Identifying the causes of di ff erences in ozone production from the CB 05 and CBMIV chemical mechanisms

Introduction Conclusions References


Introduction
The NOAA National Air Quality Forecasting Capability (NAQFC) is a modeling system established in 2004 to provide ground-level ozone (O 3 ) forecasts for the conterminous Introduction

Conclusions References
Tables Figures

Back Close
Full US (CONUS) (Eder et al., 2009).The NAQFC utilizes the North American Mesoscale (NAM) meteorological model run of the Nonhydrostatic Mesoscale Model (NMM) core of the Weather Research and Forecasting (WRF) system (WRF-NMM) (Janjic, 2003) to provide meteorological fields to drive the US Environmental Protection Agency's (EPA) Community Multiscale Air Quality (CMAQ) version 4.6 modeling system (Byun and Schere, 2006) which generates 48 h forecast fields of ground-layer O 3 .The CMAQ portion of the NAQFC system uses 12 km horizontal grid spacing and 22 vertical layers from the surface to 100 hPa.Emission inputs use the latest available EPA National Emissions Inventory with electric generating units (EGUs) updated with Continuous Emission Monitoring (CEM) data and projections to the current forecast year using data from the Department of Energy's Annual Energy Outlook.Further details of the operational NAQFC implementation and evaluations of its performance can be obtained from Eder et al. (2009), Otte et al. (2005), and Yu et al. (2010).Since its inception, the operational version of the NAQFC has used a modified implementation of the Carbon Bond Mechanism version IV (CBMIV, Gery et al., 1989) as its gas-phase chemical mechanism.An experimental version of the NAQFC has been run in parallel with the operational track to provide a testbed for planned upgrades to the operational system.Since 2008, the experimental NAQFC has employed the updated 2005 version of the Carbon Bond Mechanism (CB05, Yarwood et al., 2005) as its gas-phase mechanism.The two parallel NAQFC tracks use identical emissions base inventories (with volatile organic carbon compounds speciated appropriately for each mechanism), identical meteorological fields from NAM, and similar initial and boundary conditions.Figure 1 presents a comparison of domain-wide daily average O 3 biases for 2009 from the two parallel NAQFC systems as compared to US Environmental Protection Agency Air Quality System (AQS) measurements (http://www.epa.gov/ttn/airs/airsaqs/).Both versions of the NAQFC exhibit similar biases throughout the year, climbing from near zero in April to a peak positive bias (i.e., model -measurement) in September.The inherent, similar biases of both NAQFC versions are likely due to a variety of factors, possibly involving anthropogenic Introduction

Conclusions References
Tables Figures

Back Close
Full or biogenic emissions uncertainties, meteorological input biases, dry or wet deposition processes, or chemical processes common to both CB05 and CBMIV.However, the motivation of the work described here is to understand the causes of the relatively consistent difference between the operational and experimental NAQFC versions, given that the emissions and input data for both systems are identical and that they differ only in the gas-phase chemical mechanisms employed.
Previous investigations (Luecken et al., 1999;Sarwar et al., 2008;Luecken et al., 2008) have also observed the differences in O 3 produced from CMAQ simulations for the two mechanisms.These studies have identified mechanistic differences that might have some potential impact on O 3 production, but no overriding explanatory cause emerged.As noted in Luecken et al. (2008), O 3 photochemistry is complex and in the framework of a three-dimensional model simulation, numerous temporally-and spatially-varying factors may contribute to the observed differences in O 3 fields between the two mechanisms.Consequently, in this work we attempt to focus solely on mechanism differences without the confounding effects of additional model processes (e.g., advection, turbulent diffusion, wet and dry deposition, etc.) that may obscure the underlying mechanistic reasons for the observed differences in O 3 concentrations.Using this approach, and recognizing that CB05 is, in some sense, a "modified" version of CBMIV, a relatively simple comparison of the mechanisms was undertaken by performing sensitivity studies with box model versions of each to identify which mechanistic differences between CB05 and CBMIV may account for the O 3 differences observed in full CMAQ simulations.In subsequent sections of this paper, we describe the box model implementation of the mechanisms and the sensitivity studies that were designed to probe mechanism differences and then we present selected results from those sensitivity runs to demonstrate our findings.Introduction

Conclusions References
Tables Figures

Back Close
Full

Box model description
A box model of the convectively well-mixed boundary layer was constructed for each mechanism, CB05 and CBMIV, using a modified implementation of the Kinetic Preprocessor of Sandu and Sander (2006).It should be noted that the versions of CB05 and CBMIV that are used in the NAQFC have been altered from the mechanisms as they were originally published.In this paper, we are only concerned with and refer to CB05 and CBMIV as they have been implemented in the NAQFC system.The mechanism definition files as implemented in the NAQFC are provided as a Supplement to this paper.
After Seinfeld and Pandis (1998), but neglecting surface removal by dry deposition and mixing with background air, the concentration of a gas-phase species in a wellmixed, constant depth boundary layer is given by where C i is the concentration of species i in the boundary layer, q i is the emission rate of species i into the boundary layer, H is the boundary layer depth, and R i is the chemical production (or destruction) rate of species i .The box model for each mechanism consists of a set of equations of the form (1) for each mechanism species.
In an approach fashioned after that of Zaveri and Peters (1999), a suite of 45 box model simulation scenarios was created to exercise the mechanisms across a broad range of chemical regimes, with varying NO x (NO + NO 2 ) and hydrocarbon (HC) emission rates.Environmental conditions and initial chemical concentrations were the same for all scenarios (Table 1) so that the results from each are driven primarily by the specified emission rates (Table 2).The scenarios were designed to have emission rates and HC to NO x ratios that span conditions from urban to rural environments that might be encountered in a CONUS domain simulation of a three-dimensional model.Introduction

Conclusions References
Tables Figures

Back Close
Full For simplicity, emission rates were constrained to vary diurnally in proportion to the cosine of the zenith angle (night = 0 and emission rate maximum, E 0 , at solar noon) for a summer, mid-latitude location near ground level.Each simulation scenario was run for ten days to allow the system to obtain a quasi-stationary chemical state.The total anthropogenic non-methane hydrocarbon (NMHC) emission rate of each scenario was calculated according to the NMHC : NO x ratio given in Table 2 and then apportioned among specific NMHC species in each mechanism as presented in Table 3. Care was taken to ensure that emissions rates of NMHC species were consistent for both CB05 and CBMIV versions of the box models.For species that are unchanged between the two mechanisms, the apportioning is the same in each (e.g., ETH, OLE, TOL, XYL and HCHO).In CB05, the two carbon ethane (ETHA) species was explicitly included, so PAR emissions in CB05 simulations were reduced equal to 2 × the ETHA emissions as compared to the base CBMIV simulation.In CB05, the ALD2 species explicitly represents acetaldehyde and ALDX represents all other higher aldehydes; however, in CBMIV the ALD2 species alone represents acetaldehyde and other higher aldehydes.Consequently, the ALD2 emissions used in the base CBMIV simulations account for both the ALD2 and ALDX emissions included in CB05 simulations.Furthermore, emissions of the internal olefin species, IOLE, are treated as 2 × ALD2 in the base CBMIV simulation (assuming 2-butene as the internal olefin).As a result, ALD2 emissions in the base CBMIV simulation are calculated as Figure 2 presents results from several example scenarios from the box model simulation suite, illustrating how for a given set of emissions rates, the CB05 mechanism almost invariably produces higher O 3 values by the end of the 10-day simulation period.The behavior illustrated here from the box model versions of CBMIV and CB05 is consistent with the behavior observed from the parallel versions of the NAQFC system.Introduction

Conclusions References
Tables Figures

Back Close
Full

Sensitivity tests
Beyond routine updates of reaction rate coefficients, the mechanistic differences between CB05 and CBMIV are extensive but can be summarized briefly into 11 conceptual groups: (1) addition of molecular hydrogen reactions to improve odd-hydrogen chemistry in the upper troposphere; (2) addition of reactions involving odd-oxygen and odd-hydrogen to provide a more complete description of hydroxyl radical (OH) chemistry; (3) addition of nitrate radical (NO 3 ) reactions to improve the representation of nocturnal chemistry in the boundary layer; (4) addition of inorganic NO x recycling reactions that slowly recycle HNO 3 , HO 2 NO 2 , and N 2 O 5 back into NO 2 ; (5) addition of reactions that recycle the lumped organic nitrate species (NTR) back to HNO 3 and NO 2 ; (6) substitution of explicit chemistry for methane (CH 4 ) oxidation and the introduction of the methylperoxy radical (CH 3 O 2 ) instead of the highly parameterized representation in CBMIV; (7) addition of a reaction for OH oxidation of ethane (ETHA) rather than lumping with the PAR species; (8) introduction of a new higher aldehyde species that separates the lumped CBMIV aldehyde species into acetaldehyde (ALD2) and the higher aldehyde (ALDX) and introduces related species such as peroxyacyl radicals (CXO3), peroxynitrates (PANX), carboxylic acids (AACD) and peroxycarboxylic acids (PACD); (9) separation of the CBMIV alkene species into a lumped internal olefin (IOLE) and other olefins (OLE); (10) addition of higher organic peroxides (ROOH) were added to improve the overall representation of peroxide species which are important in aqueous sulfate production; and (11) addition of a new species (TERP) to represent terpene chemistry more explicitly.
The full CB05 mechanism represents the combined impact of these 11 mechanistic updates.Based upon these groups, 12 sensitivity simulations, described in the following sections, were designed to probe the impact of individual differences on overall a regional-to-continental scale simulation.As described below, each sensitivity test attempts to return one of the mechanistic updates to its previous representation in CB-MIV.By then comparing O 3 concentrations from the base CB05, the base CBMIV, and the sensitivity test (i.e., the CB05 mechanism with one particular update rolled back to its CBMIV representation) for each scenario in the simulation suite, we obtain an indication of which particular mechanism change or changes may account for the observed larger O 3 produced by CB05.

xH2
Two reactions were added to the CB05 mechanism that do not appear in CBMIV to account for molecular hydrogen reactions (all reaction numbers refer to the mechanisms as implemented in the NAQFC as provided in the Supplement) For sensitivity test xH2, these reactions were removed from CB05, leaving all other reactions as in the base CB05 mechanism.No emissions changes were made in the 10 scenario suite.

xoddH&O
Six reactions were included in the CB05 mechanism which do not appear in CBMIV in order to better represent odd-oxygen and odd-hydrogen chemistry

Conclusions References
Tables Figures

Back Close
Full For sensitivity test xoddH&O, these reactions were removed from CB05, leaving all other reactions as in the base CB05 mechanism.No emissions changes were made to the scenario suite.

xNO3night
Six reactions were introduced into the CB05 mechanism that are not included in CBMIV in order to better represent NO 3 radical chemistry in the nocturnal boundary layer For sensitivity test xNO3night, these reactions were removed from CB05, leaving all other reactions as in the base CB05 mechanism.No emissions changes were made to the scenario suite.

xETHA
A reaction was introduced in CB05 to account for the explicit reaction of ethane (ETHA) with OH ETHA + OH → 0.991ALD2 + 0.991XO2 + 0.009XO2N + HO 2 (R155) In sensitivity test xETHA, this reaction was removed and all ETHA emissions in the scenario suite were converted to the equivalent PAR emissions as in CBMIV.Introduction

Conclusions References
Tables Figures

Back Close
Full

xALDX
In CBMIV, the ALD2 species represents acetaldehyde (CH 3 CHO) and all C 3 and higher aldehydes.However, in CB05, ALD2 explicitly represents only acetaldehyde while a new species, ALDX, is included to represent C 3 and higher aldehydes.In CB05, ALDX produces C 3 and higher acylperoxy radicals, CXO3, in reactions which are analogous to the explicit production of acetylperoxy, C2O3. ALD2 The acylperoxy radicals may then proceed to react with NO 2 to produce a C 3 and higher peroxyacyl nitrate species, PANX, again in a manner analogous to the explicit acetylperoxy radical, allowing PAN to explicitly represent only peroxyacetyl nitrate. C2O3 In CB05, both PAN and PANX may decompose back to the peroxy radical and NO 2 via thermal or photolytic processes.Introduction

Conclusions References
Tables Figures

GMDD Introduction Conclusions References
Tables Figures

Back Close
Full

Conclusions References
Tables Figures

Back Close
Full

xROOH
In CB05, a lumped organic peroxide species (ROOH) was introduced.It is formed as a product of each of the peroxy operator species, XO2 and XO2N, reacting with hydroperoxyl radical (HO 2 ) In CBMIV, these two reactions produced only an inert product; however, in CB05 the ROOH species can react via to produce photochemically active products.In sensitivity test xROOH, the Reactions (R63) and (R64) were removed from CB05, leaving all other reactions as in the base CB05 mechanism.By removing these two reactions from the mechanism, Reactions (R56) and (R57) produce an inert product as in CBMIV.No emissions changes were made to the scenario suite.

xIOLE
In CB05, an internal alkene species (IOLE) was introduced to better represent the chemistry of these species.

Conclusions References
Tables Figures

Back Close
Full In CBMIV, the chemistry of internal alkenes is approximated by the direct emission of the aldehyde products (as ALD2) that are formed in their oxidation reactions.In sensitivity test xIOLE, the IOLE Reactions (R124), (R125), (R126) and (R127) are removed and the oxidation reactions of the original alkene species (OLE) (Reactions R116, R117, R118, and R119) are replaced by their CBMIV versions and the emissions of IOLE are replaced by emissions of 2 × ALD2 (assuming that all of the IOLE represented 2-butene).

xTERP
In CB05, the TERP species was introduced to represent terpene chemistry more explicitly.Four reactions of TERP were included In the sensitivity test xTERP, these reactions were removed from CB05, leaving all other reactions as in the base CB05.TERP emissions of the scenario suite were converted to the equivalent CBMIV representation as 1TERP = 1.5ALD2 + 0.5OLE + 6.0PAR.

xiNOxrecycle
Three reactions were introduced into the CB05 mechanism that are not included in CBMIV which provide a pathway for recycling of reactive nitrogen species For sensitivity test xiNOxrecycle, these reactions were removed from CB05, leaving all other reactions as in the base CB05 mechanism.No emissions changes were made to the scenario suite.

xNTRrecycle
The chemistry of lumped organic nitrate (NTR) was altered substantially in CB05.The yield of NTR production from the reaction of toluene-hydroxyl radical adduct (TO2) with NO was reduced from 1 to 0.1, while an additional production reaction of NTR involving the new species TERP with NO 3 was introduced.Other reactions producing NTR were unchanged and so the overall production of this species is similar between the two mechanisms.However, in CB05 two NTR recycling reactions were introduced In CBMIV, NTR is an irreversible sink of reactive nitrogen species, acting to permanently remove odd-oxygen from the system.In CB05, NTR is no longer a terminal species for reactive nitrogen, but can play a role similar to peroxyacetylnitrate (PAN) to transport and redistribute reactive nitrogen far from its original source location.For sensitivity test xNTRrecycle, the NTR recycling reactions (R61 and R62) were removed from CB05, leaving all other reactions as in the base CB05 mechanism.No changes in emissions were made to the scenario suite.

Sensitivity test xallNOxrecycle is a combination of tests xiNOxrecycle and xNTRrecycle
where all of the newly introduced reactive nitrogen recycling reactions (R51, R52, R53, R61 and R62) were removed from CB05, leaving all other reactions as in the base CB05 mechanism.No emissions changes were made to the scenario suite.

Results and discussion
The goal of this investigation was to understand the mechanistic reasons why the CB05 chemical mechanism tends to produce higher O 3 concentrations than the CBMIV mechanism, especially in the context of the NAQFC system.Because of this relatively narrow focus, in this discussion we concentrate only on O 3 and its precursors and how the sensitivity tests shed light on the higher O 3 concentrations produced by CB05.
Results from the sensitivity tests can be grouped into three categories.-6), but either in the wrong direction to explain the observed differences between CB05 and CBMIV or only exhibited the changes for a limited subset of scenarios.Figure 4 presents results from sensitivity test xexCH4.Removing the explicit CH 4 chemistry and returning the mechanism to the CBMIV representation resulted in most simulation scenarios, e.g., s04, s10, s16 and s26, exhibiting little change in O 3 concentrations.In a few of the scenarios, e.g., s01 and s37, O 3 concentrations actually moved further away from the CBMIV results, in cases with low NO x emissions where R66' produces higher overall concentrations of peroxy radicals than does the explicit representation of MEO2. Figure 5, for sensitivity test xIOLE, presents similar results.Scenarios s01, s10 and s37 showed little change in O 3 concentrations, while scenarios s04, s16 and s26 exhibited increases in O 3 , again moving further away from the CBMIV results.In these cases, the removal of explicit internal olefin chemistry resulted in a decrease in the amount of reactive nitrogen tied up in HNO 3 and NTR, allowing greater O 3 production.Finally, in Fig. 6, results from sensitivity test xTERP are presented.These results demonstrate that converting the explicit TERP chemistry introduced in CB05 back into the CBMIV representation (accounting for terpenes as a combination of ALD2, OLE and PAR) actually slightly increases O 3 production in most chemical regimes.Thus, the introduction of the explicit TERP chemistry in CB05 effectively reduced O 3 production as compared to CBMIV for a given terpene emissions rate and thus the higher observed O 3 produced by CB05 cannot be accounted for by the introduction of explicit TERP chemistry.and (R64) (thus returning ROOH to an inert end product species) reduces the amount of O 3 produced in the simulations.The introduction of Reactions (R63) and (R64) in CB05 increases the pool of peroxy radicals through direct production of HO 2 and XO2 and indirectly through the production of additional higher aldehydes.However, as seen in Fig. 7, this mechanism change alone only partially accounts for the higher observed CB05 concentrations.
The sensitivity tests xiNOxrecycle and xNTRrecycle (Figs. 8 and 9) are similar to each other in that in each nitrogen is recycled from more stable forms (HO 2 NO 2 , HNO 3 and N 2 O 5 and NTR) back to the more reactive forms NO 2 and/or NO 3 .As can be seen from the selected simulation scenario results for these tests, the introduction of these reactive nitrogen recycling reactions in CB05 has a significant impact on the production of O 3 .For xiNOxrecycle (Fig. 8), the urban-like scenarios (e.g., s01 and s04) exhibit the largest impacts since the concentrations of the inorganic nitrogen compounds are larger in these than in the rural-like scenarios (e.g., s10 and s16).On the other hand, for xNTRrecycle (Fig. 9), the lack of NO 2 recycling from NTR has a significant impact in most of the scenarios, except those in which little NTR is produced initially (e.g., s01).The combination of the two sets of mechanism changes was tested with xallNOxrecycle, shown in Fig. 10, and demonstrates that these nitrogen recycling pathways as a group have a large impact on O 3 production in all chemical regimes.In CBMIV, the lack of these recycling pathways (Reactions R51, R52, R53 and R62) effectively reduces the total amount of reactive nitrogen available in the system through which O 3 can be produced by tying up a fraction of emitted NO x in HO 2 NO 2 , HNO 3 , N 2 O 5 and especially NTR.Inclusion of the recycling pathways in CB05 enhances the effective pool of reactive nitrogen for a given fixed emissions scenario.These results clearly indicate that introducing these recycling pathways has a large impact on O 3 production in the system and can substantially account for the higher production of CB05 as compared to CBMIV.Introduction

Conclusions References
Tables Figures

Back Close
Full

Conclusions and implications
A box model investigation of mechanistic differences between CB05 and CBMIV was conducted to determine the underlying reasons why CB05 has been observed to consistently produce higher O 3 concentrations than CBMIV for a given set of simulation conditions.By performing sensitivity tests to isolate the conceptual groupings of mechanism differences between CB05 and CBMIV, this study has identified two sets of reactions that are the primary causes of the greater O 3 production observed in CB05.
First, the introduction of Reactions (R63) and (R64) in CB05, where the higher organic peroxide species reacts to produce peroxy radicals directly (as HO 2 and XO2) and indirectly (via production of higher aldehydes) results in higher O 3 concentrations in most chemical regimes.Second, the introduction in CB05 of recycling pathways for reactive nitrogen (via Reactions R51, R52, R53 and R62) effectively increases the amount of NO x available in the system for a given emissions rate.
In the context of a large-scale three-dimensional air quality simulation such as the NAQFC, the reactive nitrogen recycling reactions are particularly effective in producing larger O 3 concentrations throughout the model domain.In particular, the organic nitrate species, NTR, plays a role much as the PAN species, serving as a temporary reservoir of reactive nitrogen.NTR can be formed efficiently through a variety of pathways in the CB05 mechanism (Reactions R55, R115, R129, R133, R144, R147, R152, R156) and is formed at the highest rates near significant NO x sources.Since the lifetime of NTR with respect to its most significant chemical loss process (Reaction R62) is about 4 days, it can be transported relatively long distances away from major sources before NO 2 is regenerated.If this recycling process occurs in NO x -sensitive areas, significant additional O 3 can be produced.Through this process, NO x , which is emitted from large sources in NO x -saturated chemical regimes where it may not produce much additional O 3 , can be redistributed to more remote locations which may be NO x -sensitive, thereby increasing the O 3 -production efficiency of each emitted NO x molecule.Introduction

Conclusions References
Tables Figures

Back Close
Full As previous studies have indicated (Yarwood et al., 2005;Sarwar et al., 2008;Luecken et al., 2008), the CB05 mechanism is a better representation of the stateof-the-science (as of 2005) of gas phase chemistry than is the CBMIV mechanism.The fact that the NAQFC operational version, which uses the CBMIV mechanism, produces ground-level O 3 concentrations which have smaller biases with respect to measurements (Fig. 1) than does the CB05 version implies that the operational NAQFC system contains other compensating errors that allow it to produce better results for O 3 .In order to move the NAQFC system forward to further improve its forecasting capabilities, these compensating errors need to be uncovered and corrected.Introduction

Conclusions References
Tables Figures

Back Close
Full  Full  Full Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | O 3 production between the CB05 and CBMIV mechanisms.Each sensitivity test was designed to probe how (or if) each particular mechanistic update of CBMIV impacted O 3 concentrations over the wide range of chemical regimes typically encountered in Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 3. PAN chemistry is converted to the CBMIV representation by removing reactionswhere PANX or CXO3 (C3 and higher acylperoxy radicals) are reactants (R102, R103, R104, R105, R106, R107, R108, R109, R110, and R111) and all CXO3 products are replaced by the CBMIV representation of higher acylperoxy radicals, C2O3 (R141, R143, R145, R147, and R151); and, 4. all ALDX emissions are replaced by ALD2 emissions.
Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | R151) + 0.18XO2N + 0.24HCHO + 0.001CO + 7.0PAR + 0.21ALDX + 0.39CXO3 TERP + NO 3 → 0.47NO 2 + 0.28HO 2 + 1.03XO2 (R152) + 0.25XO2N + 0.47ALDX + 0.53NTR Figures Back Close Full Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | NTR + OH → HNO 3 + HO2 + 0.33HCHO + 0.33ALD2 (R61) + 0.33ALDX − 0.66PAR NTR + hν → NO 2 + HO 2 + 0.33HCHO + 0.33ALD2 (R62) + 0.33ALDX − 0.66PAR Discussion Paper | Discussion Paper | Discussion Paper | First, five of the sensitivity tests (xH2, xoddH&O, xNO3night, xETHA and xALDX) demonstrated little or no impact on O 3 concentrations.In other words, the mechanism updates exercised by these five sensitivity tests did not account for the observed larger O 3 produced by CB05.An example of this is presented in Fig. 3, where selected simulation scenario results for sensitivity test xALDX are compared with the base CB05 and CBMIV results for ozone.Ozone concentrations from xALDX were only slightly less for most of Discussion Paper | Discussion Paper | Discussion Paper | the scenarios than for the base CB05 runs.Results for xH2, xoddH&O, xNO3night, and xETHA were similarly negative.Consequently, none of the mechanism changes probed by these sensitivity tests account for the higher O 3 obtained from CB05.The second group of sensitivity tests (xexCH4, xIOLE, and xTERP) exhibited a more substantial change in O 3 concentrations (Figs. 4 The third group of sensitivity tests (xROOH, xiNOxrecycle, xNTRrecycle, and xall-NOxrecycle) demonstrated the most relevant changes in O 3 concentrations to the observed differences between CB05 and CBMIV.First, results for sensitivity test xROOH are presented in Fig. 7.In most simulation scenarios, removing Reactions (R63) Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Fig. 1 .Fig. 2 .
Fig. 1.Mean bias of ground-level ozone from parallel versions of the NAQFC.Ozone measurements are from the US EPA Air Quality System (AQS) measurement network.Operational NAQFC (red) and experimental NAQFC (black) mean biases (ppbv) calculated as a CONUSwide average of (model -AQS) site biases.

Table 1 .
Environmental and initial conditions for each simulation scenario.

Table 4 .
Description of box model sensitivity tests.