Global chemical transport model study of ozone response to changes in chemical kinetics and biogenic volatile organic compounds emissions due to increasing temperatures: Sensitivities to isoprene nitrate chemistry and grid resolution

Authors


Abstract

[1] Global modeling studies show a wide variability in the response of the O3 budget to climate change as projected by applying Intergovernmental Panel on Climate Change scenarios in climate models. We employ sensitivity studies to elucidate the major uncertainties in the response of tropospheric O3 to perturbations in biogenic volatile organic compounds (BVOC) emissions and reaction rate coefficients due to changes in temperature. The change in global O3 burden due to an increase in BVOC emissions associated with a +5 K depends critically on the assumed treatment for the fraction of NOx recycled (0–100%) from isoprene nitrate (+9 to +34 Tg), in contrast to the chemical reaction rate coefficients response (−8 to −9 Tg). The model O3 burden shows sensitivity (40 Tg) to the NOx recycling efficiencies (0–100%) similar to the burden's sensitivity to the grid resolution (4° × 5°–1° × 1°). The correlation of O3 with total alkyl nitrates (ΣANs) in the surface air at a California forest site shows sensitivity to the NOx recycling (40–100%) similar to the correlation's sensitivity to the horizontal resolution (4° × 5°–1° × 1°). The results of the sensitivity simulations imply that the slope of O3 to ΣANs might be used to constrain the yield of isoprene nitrate and NOx recycling fraction, but better agreement could be achieved by using a higher-resolution model with even higher NOx recycling from isoprene nitrate. Our results suggest that the reduction of NOx recycling from isoprene nitrate be set apart from that due to the effect of the grid resolution in the chemical transport model.

1. Introduction

[2] A number of global modeling studies have suggested that projected climate changes over the next century will result in a decrease in the overall global tropospheric burden of ozone (O3) because of the increased destruction of O3 associated with increased water vapor [e.g., Brasseur et al., 1998; Johnson et al., 1999; Stevenson et al., 2000; Dentener et al., 2006a]. Stevenson et al. [2006] have intercompared a total of 26 different global atmospheric chemistry models. Simulations for the assessment included the influence of climate change from 2000 and 2030 with global annual mean surface temperatures of between 0.31 and 0.95 K projected by applying Intergovernmental Panel on Climate Change (IPCC) scenarios in climate models. The models show a wide variability in the response of the global tropospheric O3 budget when a climate change scenario is considered. These differences highlight significant imperfections in our current understanding of the key factors involved (e.g., deep tropical convection, stratosphere-troposphere exchange, and water vapor response).

[3] Several 3-D chemical transport model (CTM) sensitivity studies have been performed to elucidate the major uncertainties in the response of O3 to climate change. Wild [2007] and Wu et al. [2007] have demonstrated that the magnitudes of the emissions of nitrogen oxides (NOx = NO + NO2) from lightning and of isoprene emissions from vegetation contribute to major uncertainties in model calculations of O3 production. Both of them recognized that the different isoprene chemistry schemes might be an important source of differences between model studies. Our 3-D model study follows these comprehensive studies and aims to elucidate the role of the organic hydroxynitrates (RONO2, “isoprene nitrates”) from biogenic volatile organic compounds (BVOC) emissions in the response of O3 to temperature increases.

[4] There has been increasing attention paid to the climate change penalty of an increase in surface O3 mixing ratios near polluted areas. However, regional and global model studies have shown that there are significant uncertainties in the response of surface O3 to temperature increases, especially over the southeastern United States [Hogrefe et al., 2004; Murazaki and Hess, 2006; Dawson et al., 2007; Tagaris et al., 2007; Racherla and Adams, 2008; Wu et al., 2008; Zhang et al., 2008]. Dawson et al. [2007] used the CBM-IV chemical mechanism [Gery et al., 1989] and showed that a uniform increase of surface temperature by 2.5 K would result in 0–4 ppb increases of the summertime daily maximum 8-h average O3. Hogrefe et al. [2004] used the CBM-IV chemical mechanism and found the average daily maximum 8-h O3 increase of 4–8 ppb during summertime in the 2050s relative to the 1990s using the IPCC SRES A2 scenario (1.5–3.5 K). Zhang et al. [2008] reported a lower O3 increase of 1–5 ppb using the IPCC A1B scenario (2–4.6 K). Racherla and Adams [2008] used 100% NOx recycling with a 12% yield of isoprene nitrate [Horowitz et al., 1998] and predicted a 5 ppb increase in the 95th percentile O3 in response to the 1.4-K increase in the surface temperatures projected using the IPCC SRES A2 scenario. Murazaki and Hess [2006] used 100% NOx recycling with an 8% yield [Horowitz et al., 2003] without considering changes in BVOC emissions and reported a smaller average O3 increase of 1–2 ppb during summertime between 1991 and 1999 and 2091–2099 using the IPCC A1B scenario (2–3.5 K) than the 3–9 ppb range reported by Racherla and Adams [2006]. On the other hand, Tagaris et al. [2007] and Wu et al. [2008] found little effect. Wu et al. [2008] speculated that different treatments of isoprene nitrate chemistry could be a major source of the difference in the model responses of the O3 to temperature increase. However, Tagaris et al. [2007] used the SAPRC mechanism [Carter, 2000], in which the NOx recycling is larger than that in the CBM-IV. Thus the attribution of the differences in the projected O3 changes to one single factor is often disturbed by multiple competing effects such as the model inputs (e.g., emissions, land cover, and boundary conditions), the BVOC emission models used, the treatments of various atmospheric meteorological and chemical processes in global versus regional models and the horizontal grid resolutions used in the simulations.

[5] There is notable uncertainty in present-day estimates of BVOC emissions. Guenther et al. [2006] developed an emissions model to estimate the range of global isoprene emissions. They found that their estimates ranged from 440 to 660 TgC a−1 and were strongly dependent on the driving variables, which included land cover (leaf area index and plant functional type distributions) and weather (solar radiation, air temperature, humidity, wind speed, and soil moisture) conditions. A wide range of biogenic emissions for isoprene has been used in different chemistry transport models (220–630 TgC a−1 [Stevenson et al., 2006]). The lowest value for isoprene emissions was used for the IPCC Third Assessment Report (TAR) [Prather et al., 2003], while significantly higher emissions were used for the Fourth Assessment Report (AR4). There is an ongoing debate about the convergence toward a common estimate of global isoprene emissions [Arneth et al., 2008; Lelieveld et al., 2008].

[6] Accurate future predictions of isoprene emissions are also uncertain, because the response of terrestrial ecosystems to human activities and climate change is highly complex. Predicted estimates of global isoprene emissions range from 640 to 890 TgC a−1 at the end of the 21st century [Sanderson et al., 2003; Lathière et al., 2005; Wiedinmyer et al., 2006], with the biggest driver being the projected increase in surface temperature. Guenther et al. [2006] estimated that isoprene emissions may increase by more than a factor of 2 using temperature distributions simulated by applying IPCC A1 scenarios in global climate models for the year 2100. In addition to the temperature changes, the long-term prediction of isoprene emissions is sensitive to land-use change [Lathière et al., 2006; Ito et al., 2008], CO2 concentration [Rosenstiel et al., 2003; Arneth et al., 2007], and water availability [Pegoraro et al., 2005; Müller et al., 2008].

[7] The spatial and temporal changes of surface O3 in response to an increase of isoprene emissions due to increased surface temperatures are strongly influenced by background NOx levels [e.g., Sanderson et al., 2003; Hauglustaine et al., 2005; Brasseur et al., 2006; Wiedinmyer et al., 2006; Tao et al., 2007; Kunkel et al., 2008; Zeng et al., 2008]. In a comprehensive modeling study, Sillman and Samson [1995] described how the partitioning of total reactive nitrogen (NOy) into total peroxy nitrates (ΣPNs), total alkyl nitrates (ΣANs), and HNO3 is expected to vary as a function of temperature. From their study of polluted rural environments, they concluded that O3 mixing ratios increase with temperature because of an increase in the thermal decomposition rate for ΣPNs, making ΣANs the dominant contributors to NOz (= NOy − NOx) at higher temperatures rather than HNO3. Their model result was consistent with observations at a rural site in the southeastern United States [Olszyna et al., 1997]. On the other hand, Day et al. [2008], who observed atmospheric reactive nitrogen compounds at a California forest site, argued that there was a larger increase in the contribution of HNO3 to NOz with temperature and a smaller increase in the contribution of ΣANs to NOz than that calculated by Sillman and Samson [1995]. They attributed the inconsistency between the model and observation to a steep increase in the hydroxyl radical (OH) inferred from the observations of temperature trends in the HNO3/NO2 ratio. These observed changes in reactive nitrogen compounds versus temperature provide a good basis for evaluating the ability of models to predict temperature-dependent responses. The temperature-dependent response of NOy partitioning is especially important for assessments of future O3, because the fate of NOx significantly affects O3 formation in the global atmosphere. Thus it is meaningful to compare the sensitivity of relatively long lived NOx reaction products (ΣPNs, ΣANs and HNO3) in a CTM to temperature changes with those observed.

[8] Previous box and 3-D model studies on regional and global scales have demonstrated that O3 mixing ratios and reactive nitrogen partitioning are sensitive to uncertainties in the chemical oxidation pathways of isoprene [e.g., Horowitz et al., 1998, 2007; Pöschl et al., 2000; von Kuhlmann et al., 2004; Fiore et al., 2005; Ito et al., 2007a]. Fiore et al. [2005] showed that the modeled response of O3 over the eastern United States to changes in isoprene emissions depends on the fate of isoprene nitrates. The reaction of isoprene nitrates with OH and O3 can lead to the net production of NOx (recycling) in the remote atmosphere or to the loss of NOx in isoprene-rich air if there are significant yields of secondary multifunctional organic nitrates [Giacopelli et al., 2005]. In an extreme case, when the recycling of NOx was changed from 0% to 100%, surface O3 mixing ratios increased by approximately 20% at almost all locations [Ito et al., 2007a], which was comparable to the difference between different chemical mechanisms considered in the box model intercomparison by Pöschl et al. [2000].

[9] Day et al. [2003] found that the ΣANs is strongly correlated with O3 with a slope of 80 O3 per ΣANs (ppb/ppb) at UC–BFRS. Horowitz et al. [2007] used measurements from the ICARTT field campaign [Fehsenfeld et al., 2006; Singh et al., 2006] to infer the yield of isoprene nitrate and the NOx recycling fraction, based in part on the model versus measured slope (82 observed, 81 simulated with 40% NOx recycling and a 4% yield of isoprene nitrate). Recently, Paulot et al. [2009] found a higher yield (12%) and recycling efficiency of 55% on the basis of an analysis using atmospheric smog chamber experiments. Perring et al. [2009] suggested much higher recycling efficiencies for different yields (67% isoprene nitrates recycling for 12% yield; 97% for 4.4%) on the basis of an observed correlation between ΣANs and HCHO. We investigate the sensitivity of O3 per ΣANs for different NOx recycling fractions to understand why the NOx recycling fraction obtained by the model calibration is lower than that obtained by Paulot et al. [2009].

[10] In addition to the effects of the treatment of isoprene chemistry, the production of O3 over isoprene-emitting regions near polluted areas is influenced by the model resolution, owing to artificial mixing of air masses of different origins when near surface emissions are added to coarse grids [e.g., Sillman et al., 1990; Liang and Jacobson, 2000; Fiore et al., 2003; Wild and Prather, 2006; Wild, 2007]. Sillman et al. [1990] calculated more O3 production in the boundary layer and less NOz export to the free troposphere (FT) at lower resolution, because dilution of precursor NOx emissions in a coarse model grid increased the O3 production efficiency, defined as the number of O3 molecules produced per unit NOx oxidized [Liu et al., 1987]. Fiore et al. [2003] compared the GEOS-Chem model with regional model simulations (horizontal resolution of 36 km2) and showed that the 4° × 5° resolution captured the large-scale features of the surface O3 distribution, although the ability to capture local O3 maxima was compromised.

[11] The purpose of this paper is to investigate the processes that could have a major impact on O3 mixing ratios when temperatures change. We focus on the link between the treatment of isoprene nitrates and the horizontal resolution in the model and on the evaluation of the model predictability of temperature-dependent responses of O3 and the relatively long lived reactive nitrogen compounds. For this purpose, a global CTM is used to calculate the O3 mixing ratios. We investigate a sensitivity study, in which we prescribe the water vapor concentration in the model and only perturb the effects of temperature on chemical reaction rates and BVOC emissions. We also examine the correlation between O3 and temperature and between reactive nitrogen species and temperature in the day-to-day record as predicted by the CTM and in comparison with observations over an isoprene-emitting region (reported by Day et al. [2008]). These day-to-day correlations with temperature can reflect the influence of a wide variety of meteorological factors (solar radiation, humidity, mixed layer depth, and transport patterns), which coincidentally vary with temperature, rather than the effect of temperature in isolation. Day-to-day correlations with temperature also may not reflect the combined changes in meteorology and emissions associated with future climate. Nonetheless, they provide a test of the ability of models to predict changes in chemically reactive species in response to temperature against ambient conditions. Section 3 describes the different scenarios examined. Section 4 examines the sensitivity of the global O3 budget and surface mixing ratios in different simulations as well as comparisons of the model with observations of O3 and other intermediate species. Section 5 presents a summary of our findings.

2. Model Approach

[12] For this study, we use the Integrated Massively Parallel Atmospheric Chemical Transport (IMPACT) model driven by assimilated meteorological fields which are obtained from the Goddard Earth Observation System (GEOS) of the NASA Global Modeling and Assimilation Office (GMAO) [Schubert et al., 1993]. The GEOS-3 meteorological data set [Zhu et al., 2003] for the year 2001 is suitable for the comparison with the observations by Day et al. [2008]. IMPACT was originally developed at the Lawrence Livermore National Laboratory (LLNL) [Rotman et al., 2004] but was extended to treat aerosols and detailed chemical reactions for a wider set of VOC at the University of Michigan [Liu et al., 2005; Feng and Penner, 2007; Ito et al., 2007a]. The GEOS-3 meteorological fields were provided with a 6-h temporal resolution on a horizontal resolution of 1° × 1° with 48 vertical layers. Since the gas-phase model does not communicate with the aerosol model interactively, two separate simulations were performed for a 1-year time period with a 6-month spin-up period for the gas-phase model and a 2-month spin-up for the aerosol model.

[13] A number of revisions to the chemistry described by Ito et al. [2007a] were made for the present study. We adopt the production of secondary multifunctional organic nitrates (XNITR) from the oxidation of isoprene and terpene nitrates by the three main oxidants, OH, O3, and NO3 in the work of Horowitz et al. [2007] to the previous IMPACT mechanism of Ito et al. [2007a]. Our yield of isoprene nitrate from the reaction of isoprene peroxy radicals with NO which ranges from 8% for 593 hPa to 12% for 1000 hPa [Sprengnether et al., 2002] is larger than the 4% yield [Chen et al., 1998] in the base simulation of Horowitz et al. [2007]. This chemistry reaction mechanism assumes that the reaction of isoprene nitrate with OH and O3 yields

equation image

where GLYC represents hydroxyacetaldehyde, sum of A and B equals 1, and A takes on the values 0, 0.4, and 1. We changed the production of hydroxyacetone in the assumption by Paulson and Seinfeld [1992] to GLYC, on the basis of the model comparison with observations of Ito et al. [2007a]. Species scavenging depends on solubility through the Henry's law coefficient, as described by Rotman et al. [2004]. The Henry's law constant of 1.7 × 104 M atm−1 for isoprene nitrate [O'Sullivan et al., 1996] is of the same order of magnitude as that of 8.3 × 104 M atm−1 for HNO3 [Schwartz and White, 1981]. XNITR is removed by wet and dry deposition at the same rates as isoprene nitrate, as its further reactions are assumed to convert it to more highly substituted organic nitrates that do not recycle NOx back into the atmosphere. We include methanol deposition following Jacob et al. [2005]. The water vapor in the troposphere is prescribed by the meteorological data. The tropopause height is defined as the lowest level at which a synthetic O3 tracer from the stratosphere is greater than 150 ppbv [McLinden et al., 2000]. Reaction rates are taken from the recent IUPAC [Atkinson et al., 2004] and JPL [Sander et al., 2006] assessments. The aerosols used in the calculation of the uptake of N2O5 are calculated using the model described by Feng and Penner [2007]. We apply the same monthly averaged NOx emission rates and the HOx (= OH + HO2) and O3 output from an initial gas-phase run with the previously fixed aerosol concentrations for the reaction of N2O5 on aerosols in the aerosol simulation. We use the aerosol emissions of Dentener et al. [2006b]. The uptake rates of N2O5, HO2, NO2 and NO3 by aerosols are calculated from the aerosol simulation driven by the 2001 meteorological fields at 2° × 2.5° resolution.

[14] The emissions inventory has been recompiled to represent the year 2001 for the comparison with the measurements by Day et al. [2008]. Table 1 lists the global emissions of O3 precursors except BVOC used in the present work. Isoprene emissions are based on a modified version of the inventory of Guenther et al. [1995] by Wang et al. [1998] and Bey et al. [2001]. Monoterpene emissions are based on the work of Guenther et al. [1995] as modified by Wang et al. [1998]. The BVOC emissions are calculated interactively with the same meteorological data. We run the model for 1 year and calculate the monthly emissions. Then, we scale the monthly emissions to normalize to a global biogenic isoprene source of 500 TgC a−1 and terpene sources of 130 TgC a−1 as in the work of Guenther et al. [1995] for all scenarios at the base case temperature. The other temperature-dependent BVOC emissions for ethene, propene, acetone, and methanol are distributed according to emissions of isoprene [Ito et al., 2007a]. We use the surface air temperature dependency from Guenther et al. [1999] following Palmer et al. [2006] to determine the resulting change in emissions. This functional dependence exponentially increases BVOC emissions with increasing surface air temperature up to a maximum that is dependent on the average temperature during the past 15 days. Lightning NOx emissions are based on the convective mass fluxes from the meteorological fields at each resolution [Allen and Pickering, 2002]. The vertical distribution of the source is based on observed profiles [Pickering et al., 1998]. We normalized the total emissions to 3 TgN a−1 as per Bey et al. [2001]. The soil NOx emissions are obtained from Yan et al. [2005] for the year 2001. Anthropogenic fossil fuel emissions are taken from M. G. Schultz et al. (RETRO emission data sets and methodologies for estimating emissions, project report, 2007, available at http://retro.enes.org/reports/D1-6_final.pdf) for 2000 except for aircraft NOx emissions, which are from the monthly mean emission inventory for 1999 of Mortlock and Van Alstyne [1998] and Sutkus et al. [2001]. The global biomass burning emissions are updated to those of Ito et al. [2007b], which are based on the emissions developed by Ito and Penner [2005] for the year 2000 and scaled to the year 2001 using the TOMS AI data [Herman et al., 1997]. The MODIS fire product [Giglio et al., 2006] is used to produce the seasonal variations in open biomass burning emissions except in southern Africa, where regional emissions from open vegetation burning are available [Ito et al., 2007c].

Table 1. Global Emissions of O3 Precursors Except Biogenic Volatile Organic Compounds for the Year 2001a
SpeciesEmission Rate
  • a

    Units are TgN a−1 for NOx, TgCO a−1 for CO, and TgC a−1 for nonmethane volatile organic compounds.

NOx46
CO1232
Methylethyl ketone8.6
C2H69.0
C3H84.5
CH3CHO6.0
HCHO6.1
C4–C5 alkanes (ALK4)15.9
≥C6 alkanes (ALK7)10.3
Ethene14.4
Propene18.9
Trans-2-butene5.1
Benzene7.2
Toluene10.0
m-xylene9.1
Acetone4.4
Methanol11.1
Formic acid2.4
Acetic acid14.7
Phenol12.5

3. Chemical Transport Model Sensitivity Studies

[15] The different simulations performed in this study are summarized in Table 2. To assess the effects of increases in temperature on the chemical composition of the atmosphere, we consider the change in both the reaction rate coefficients (section 3.1) that depend on temperature and the BVOC emissions (section 3.2) to perturbations of temperature throughout the troposphere. A fixed uniform change of temperature is useful to investigate each of these parameters separately so that the effects of each as well as the synergy can be determined. This would help identify the major factors that could have an effect on air quality as temperature changes. To assess the uncertainties in the model response, we examine the sensitivity of O3, its precursors, and its production to the treatment of the recycling of NOx from isoprene nitrates (section 3.3) and to the model resolution (section 3.4).

Table 2. Summary of Different Simulations Performed in This Study
SimulationRun NameIsoprene EmissionNOx Recycling, %Resolution
  • a

    Low resolution (4° × 5°).

  • b

    Temperature increase (+5 K) applied to chemical reaction rate coefficients and/or biogenic volatile organic compound (BVOC) emissions.

  • c

    BVOC emission perturbation.

  • d

    Temperature decrease (−5 K) applied to chemical reaction rate coefficients and/or BVOC emissions.

  • e

    Chemical reaction rate coefficient perturbation.

  • f

    BVOC emission and chemical reaction rate coefficient perturbations.

  • g

    Middle resolution (2° × 2.5°).

  • h

    High resolution (1° × 1°).

1La_050004° × 5°
2L_40500404° × 5°
3L_1005001004° × 5°
4L_0_PbEc94004° × 5°
5L_0_MdE27004° × 5°
6L_40_PE940404° × 5°
7L_40_ME270404° × 5°
8L_100_PE9401004° × 5°
9L_100_ME2701004° × 5°
10L_0_PCe50004° × 5°
12L_0_MC50004° × 5°
13L_40_PC500404° × 5°
13L_40_MC500404° × 5°
14L_100_PC5001004° × 5°
15L_100_MC5001004° × 5°
16L_0_PECf94004° × 5°
17L_0_MEC27004° × 5°
18L_40_PEC940404° × 5°
18L_40_MEC270404° × 5°
19L_100_PEC9401004° × 5°
20L_100_MEC2701004° × 5°
21Mg_40500402° × 2.5°
22Hh_40500401° × 1°
23Hh_1005001001° × 1°

3.1. Effects of Temperature on Chemistry

[16] The effects of temperature on oxidant photochemistry are examined by applying globally uniform changes in temperature (±5 K) to the chemical reaction rate coefficients only after Wild [2007]. Increased temperatures affect chemical reaction kinetics and thus O3 production and loss rates. The reactions responsible for this temperature effect are mainly associated with the chemistry of peroxyalkyl nitrates such as peroxyacetyl nitrate (PAN, CH3C(O)O2NO2), because thermal decomposition is the main loss reaction for these compounds [Carter et al., 1979]. Although we focus on the sensitivity of O3 in this paper, we note that the sensitivity of CH4 oxidation to temperature has also been shown to have temperature dependence relevant for chemistry-climate links [Johnson et al., 1999; Fiore et al., 2006].

3.2. Effects of Temperature on Emissions

[17] The effects of temperature on BVOC emissions are examined by applying a perturbation based on changes in temperature (±5 K) to the calculation of BVOC emission rates. The BVOC whose emission is sensitive to temperature include isoprene, ethene, propene, terpenes, acetone, and methanol. Our estimates of isoprene emission range from 270 (−5 K) to 940 TgC a−1 (+5 K). In section 4.1.3, we discuss the implications of the findings from recent studies by Arneth et al. [2007] and Heald et al. [2009].

3.3. Effects of NOx Recycling From Isoprene Nitrates

[18] The response of O3 to a change in temperatures is sensitive to the NOx recycling rate from the oxidation products of biogenic emissions, which are related to the temperature. The treatment of the recycling of NOx when isoprene nitrates and nitrates from the oxidation of terpene react with OH and O3 is varied from 0% to 40% to 100% following Horowitz et al. [2007].

3.4. Effects of Model Resolution

[19] The response of O3 to the NOx recycling rate is influenced by the model resolution when biogenic emissions are mixed with anthropogenic sources within the same coarse grids. We examine the sensitivity of O3, its precursors, and its production to horizontal resolution by running the model at three different horizontal resolutions: low resolution (4° × 5°), middle (2° × 2.5°), and high (1° × 1°). The middle and high resolutions used here may be comparable with those used by Wild and Prather [2006] at T42 (2.8° × 2.8°) and T106 (1.1° × 1.1°) resolution for March and April 2001.

4. Model Results and Discussion

4.1. Global Ozone Budget

[20] The results of the sensitivity studies for the low-resolution simulations (numbers 1–20) listed in Table 2 are shown in Figure 1. In the analysis of the global O3 budget, “O3” represents “O3 + NO2 + 2 × NO3 + PAN + MPAN + HNO4 + HNO3 + 3 × N2O5 + npan.” The lifetime of O3 is determined from its chemical removal and deposition rates. In sections 4.1.1 and 4.1.2, we investigate each of these parameters separately. In section 4.1.3, we examine the combined effect of these parameters. In section 4.1.4, we explore the sensitivity to model resolution.

Figure 1.

The relationship between the tropospheric burden of O3 and its lifetime from chemical removal and deposition from the results of the sensitivity studies listed in Table 2. The black line represents the sensitivity to the assumed recycling of NOx from isoprene nitrates. The red lines denote the sensitivity to changes in both chemical reaction rate coefficients and emissions from changes in temperature. The blue lines indicate the sensitivity to changes in emissions from changes in temperature. The green lines show the sensitivity to changes in chemical reaction rate coefficients from changes in temperature. The circles represent results for a temperature increase (+5 K) applied to the chemistry and/or emissions. The squares represent results for a temperature decrease (−5 K) applied to the chemistry and/or emissions. PC and MC represent temperature increase (+5 K) and decrease (−5 K), respectively, applied to chemical reaction rate coefficients. PE and ME represent temperature increase (+5 K) and decrease (−5 K), respectively, applied to emissions. PCE and MCE represent temperature increase (+5 K) and decrease (−5 K), respectively, applied to chemical reaction rate coefficients and emissions.

4.1.1. Sensitivities to BVOC Emissions and NOx Recycling

[21] The response of the calculated O3 budget to the perturbation of BVOC emissions due to temperature changes is sensitive to the treatment of the recycling of NOx. Increased recycling of NOx from isoprene nitrates provides a direct increase in the model mixing ratios of both NOx and organics (which are also recycled), which lead to increased O3 formation and thus its burden from 330 Tg for (L_0) to 370 Tg for (L_100). Higher NOx recycling could transport more NOx to remote regions than lower recycling, because the former does not remove NOx efficiently over isoprene-emitting regions (see section 4.2 for further discussion).

[22] Stevenson et al. [2006] showed the large spread in O3 burden (274–407 Tg) among 25 atmospheric models, although it is not clear how many previous studies have included the isoprene nitrate chemistry. Wild [2007] did not include isoprene nitrates in the simplified hydrocarbon oxidation scheme and showed that differences in isoprene emissions (220–630 TgC a−1) might account for 20 Tg in O3 burden. Wu et al. [2008] used 0% NOx recycling with the yield of Sprengnether et al. [2002] and showed little changes in O3 burden (315–319 Tg) by varying the isoprene emissions (0–400 TgC a−1). The effect of increased isoprene emissions from 270 to 940 TgC a−1 on the increase in O3 burden depends critically on the assumed treatment for the fraction of NOx recycled and ranges from 17 Tg for (L_0_PE) − (L_0_ME) to 57 Tg for (L_100_PE) − (L_100_ME). This interrelationship can be explained by the synergy between additional BVOC and increased ambient NOx associated with recycling [McKeen et al., 1991; Tao et al., 2003]. As BVOC emissions are increased, ambient NOx in biogenic source regions is increased with increasing NOx recycling owing to the increase of the organic nitrates except XNITR. The increase in O3 is driven largely by the increased role of PAN as a transporter of NOx and the rerelease of NOx from isoprene nitrates is determined by the recycling fraction of NOx. This increases the transport of NOx to the remote troposphere [e.g., Singh et al., 1998, 2000], which increases the overall net production of O3 because O3 production in the remote troposphere is more efficient [e.g., Liu et al., 1987].

[23] The slope of the lifetime versus O3 burden due to the effect of a temperature increase on the BVOC emissions is smaller when the recycling of NOx is lower. The steeper slope represents a higher chemical production of O3. Increases in BVOC emissions lead to increased net chemical O3 production for (L_40_PE) − (L_40) (17 Tg a−1) and (L_100_PE) − (L_100) (58 Tg a−1) recycling of NOx but decreased net production for (L_0_PE) − (L_0) (−2 Tg a−1). The slope for the low recycling of NOx due to the effects of a decrease in temperature of −5 K on emissions is steeper than that owing to an increase of +5 K. Wu et al. [2007] found that net chemical O3 production became saturated for nonmethane volatile organic compounds (NMVOC) emissions in the range of 200 and 500 (Tg C a−1). This is consistent with our results. A further increase in the BVOC emissions to 940 (Tg C a−1) changes the sign of the net chemical production. These results indicate that a lower recycling fraction for NOx introduces a larger chemical loss rate for O3 for the effect of a temperature increase of 5 K on BVOC emissions. However, the O3 burden (17 Tg O3) is increased for (L_0_PE) − (L_0_ME), mainly because the wet deposition of HNO3 (−13 Tg O3 a−1) is decreased.

4.1.2. Sensitivity to Chemical Reaction Rate Change due to Temperature Increases

[24] The effect of changes in chemical reaction rates due to increased temperatures by 5 K causes slight decreases (−8 to −9 Tg) in the total tropospheric O3 burden in our model. Previously, Sillman and Samson [1995] suggested that total tropospheric O3 would decrease in this case because precursors are processed and removed more rapidly in polluted regions, where the O3 production efficiency per NOx molecule is lower. Wild [2007] showed that the tropospheric burden of O3 dropped by less than 1% for a temperature rise of 5 K on oxidant photochemistry. Most of the decreases in O3 are seen in the FT (global averaged differences for (L_40_PC) − (L_40) are −1.1 ppb in FT versus −0.5 ppb in surface air) and remote sites where NOx from the decomposition of PAN affects O3 owing to the higher O3 production efficiency per unit NOx (OPE) in these regions. On the other hand, the effect of increased temperatures on chemistry leads to increased O3 in source regions (see section 4.2).

[25] Here, we find that the change in the O3 burden due to the change in reaction rate coefficients associated with a temperature change is largely independent from the assumed NOx recycling fraction, in contrast to the BVOC emission response. Gross chemical O3 production decreases (−13.6 Tg a−1) for (L_40_PC) − (L_40) in response to reduced NOx (the global averaged differences are −1.8 ppt in the FT versus +1.6 ppt in surface air) resulting from less efficient transport by PAN (as suggested by Singh et al. [1998, 2000]). The decrease in chemical O3 production (−23.4 Tg a−1) for (L_100_PC) − (L_100) is larger than that (−6.7 Tg a−1) for (L_0_PC) − (L_0) in response to larger decrease in NOx (the global averaged differences in the FT are −2.4 ppt for (L_100_PC) − (L_100) versus −1.4 ppt for (L_0_PC) − (L_0)). However, this is counterbalanced by larger decrease in total loss (−23.5 Tg a−1) for (L_100_PC) − (L_100) than that (−6.5 Tg a−1) for (L_0_PC) − (L_0), which is reflected in larger decrease in the chemical loss via the reaction of O3 with OH and photochemical destruction (the global averaged differences in the FT are −1.3 ppb for (L_100_PC) − (L_100) versus −0.98 ppb for (L_0_PC) − (L_0)).

4.1.3. Sensitivity to Combined Simulations

[26] The effect of changes in chemical reaction rates due to increased temperatures (+5 K) causes slight decreases in the total tropospheric O3 burden (−8 to −9 Tg). Increased emissions of BVOC (+5 K), however, lead to increases in O3 burden (+9 to +34 Tg) and this effect is larger than the net decrease in global O3 burden. Overall, global mean O3 increases with temperature (+5 K), ranging from 1 Tg for 0% NOx recycling to 22 Tg for 100%.

[27] Arneth et al. [2007] found that the increase in isoprene emissions induced by temperature and vegetation changes in the next century would be offset when the inhibition of leaf isoprene emissions by increasing atmospheric CO2 concentration is accounted for in their model. In this case which maintains the isoprene emissions within ±15% of present values, the response of O3 to temperature increase would be insensitive to the uncertainties in the NOx recycling rate, because the response of the O3 burden to the change in reaction rate coefficients associated with a temperature change is largely independent from the assumed NOx recycling fraction. Recently, Heald et al. [2009] found that future isoprene emissions increased by more than a factor of 2 in 2100 (to 1242 TgC a−1) owing to temperature and vegetation changes even when the effect of CO2 inhibition is included. In this case, the response of O3 to future climate would be sensitive to the uncertainties in the NOx recycling rate, because the increase in global O3 burden due to an increase in BVOC emissions depends critically on the fraction of NOx recycled from isoprene nitrate. The divergence of the future predictions of isoprene emissions highlights the need for further study of the effects of changes in isoprene emissions in the future climate.

4.1.4. Sensitivity to Model Resolution

[28] Model results at the low resolution (4° × 5°) which use lower recycling fractions for NOx (between 0% and 40%) compare better with the tropospheric O3 burden of 335 ± 10 Tg determined by Wild [2007] from available O3 climatologies. However, the global mean O3 burden for 2001 is 13% smaller at the high resolution (302 Tg) than at the low with the 40% NOx recycling fraction (340 Tg), reflecting less net chemical O3 production at the higher resolution. This amount is comparable to the difference in O3 burden between 0% and 100% NOx recycling of isoprene nitrates (40 Tg). The global mean O3 burden during March and April decreases by 3.0% as the resolution decreases from 2° × 2.5° to 1° × 1°, which is comparable to the decrease (−3.2%) from a resolution change from T42 to T106 calculated by Wild and Prather [2006]. On the basis of the tendency calculated using the low-resolution model, the burden is expected to be increased by 8% when 100% recycling of NOx is assumed rather than 40%. These results suggest that the assumptions of lower BVOC emissions and lower NOx recycling fractions are not needed for higher-resolution simulation to match the burden of O3 derived by Wild [2007].

4.2. Changes in Surface O3 and Its Precursors due to Increases in Temperature

[29] The sensitivity of surface O3 and its precursors to changes in BVOC emissions (section 4.2.1) and chemical reaction rates (section 4.2.2) due to increases in temperature are analyzed here for July. In section 4.2.3, we examine the combined effect of these parameters. In section 4.2.4, we investigate the NOx recycling fraction associated with model resolution.

4.2.1. Sensitivity to BVOC Emissions

[30] The geographical distribution of the changes in isoprene mixing ratios (ppbv) for (L_40_PE) − (L_40) is shown in Figure 2a. Enhancements of more than 2 ppbv are found over the main isoprene-emitting regions in the temperate and boreal forests. In the tropical forests, the changes reach more than 8 ppbv. The largest changes in tropical forest regions occur because of a large decrease in OH and increase in the photochemical lifetime of isoprene [e.g., Houweling et al., 1998; Guenther et al., 1999]. However, Lelieveld et al. [2008] proposed that BVOC oxidation recycles OH efficiently (40–80%) at low-NOx mixing ratios through reactions of HO2 with organic peroxy radicals (RO2), on the basis of the aircraft measurements of OH and related species in unpolluted air over the Amazon rain forest, laboratory measurements, and numerical modeling. Subsequently, Butler et al. [2008] demonstrated that a recycling of OH of 40–50% was needed to match the model results with the observations, although the resulting high-OH concentrations lead to unrealistically low mixing ratios of isoprene. In section 4.3.2.4, the isoprene mixing ratios and OH concentrations from the model results are compared to the observations.

Figure 2.

The changes in mixing ratios (ppbv) for (a) isoprene, (b) peroxyacetyl nitrate (PAN), and (c) NOx, for the difference between the case with increased emissions caused by a 5-K increase in temperature and the base case, (L_40_PE) − (L_40), in the surface air in July. The white box in Figure 2a shows the location of the University of California–Blodgett Forest Research Station (UC–BFRS).

[31] The changes in the PAN mixing ratios (ppbv) are shown for (L_40_PE) − (L_40) in Figure 2b. Increases in the surface PAN mixing ratios are calculated over polluted regions (0.1–0.5 ppbv), reflecting increases in the organic compounds available for the PAN formation.

[32] The change in the NOx mixing ratio (ppbv) is shown for (L_40_PE) − (L_40) in Figure 2c. Decreases of the surface NOx mixing ratios (−0.05–0.2 ppbv) are calculated over both the isoprene- and NOx-emitting regions for the case when 40% recycling fraction for NOx is assumed. Increases (0.01–0.04 ppbv) are found over remote locations where nitrogen-containing compounds can be transported from the source regions [e.g., Sanderson et al., 2003; Hauglustaine et al., 2005; Wiedinmyer et al., 2006; Zeng et al., 2008]. Increased NOx is especially noteworthy in the desert and semidesert regions of central Asia and the western United States. In each of these regions there is an apparent release of NOx imported from neighboring regions that have high emissions of both anthropogenic NOx and BVOC. These increases in NOx occur even with the assumed 40% recycling fraction for NOx. Over the same regions, smaller NOx increases (up to 0.03 ppbv) occur in the cases with 0% NOx recycling, while larger NOx increases (up to 0.09 ppbv) occur in the cases with 100% NOx recycling.

[33] The resulting changes in surface O3 mixing ratios due to the effect of a +5-K increase in temperature on emissions are shown for the 0, 40, and 100% recycling fractions for NOx in Figures 3a3c. The O3 changes due to the effects of +5 K on emissions are larger than those due to −5 K (the global averaged differences are 1.3 ppb for (L_40_PE) − (L_40) versus −1.1 ppb for (L_40_ME) − (L_40)), because the responses of O3 to the temperature changes are subject to the effect of nonlinearities in the BVOC emissions which increase exponentially with increasing surface air temperature.

Figure 3.

The changes in surface O3 mixing ratios (ppbv) in July due to the difference between the cases with increased emissions caused by a 5-K increase in temperature (L_0_PE, L_40_PE, and L_100_PE) and the base cases (L_0, L_40, and L_100) for the assumption of (a) 0% recycling of NOx from isoprene nitrates, (b) 40% recycling of NOx from isoprene nitrates, and (c) 100% recycling of NOx from isoprene nitrates.

[34] Decreases in surface O3 (up to −4.5 ppbv) are found over tropical forests because more isoprene directly reacts with O3 to increase its loss and more isoprene nitrates act as a sink of NOx to decrease its production. Under low-NOx conditions, where O3 formation is sensitive to the level of NOx, more NOx is removed as isoprene nitrates for increased emissions of isoprene as well as other BVOC and more is scavenged as XNITR in the case with a lower recycling fraction for NOx. The largest increases in surface O3 (up to 11.4 ppbv) are found near or downwind from polluted areas. These increases occur in two types of locations: (1) source regions in which O3 formation is likely to be sensitive to VOC emissions, such as over north China and northern Europe, and (2) downwind locations that have increased NOx (Figure 2c) over the Middle East and central Asia [e.g., Li et al., 2001; Lelieveld et al., 2002; Wild et al., 2004]. Our model results suggest that additional measurements for the downwind regions would be helpful in providing constraints on the temperature response of O3 and the recycling fraction for NOx.

[35] The response of O3 to the effect of increases in temperature on emissions shows that the changes in O3 are largely determined by the assumed recycling fraction for NOx. The O3 increase is larger in the case with higher recycling of NOx. In the case with 0% recycling of NOx, O3 mixing ratios mostly increase (up to 7.3 ppbv) over high-NOx locations such as Germany, Beijing, Los Angeles, and the New York area. On the other hand, high BVOC emissions regions near polluted areas (e.g., the southeastern United States) result in a decrease in O3 (up to −4.5 ppbv). This result indicates that the increased BVOC emissions near polluted areas efficiently remove O3 precursors in the case with lower recycling fractions for NOx.

[36] Zhang et al. [2008] used CMAQ [Byun and Schere, 2006] with the CBM-IV chemical mechanism at the horizontal resolution of 36 km and found that the increase in BVOC emissions due to temperature increases caused a decrease in surface O3 by up to 1 ppb over the southeastern United States. Their results are consistent with our results using a 0% recycling fraction for NOx. The lumping approach used in the CBM-IV to reduce the complexity of the degradation schemes for higher hydrocarbons is different from that used in this work. A structural lumping, which groups species according to their bond type is used, as opposed to the molecular lumping approach used in this work, where groups of reactions of entire molecules are combined. Pöschl et al. [2000] and von Kuhlmann et al. [2004] have discussed the differences between these two approaches and attributed the major cause of differences to the approximate 50% loss of carbon in the initial reaction of isoprene with OH in the CBM scheme. Our results suggest that the more likely cause for the decrease in surface O3 is the implicit assumption of large losses of NOx in the CBM scheme, which lumps isoprene nitrate into one organic nitrate. The reaction of organic nitrate with OH in the CBM mechanism yields HNO3 and thus a 0% recycling fraction for NOx.

4.2.2. Sensitivity of Chemical Reaction Rates to Temperature Increases

[37] The changes in surface O3 mixing ratios due to the effect of changes in chemical reaction rates associated with temperature increases are shown for the 40% recycling fraction for NOx in Figure 4. Surface O3 increases (up to 4.2 ppbv) with increased temperature at most continental surface locations regardless of the assumed NOx recycling fraction. The response of O3 to temperature through changes in the chemical reaction rate coefficients occurs mainly because higher temperatures increase PAN decomposition, and the resulting release of NOx, HOx, and organics causes increased O3 formation [Carter et al., 1979]. Steiner et al. [2006] used CMAQ with SAPRC99 chemical mechanism at the horizontal resolution of 4 km and found the average afternoon O3 increase from 0 to 1 ppb (rural) up to 2–4 ppb (urban) owing to the effect of 2–3 K increase on chemical reaction rates during summertime. This may be comparable to our results over land areas (0–2.5 ppbv) owing to the effect of a 5-K increase on a per-degree basis.

Figure 4.

The change in surface O3 mixing ratios (ppbv) in July due to the effects of a 5-K increase in temperature on the chemical reaction rate coefficients (L_40_PC) for the assumption of 40% recycling of NOx from isoprene nitrates.

4.2.3. Sensitivity to Combined Simulations

[38] Large uncertainties have been found in the response of surface O3 over the southeastern United States, when the effects of climate change on isoprene emissions from the biosphere are considered in modeling experiments [e.g., Hogrefe et al., 2004; Tagaris et al., 2007; Racherla and Adams, 2008; Wu et al., 2008; Zhang et al., 2008]. Our results show decreases in O3 (down to −2.5 ppbv) for a 0% NOx recycling fraction but increases in O3 (up to 10 ppbv) with a 100% recycling fraction and a uniform 5-K increase in temperature over the southeastern United States (Figure 5).

Figure 5.

The change in surface O3 mixing ratios (ppbv) in July due to the effects of a 5-K increase in temperature on the emission and chemical reaction rate coefficients for the assumption of (a) 0% recycling of NOx from isoprene nitrates, (b) 40% recycling of NOx from isoprene nitrates, and (c) 100% recycling of NOx from isoprene nitrates.

[39] The rates of increase of O3 in polluted source regions due to chemistry and emissions are less than 10 ppbv for 0% and 40% NOx recycling fractions and are 10–14 ppbv for a 100% NOx recycling fraction, respectively (Figure 5). A similar analysis by Steiner et al. [2006] for the San Joaquin Valley in California found an average afternoon O3 increase of 0–2 ppb in rural locations and 3–5 ppb near urban areas during summertime due to the effect of a 2-K temperature increase on biogenic emissions alone. The SAPRC scheme used by Steiner et al. [2006] assigns isoprene nitrate into a lumped organic nitrate with a 65% recycling fraction for NOx, and thus falls somewhere between our cases with 40% and 100% recycling. Although our results lack the sharp distinction between geographical regions shown by Steiner et al. [2006] (who used 4-km horizontal resolution in a regional simulation), our result for California is comparable on a per-degree basis.

4.2.4. Sensitivity to Model Resolution and NOx Recycling Fraction

[40] The surface O3 mixing ratios for H_100, L_100, and L_40 in July are compared to investigate the relationship between the change in O3 due to increases in the horizontal resolution and the change due to the NOx recycling fractions (Figure 6). The high-ozone region (54–63 ppb) for L_100 is extended further from the source regions to east than that for H_100. Better agreement can be achieved by reducing the NOx recycling fraction, as shown in L_40. This result illustrates that the treatment of isoprene nitrates as a loss for NOx in a coarse-resolution model compensates for the higher OPE that occurs in spreading NOx emissions over a larger box. Thus the lower NOx recycling fraction inferred in the study by Horowitz et al. [2007] than that deduced by Paulot et al. [2009] likely includes the effect of horizontal resolution in the model calibration. The lower recycling required at the coarse model grid is probably related to the conversion of NO2 to HNO3 in high-OH and low-VOC conditions because the artificial mixing of higher BVOC leads to more formation of ΣANs (more recycling of NOx and thus more production of O3) than actual partitioning into HNO3. In this case, the decrease of NOx availability for O3 production with temperature in rural areas should not be related to BVOC emissions in rural areas but to those in urban areas. The reasons for the specific model calibration should be elucidated before implementing it these calibrations in future predictions. In section 4.3.2.3, the correlation of O3 with ΣANs is compared between model results and observations.

Figure 6.

Ozone modeled surface mixing ratios (ppb) for (a) H_100, (b) L_100, and (c) L_40. All maps are monthly averages for July.

4.3. Comparison With Observations

[41] The performance of the model in simulating the observed O3 and other intermediate species has been presented in earlier studies [Rotman et al., 2004; Ito et al., 2007a]. Ito et al. [2007a] have shown the comparison between the model and observations for ΣPNs and ΣANs with different NOx recycling fractions for isoprene nitrates. In section 4.3.1, we provide some additional discussion on the comparison of O3 mixing ratios with measurements. Further, the responses of O3, OH and reactive nitrogen compounds to temperature changes are discussed in section 4.3.2.

4.3.1. Comparisons With Observed O3

[42] In order to identify whether the changes in the NOx recycling fraction affect the agreement of the vertical profile between models and measurements, the average O3 mixing ratios from the simulations with low, middle, and high NOx recycling fractions are compared with ambient measurements (Figure 7). The observational data set was compiled from field campaigns over selected regions and dates by Emmons et al. [2000].

Figure 7.

Comparison between measured O3 [Emmons et al., 2000] and model results at various sites. The blue lines with squares represent the mixing ratios of O3 with 0% NOx recycling. The red lines with circles denote the mixing ratios of O3 with 40% NOx recycling. The green lines with crosses show the mixing ratios of O3 with 100% NOx recycling. Observations are indicated by the black lines. The black dashed lines show the standard deviations of measured values.

[43] The uncertainty ranges in O3 mixing ratios due to different NOx recycling fractions over tropical regions (10 ppbv) are larger than those at northern midlatitudes (8 ppbv) during summer. The ranges over East Brazil (14 ppbv) during the Amazonian forest fire season for September and October are larger than those for other seasons. The ranges are larger at higher altitudes (around 8 km) than those at lower altitudes around 2 km (10ppbv). Ito et al. [2007a] showed that the surface O3 mixing ratios for these regions were sensitive to different emission estimates from biomass burning but that those at higher altitudes (around 8 km) were less sensitive. These results suggest that measurements in the FT over isoprene-rich locations in regions with high-NOx emissions may provide constraints on the NOx recycling fraction.

4.3.2. Comparison of Sensitivities to Temperature

[44] Even if models agree with the current O3 measurements, it would not necessarily guarantee that the models could predict the future O3 well, as we showed by examining the response of O3 to temperature changes due to different NOx recycling rates. However, the model should represent the response of O3 to short-term changes before extrapolating it to long-term changes. The ability of models to predict the response of O3 to short-term changes in temperature can be evaluated in part by examining how ambient concentrations of O3 and other related species respond to day-to-day variations in temperature. It is well known that O3 in polluted regions shows a strong correlation with temperature [e.g., Sillman and Samson, 1995]. Recently, Day et al. [2008] reported measurements for the variation of O3 and various reactive nitrogen species with temperature at the University of California–Blodgett Forest Research Station (UC–BFRS) (1315 m above sea level, 38.9° N, 120.6° W) during the summer of 2001. This location is a forested in the Sierra foothills and is removed from immediate anthropogenic influence, but is strongly influenced by urban emissions from Sacramento (80 km to the northwest) and other urban centers in the San Joaquin valley. Here we examine the predicted correlations with temperatures from model simulations for the same time period in comparison with measured values. These day-to-day variations with temperature are largely influenced by changes in reaction rates and in biogenic emissions associated with temperature [e.g., Steiner et al., 2006; Day et al., 2008]. Correlations with temperature can also be influenced by meteorological patterns, such as correlations between winds, vertical diffusion or cloudiness with temperature [e.g., Jacob et al., 1993], although Day et al. [2008] concluded that these had a relatively minor influence at UC–BFRS during the time period.

[45] Table 3 shows the response of O3, OH, and the reactive nitrogen species to day-to-day temperature changes based on model results for low and high horizontal resolution with different NOx recycling fractions (L_0, L_40, H_40, and H_100) and those reported by Day et al. [2008] at UC–BFRS from June to September. Day et al. [2008] observed the sum of gas-phase HNO3 and thermally labile HNO3 aerosols such as NH4NO3. However, Day et al. [2008] concluded that any aerosol NO3 in the region was nonvolatile and would not have been detected as part of their measurements of HNO3. Similarly, Day et al. [2008] observed the sum of all alkyl and multifunctional nitrates present in the gas or aerosol phase. Thus we calculate the total alkyl nitrate in our model for comparison with their observations. The afternoon (1200–1600 LT) medians of O3, OH, and the reactive nitrogen species are calculated for each day of summer 2001. The slope is calculated from a least squares fit to the median and the daily maximum temperature. The mean value for each species at 298 K is estimated from the function for each scenario. Day et al. [2008] excluded the observations on Saturday through Monday and days where trace gas species were influenced by forest fires from the measured values. We present here all days from the model results of L_0, L_40, H_40, and H_40, since we did not include weekday-weekend differences in emissions or event-specific forest fires in the model (Figures 810). Thus the total number of days included in our analysis (122) is larger than that for measured values (63).

Figure 8.

Modeled (red) and observed (black) ΣPNs/NOz, ΣANs/NOz, HNO3/NOz, and O3 versus daily maximum temperature (K) for 1200–1600 LT, June to September 2001 (L_40 and H_40 simulations) at UC–BFRS.

Figure 9.

Modeled (red) and observed (black) O3 versus ΣANs for 1200–1600 LT, June to September 2001 (L_40 and H_40 simulations) at UC–BFRS.

Figure 10.

Modeled (red) and observed/inferred (black) HNO3/NO2 and OH versus daily maximum temperature (K) for 1200–1600 LT, June to September 2001 (L_40 simulation) at UC–BFRS.

Table 3. Model Results Compared With O3 and Reactive Nitrogen Compounds Measured and OH Inferred From the Observations of Temperature Trends in the HNO3/NO2 Ratio at Blodgett Forest Research Station by Day et al. [2008]a
  • a

    The model results of L_0, L_40, H_40, and H_100 and the measurements are for 1200–1600 LT, 1 June to 30 September 2001.

 L_0L_40H_40H_100Day et al. [2008]
Mean Value at 298 K
O3 (ppbv)5658515764
NOx (ppbv)0.920.980.240.350.7
NOz (ppbv)2.82.92.02.32.3
OH (106 molecules cm−3)8.89.32.22.818
ΣPNs/NOz0.260.260.190.220.49
ΣANs/NOz0.130.100.150.100.20
HNO3/NOz0.610.640.660.680.42
HNO3/NO22.62.67.46.32.0
 
Slope K−1
O3 (ppbv)+2.5+2.8+1.4+2.0+2.2
NOx (ppbv)+0.011+0.015+0.006+0.015−0.0007
NOz (ppbv)+0.11+0.13+0.11+0.14+0.036
OH (106 molecules cm−3)−0.29−0.27−0.031−0.008+1.1
ΣPNs/NOz−0.014−0.013−0.019−0.018−0.023
ΣANs/NOz+0.0037+0.0022−0.0001−0.0009+0.0036
HNO3/NOz+0.0098+0.011+0.019+0.019+0.019
HNO3/NO2+0.081+0.084+0.35+0.22+0.11
4.3.2.1. ΣPNs, ΣANs, and HNO3

[46] Comparisons between model results and measurements at the UC–BFRS site are difficult because the coarse horizontal resolution does not account for the distinctive local geography, in which anthropogenic emissions from urban sources are transported to the measurement site in a region with high-BVOC emissions. The model includes both the Sacramento urban area (80 km upwind from UC–BFRS) and the UC–BFRS site within the same horizontal grid for the low and middle resolutions but not for the high-resolution model (1° × 1°). However, it is noteworthy that the ratios of the relatively long lived NOx reaction products (ΣPNs, ΣANs, and HNO3) to NOz show variations with temperature that are comparable with measurements (Figure 8). The correlation coefficients of ΣPNs/NOz (0.46), ΣANs/NOz (0.17), and HNO3/NOz (0.27) with temperature are comparable to that of Day et al. [2008] (0.52, 0.069, and 0.37, respectively). Despite the low R2 values, Figure 8 shows that the regression lines capture the major patterns present in the data. The model slopes versus temperature are all statistically significant at a 95% confidence interval except for ΣANs/NOz, which shows a weak dependence on temperature in both the model and measurements. The match of the model trend to observed trend varies with the model simulation, but the H_40 trends versus temperature for ΣPNs/NOz (−0.02 ppb K−1), ΣANs/NOz (0.00 ppb K−1), and HNO3/NOz (0.02 K−1) are noticeably closer to the observed trends. The ratio of ΣPNs to NOz decreases, the ratio of ΣANs to NOz does not change, and the ratio of HNO3 to NOz increases with increasing temperature both in the model and in the measurements. The mixing ratios of NOz in model are in good agreement (within 26%) with observations, while the ratios of ΣPNs, ΣANOs, and HNO3 to NOz in the model differ significantly from the measured values at UC–BFRC. The model underestimates ΣPNs/NOz and ΣANs/NOz but overestimates HNO3/NOz, possibly suggesting that the anthropogenic NOx is more quickly converted to oxidized products by reactions with OH than by those with RO2 under urban area conditions in the model where OH is high and BVOC emissions are low.

4.3.2.2. O3

[47] The model scenarios also estimate that day-to-day ambient O3 correlates with temperature (R2 = 0.55 for L_0, 0.56 for L_40, 0.25 for H_40 and 0.34 for H_100) and that the rate of increase is comparable to the observed increase at UC–BFRC. The model scenarios that use a 0% NOx recycling fraction in the low-resolution model and 100% NOx recycling in the high-resolution model show increased O3 (2.5 and 2.0 ppbv K−1) with temperature that is comparable to the observed slope in O3 versus temperature (2.2 ppbv K−1). Our O3 mixing ratios (58 ppbv at 298 K for L_40) are also in good agreement with the observations (Figure 8).

4.3.2.3. ΣANs and O3

[48] Horowitz et al. [2007] found higher slope with higher NOx recycling (100%; 257 O3 per ΣANs) but lower slope with higher yield of isoprene nitrate (8%; 62 O3 per ΣANs). Our results in the surface air using the high yield of isoprene nitrate show lower slope with higher resolution (106 for L_40 versus 69 for H_40) (Figure 9). The slope between model O3 and ΣANs shows similar sensitivity to the model resolution as to NOx recycling from isoprene nitrates (69 for H_40 versus 106 for H_100). These results suggest that the assumptions of higher yield of isoprene nitrate and higher NOx recycling fractions using higher-resolution simulation might be able to match the slope of O3 per ΣANs.

4.3.2.4. OH

[49] Typically, models underestimate OH concentrations in regions with high-VOC and low-NO mixing ratios [e.g., Carslaw et al., 2001; Tan et al., 2001; Thornton et al., 2002; Karl et al., 2007; Kuhn et al., 2007; Farmer and Cohen, 2008; Ren et al., 2008]. A key question here with regard to the temperature dependency of OH is whether OH increases with temperature owing to the ozonolysis of unidentified reactive BVOC as a temperature-dependent source of OH in the atmosphere [Kurpius and Goldstein, 2003; Di Carlo et al., 2004] or owing to the efficient OH recycling via HO2 + RO2 reactions which occur in an unpolluted low-NOx environment [Thornton et al., 2002; Lelieveld et al., 2008; Butler et al., 2008]. In the former case, the sensitivity of O3 to the assumed NOx recycling fraction may be less important, because the additional OH sources would partition anthropogenic NOx into HNO3 with warmer temperatures. In the latter case, the assumed NOx recycling fraction can have a large influence on anthropogenic NOx, but it depends on how much OH forms relative to BVOC and NO levels.

[50] Day et al. [2008] infer that OH rapidly increases with temperature in the mixed layer, on the basis of the observed increase in the ratio of HNO3/NO2 with increasing temperature. The model OH concentrations in the high-resolution version (2–3 × 106 molecules cm−3) are lower than inferred value at UC–BFRC (18 × 106 molecules cm−3). The low OH at high resolution is probably due to the low NOx (0.2–0.4 ppb). The measured NOx (0.7 ppb) is influenced by transport of NOx from urban source regions to the UC–BFRC site by a strong mountain/valley wind over the western slope of the Sierra Nevada [Murphy et al., 2006]. In the low-resolution simulation, our OH (9 × 106 molecules cm−3) is in better agreement with the inferred value than the model at the high resolution, probably because the conditions in the low-resolution version represent relatively high-NOx conditions (0.98 ppb at 298 K for L_40), which are in much better agreement with the observations. In the low-resolution model simulations (Figure 10), OH in the afternoon is calculated to decrease with temperature (−0.27 × 106 molecules cm−3 K−1 for L_40), in contrast to the behavior of OH inferred. Thus one might argue that the additional OH required to match the inferred increase in OH with temperature might be explained by the ozonolysis of unidentified reactive BVOC as a temperature-dependent source of OH in the atmosphere [Kurpius and Goldstein, 2003; Di Carlo et al., 2004]. However, the model results for HNO3/NO2 (2.6 at 298 K) and its increase with temperature (0.08 K−1) are in good agreement with the measurements without the assumption of a temperature-dependent source of OH.

[51] The increases in HNO3 with temperature can be associated with changes in the PAN thermal equilibrium providing increased NO2, which reacts with OH to form HNO3. The increases in both OH and NO2 versus temperature can be seen earlier in the day, but are no longer visible in the model afternoon values. A similar tendency appeared in the calculations reported by Sillman and Samson [1995]. The observed increase in HNO3/NO2 with temperature reported by Day et al. [2008] may also reflect increases in HNO3 in upwind urban areas rather than changes in OH at the site. Murphy et al. [2006] suggested that the observed NOy at noon at UC–BFRS reflected urban emissions during the morning rush hour located 5–6 h upwind. As noted by Day et al. [2008] and Farmer and Cohen [2008], HNO3 is produced more rapidly upwind where NO2 mixing ratios are much higher [Murphy et al., 2006]. Thus the overestimate of OH and its increase with temperature in the steady state calculation from the HNO3/NO2 ratio is probably associated with the content of the transported air (i.e., increased HNO3 with temperature in the upwind urban areas). In this context, Sillman and Samson [1995] reported that OH and HNO3/NO2 both increase sharply with temperature in calculations for an urban location. Further, Lelieveld et al. [2008] and Butler et al. [2008] calculated an insignificant change in OH (±5%) during June, July, and August over this region when they changed the fraction of OH recycling associated with isoprene oxidation. Thus, at moderately high NOx conditions, OH concentrations are not expected to increase significantly as a result of OH recycling. In this context it is noteworthy that our OH in the low-resolution model is in fairly good agreement with that of Dillon et al. [2002], who calculated OH (11 ± 5 × 106 molecules cm−3) in the Sacramento urban plume using a Lagrangian model and VOC measurements.

[52] Finally, model results using the low- and high-resolution versions for July are also compared with measurements of isoprene, OH, and NO from three summer field campaigns: AEROBIC97 (39.9° N, 21.7° E), PROPHET98 (45.6° N, 84.7° W), and PMTACS-NY (44.4° N, 73.9° W) (Table 4). The model results at rural sites are in good agreement with the daytime peak OH levels in the Northern Hemisphere [Carslaw et al., 2001; Tan et al., 2001; Ren et al., 2006]. These results might justify the predictability of the model at the rural sites to support our interpretation over isoprene-emitting regions near polluted source areas. On the other hand, Butler et al. [2008] have shown that the modeled OH concentrations over the three sites were too high with artificial OH recycling. Clearly, further work is needed to investigate the role of OH relative to BVOC and NO mixing ratios before the practical application of OH recycling to CTM.

Table 4. Measured OH Concentrations and Mixing Ratios of Isoprene and NO at Local Noon From Field Measurements Compared With Model Simulationsa
 L_0L_40H_40H_100AEROBIC97
Isopreneb0.590.521.91.61
NOc334385113<100
OHc3.44.12.33.14
 L_0L_40H_40H_100PROPHET98
Isoprened0.580.512.21.72
NOe1481574060<100
OHe4.75.31.42.04
 L_0L_40H_40H_100PMTACS-NY
Isoprenef0.930.852.01.6<1
NOf178187130150≈100
OHf4.14.52.22.73

5. Summary and Conclusions

[53] We investigated the effects of temperature changes (±5 K) on O3 mixing ratios by examining changes in BVOC emissions and chemistry reaction rate coefficients. Our focus was to elucidate the relationship between the treatment of isoprene nitrates and the horizontal resolution in the model and to evaluate the ability of the model to predict temperature-dependent responses of O3 and long-lived NOx reaction products (ΣPNs, ΣANs, and HNO3) on the basis of observed response.

[54] Global CTM calculations showed that the increases in the tropospheric O3 burden as a result of the increases in BVOC, in particular isoprene (270–940 TgC a−1), emissions associated with temperature increases strongly depended on the assumption of the amount of recycling of NOx from isoprene nitrates (17–57 Tg for 0–100% NOx recycling). Substantial differences in the global mean O3 burden (40 Tg) were caused by the differences in the assumption of the amount of recycling of NOx from isoprene nitrates between 0% and 100%. These differences were equivalent to the differences in the model resolution between the high resolution (302 Tg) and the low with the 40% NOx recycling fraction (340 Tg). Our results in the surface air at a forested site in California showed lower slope with higher resolution (106 for L_40 versus 69 for H_40). The slope between model O3 and ΣANs showed similar sensitivity to the model resolution as to NOx recycling from isoprene nitrates (69 for H_40 versus 106 for H_100). The model versus measured slope might be used to infer the yield of isoprene nitrate and NOx recycling fraction. However, our results indicate that the model slope and the recycling efficiencies of NOx in a CTM are related to the isoprene nitrate chemistry as well as the grid resolution. It suggests that better agreement in the response of O3 to temperature increases with observations could be achieved by using higher-resolution model with higher yield of isoprene nitrate and higher NOx recycling fraction from isoprene nitrate.

[55] Model results have been compared with the observed responses in O3 and reactive nitrogen compounds to temperature at a forested site in California observed by Day et al. [2008]. The model result showed a rapid increase in O3 (2.0 ppb K−1) and HNO3/NOz (0.02 K−1), little change in ΣANs/NOz (0.00 ppb K−1) and a decrease in ΣPNs/NOz (−0.02 ppb K−1) with increasing temperatures in the surface air. These compare closely with the measured variations, although the agreement varies with scenarios and species considered (see Table 3 and Figure 8). These results might justify the predictability of the model at the rural sites to support our interpretation over isoprene-emitting regions near polluted source areas. However, the interactions of terrestrial ecosystems with human activities and climate change are highly complex and the temperature sensitivities of O3 may vary in different conditions. Further work is needed to provide constraints on the response of O3 to temperature increases for a more accurate model prediction of effect of climate change on air quality.

Acknowledgments

[56] We thank R. C. Cohen and his colleagues for kindly providing the observational data set at UC–BFRS. Support for this research was provided to A. Ito by the Innovative Program of Climate Change Projection for the 21st Century (MEXT) and to J. E. Penner and S. Sillman by the U.S. National Science Foundation (grant ATM 0454838) and the Environmental Protection Agency (STAR grant RD-83337701-0). All of the global simulations in this study were performed using the SGI Altix4700 at the JAMSTEC. We wish to thank X. Liu for making the AeroCom emission data available for use in IMPACT.