2.3.6.2 Tropospheric Ozone

The TAR identified large regional differences in observed trends in tropospheric ozone from ozonesondes and surface observations. The TAR estimate of RF from tropospheric ozone was +0.35 ± 0.15 W m^–2. Due to limited spatial and temporal coverage of observations of tropospheric ozone, the RF estimate is based on model simulations. In the TAR, the models considered only changes in the tropospheric photochemical system, driven by estimated emission changes (NO_x, CO, non-methane volatile organic compounds (NMVOCs), and CH₄) since pre-industrial times. Since the TAR, there have been major improvements in models. The new generation models include several Chemical Transport Models (CTMs) that couple stratospheric and tropospheric chemistry, as well as GCMs with on-line chemistry (both tropospheric and stratospheric). While the TAR simulations did not consider changes in ozone within the troposphere caused by reduced influx of ozone from the stratosphere (due to ozone depletion in the stratosphere), the new models include this process (Gauss et al., 2006). This advancement in modelling capabilities and the need to be consistent with how the RF due to changes in stratospheric ozone is derived (based on observed ozone changes) have led to a change in the definition of RF due to tropospheric ozone compared with that in the TAR. Changes in tropospheric ozone due to changes in transport of ozone across the tropopause, which are in turn caused by changes in stratospheric ozone, are now included.

Trends in anthropogenic emissions of ozone precursors for the period 1990 to 2000 have been compiled by the Emission Database for Global Atmospheric Research (EDGAR) consortium (Olivier and Berdowski, 2001 updated). For specific regions, there is significant variability over the period due to variations in the emissions from open biomass burning sources. For all components (NO_x, CO and volatile organic compounds (VOCs)) industrialised regions like the USA and Organisation for Economic Co-operation and Development (OECD) Europe show reductions in emissions, while regions dominated by developing countries show significant growth in emissions. Recently, the tropospheric burdens of CO and NO₂ were estimated from satellite observations (Edwards et al., 2004; Richter et al., 2005), providing much needed data for model evaluation and very valuable constraints for emission estimates.

Assessment of long-term trends in tropospheric ozone is difficult due to the scarcity of representative observing sites with long records. The long-term tropospheric ozone trends vary both in terms of sign and magnitude and in the possible causes for the change (Oltmans et al., 2006). Trends in tropospheric ozone at northern middle and high latitudes have been estimated based on ozonesonde data by WMO (2003), Naja et al. (2003), Naja and Akimoto (2004), Tarasick et al. (2005) and Oltmans et al. (2006). Over Europe, ozone in the free troposphere increased from the early 20th century until the late 1980s; since then the trend has levelled off or been slightly negative. Naja and Akimoto (2004) analysed 33 years of ozonesonde data from Japanese stations, and showed an increase in ozone in the lower troposphere (750–550 hPa) between the periods 1970 to 1985 and 1986 to 2002 of 12 to 15% at Sapporo and Tsukuba (43°N and 36°N) and 35% at Kagoshima (32°N). Trajectory analysis indicates that the more southerly station, Kagoshima, is significantly more influenced by air originating over China, while Sapporo and Tsukuba are more influenced by air from Eurasia. At Naha (26°N) a positive trend (5% per decade) is found between 700 and 300 hPa (1990–2004), while between the surface and 700 hPa a slightly negative trend is observed (Oltmans et al., 2006). Ozonesondes from Canadian stations show negative trends in tropospheric ozone between 1980 and 1990, and a rebound with positive trends during 1991 to 2001 (Tarasick et al., 2005). Analysis of stratosphere-troposphere exchange processes indicates that the rebound during the 1990s may be partly a result of small changes in atmospheric circulation.

Trends are also derived from surface observations. Jaffe et al. (2003) derived a positive trend of 1.4% yr^–1 between 1988 and 2003 using measurements from Lassen Volcanic Park in California (1,750 m above sea level), consistent with the trend derived by comparing two aircraft campaigns (Parrish et al., 2004). However, a number of other sites show insignificant changes over the USA over the last 15 years (Oltmans et al., 2006). Over Europe and North America, observations from Whiteface Mountain, Wallops Island, Hohenpeisenberg, Zugspitze and Mace Head (flow from the European sector) show small trends or reductions during summer, while there is an increase during winter (Oltmans et al., 2006). These observations are consistent with reduced NO_x emissions (Jonson et al., 2005). North Atlantic stations (Mace Head, Izana and Bermuda) indicate increased ozone (Oltmans et al., 2006). Over the North Atlantic (40°N–60°N) measurements from ships (Lelieveld et al., 2004) show insignificant trends in ozone, however, at Mace Head a positive trend of 0.49 ± 0.19 ppb yr^–1 for the period 1987 to 2003 is found, with the largest contribution from air coming from the Atlantic sector (Simmonds et al., 2004).

In the tropics, very few long-term ozonesonde measurements are available. At Irene in South Africa (26°S), Diab et al. (2004) found an increase between the 1990 to 1994 and 1998 to 2002 periods of about 10 ppb close to the surface (except in summer) and in the upper troposphere during winter. Thompson et al. (2001) found no significant trend during 1979 to 1992, based on Total Ozone Mapping Spectrometer (TOMS) satellite data. More recent observations (1994 to 2003, in situ data from the Measurement of Ozone by Airbus In-service Aircraft (MOZAIC) program) show significant trends in free-tropospheric ozone (7.7 to 11.3 km altitude) in the tropics: 1.12 ± 0.05 ppb yr^–1 and 1.03 ± 0.08 ppb yr^–1 in the NH tropics and SH tropics, respectively (Bortz and Prather, 2006). Ozonesonde measurements over the southwest Pacific indicate an increased frequency of near-zero ozone in the upper troposphere, suggesting a link to an increased frequency of deep convection there since the 1980s (Solomon et al., 2005).

At southern mid-latitudes, surface observations from Cape Point, Cape Grim, the Atlantic Ocean (from ship) and from sondes at Lauder (850–700 hPa) show positive trends in ozone concentrations, in particular during the biomass burning season in the SH (Oltmans et al., 2006). However, the trend is not accompanied by a similar trend in CO, as expected if biomass burning had increased. The increase is largest at Cape Point, reaching 20% per decade (in September). At Lauder, the increase is confined to the lower troposphere.

Changes in tropospheric ozone and the corresponding RF have been estimated in a number of recent model studies (Hauglustaine and Brasseur, 2001; Mickley et al., 2001; Shindell et al., 2003a; Mickley et al., 2004; Wong et al., 2004; Liao and Seinfeld, 2005; Shindell et al., 2005). In addition, a multi-model experiment including 10 global models was organised through the Atmospheric Composition Change: an European Network (ACCENT; Gauss et al., 2006). Four of the ten ACCENT models have detailed stratospheric chemistry. The adjusted RF for all models was calculated by the same radiative transfer model. The normalised adjusted RF for the ACCENT models was +0.032 ± 0.006 W m^–2 DU^–1, which is significantly lower than the TAR estimate of +0.042 W m^–2 DU^–1.

The simulated RFs for tropospheric ozone increases since 1750 are shown in Figure 2.9. Most of the calculations used the same set of assumptions about pre-industrial emissions (zero anthropogenic emissions and biomass burning sources reduced by 90%). Emissions of NOx from soils and biogenic hydrocarbons were generally assumed to be natural and were thus not changed (see, e.g., Section 7.4). In one study (Hauglustaine and Brasseur, 2001), pre-industrial NO_x emissions from soils were reduced based on changes in the use of fertilizers. Six of the ACCENT models also made coupled climate-chemistry simulations including climate change since pre-industrial times. The difference between the RFs in the coupled climate-chemistry and the chemistry-only simulations, which indicate the possible climate feedback to tropospheric ozone, was positive in all models but generally small (Figure 2.9).

Figure 2.9. Calculated RF due to tropospheric ozone change since pre-industrial time based on CTM and GCM model simulations published since the TAR. Estimates with GCMs including the effect of climate change since 1750 are given by orange bars (Adjusted RF, CC). Studies denoted with an (*) give only the instantaneous RF in the original publications. Stratospheric-adjusted RFs for these are estimated by reducing the instantaneous RF (indicated by the dashed bars) by 20%. The instantaneous RF from Mickley et al. (2001) is reported as an adjusted RF in Gauss et al. (2006). ACCENT models include ULAQ: University of L’Aquila; DLR_E39C: Deutsches Zentrum für Luft- und Raumfahrt European Centre Hamburg Model; NCAR_MACCM: National Center for Atmospheric Research Middle Atmosphere Community Climate Model; CHASER: Chemical Atmospheric GCM for Study of Atmospheric Environment and Radiative Forcing; STOCHEM_HadGEM1: United Kingdom Meteorological Office global atmospheric chemistry model /Hadley Centre Global Environmental Model 1; UM_CAM: United Kingdom Meteorological Office Unified Model GCM with Cambridge University chemistry; STOCHEM_HadAM3: United Kingdom Meteorological Office global atmospheric chemistry model/Hadley Centre Atmospheric Model; LMDzT-INCA: Laboratoire de Météorologie Dynamique GCM-INteraction with Chemistry and Aerosols; UIO_CTM2: University of Oslo CTM; FRSGC_UCI: Frontier Research System for Global Change/University of California at Irvine CTM.

A general feature of the models is their inability to reproduce the low ozone concentrations indicated by the very uncertain semi-quantitative observations (e.g., Pavelin et al., 1999) during the late 19th century. Mickley et al. (2001) tuned their model by reducing pre-industrial lightning and soil sources of NO_x and increasing natural NMVOC emissions to obtain close agreement with the observations. The ozone RF then increased by 50 to 80% compared to their standard calculations. However, there are still several aspects of the early observations that the tuned model did not capture.

The best estimate for the RF of tropospheric ozone increases is +0.35 W m^–2, taken as the median of the RF values in Figure 2.9 (adjusted and non-climate change values only, i.e., the red bars). The best estimate is unchanged from the TAR. The uncertainties in the estimated RF by tropospheric ozone originate from two factors: the models used (CTM/GCM model formulation, radiative transfer models), and the potential overestimation of pre-industrial ozone levels in the models. The 5 to 95% confidence interval, assumed to be represented by the range of the results in Figure 2.9, is +0.25 to +0.65 W m^–2. A medium level of scientific understanding is adopted, also unchanged from the TAR (see Section 2.9, Table 2.11).