Production of nitrogen oxides (NOx = NO + NO2) by lightning (LtNOx) is the most uncertain among the global NOx sources, with recent estimates ranging from about 1–20 Tg(N)/yr. Previous studies of LtNOx have focused mainly on its role in the tropospheric NOy (reactive nitrogen) and O3 budgets. We show that the global mean OH concentration is also very sensitive to LtNOx. Furthermore, despite the fact that the largest changes in NOx due to lightning are in the upper troposphere, where reactions with OH are generally slower, we find that the sensitivity of the mean tropospheric lifetime of methane (CH4) and methylchloroform (CH3CCl3) to assumptions about LtNOx are as large as the sensitivity of the tropospheric O3 burden. Thus, an improved understanding of LtNOx will be important for our ability to accurately simulate the tropospheric oxidizing efficiency and its changes over time.
 Lightning plays an important role in atmospheric chemistry by producing nitric oxide (NO), a precursor of ozone. However, the violent nature of this phenomenon makes it difficult to accurately determine the global amount of lightning-produced NOx(LtNOx). LtNOx still has the largest uncertainty of all NOx sources, with estimates ranging from about 1–20 Tg(N)yr−1 [Lawrence et al., 1995; Price et al., 1997], or ∼2–30% of the total global NOx emissions. Many factors contribute to this large difference in the estimates, among them, uncertainties in the current understanding of the fundamental processes of lightning generation itself, such as the charge separation mechanism and the partitioning between intra-cloud, inter-cloud and cloud-to-ground flashes, as well as LtNOx-specific processes, such as the NO formation (“freeze-out”) mechanism, the number of NO molecules produced per flash, and its vertical distribution. Unlike ground sources, lightning produces NOx in and around cumulonimbus clouds. Convective updrafts can then carry them to the upper troposphere (UT) where they have a longer lifetime than near the surface. LtNOx has been shown to enhance O3 levels by up to 30% [Stockwell et al., 1999]. LtNOx also affects the budget of the hydroxyl radical (OH), and in turn the tropospheric oxidizing efficiency. These effects, however, have not received as much attention as the effect on O3, since a large effect on the oxidizing efficiency would be somewhat counterintuitive due to the strong temperature dependence of the oxidizing reactions of gases such as CH4 and CH3CCl3. In this study we investigate the effects of LtNOx on NOx, OH, the tropospheric oxidizing efficiency and O3.
2. Model Description and Runs Setup
 For this study we employ the Model of Atmospheric Transport and Chemistry, Max-Planck Institute for Chemistry version 3.1 (MATCH-MPIC hereafter). MATCH-MPIC is an offline chemistry transport model based on the NCAR CCM2 (Community Climate Model-Version 2). It consists of a meteorology module and a chemistry module which comprises 141 gas phase reactions of 56 species, describing the major known sources and sinks of O3 and its associated chemistry. For a comprehensive description of the model, see von Kuhlmann et al. [2003a], Lawrence et al. , Rasch et al.  and references therein. The runs for this study use the same version as Lawrence et al. [2003a] (except for the LtNOx source), with input data from the NCEP/NCAR reanalysis on a reduced horizontal resolution of T21 (about 5.6° × 5.6°), 28 vertical sigma levels from the surface to ∼0.2 hPa, and a timestep of 30 minutes. The runs are done for 1997, with a four month spin-up time. The deep convection scheme, on which the modeled lightning is based, is from Zhang and McFarlane , and accounts for the vertical transport of a wide range of inorganic and organic precursors of OH and O3 [Lawrence et al., 2003a]. Evaluations of the vertical distribution of a number of species including CO, O3, alkenes and CH3OOH have not shown a clear indication of a substantial over- or underestimate of vertical transport in the model [Lawrence et al., 1999; von Kuhlmann et al., 2003a, 2003b; Lawrence et al., 2003b]. Evaluations using radon and methyliodide have not yet been carried out for MATCH with this scheme.
 The parameterization for the horizontal distribution of lightning used in MATCH-MPIC is based on Price and Rind : F = 3.44 × 10−5H4.9 for continental convective clouds, and F = 6.4 × 10−4H1.73 for marine clouds, where F is the flash frequency (flashes/min/8 × 10 box) and H is the modeled cloud top height in km. Although widely used, deficiencies with this simple approach have to be acknowledged. In preliminary results of a parallel study, comparing to satellite observations, we have found that the parameterization captures the main features of lightning activity (land/sea, tropics/extratropics and seasonal patterns); however, it also tends to overestimate (underestimate) flash activity for the tropical (extratropical) latitudes, and within the tropics, modeled flash densities over Africa are lower than over South America. It is not yet clear whether this is due to the lightning parameterization, including neglected effects such as aerosols, or the convection parameterization. The vertical distribution of LtNOx in MATCH-MPIC for this study is based the C-shaped vertical profiles developed by Pickering et al. .
 We have carried out 5 different runs in which we have only varied the LtNOx source strength and kept all other model parameters the same. The LtNOx production rates are 0, 2, 5, 10 and 20 Tg(N)yr−1(hereafter the L0, L2, L5, L10 and L20 runs, respectively), which reflects the extremes of the uncertainty range (the L0 run is not realistic, but rather is used to help illustrate the net effect of any other given LtNOx source strength). We will focus our discussion on the L5 run, since it best reflects the currently most accepted production estimate for the LtNOx source.
3.1. Effects on NOx
 We first examine how NOx itself is affected by LtNOx by comparing the L5 and L0 runs. The largest effects of LtNOx are in the tropical UT (Figure 1), particularly over equatorial South America and eastern Indonesia and, to a lesser extent, over central Africa, which is consistent with observations that show that lightning activity is heavily weighted towards the tropics and continental regions [Christian et al., 2003], though as noted above, the effect over Africa is likely underestimated and over South America and Indonesia overestimated. Despite the fact that 20% of the total LtNOx mass is released in the first 2 km above the continental landmasses, the increase in the tropical lower troposphere (LT) is relatively small, since there LtNOx must compete against the other major surface NOx sources, namely soil emissions, biomass burning and industrial emissions.
3.2. Effects on OH, The Tropospheric Oxidizing Efficiency and O3
 LtNOx has a substantial effect on the global mean OH mass and volume-weighted concentrations ([OH]m and [OH]v hereafter) [Lawrence et al., 2001]. Table 1 shows that [OH]m ([OH]v) increases by about 50% (>63%) from the L0 to the L20 run. For every 5Tg(N)yr−1 increase in the LtNOx source strength, [OH]m ([OH]v) increases by about 22% (29%) between the L0 and L5 runs, 10% (12%) between the L5 and L10 runs, and 6% (6%) between the L10 and L20 runs, indicating a tendency towards saturation. The vertical distribution of OH and its regional changes due to LtNOx are depicted in Figures 2–4. While the largest absolute increases in the OH concentration occur in the middle troposphere (MT) to UT (∼300–500 hPa) (Figure 3), the largest relative enhancement occurs higher up (∼200–300 hPa) (Figure 4). The reason for this is twofold; first, the extremely low OH concentrations in the cold and dry tropical UT in the L0 run cause even a small absolute increase in OH to translate into a large relative increase in that region. Second, near the tropopause, OH production by O(1D) + H2O is slow, so that secondary sources such as the reaction of NO with HO2 become relatively more important than at lower altitudes. Furthermore, HOx yields for reactions such as photolysis of acetone and oxidation of methane are enhanced in the presence of NOx [Folkins and Chatfield, 2000].
Table 1. OH Mass- and Volume-Weighted Tropospheric Mean Concentrations and Lifetimes of CH4 and MCF for the 5 Different Model Runs
Calculated using online budgets.
Approximated with monthly mean model output (see text).
 How do these changes in OH affect the oxidizing efficiency of the troposphere? We examine this with respect to two long-lived trace gases, methane (CH4) and methyl-chloroform (CH3CCl3, or MCF for short). Their lifetimes and τ(MCF)) for the five runs are given in Table 1. While = 10.26 years is computed for the L0 run, the L5 run yields = 8.79 years, a reduction of nearly 15%. A similar relative reduction in τ(MCF) is also computed. Note that since CH4 is simulated in the model, we can use the online model budget routines to compute (using the tropopause based on the temperature lapse rate criterion). However, since we do not simulate MCF, we approximate τ(MCF) using monthly mean OH and temperature fields from the model output, a climatological tropopause, and a uniform distribution in the troposphere. We verify that the error in this approximation is small by doing the same for and comparing to the online budget values (Table 1), showing that the offline estimates are biased ∼5.5% and ∼3.9% high in the L0 and L5 run, respectively; thus, τ(MCF) values in Table 1 are also expected to be slightly overestimated.
 Comparing our computed lifetimes to recent estimates given in the literature (Table 2), we see that at both extremes of the LtNOx source magnitude range, the computed and τ(MCF) values are still barely within the current range of uncertainty. A tendency is hard to discern, indicating the uncertainty in factors other than LtNOx that also affect OH and therefore and τ(MCF). However, we notice that in all five references corresponds to or includes the 2–5 Tg yr−1 LtNOx production range. τ(MCF) values correspond to the higher end of the spectrum, even after accounting for the slight high bias in our offline estimates. The inconsistency between the implied LtNOx source magnitudes based on and τ(MCF) is related to the different OH spatial distributions used in the different studies as well as to the different temperature dependencies of the reaction rates of the two gases with OH.
Table 2. Lifetimes of CH4 and MCF and Their Corresponding LtNOx Production Range According to Various Studies
Lower atmospheric lifetime values, rounded to one decimal point.
 Which part of the atmosphere is mainly responsible for the computed changes in the lifetime? We divided the troposphere into compartments (following Lawrence et al. ) and computed the amounts of oxidized CH4 as well as the fraction (in %) of the total tropospheric CH4 burden oxidized in each compartment for both the L0 and L5 runs (Figure 5). In absolute as well as relative terms, most of the CH4 is oxidized in the tropical troposphere, especially in the LT. The largest absolute increase in the amount of CH4 oxidized occurs in the two lowermost compartments of the tropical troposphere due to the strong temperature dependence of the oxidation reaction despite the only moderate increase in OH there, whereas the largest relative increases occur above 500 hPa (with increases of ∼65 and ∼47%, respectively, in the upper two tropical compartments) due to the larger relative increase in OH there (Figure 4).
 Finally, we can compare these effects to the effects of LtNOx on O3. In our runs, we find that the tropospheric burden of O3 increases from 377 to 430 Gg, or ∼14%, between the L0 and L5 runs. Thus, the effects of LtNOx on the tropospheric oxidizing efficiency are similar in magnitude to the effects on O3, while the effects on [OH]m and [OH]v are even larger.
4. Discussion and Conclusions
 Our results show that the OH concentration and the tropospheric oxidizing efficiency are strongly sensitive to the global LtNOx source magnitude, comparable to, or larger than, the effect of LtNOx on O3. For our model, the uncertainty in LtNOx alone translates into an uncertainty in our simulated and τ(MCF) which is about as large as their overall uncertainty based on other recent studies using various approaches to estimate their lifetimes. Based on this, it is tempting to suggest that improved estimates of the lifetimes of traces gases such as CH4 and MCF could be used to constrain the global source of NOx from lightning. However, this is currently impractical, due to the many other uncertainties involved. For instance, the LtNOx parameterization introduces an unquantified error in both its horizontal and vertical distribution, which is important because the sensitivity of OH to additional NOx varies regionally. The competition of LtNOx with other NOx sources (e.g., biomass burning) will be influenced by uncertainties in their magnitudes and in their convective lofting to the upper troposphere; this also applies to other competing HOx precursors (e.g., acetone). Finally, there are still uncertainties in many key reaction rates, such as PAN formation and thermal degradation. Reduction in these uncertainties will lead to quantitative changes in the effects of LtNOx on OH and on the oxidizing efficiency computed with improved models in future studies. We do not expect, however, that this will result in a qualitative change in our main conclusion, that is, that the strong effect of LtNOx on OH and the tropospheric oxidizing efficiency should be considered alongside its effects on O3, and reinforces the need for a more accurate determination of the source of NOx from lightning in future studies.
 The authors would like to thank Phil Rasch for the support of MATCH. This work was supported by funding from the German Ministry of Education and Research (BMBF), project 07-ATC-02.