimation of the PM number emission rates (by a factor of 3–5.3) and several other model assumptions/treatments may contribute to the PM number overpredictions.Among the factors that we studied, the floor value of Kzz , the boundary conditions (BCONs) of O3, the emissions of gaseous precursors such as NO x and NH3 and primary PM species such as POM, and the assumed initial PM size distribution and emission fractions have been identified to be the most influential factors that affect the overall model performance. Sensitivity simulations with a floor value of Kzz of 0.1 cm2  s−1, adjusted emissions of NO x , NH3, and POM, and adjusted initial PM size distribution and emission fractions provide moderate-to-significant improvements. Further investigation into the uncertainties/deficiencies in model treatments for PM such as gas-to-particle mass transfer will identify additional causes for discrepancies between observations and predictions.
Keywords: CMAQ; SOS; Sensitivity study; Ozone; PM2.5; Particle size distribution;

The Models-3 Community Multi-scale Air Quality (CMAQ) modeling system with meteorological fields calculated by the Regional Atmospheric Modeling System (RAMS) was applied to East Asia to investigate the transport and photochemical transformation of tropospheric ozone during the Transport and Chemical Evolution over the Pacific (TRACE-P) field campaign. Modeled concentrations of hydroxyl radical, hydroperoxyl radical, nitric oxide, nitrogen dioxide, ethene, ethane, carbon monoxide, and ozone were compared with observations obtained onboard of two aircrafts in order to evaluate the model performance. Comparison indicates that the model reproduced the tempo-spatial distributions of ozone and its related chemical species reasonably well, and most model results were within a factor of 2 of the observations.
Keywords: Chemical transport model; Long-range transport; Tropospheric ozone; Hydroxyl radical; East Asia;

A Bayesian statistical approach for the evaluation of CMAQ by Jenise L. Swall; Jerry M. Davis (4883-4893).
Bayesian statistical methods are used to evaluate Community Multiscale Air Quality (CMAQ) model simulations of sulfate aerosol over a section of the eastern US for 4-week periods in summer and winter 2001. The observed data come from two U.S. Environmental Protection Agency data collection networks. The statistical methods used here address two problems that arise in model evaluation: the sparseness of the observational data which is to be compared to the model output fields and the comparison of model-generated grid cell averages with point-referenced monitoring data. A Bayesian hierarchical model is used to estimate the true values of the sulfate concentration field. Emphasis is placed on modeling the spatial dependence of sulfate over the study region, and then using this dependence structure to estimate average grid cell values for comparison with CMAQ. For the winter period, CMAQ tends to underpredict the sulfate concentrations over a large portion of the region. The CMAQ simulations for the summer period do not show this systematic underprediction of sulfate concentrations.
Keywords: Spatial analysis; Bayesian methods; Air quality model; Aerosol sulfate; Model validation;

An operational evaluation of the Eta–CMAQ air quality forecast model by Brian Eder; Daiwen Kang; Rohit Mathur; Shaocai Yu; Ken Schere (4894-4905).
The National Oceanic and Atmospheric Administration (NOAA), in partnership with the United States Environmental Protection Agency (EPA), are developing an operational, nationwide Air Quality Forecasting (AQF) system. An experimental phase of this program, which couples NOAA's Eta meteorological model with EPA's Community Multiscale Air Quality (CMAQ) model, began operation in June of 2004 and has been providing forecasts of ozone (O3) concentrations over the northeastern United States. An important component of this AQF system has been the development and implementation of an evaluation protocol. Accordingly, a suite of statistical metrics that facilitates evaluation of both discrete- and categorical-type forecasts was developed and applied to the system in order to characterize its performance. The results reveal that the AQF system performed reasonably well in this inaugural season (mean domain wide correlation coefficient=0.59), despite anomalously cool and wet conditions that were not conducive to the formation of O3. Due in part to these conditions, the AQF system overpredicted concentrations, resulting in a mean bias of +10.2 ppb (normalized mean bias=+22.8%). In terms of error, the domain-wide root mean square error averaged 15.7 ppb (normalized mean error=28.1%) for the period. Examination of the discrete and categorical metrics on a daily basis revealed that the AQF system's level of performance was closely related to the synoptic-scale meteorology impacting the domain. The model performed very well during periods when anticyclones, characterized by clear skies, dominated. Conversely, periods characterized by extensive cloud associated with fronts and/or cyclones, resulted in poor model performance. Subsequent analysis revealed that factors associated with CMAQ's cloud cover scheme contributed to this overprediction. Accordingly, changes to the cloud schemes are currently underway that are expected to significantly improve the AQF system's performance in anticipation of its second year of operation.
Keywords: Air quality forecasting; Ozone; Model evaluation; Community Multiscale Air Quality (CMAQ) model; Eta model;

CMAQ/CAMx annual 2002 performance evaluation over the eastern US by T.W. Tesche; Ralph Morris; Gail Tonnesen; Dennis McNally; James Boylan; Patricia Brewer (4906-4919).
Operational, diagnostic, and comparative evaluations of two one-atmosphere regional models were performed for the full calendar year 2002 in support of regional haze regulatory applications in the eastern US. Using consistent emissions, meteorological and air quality data sets, the community multi-scale air quality and comprehensive air quality model with extensions models were exercised on a nested 36/12 km grid system and evaluated across a broad range of time and space scales for numerous gas-phase and fine particulate species derived from routine and research-grade ambient measurements at six monitoring networks. Performance by both models for speciated fine particulate matter (PM) across the eastern US ranged from quite good (e.g., SO4 2−) to poor (e.g., soil). For most species, model bias was higher in the winter and lower (usually negative) in the summer suggesting potential issues related to vertical mixing (e.g., too little in winter), temporal allocation of emissions, and/or other model science processes or inputs. These results may be used to (a) guide one-atmosphere model refinements, (b) improve data input preparation procedures, (c) evaluate methods for rigorous, stressful performance testing, and (d) clarify the uncertainty in model estimates for regional haze and PM2.5 control strategy programs.
Keywords: Model evaluation; Visibility; CMAQ; CAMx; Fine particulate; VISTAS;

An objective comparison of CMAQ and REMSAD performances by Edith Gégo; P. Steven Porter; Christian Hogrefe; John S. Irwin (4920-4934).
Photochemical air quality modeling systems are the primary tools used in regulatory applications to assess the impact of different emission reduction strategies aimed at reducing air pollutant concentrations to levels considered safe for public health. Two such modeling systems are the community multiscale air quality (CMAQ) model and the regional modeling system for aerosols and deposition (REMSAD). To facilitate their inter-comparison, the United States Environmental Protection Agency performed simulations of air quality over the contiguous United States during year 2001 (horizontal grid cell size of 36×36 km) with CMAQ and REMSAD driven by identical emission and meteorological fields. Here, we compare the abilities of CMAQ and REMSAD to reproduce measured aerosol nitrate and sulfate concentrations. Model estimates are compared to observations reported by the interagency monitoring of protected visual environment (IMPROVE) and the clean air status and trend network (CASTNet). Root mean squared errors are calculated for simulation/observation pairs from ten geographic regions and 12 seasons (months). Following the application of the Wilcoxon matched-pair signed rank test, we conclude that CMAQ is more skillful than REMSAD for simulation of aerosol sulfate. Simulations of particulate nitrate concentrations by CMAQ and REMSAD can seldom be differentiated, leading to the conclusion that both models perform equally for this pollutant specie.
Keywords: Aerosol sulfate; Aerosol nitrate; Wilcoxon signed rank test; Evaluation metric; Photochemical model;

The general situation (but exemplified in urban areas), where a significant degree of sub-grid variability (SGV) exists in grid models poses problems when comparing grid-based air-quality modeling results with observations. Typically, grid models ignore or parameterize processes and features that are at their sub-grid scale. Also, observations may be obtained in an area where significant spatial variability in the concentration fields exists. Consequently, model results and observations cannot be expected to be equal. To address this issue, we suggest a framework that can provide for qualitative judgments on model performance based on comparing observations to the grid predictions and its SGV distribution. Further, we (a) explore some characteristics of SGV, (b) comment on the contributions to SGV and (c) examine the implications to the modeling results at coarse grid resolution using examples from fine scale grid modeling of the Community Multi-scale Air Quality (CMAQ) modeling system.
Keywords: Neighborhood-scale models; CMAQ fine scale modeling; Sub-grid distributions; Sub-grid variability; Multiscale air-quality modeling;

In order to use an air quality modeling system with confidence, model performance must be evaluated against observations. While ozone modeling and evaluation is fairly developed, particulate matter (PM) modeling is still an evolving science. EPA has issued minimal guidance on PM and visibility model performance evaluation metrics, goals, and criteria. This paper addresses these issues by examining various bias and error metrics and proposes PM model performance goals (the level of accuracy that is considered to be close to the best a model can be expected to achieve) and criteria (the level of accuracy that is considered to be acceptable for modeling applications) that vary as a function of concentration and extinction. In this paper, it has been proposed that a model performance goal has been met when both the mean fractional error (MFE) and the mean fractional bias (MFB) are less than or equal to +50% and ±30%, respectively. Additionally, the model performance criteria has been met when both the MFE⩽+75% and MFB⩽±60%. Less abundant species would have less stringent performance goals and criteria. These recommendations are based upon an analysis of numerous PM and visibility modeling studies performed throughout the country.
Keywords: Bias; Error; Particulate matter; Aerosols; Regional haze;

Model sensitivity evaluation for organic carbon using two multi-pollutant air quality models that simulate regional haze in the southeastern United States by Ralph E. Morris; Bonyoung Koo; Alex Guenther; Greg Yarwood; Dennis McNally; T.W. Tesche; Gail Tonnesen; James Boylan; Patricia Brewer (4960-4972).
Photochemical grid models are being used in technical analyses by the Visibility Improvement State and Tribal Association of the Southeast (VISTAS), a regional air quality planning organization in the southeastern United States, to support state implementation plans for regional haze and related air quality issues. VISTAS has embarked on a multi-phase process of testing and evaluating regional meteorological, emissions and air quality models that will be used to project visibility improvements as required by the regional haze rule. VISTAS has generated 2002 annual emissions and meteorological inputs for two photochemical grid models, the community multi-scale air quality (CMAQ) and the comprehensive air-quality model with extensions (CAMx), at a 36 km resolution for the continental US and at 12 km resolution for the eastern US. The two models were evaluated using speciated PM measurements from various monitoring networks and detailed analysis was performed for organic carbon (OC) mass using the IMPROVE, STN, and SEARCH networks. The differences in model performance between CMAQ and CAMx were used as a diagnostic tool to investigate performance issues for several compounds. CAMx performed substantially better than CMAQ for OC (defined as 1.4×measured organic carbon) which led to investigations into methods for improving the CMAQ OC model performance. The treatment of secondary organic aerosol (SOA) was identified as an area needing improvements in both models. The impact of replacing the CMAQ SOA parameters with those from CAMx was investigated. Further analysis identified several processes that are potentially important for SOA formation that are not treated in either model including, polymerization of the SOA into non-volatile particles and SOA formation from sesquiterpene, isoprene and other biogenic VOCs. A prototype mechanism for several of these missing processes was developed and the CMAQ SOA module was enhanced to include these SOA formation processes. SOA yields, specifically from biogenic emissions, were increased by the modified SOA module and CMAQ model performance for particulate OC at the IMPROVE, SEARCH, and STN sites in the VISTAS region was improved.
Keywords: Visibility; Fine particulate; Organic carbon; Secondary organic aerosol; CMAQ; CAMx;

This paper calculates PM2.5 mass concentrations (PM2.5) from CMAQ output quantities, and focuses on analysing the uncertainty caused by using mass concentrations of particles in the two CMAQ fine modes (PM i+j ) as approximations of PM2.5 in model evaluations. Conceptually, CMAQ fine mode and PM2.5 particles are different in terms of both concentrations and compositions. Quantitatively, the modelling results of the Pacific 2001 scenario at the Lower Fraser Valley shows that PM i+j and PM2.5 can be substantially or even extremely different, depending on relative humidity (RH) values. Under low RHs, PM i+j and PM2.5 generally correlate well and their quantitative differences are mostly moderate, although their maximum differences can still be substantial. Under high RHs, the correlation between PM i+j and PM2.5 deteriorates considerably and the quantitative differences increase dramatically. This is true whether the analysis is conducted on an all-component or a dry-component-only basis. Therefore, PM i+j could be used as an approximation of PM2.5 only on an average basis when RHs are low, but not under more general conditions. When compared with measured PM2.5 concentrations, the modelled concentrations of PM2.5 dry components (PM2.5,dry) performed much better than the modelled concentrations of fine mode dry components (PM i + j ,dry) for the modelling domain and period, since the overall positive bias of the modelled PM i + j ,dry was partially compensated by the lower PM2.5,dry values in comparison with PM i + j ,dry. In addition, by using PM2.5,dry, the model demonstrated a better skill in simulating 24-h moving averages (MA) of measured concentrations in comparison with simulating hourly concentrations. This is different from the case of using PM i + j ,dry, where the model could not clearly show a better skill in simulating 24-h MAs. Therefore, it is highly desirable to calculate PM2.5 values from CMAQ output and use them instead of PM i+j in model evaluations, especially under situations when RHs can be high. The method outlined in this paper can also be readily used for the calculations of PM concentrations at any cut-off diameter of interest, in addition to PM2.5 discussed here.
Keywords: Particulate matter; Air quality; Modelling; Aerosol; Size distribution;

Seasonal NH3 emissions for the continental united states: Inverse model estimation and evaluation by Alice B. Gilliland; K. Wyat Appel; Robert W. Pinder; Robin L. Dennis (4986-4998).
Significant uncertainty exists in the seasonal distribution of NH3 emissions since the predominant sources are animal husbandry and fertilizer application. Previous studies that estimated bottom–up and top–down NH3 emissions have provided the most comprehensive information available about the seasonality of NH3 emissions. In this study, this bottom–up and top–down emission information is combined with the most recent 2001 USEPA National Emission Inventory (NEI) to construct a best prior estimate of seasonal NH3 emissions. These emission estimates are then used in an annual 2001 USEPA Community Multiscale Air Quality (CMAQ) model simulation for the continental United States. A key objective of this study is to evaluate these prior NH3 emission estimates and test the top–down inverse modeling method for a different year and a larger modeling domain than used previously. Based on the final posterior NH3 emission estimates, the inverse modeling results suggest that the annual total NEI NH3 emissions are reasonable and that a previous high bias in older USEPA emission inventories has been addressed in the updated inventory. Inverse modeling results suggest that the prior NH3 emission estimates should be increased in the summer and decreased in the winter, while results for the spring and fall are questionable due to precipitation prediction biases. A final conclusion from this study is that total NH x (NH3 and aerosol NH4 +) air concentration data are essential for quantitative top–down analyses of NH3 emissions that can extend beyond what is possible using precipitation chemistry data.
Keywords: NH3 emissions; Inverse modeling; Top–down emission estimates; Air quality modeling; Seasonal variability;

Comparison of spatial patterns of pollutant distribution with CMAQ predictions by Sharon B. Phillips; Peter L. Finkelstein (4999-5009).
To evaluate the Models-3/Community Multiscale Air Quality (CMAQ) modeling system in reproducing the spatial patterns of aerosol concentrations over the country on timescales of months and years, the spatial patterns of model output are compared with those derived from observational data. Simple spatial interpolation procedures were applied to data from the Clean Air Status and Trends Network (CASTNet) and Speciation Trends Network (STN) monitoring networks. Species included sulfate PM, total nitrate ( NO 3 - + HNO 3 ) , and ammonium PM. Comparisons were made for the annual average concentrations for 2001, and for one lunar month (4 weeks), where the month chosen for each species represents the highest concentrations of the year. Comparisons between the modeled and interpolated spatial patterns show very good agreement in the location and magnitude of the maxima and minima, as well as the gradients between them. Some persistent biases are identified and noted. Limitations on our ability to describe the spatial pattern from sparse data as well as the limitations of the networks are briefly discussed.
Keywords: CMAQ; Spatial statistical analysis; Model evaluation; Air pollution; Air quality;

A quantitative assessment of the influence of grid resolution on predictions of future-year air quality in North Carolina, USA by Saravanan Arunachalam; Andrew Holland; Bebhinn Do; Michael Abraczinskas (5010-5026).
Increased focus has been directed at fine-scale modeling for improving the ability of air quality modeling systems to capture local phenomena. While numerous studies have investigated model performance at finer resolution (4–5 km), there is relatively limited information available for choosing the optimum grid resolution for predicting future air quality in attainment demonstration studies. We demonstrate an evaluation of the MM5–SMOKE–MAQSIP modeling system for four 8-h ozone episodes in the summers of 1995, 1996 and 1997 in North Carolina using a one-way nested 36/12/4-km application. After establishing acceptable base-case model performance for ozone predictions during each episode, we developed future-year emissions control scenarios for 2007 and 2012, and finally computed relative reduction factors (RRFs) using model outputs from each of the three grid resolutions. Our analyses, based upon qualitative as well as quantitative approaches like the Student's t-test, indicate that RRFs computed at specific monitoring locations—and hence predicted future-year air quality—are not very different between the 4- and 12-km results, while the differences are slightly larger between the 4- and 36-km results. The results imply that grid resolution contributes to a variability of about 1–3 ppb in the projected future-year design values; this variability needs to be incorporated into policy-relevant decision-making. Since this assessment was performed for four different episodes under diverse meteorological, physical and chemical regimes, one can generalize the results from this study. They are also relevant for regional modeling applications that are currently ongoing for studying PM2.5 nonattainment issues, where the need for annual base-year and future-year simulations for demonstrating attainment may place a large demand on computing resources. Based upon the results from this study, future studies may consider using results from 12-km modeling to address future-year air quality goals for ozone and PM2.5 and its components, and then incorporate grid-resolution uncertainties into the computed results.
Keywords: Emissions strategies; CMAQ; Student's t-test; O3 NAAQS; Attainment;

CMAQ was run to simulate urban and regional tropospheric conditions in the southeastern US over 14 days in July 1999 at 32, 8 and 2 km grid spacings. Runs were made with either of two older mechanisms, Carbon Bond IV (CB4) and the Regional Acid Deposition Model, version 2 (RADM2), and with the more recent and complete California Statewide Air Pollution Research Center, version 1999 mechanism (SAPRC99) in a sensitivity matrix with a full emissions base case and separate 50% control scenarios for emissions of nitrogen oxides (NO X ) and volatile organic compounds (VOC). Results from the base case were compared to observations at the Southeastern Aerosol Research and Characterization Study (SEARCH) site at Jefferson Street in Atlanta, GA (JST) and the Southern Oxidant Study (SOS) Cornelia Fort Airpark (CFA) site downwind of Nashville, TN. In the base case, SAPRC99 predicted more ozone (O3) than CB4 or RADM2 almost every hour and especially for afternoon maxima at both JST and CFA. Performance of the 8 km models at JST was better than that of the 32 km ones for all chemistries, reducing the 1 h peak bias by as much as 30 percentage points; at CFA only the RADM2 8 km model improved. The 2 km solutions did not show improved performance over the 8 km ones at either site, with normalized 1 h bias in the peak O3 ranging from 21% at CFA to 43% at JST. In the emissions control cases, SAPRC99 was generally more responsive than CB4 and RADM2 to NO X and VOC controls, excepting hours at JST with predicted increased O3 from NO X control. Differential sensitivity to chemical mechanism varied by more than ±10% for NO X control at JST and CFA, and in a similar range for VOC control at JST. VOC control at the more strongly NO X - limited urban CFA site produced a differential sensitivity response of <5%. However, even when differential sensitivities in control cases were small, neither their sign nor their magnitude could be reliably determined from model performance in the full emissions case, meaning that the degree of O3 response to a change in chemical mechanism can differ substantially with the level of precursor emissions. Hence we conclude that properly understanding the effects of changes in a model's chemical mechanism always requires emissions control cases as part of model sensitivity analysis.
Keywords: Model evaluation; Sensitivity analysis; Chemical mechanism; Emissions control strategies; Differential sensitivity;

Temporal features in observed and simulated meteorology and air quality over the Eastern United States by C. Hogrefe; P.S. Porter; E. Gego; A. Gilliland; R. Gilliam; J. Swall; J. Irwin; S.T. Rao (5041-5055).
Over the next several years, grid-based photochemical models such as the community multiscale air quality (CMAQ) model, the regional modeling system for aerosols and deposition (REMSAD), the comprehensive air quality model with extensions (CAMx), and other regional models will be used by regulatory agencies in the United States for designing emission control strategies to meet and maintain the National Ambient Air Quality Standards (NAAQS) for O3, PM2.5, and regional haze. In this study, temporal scale analysis is applied as a technique to evaluate an annual simulation of meteorology, O3, and PM2.5 and its chemical components over the continental US utilizing two modeling systems. The spectral decomposition of total PM2.5 mass from hourly observations and CMAQ and REMSAD model predictions revealed that days of high PM2.5 concentrations are generally characterized by positive forcing from fluctuations having periods equal to or greater than a day (i.e., the diurnal, synoptic, and longer-term components) while the magnitude of intra-day fluctuations showed only small differences between average and episodic conditions. Both modeling systems did not capture most of the variability of the high-frequency, intra-day component for all variables for which hourly measurements were available. Furthermore, it is illustrated that correlations were insignificant on the intra-day time scale for all variables, suggesting that these models in the setup used for this study were not skillful in simulating the higher-frequency variations in meteorological variables and the levels of all pollutants. The models exhibited greatest skills at capturing longer-term (seasonal) fluctuations for temperature, wind speed, O3, sulfate and nitrate. Correlations for total PM2.5, ammonium, elemental carbon (EC), organic carbon (OC) and crustal PM2.5 correlations were highest for the synoptic time scale implying problems with factors other than meteorology, such as emissions or lateral chemical boundary conditions, in capturing the baseline fluctuations.
Keywords: Model evaluation; Air quality modeling; Particulate matter; Spectral analysis; CMAQ;

The complex configuration of the northeastern Iberian Peninsula (NEIP) provokes a complex behavior of photochemical pollutants, which demands a high spatial resolution when applying an air quality model. CMAQ has been used for air quality assessment in the NEIP coupled with the MM5 meteorological model and EMICAT2000 emission model, and has been extensively evaluated against available ambient data during a typical summertime photochemical pollution episode. Simulations with different resolutions were evaluated to select the needed grid resolution. Meteorological inputs are sensitive to the degree of topographical smoothing. Fine-resolution simulations present the best scores during the development of the sea breeze. The performance of statistical parameters for ground-level O3 greatly improves when decreasing the horizontal and vertical grid spacing. Statistical parameters indicate that decreasing the horizontal grid spacing to 2 km greatly improves the critical success index, the false alarm ratio and the probability of detection. Furthermore, sensitivity studies provide the opportunity to check whether O3 values react consistently to similar changes in emissions. The model sensitivity was evaluated by performing simulations to represent O3 formation with baseline emission rates for VOCs and NO x , and reducing anthropogenic VOC and NO x emissions by 35%. Evaluation of ground-level O3 shows a good agreement when the model predicts dominant VOC-sensitive chemistry. Statistical parameters of O3 evaluation worsen when reducing VOCs emissions and improve in the—35% NO x case, indicating that the O3-production chemistry may not be sufficiently reactive.
Keywords: Ozone; Air quality modeling; Model evaluation; Photochemistry; Air pollution;

New methods for evaluating meteorological models used in air quality applications by Robert C. Gilliam; Christian Hogrefe; S.T. Rao (5073-5086).
Meteorological models in conjunction with air quality models are being used to simulate the transport and fate of pollutants in the atmosphere. Hence, there is a need for an extensive evaluation of the entire modeling system. In this study, several new techniques to assess the performance of mesoscale meteorological models are introduced with an emphasis on evaluating the variables and processes that have the potential to influence the air quality predictions, since errors in the meteorological fields are passed on to the air quality model.Model performance was diagnosed by examining the inter-correlation of observable variables in the atmosphere on distinct time scales: intraday, diurnal, and synoptic. It was found that the Mesoscale Model version 5 (MM5) model did replicate the observed relationship between intraday wind speed and temperature, intraday surface pressure and temperature, diurnal surface pressure and temperature as well as most of the correlations between variables on the synoptic timescale. However, a negative correlation between temperature and precipitation was evident in the observations on the intraday scale, but such relationship was not evident in the model output. Furthermore, the diurnal response of increasing wind speed with temperature was strong in the observed time series, but it was much weaker in the model. The correlation between diurnal changes in temperature and cloud fraction was consistently negative in the model whereas it was slightly positive in the observations.Wind profilers were used to examine the simulated boundary layer wind structure. Of the twelve sites examined, the average distance error between the 24-h observed and modeled trajectory was approximately 150 km at height of 100 m above the surface. Errors in transport of this magnitude (100–200 km) can produce errors in air quality predictions. It is not the intent of this study to establish quantitative links between the performance of the specific meteorological simulation analyzed here and subsequent air quality simulations. Rather, the results presented here draw attention to errors and inconsistencies in the meteorology that are passed on to the air quality model which, in turn, have the potential to cause errors in air quality model predictions.
Keywords: Model evaluation; MM5; K–Z filter; Trajectories; Correlation analysis;

The concentrations of five hazardous air pollutants were simulated using the community multi-scale air quality (CMAQ) modeling system. Annual simulations were performed over the continental United States for the entire year of 2001 to support human exposure estimates. Results are shown for formaldehyde, acetaldehyde, benzene, 1,3-butadiene and acrolein. Photochemical production in the atmosphere is predicted to dominate ambient formaldehyde and acetaldehyde concentrations, and to account for a significant fraction of ambient acrolein concentrations. Spatial and temporal variations are large throughout the domain over the year. Predicted concentrations are compared with observations for formaldehyde, acetaldehyde, benzene and 1,3-butadiene. Although the modeling results indicate an overall slight tendency towards underprediction, they reproduce episodic and seasonal behavior of pollutant concentrations at many monitors with good skill.
Keywords: Air toxics; HAPs; Benzene; Formaldehyde; Acrolein;

Plume-in-grid modeling of summer air pollution in Central California by Krish Vijayaraghavan; Prakash Karamchandani; Christian Seigneur (5097-5109).
CMAQ-APT (Community Multiscale Air Quality model with “Advanced Plume Treatment”), a state-of-the-science implementation of sub-grid scale reactive plumes in CMAQ, was used to simulate ozone (O3) and nitric acid (HNO3) formation during a 4-day July/August 2000 episode in central California. The top ten NO x emitting plants in the Central California Ozone Study (CCOS) domain were selected for explicit plume treatment. The VOC- vs. NO x -limited nature of the background environment, as determined from air quality data in different parts of the CCOS domain, was used to understand the differences in ozone production and destruction between the APT and base results. Use of the plume-in-grid treatment results in up to 10 ppb less O3 than the base under some O3 production conditions and up to 6 ppb higher O3 under others. Over most of the areas impacted by the top ten NO x emitting plants, the surface HNO3 concentrations in the APT simulation are about 0.1–1 ppb (1–25%) lower than those in the base simulation. The low NO x emissions from point sources in central California explain these results.
Keywords: Air quality model; Sub-grid; Ozone; Nitric acid; VOC- vs. NO x -limited;