CTM study of changes in tropospheric hydroxyl distribution 1990–2001 and its impact on methane



[1] Impacts of emission changes on hydroxyl (OH) and methane lifetime in the troposphere are studied using an emission inventory for the period 1990–2001 as input to a global Chemical Transport Model (CTM) run with repeated year 2000 meteorology. Global OH is estimated to increase, in average by 0.08% per year, with exception of 1997 and 1998 when deviations from the trend result from unusually large biomass burning emissions. The global OH evolution is highly dependent on the CO to NOx emission ratio, and a statistically significant linear relation between this ratio and the OH concentration is found. However, we find large regional variations of OH distributions and methane loss. One third of a calculated increase in global average methane loss (0.50% per year) can be explained by OH changes resulting from emission changes of CO, NOx and NMVOCs, while two third is due to increases in methane itself.

1. Introduction

[2] The hydroxyl radical is the main oxidant in the troposphere, initiating chain reactions removing a number of pollutants and climate gases. Primary production of OH takes place when ozone is photolysed by ultraviolet (UV) sunlight yielding electronically excited oxygen O(1D) followed by reaction with H2O [Levy, 1971]. OH reacts rapidly with CO and hydrocarbons, primarily CH4. A following reaction chain may be an efficient sink of OH although there is also a possibility of recycling through reaction of HO2 with NO or O3, or photolysis of peroxides. Lelieveld et al. [2002a, 2004] calculates that this secondary formation is of similar magnitude as the primary and that nearly half of the OH initially lost in the oxidation of CH4 and CO is recycled. The degree of recycling is very dependent on the NOx (NO + NO2) levels. At very high NO2, reaction of OH with NO2 is an important OH loss reducing the recycling. At moderate to high NOx recycling is efficient, a perturbation due to increased emissions of CO would not be efficient in reducing OH. Efficient recycling also implies formation of O3 which leads to more primary OH production. In regions with low NOx an increase in CO would enhance OH loss, formation of peroxides like HO2 recombination into H2O2 dominates and recycling is inefficient.

[3] In this study we focus on how emission changes of NOx and CO during the last decade determined the tropospheric OH distribution and how this affected methane oxidation in the troposphere. Oxidation by OH in the troposphere is the major loss of atmospheric methane. Chemical reactions in the stratosphere constitute 7–11% and dry deposition 1–10% of the atmospheric methane loss [Lelieveld et al., 1998]. Despite the large sensitivity of OH to precursor emissions, which have changed significantly over historic time, CTM calculations give only a small trend in the global averaged concentration since pre-industrial times [Berntsen et al., 1997; Wang and Jacob, 1998; Lelieveld et al., 2002a] and also through the decades up to the nineties [Karlsdóttir and Isaksen, 2000; Dentener et al., 2003; Wang et al., 2004]. Concurrent increases in NOx, CO and hydrocarbons led to increases in both OH production and loss. The current global averaged recycling probability of about 0.5 makes the system quite stable to perturbations [Lelieveld et al., 2002a]. Despite the apparent stability there have been significant regional changes. An increase in OH levels over the continents is found and a decrease over the oceans away from NOx sources. In general the recycling probability has increased, but many NOx scarce regions over tropical oceans have low recycling probability. OH in these areas is therefore sensitive to perturbations of CO and CH4 [Lelieveld et al., 2002a]. These regions have the highest OH concentrations and are also important areas for oxidation of CH4 by OH due to the temperature dependency of the reaction rate. Another interesting aspect is the large variations in the growth rate of tropospheric methane during the last decade [Dlugokencky et al., 2003]. Several suggestions for these variations have been given [Wang et al., 2004, and references therein].

[4] Interestingly, indirect methods of OH determination through measurements of methyl chloroform and 14CO [Krol and Lelieveld, 2003; Prinn et al., 2005; Bousquet et al., 2005; Manning et al., 2005] predict large interannual variations and multiyear trends in OH during the last quart century. These patterns are at odds with CTM calculations and the current understanding of the recent trend in methane [Krol and Lelieveld, 2003].

2. Model and Experiment

[5] The tropospheric version of the OsloCTM2 model was run in T42 resolution (approximately 2.8° × 2.8°) with 40 vertical layers to study the chemical development between 1990 and 2001 using a spin-up time of 8 months. The model and its performance are described in more detail by Berglen et al. [2004], Isaksen et al. [2005] and Dalsøren et al. [2006]. Since we focus on changes in emissions of constituents involved in tropospheric chemistry and their effect on oxidants, stratospheric boundary conditions and meteorology for the same year (2000) were used for all the years in the simulation. In order to study regional and intercontinental transport due to shifting year to year amounts and positions of vegetation fires, CO from these emissions in 1997 and 1998 were used as a tracer. For Siberia plus the 10 regions in Figure 1, the tracer was given an e-folding lifetime of 30 days.

Figure 1.

Trends in non-biomass burning anthropogenic emissions of CO and NOx, CTM calculated trends in tropospheric OH and CH4 removal, and correlation coefficient for linear regression of OH dependency on total regional CO/NOx emission ratio. Blue numbers denote an increase, red a decrease. Regions marked purple are characterized by a high correlation coefficient, regions marked green intermediate correlation coefficient and regions typed in black low or negative correlation coefficient.

[6] We used the emission inventory from the EU-funded project POET [Olivier et al., 2003]. The database contains emissions of CO, NOx and NMVOCs as well as methane surface fields based on observations. The biomass burning emissions have month to month temporal variation. Outside the tropics the biomass burning emissions are based on UN-ECE/FAO (United Nations Economic Commission for Europe and Food and Agriculture Organisation of the United Nations) forest fire activity reports. There are limited historical statistics available for the tropics and there the use of a climatological mean results in repeated yearly total emissions and seasonal variation until 1997 when satellite data become available. The rest of the anthropogenic emissions include inter-annual variations. Figure 1, which gives the trends for CO and NOx in different regions, shows that there are large regional differences. Regulations reduced emissions in some regions but sometimes this has been more successful or only taking place for one of these key components. The anthropogenic NMVOC emissions show smaller change throughout the period. The natural emissions datasets contain monthly variation and we assumed no trend in these emissions. All sectors in the used emission inventory have inherent uncertainties. The reported numbers [Olivier et al., 2003] only include qualitative (small, medium, large, very large) and not quantitative measures making sensitivity studies on the uncertainties difficult. The largest uncertainties are likely found for biomass burning (large) and sectors in regions like Asia (medium-large) going through rapid economical development.

3. Results

3.1. Global Tropospheric OH Distribution

[7] In agreement with most CTM studies and opposite to methyl chloroform inversion methods we find higher OH concentrations in the Northern Hemisphere (NH) than in the Southern Hemisphere (SH). In the time period 1990–2001 the yearly mean hemispheric ratio varies in the range 1.08 to 1.17. A large hemispheric difference is found between the summer seasons when higher levels of pollutants in the NH result in both more primary production and recycling. Our calculations indicate an average increase in global tropospheric OH (volume integrated) of 0.08%/yr (Figure 2). The small positive trend is driven by changes in the NH (0.15%/yr) whereas there is no significant trend (0.01%/yr) in the SH. There are large deviations from the positive trends (Figure 2) in the ENSO (El Niño Southern Oscillation) years 1997 and 1998 which had large emissions from vegetation fires in both hemispheres [Duncan et al., 2003a; van der Werf et al., 2004]. The results are consistent with indications of high biomass burning emissions in 1997 and 1998 reducing OH on regional [Williams et al., 2001; Duncan et al., 2003b] and hemispheric scales [Prinn et al., 2005; Manning et al., 2005]. Biomass burning is characterized by rather inefficient combustion yielding high CO/NOx emission ratio. The result is a reduction in global averaged OH [Dalsøren, 2001]. The importance of this emission ratio for the global averaged OH is demonstrated in Figure 2 where the predicted OH concentration from a simple linear regression is shown. The relation found from least squares fit is OH = ab · equation image with a = 1.25 · 106equation image and b = 8.6·103equation image. The relation is statistically significant with a correlation coefficient of 0.96. The relation could make it possible to predict changes in global tropospheric OH due to changes in emissions of gases involved in tropospheric chemistry. As discussed in more detail in section 4, prerequisites are that the CO and NOx emissions are in a certain range and that changes in hydrocarbon emissions are modest since this was the case for our study period. In an analytical solution to a box model Wang and Jacob [1998] indicated that global tropospheric OH varied proportionally to Sn/(Sc)3/2, with Sn being tropospheric sources of NOx and Sc tropospheric sources of CO and hydrocarbons. A weaker dependency on the same ratio was found with a global 3D model.

Figure 2.

Yearly global averaged tropospheric OH from the CTM, predicted OH from linear regressions analysis and CO/NOx emission ratio.

3.2. Regional OH Changes

[8] In this section we take a closer look at regional OH changes. Figure 1 summarizes regional OH trends and correlation coefficients from linear regression analysis for dependency on regional CO/NOx emission ratios.

[9] For the regions marked with purple in Figure 1 (Latin America, South America, Southeast Asia) the correlation coefficients (r = 0.72–0.94) indicate that much of the changes in OH can be explained by variations in the region's CO/NOx emission ratio. The CO emission decrease in Latin America is driven by decreases in the most important sectors in Mexico. NOx emissions increase over the whole area. A deviation from a positive OH trend is found in 1998 when there were large emissions from regional vegetation fires. Southern America has an increase in man made emissions (Figure 1) except for biofuel use in Brazil. However, natural sources and vegetation fires are dominating factors on this continent and the evolution of OH is dependent on inter-annual variation of fires. In Southeast Asia most CO emissions increase. The increases in NOx emissions are dramatic (>60%). Emissions from road traffic are more than doubled in many regions. Except for the ENSO years 1997–98 there is a large steady increase in OH.

[10] In the regions marked green (N. America, Southern Europe + Mediterranean) OH shows some dependency on the regional CO/NOx emission ratio. For Northern America the overall small changes in the pure anthropogenic emissions result in small OH changes. A simulated decrease in OH due to efficient transport from many forest fires within the region [Wotawa and Trainer, 2000] in 1994–1995 is captured by the regression. The regression does not fully copy a 3.8% decrease in OH from 1997 to 1998. Our tracer study suggests that this is due to transport of CO and hydrocarbons from regions in Siberia and Southeast Asia with strongly enhanced burning in 1998 (Figures 3a–3d). A detailed study with 1997 and 1998 meteorology [Spichtinger et al., 2004] and CO as a tracer shows similar patterns but a stronger signal. They concluded that in 1998 a larger fraction of Siberian forest fire CO was subject to intercontinental transport due to ENSO. Spichtinger et al. [2004] found that CO from Siberian fires contributed more to Canadian CO concentrations than Canada's own biomass burning. Our vegetation fire tracers (not shown) and previous studies [Lelieveld et al., 2002b] found that Southern Europe plus Mediterranean is influenced by transport of pollutants from several regions. An OH drop in years with many fires in temperate regions (1994–1995) and globally (1997–1998) is not reproduced with the regional regression. The larger decrease in CO than NOx emissions in the period 1990–2001 could explain the small overall positive trend in OH.

Figure 3.

Average tropospheric CO tracer column (molecules/cm2) for the period July–September in (left) 1997 and (right) 1998. (a and b) Tracer from vegetation fires in Siberia, (c and d) Southeast Asia, (e and f) Northern America, and (g) South America.

[11] The regions typed in black (Figure 1) show no simple correlation with the regional CO/NOx emission ratio. This is the case for several regions and in particular for Northern Europe. In Europe the emissions declined due to regulations, change from coal burning to gas or nuclear power, restructuring and cleaning technology in Eastern Europe, and economical decline after breaking up of the former SU [Lövblad et al., 2004]. However, emissions increased from some sectors like oil and gas related activity and international shipping. The region is dominated by anthropogenic activities, low CO/NOx emission ratio and quite high NOx concentration. Non-linear chemistry is therefore important and could explain why OH decreased even if the CO/NOx emission ratio decreased. Transport from large forest fire emissions in 1998 (Figures 3e and 3f) in North America is also a factor that disturbs a simple OH dependency on regional emissions. Forster et al. [2001] found that import from North American fires made a larger contribution to their CO tracer than all European emissions in the second half of August 1998. In both Northern and Southern Africa the trends in pure anthropogenic emissions are similar for CO and NOx due to large increases in CO emissions from biofuel. Small trends in OH are found though there are significant fluctuations for the years 1997–2001 when the emission data include inter-annual variation in biomass burning for the tropics. For Southern Africa there is a clear decrease in OH during 1997–98 even if fewer fires resulted in a declining CO/NOx emission ratio. This suggests an influence on OH by intercontinental transport of CO and hydrocarbons from regions with enhanced burning during ENSO. As can be seen in Figure 3g the region is heavily influenced by transport from fires in South America. The Middle East has a growth in emissions for almost all sectors. Except for 1998 there is a steady increase in OH. However, no or slightly negative correlation with the CO/NOx emission ratio is found. The increase is likely explained by an increase in primary production of OH as we find large increases in ozone particularly for the monsoon season when transport from Asian emissions is important. Vegetation fires dominate the total emissions in Australia and New Zealand. There seems to be an upward trend in OH due to the increased anthropogenic NOx emissions, but for the last years when our emission dataset for the region indicates generally large but variable biomass burning emissions the pattern is getting more complex.

3.3. Trend in Tropospheric Methane Loss Rate

[12] The average increase in tropospheric methane concentration in the model is 0.32%/year reflecting adopted measured concentrations from a network of surface stations. In the CTM simulation we calculated the methane loss, (1) k • [OH] • [CH4] and the methane loss rate due to OH (2) k • [OH], k is the reaction rate and [OH] and [CH4] are the concentrations of hydroxyl and methane respectively. We used the global methane burden and the global loss to calculate a decrease in the global averaged tropospheric lifetime of methane from 8.33 years in 1990 to 8.05 years in 2001. The global average tropospheric methane loss is increasing quite steadily with an average of 0.5%/year with the exception of 1997 and 1998, corresponding to years with substantial OH decrease (Figure 2) from biomass burning. The methane loss is influenced by the adopted increase in methane concentration which is decoupled from the CO/NOx emission ratio and its suggested relation with OH. However, changes in OH especially in the lower troposphere also affect the methane loss. From a regression analysis we find a correlation coefficient of 0.6 for dependency of the methane loss to the CO/NOx emission ratio. (1) gives an increase in global average tropospheric methane loss of 0.5%/year and (2) gives a 0.16%/year increase in global average tropospheric methane loss rate. The relative contribution to methane loss of changes in methane and hydroxyl concentrations can be found by the information in these numbers. The increase in methane itself is responsible for about two third of the increase in its loss and changes in OH through emissions of CO, NOx and NMVOCs explain about one third.

4. Discussion and Conclusions

[13] CTM studies show that emission changes in the period 1990–2001 caused a global average increase in tropospheric OH of 0.08%/yr. The global increase in OH is driven by changes in the Northern Hemisphere. Deviations from the trend were found in years with much biomass burning. The inefficient combustion of most vegetation fires leads to a high CO/NOx emission ratio that tends to reduce OH. Given the relative small changes in emissions of VOCs we find that global OH is highly dependent on the CO/NOx ratio. A simple statistically significant linear relation of the form OH = ab · equation image was found.

[14] From the discussion in the introduction a dependency of OH on CO and NOx seems plausible. In theory the system could show nonlinear behavior to perturbations of these and other constituents and the chemistry scheme of the CTM has shown ability to reproduce nonlinear chemistry in other studies [Isaksen et al., 1978, 2005].

[15] Nevertheless, we find a linear relation on the global scale and also for some regions. Large regional differences in emission trends and different lifetimes of decisive components result in regional differences in OH changes. The OH abundance within a region is not only dependent on the in-situ emissions but also long range transport of CO, NOx-reservoirs and ozone from other regions. The regions where we found a linear relation were characterized by relative large emissions and not so much influence from transboundary pollutant transport. The CO/NOx emission ratios in these regions span a certain but quite wide range of 20–40 Tg(CO)/Tg(N). Though smaller areas with nonlinear chemistry could be masked when averaging over large scales, Northern Europe which is furthest away from fitting to a linear relation has a low CO/NOx emission ratio (around 11). A low ratio resulting from high NOx emissions could indicate that nonlinear chemistry is more important.

[16] The global averaged tropospheric methane loss increases with 0.5%/yr. The increase is larger than the increase in methane itself and we find that equation image is explained by the increase in methane and that equation image is due to enhancing OH levels related to changed emissions of CO, NOx and NMVOCs. The methane loss is particularly sensitive to abundance and changes of OH at low latitudes. We calculate that the Middle East and Southeastern Asia contribute almost 20% (2.8% and 15.9% respectively) to the total loss in 2001. Over the period 1990–2001 we find a large increase in OH levels and corresponding methane loss in these regions (Figure 1). A continuing emission increase in these regions with a shift towards a lower CO/NOx emission ratio, representative of fossil fuel combustion sources, could have important implications for the global methane budget. To study this we did a sensitivity study where all 2001 emissions were kept unchanged except the anthropogenic emission of CO and NOx in Southeast Asia which were allowed to increase by the same relative amount as in the 1990–2001 period (14.8% and 61.3% respectively, see Figure 1). This increased the OH concentration in Southeast Asia with 3.69% and the methane loss in the region with 6.26%. The resulting global averaged tropospheric changes were 0.84% for OH and 1.33% for methane loss. Southeast Asia's contribution to total global tropospheric methane loss increased by almost one percent.

[17] We like to repeat that we solely considered anthropogenic emission influence of CO, NOx and VOCs on OH and methane lifetime and that there are other influencing factors like changes in stratospheric ozone, aerosols, water vapor concentrations, temperature and meteorology that are important [Dentener et al., 2003; Wang et al., 2004]. Even if it is not a straightforward task, anthropogenic emissions are one of the factors possible to control. Though it is important to have the previous mentioned emission uncertainties (section 2) in mind this study reveals that even over the time frame of a decade emission changes and regulations have significant effects on the OH and methane budget, both globally and regionally. We believe that our findings have further policy implications as the derived relation of OH with the CO/NOx emission ratio in some regions could be used as a predicator of OH changes due to emission changes and measures taking place in the coming decade.