INTERCONTINENTAL TRANSPORT OF OZONE:
LINKING AIR QUALITY AND CLIMATE ISSUES
Daniel J. Jacob, Harvard University
http://www-as.harvard.edu/chemistry/trop
presented at Swiss-Japanese Ozone Symposium
Interlaken, Switzerland, July 1-4, 2002
There is ample evidence from observations that tropospheric ozone concentrations at northern midlatitudes have increased considerably over the past century (SLIDE 2). Interestingly, current tropospheric chemistry models underestimate this increase. They cannot reproduce the very low (5-10 ppbv) concentrations observed in the late 19th century (SLIDE 3, from Mickley et al. [2001]), and they also cannot reproduce the robust trend in ozone observed over the past 50 years. This difficulty in the models could reflect an overestimate of the natural component of ozone, or a long-term trend in natural sources that the models do not account for. The implications for radiative forcing are large; if one adjusts the natural ozone sources in a model in such a way as to fit the 19th century observations, the radiative forcing from tropospheric ozone becomes large, 0.80 W m-2 as compared to 0.4 W m-2 in the standard models used for the recent IPCC assessment (SLIDE 3).
Global dispersion of anthropogenic ozone produced over the industrial continents is important not only for climate forcing but also for intercontinental transport of pollution (SLIDE 4, from Li et al. [2002a]). Mean surface ozone enhancements in the northern midlatitudes continents due to anthropogenic sources in the continents upwind are of the order of 5 ppbv, as simulated in our GEOS-CHEM global model of tropospheric chemistry. These mean enhancements may overestimate the contribution of intercontinental transport during regional pollution episodes, when stagnation restricts ventilation and hence the import of ozone. This point is illustrated in SLIDE 5 (from Fiore et al. [2002a]), which shows the observed ozone vs. (NOy –NOx) relationship in summer at the rural site of Harvard Forest (Massachusetts) and compares to GEOS-CHEM results. Also shown in the figure is the contribution from background ozone, calculated in the model with a tagged ozone tracer produced only outside the North American boundary layer. We see that this background contribution to ozone decreases from typically 20 ppbv under clean conditions to less than 10 ppbv under polluted conditions.
Another perspective on this result is SLIDE 6 (from Fiore et al. [2002a]), which shows the probability distribution of background contribution to surface ozone in the United States in summer afternoons in the GEOS-CHEM model, for the ensemble of conditions in the United States and for severe pollution episodes (ozone > 80 ppbv). The mode of the probability distribution decreases from 23 ppbv for the ensemble of conditions to 12 ppbv under polluted conditions. However, there are still occasional cases where the background contributes more than 30 ppbv during pollution episodes (high tail of the distribution). These cases reflect contributions from convective events delivering elevated background ozone from the free troposphere to the surface, with subsequent additional production taking place in the U.S. boundary layer.
The U.S. air quality standard (80 ppbv) is particularly loose. Air quality standards in Europe are tighter (45-65 ppbv). We find that while the background contribution to ozone tends to be low in contributing to exceedances of the former standard, it is very important for contributing to exceedances of the latter standard. This is illustrated in SLIDE 7, which shows the enhancements of surface ozone over the United States due to anthropogenic emissions in Asia and Europe, as calculated with the GEOS-CHEM model, for the ensemble of summer afternoon 1995 conditions simulated by the model. The maximum enhancements (up to 14 ppbv) occur when ozone is in the 50-65 ppbv range. Again, this range reflects a high contribution to ozone by subsidence from the free troposphere, where ozone has a long lifetime and is zonally elevated due to anthropogenic emissions from all northern midlatitude continents.
This importance of intercontinental transport in ozone air quality is likely to increase in the future as air quality standards become tougher and as the ability to meet these standards through domestic emission controls is thwarted by rising emissions from the developing world. This need for a global perspective offers opportunities for linking policy efforts in air quality and climate change. SLIDE 8 (from Fiore et al. [2002b]) illustrates this point with a GEOS-CHEM simulation of the effects of future global changes in anthropogenic emissions on the frequency of U.S. air pollution events (diagnosed as the number of model grid-square days per summer with surface concentrations higher than 80 ppbv) as well as radiative forcing. We find that a 50% global decrease in anthropogenic NOx emissions would greatly improve U.S. air quality but would be near-neutral for climate forcing. However, a decrease in methane emissions (reducing the global ozone background) would reap substantial benefit for both air quality in climate. Simulations with realistic future projections (IPCC scenarios A1 and B1) indicate in the A1 scenario a worsening of air quality in the U.S. despite domestic decreases in NOx emissions; this worsening is largely due to the global rise in methane. Controlling methane emissions thus offers significant benefits for both air quality and mitigating climate change.
Our discussion of intercontinental transport of pollution has focused so far on a U.S. air quality perspective, but transatlantic transport of U.S. pollution also has a significant impact on European air quality according to our GEOS-CHEM model. The effect is largest in spring (SLIDE 9, from Li et al. [2002a]). Evaluating the reliability of the model is difficult, however, because the intercontinental pollution enhancements are small compared to the effects of regional pollution sources. We examined this problem recently by evaluating the ability of the GEOS-CHEM model to reproduce events of elevated ozone at Bermuda during spring. There has been considerable controversy in the literature as to whether these events are due to transport from the stratosphere or to pollution outflow from North America. As shown in SLIDE 10 (from Li et al. [2002b]), GEOS-CHEM has considerable success in reproducing the observed structure of the high-ozone episodes in surface air at Bermuda, and we find through tagged tracers that this structure is associated almost exclusively with boundary layer transport of U.S. pollution behind cold fronts. We also have considerable success in reproducing the ozonesonde observations at Bermuda and at sites in North America (SLIDE 11 and SLIDE 12, from Li et al. [2002b]), and this successful simulation implies negligible stratospheric influence on ozone throughout the troposphere.
Previous arguments for a stratospheric source of surface ozone at Bermuda in spring had relied on subsiding back-trajectories and correlation with beryllium-7 as evidence (SLIDE 13). We reproduce these observed features (SLIDE 14, SLIDE 15, and SLIDE 16, from Li et al. [2002b]) and explain them as reflecting the subsidence behind cold fronts mixing with the continental boundary layer outflow. We similarly reproduce the ozone-7Be correlation observed at Tenerife in the Canary Islands (SLIDE 16) and previously attributed to a stratospheric or upper tropospheric source of ozone; we find in the model that ozone at Tenerife is mostly produced in the middle troposphere.
Bermuda thus offers an unambiguous signal for long-range transport of ozone pollution from the U.S. to the North Atlantic. Intercontinental transport to Europe is more difficult to verify. SLIDE 17 (from Li et al. [2002a]) shows simulated and observed time series of surface ozone at Mace Head (west coast of Ireland) in spring. The model reproduces well the observed ozone structure, but most of this structure is due to European pollution sources of ozone. North American pollution events at Mace Head elevate ozone by up to 10 ppbv but the signal is faint relative to the European signal and would have to be teased through correlations with chemical tracers. That work has not been done so far. It needs to be done because we do find in the model that transatlantic transport of pollution from U.S. sources has significant implications for European air quality standards, increasing the number of days in exceedance of the 55 ppbv 8-hour standard by about 20% (SLIDE 18, from Li et al. [2002a]). We also find from a 6-year simulation that this U.S. pollution influence over Europe is strongly correlated with the North Atlantic Oscillation (NAO) (SLIDE 19, from Li et al. [2002a]), which is to be expected since the NAO affects the strength of the transatlantic westerly winds. There is presently much interest in studying the effect of climate change on the NAO because of the implications for regional climate in western Europe; we find here that air pollution would be an additional implication.
Our study of intercontinental transport of pollution has also included an application of GEOS-CHEM to examine the effect of European pollution sources on surface ozone in Asia. SLIDE 20 (from Liu et al. [2002]) shows that the effect is maximum (3-5 ppbv) in spring. We have worked recently on using the ozonesonde data along the Asian Pacific rim to better understand the sources of tropospheric ozone in this region. Comparison of GEOS-CHEM results to sondes in Hong Kong (SLIDE 21, from Liu et al. [2002]) shows that the model can reproduce two prominent features in the observations: (1) the high upper tropospheric variability in winter-spring, due to alternance of tropical and stratospheric influences (SLIDE 22, from Liu et al. [2002]), and (2) the springtime enhancement due to biomass burning emissions from Southeast Asia. Surface ozone over eastern Asia is low in summer because of frequent convection associated with the summer monsoon. However, we find in the ozonesonde data some large enhancements of ozone in the middle and upper troposphere from Asian pollution (SLIDE 23, from Liu et al. [2002]). Our model analysis shows that part of this high-altitude pollution is transported to the Pacific in the westerly flow, while part circulates around the upper tropospheric Tibetan High and heads towards the Middle East (SLIDE 23).
Our most flagrant problem at present in simulating tropospheric ozone with GEOS-CHEM is a severe overestimate over India (SLIDE 24, from Martin et al. [2002a]). Including non-conventional aerosol effects in the model reduces the overestimate slightly but is far from sufficient. The cause of the remaining discrepancy is not clear. It is particularly difficult to explain why the overestimate extends over the depth of the tropospheric column. This current problem prevents us from saying much about the outflow of Indian pollution, which is expected to be a major driver for future global change in atmospheric composition.
There is hope for better simulation and understanding of tropospheric ozone in the future through top-down mapping of ozone precursor emissions from space. SLIDE 25 (from Martin et al. [2002b]) shows tropospheric NO2 columns retrieved from solar backscatter measurements by the GOME instrument, and compares to the corresponding results from the GEOS-CHEM model driven by our best emission estimates. There is significant correspondence but also large discrepancies that need to be explained either through errors in the GOME retrievals or in the bottom-up emission inventories. To explore this question further, we focused on the United States where bottom-up emission estimates are thought to be reliable (SLIDE 26, from Martin et al. [2002b]). Correlation between GOME and GEOS-CHEM is strong, and the national bias is small (+18%), but there are intriguing differences such as over eastern Texas that point to the need for further study.
We have similarly used HCHO column measurements from GOME as a proxy for top-down mapping of emissions from volatile organic compounds (VOCs), the second precursor for ozone. HCHO is a high-yield product from atmospheric oxidation of VOCs. SLIDE 27 (from Palmer et al. [2001]) compares HCHO column measurements from GOME in July 1996 to GEOS-CHEM model results. The maximum in the southeastern United States, found in both the observations and in the model, is due to biogenic emission of isoprene. Isoprene emission estimates in the model, taken from the Global Emissions Inventory Activity (GEIA) inventory, seem overall successful in reproducing the HCHO columns observed by GOME. As a next step towards using the GOME observations for VOC mapping, we performed an indirect validation of the GOME data by comparing to in situ surface observations of HCHO in the United States, using the model as an intermediary. We find that the model driven by GEIA isoprene emissions can reproduce both the GOME data and the in situ data (SLIDE 28, from Palmer et al. [2002]) with little bias, supporting the validity of the GOME observations.
Isoprene emissions in the U.S. in the summer are thought to peak in the Ozark mountains (southeastern Missouri) due to high temperature and a high density of oaks. GOME observes this peak but not on all days (SLIDE 29, from Palmer et al. [2002]). We examined whether this variability was correlated with temperature in a manner consistent with isoprene emissions (SLIDE 30, from Palmer et al. [2002]). Results are encouraging but the GOME data show a drop-off at very high temperatures (> 300 K) that would not be expected from current process models of isoprene emission.
Relating the HCHO columns observed by GOME to the underlying isoprene emissions depends on the HCHO yield from isoprene oxidation and on the loss rate constant for HCHO loss by photolysis and oxidation (SLIDE 31). We can use GEOS-CHEM to derive this relationship and its geographical variation (SLIDE 32, from Palmer et al. [2002]) and from there to construct a “GOME inventory of isoprene emission” (SLIDE 33, from Palmer et al. [2002]). This inventory is not inconsistent with GEIA but is about a factor of 2 higher than the BEIS-2 isoprene inventory presently recommended by the U.S. Environmental Protection Agency for ozone modeling. Our next step is to apply the method globally. SLIDE 34 shows a global map of HCHO columns from GOME in July 1996; elevated values are found both in regions of high biogenic emissions and in regions of large fires. Because of aerosol interferences, it is not clear at present that the HCHO columns in burning areas can be related to the underlying VOC emissions. More research is needed!