2.3. Uncertainties in the Impact Assessment of Present Aviation: Implications for Use of Models in Predicting Future Change

2.3.1. Uncertainties in Modeling Aviation Impacts

2.3.1.1. Key Issues and Processes for Tropospheric Models

In performing assessments of the impacts of subsonic aircraft NOx emissions on tropospheric composition, the modeling studies have pointed out several key issues and processes that have to be adequately addressed:

Model spatial and time resolution
Time resolution of meteorological data
Subgrid-scale processes (e.g., plume processes)
Tropospheric NOx and NOy sources
Tropospheric gas-phase and heterogeneous chemistry
Stratospheric gas-phase and heterogeneous chemistry
Stratosphere-troposphere exchange
Upper troposphere-lower stratosphere dynamics and convective transport
Sources and sinks of water and NOy.

These issues and processes are discussed in general terms in the paragraphs that follow. A detailed description of process representation in the assessment models is given in Chapter 4.

Over the years, there has been a steady increase in the spatial resolution of the models used in the assessment of subsonic aircraft impacts on tropospheric composition. Initially, the assessment models were one-dimensional (altitude), averaged around latitude circles and from north to south poles. Relatively quickly, researchers realized that interhemispheric gradients were crucial for ozone and anthropogenic trace gases, so much of the assessment work has been carried out with 2-D (altitude and latitude) models that average around latitude circles. With NOx lifetimes of days or less and transport times around latitude circles of up to 2 weeks, researchers appreciated that 2-D models needed parameterizations to represent the main features of the distributions of NOx and other short-lifetime species. Some research teams investigated 2-D channel models (altitude and longitude) models, but the development of 3-D chemistry transport models (CTMs) has been the main thrust of tropospheric modeling efforts.

Global 3-D CTMs are now the main tools for the assessment of subsonic aircraft impacts in the troposphere. Typically, these models have horizontal resolutions of 2-6° by 2-6°, limited by the resolution of the general circulation model (GCM) from which they have been derived. Emissions databases are now available with higher spatial resolution, so model performance is limited by the meteorological databases used in CTMs and the necessary computing time. Vertical resolution is a major limitation with current CTMs, in terms of the height taken as the top of the model and the number of layers into which the model domain is divided. Few models have enough vertical resolution to fully resolve the atmospheric boundary layer and tropopause domain and to describe the exchange of trace gases between the UT and the LS.

There is a major concern with 3-D CTMs regarding the adequacy of time resolution required in emission inventories and in meteorological data used to transport trace gases from their sources to their sinks. Initially, some CTMs used monthly averaged fields of horizontal and vertical winds, temperatures, clouds, and humidities. To resolve major storm systems and convective events, the meteorological data have been updated in the CTMs on a steadily increasing frequency; fields are now usually updated every 6 hours. On this basis, it is possible to resolve the changing stability of the atmospheric boundary layer, the developing behavior of major weather systems, and large-scale convective events.

Time and spatial resolution are crucial issues in evaluating the impacts of subsonic aircraft. To evaluate whether the chosen time and spatial resolutions are adequate in each of the tropospheric assessment tools, a number of sensitivity studies should be carried out in the near future.

Concentration changes occurring in the aircraft plume and wake take place on a spatial scale (i.e., < 20 km) that is less than the smallest global atmospheric model scale (i.e., > 100 km). Consequently, global models do not typically treat aircraft near-field processes in detail. In fact, most current global model studies have input aircraft emissions inventories (i.e., emissions indexes, or EI) by simple dilution of the aircraft plume at the altitude of injection, with no chemical changes taking place in the near field. As a possible means of connecting near-field processes to the global model grid scale, Petry et al. (1998) and Karol et al. (1998) have proposed the concept of effective emissions index (EEI) to account for changes in species concentrations in the plume dispersion region resulting from photochemical reactions. As an example, EEI(NOx) will be less than the corresponding EIs because of plume processes that convert NOx to NOy. However, first estimates show that EEIs are very sensitive to temperature and light intensity, which results in a large variation of EEIs in latitude, altitude, and season (Karol et al., 1997; Meijer et al., 1997).

Recent ozone model studies have also pointed out the importance of background NOx sources in understanding ozone tendencies with respect to increasing, or additional, NOx sources. In a sense, the aircraft case is one example of a broader issue regarding nonlinearity between ozone impacts and NOx levels. Model studies have shown that the magnitude of ozone changes from aircraft NOx emissions may depend significantly on the amount of NOx from non-aircraft sources. One difficulty with modeling background NOx sources, however, is the short lifetime of NOx (typically 1-5 days). In general, aircraft NOx impacts on upper tropospheric and lower stratospheric ozone will be overstated if background NOx concentrations are underestimated, and vice versa.

Over the past 2 decades, a significant amount of research has been committed to improving our understanding of background NOx sources through model studies and observations of aircraft NOx emissions. Much of the NOx in the troposphere comes from surface NOx sources, either via fast vertical transport as NOx (Ehhalt et al., 1992) or by conversion to temporary reservoir NOy carriers such as PAN or HNO3, followed by subsequent conversion back to NOx. An accurate representation of the contribution made by surface NOx sources to the UT and LS requires full treatment of boundary layer chemistry, exchanges between the boundary layer and the free troposphere, deposition and wet scavenging, free tropospheric chemistry and transport to the upper troposphere by convection, atmospheric circulation, and synoptic-scale weather systems. An estimate of annual flux into the free troposphere from European surface NOx sources, as a fraction of the total surface NOx source, can be made from European Monitoring and Evaluation Program (EMEP) modeling studies (Tuovinen et al., 1994). For NOx, 52% of the emitted NOx was deposited within the EMEP area of Europe and 48% was exported out of the model region during 1985-93. Approximately half of this material is vented into the free troposphere as NOy (1.7 Tg N yr^-1), with the remainder deposited elsewhere, without reaching the free troposphere.

In North America, a similar picture applies. Model calculations (Brost et al., 1988) have estimated that about 1.8 Tg yr^-1 (of total North American emissions of 7.9 Tg N yr^-1) is transported east to the Atlantic Ocean between the surface and 5.5-km altitude. That is, about 25% of the North American NOx emissions remained airborne in the boundary layer or free troposphere as the air left North America. More recent studies (Jacob et al., 1993; Horowitz et al., 1998; Liang et al., 1998) have derived a lower estimate of the transport out of North America (on the order of 6%).

A detailed model study of advective and convective venting of ozone, NOx, and NOY out of the boundary layer over northwest Europe during July and October-November 1991 showed that, of surface NOx emissions, 7% were brought to the free troposphere as NOx and 20% as NOy during the summer (Flatoy and Hov, 1996), with slightly smaller percentages during the fall.

Transport from the surface is therefore undoubtedly an important contributor to background NOx. However, it is difficult to evaluate how well this process is handled in each of the tropospheric assessment models used for aviation impact calculations.

Lightning is an important NOx source in the UT (Chameides et al., 1987; Lamarque et al., 1996). Because of its sporadic nature and small spatial scale (tens of km), it is exceedingly difficult to represent quantitatively in even the most complex of tropospheric models. Most model studies include some representation of lightning NOx, with global total emissions in the range 1-10 Tg N yr^-1. However, there is no consensus on how to represent this source with time of day, season, altitude, or spatially, nor how to treat lightning in concert with convection, cloud processing, and wet scavenging.

Stratospheric NOy is a further important source of NOx in the UT and LS, through the photolysis of HNO3 (Murphy et al., 1993). There is a downward flux of NOy from the stratosphere to the troposphere that globally balances the stratospheric NOx source produced by the reaction of N2O with O(1D). Few model studies of aircraft NOx emissions extend high enough in altitude to include a full treatment of stratospheric NOx and NOy. Moreover, most do not include an explicit representation of stratospheric chemistry (i.e., halogen chemistry) from which to realistically calculate stratospheric NOx. Instead, most models use a constant ratio of NOy to ozone and describe the stratospheric NOy source in the same way as the stratosphere-troposphere exchange of ozone. Typical ozone to NOy ratios are assumed to be about 1000:1, giving a stratospheric NOy source in the UT of about 0.5 Tg N yr^-1.

Although the issue of the magnitude of background NOx levels has been clearly identified and much work has been performed to characterize surface, lightning, and stratospheric sources, there are still too few measurements of NOx and NOy in the UT and LS with which to assess quantitatively representations of background NOx in the models summarized in Table 2-1. Recently published data (Emmons et al., 1997) have begun to be used for evaluation of model performance (Wang et al., 1998b).

As discussed in Section 2.1.2.4, the response of ozone to increasing NOx depends on the strength of the HOx source. Recent evidence (Brune et al., 1998; Wennberg et al., 1998) supports the presence of additional upper tropospheric HOx sources from organic precursors that are not included in many current models. Improved model treatments of HOx production from precursors such as acetone, peroxides, and aldehydes will require additional data on the mechanisms and kinetics of a number of NMHC reactions. The role of heterogeneous chemistry in influencing HOx and NOx levels in the UT has not been investigated fully yet and is expected to become an increasingly important issue for tropospheric models.

2.3.1.2. Tropospheric Model Evaluation

Previous IPCC reports (IPCC, 1996) identified tropospheric ozone modeling as one of the more difficult tasks in atmospheric chemistry. Difficulties arise, in part, from the large number of processes that control tropospheric ozone and its precursors and, in part, from the large range of spatial and temporal scales that must be resolved. Global 3-D CTMs attempt to simulate the life cycles of many trace gases and the impacts of subsonic aircraft NOx emissions on them. We need to understand the level of confidence that is to be ascribed to these model studies.

There is a significant amount of scatter in current model assessments of the impacts of subsonic aircraft NOx emissions on all aspects of tropospheric composition. With respect to reducing the range of uncertainty, it would be helpful if we could point to particular aspects of model performance and gauge models against specified benchmarks. Some model evaluation studies have begun the difficult task of identifying the current level of model performance and defining the level of confidence that should be placed in them. To this end, a number of model intercomparison exercises have been completed; some are in hand, and some are only at the planning stage. These exercises have involved the following elements:

Transport of 222Rn

Fast photochemistry

Transport of NOx

Comparison of model data and observations of tropospheric ozone.

2.3.1.2.1. Transport of 222Rn

Twenty atmospheric models participated in the 222Rn intercomparison for global CTMs (IPCC, 1996; Jacob et al., 1997). Differences between model-calculated distributions of this short-lived (e-folding lifetime of 5.5 days) radioactive decay product emitted at the surface from soils were large, which enabled the drawing of conclusions about the general adequacy of transport schemes in CTMs. Owing to the lack of extensive observations, evaluation efforts to date have been restricted mainly to model-model intercomparisons.

The 222Rn model intercomparison concluded that tropospheric CTMs based on 2-D models and monthly averaged 3-D models have a fundamental flaw in transporting tracers predominantly by diffusion; thus, these models cannot be viewed as reliable in simulating the global transport of tracers. Synoptic 3-D models need significantly improved representations of boundary layer processes, clouds, and convection. Large differences are found among established 3-D CTMs in the rates of global-scale meridional transport in the UT-particularly, interhemispheric transport. These latter differences are particularly relevant to the current issue of subsonic aircraft impacts.

2.3.1.2.2. Fast photochemistry

More than 20 model groups participated in the tropospheric photochemical model intercomparison exercise, PhotoComp-a tightly controlled experiment in which consistency was determined among models used to predict tropospheric ozone changes (IPCC, 1996; Olson et al., 1997). A similar study, involving fewer models, was carried out as part of a U.S. National Aeronautics and Space Administration (NASA) assessment (Friedl, 1997). As with the radon case, there are no easy observational tests of model fast photochemistry, so model-model intercomparison exercises were carried out in both cases.

Over the intercomparison tests for fast photochemistry of the sunlit troposphere, modeled OH concentrations fell within a ±20% band, and ozone changes fell within a ±30% band. These obvious variations between model results did not correlate with other model differences, and no single model input parameter appeared to account for all of the spread in the results. Nevertheless, ozone photolysis rates used in the models accounted for about half of the root mean square (RMS) differences; further investigation of these parameters and their comparison with observations is called for. The results also became more uncertain in model experiments involving NMHCs. Further CTM development is required so that models have the required grid and time resolutions to simulate accurately the scales of chemistry required to describe the removal of NOx and NMHCs while producing and destroying ozone, quantitatively.

2.3.1.2.3. Transport of NOx

Passive transport of subsonic aircraft NOx emissions has been studied with a hierachy of global CTMs (Friedl, 1997; van Velthoven et al., 1997). The 3-D CTMs showed that the monthly mean NOx concentrations varied by a factor of three longitudinally and that the temporal variability of background NOx in the air traffic corridor was about ±30% on synoptic time scales. Vertical redistribution by convection strongly affected the maximum NOx concentrations at subsonic aircraft cruise altitudes.

A number of model deficiencies and biases were found, including the oscillatory nature of NOx distributions obtained with a spectral advection scheme, the strong diffusion of GCMs into polar regions, and the too-intense interhemispheric exchange found in some 2-D CTMs. The intercomparisons concluded that assessment of the tropospheric impacts of subsonic aircraft NOx emissions could be performed better with 3-D CTMs.

2.3.1.2.4. Comparison of model data and observations of tropospheric ozone

An increasing number of activities are aimed at evaluating global model results in relation to ozone observations (Wang et al., 1998b; Wauben et al., 1998). However, there are too few ozone data, especially in the tropics, to allow for comprehensive evaluations. Comparisons are showing that model simulations are reproducing the broad features of monthly mean measured ozone concentrations. Some models do not produce the observed seasonality in the northern mid-latitude troposphere. Differences are most pronounced in the free troposphere, especially close to the tropopause (see Figure 2-7).

As part of the International Global Atmospheric Chemistry Project/Global Integration and Modeling Activity (IGAC/GIM) study, an intercomparison exercise is currently being attempted of ozone concentrations calculated by 12 global 3-D CTMs (Kanikidou et al., 1998). Many of these CTMs have already performed assessments of the impacts of subsonic aircraft NOx emissions on tropospheric ozone; their results have been included in Table 2-1. Furthermore, all of the tropospheric assessment models employed in Chapter 4 have submitted results to the IGAC/GIM intercomparison.

The GIM intercomparison extends the intercomparisons described above in that it employs some of the available observational database to evaluate intermodel differences. Figure 2-7 presents some of the model intercomparison results for seasonal cycles of ozone at 300 mb at three widely separated sites.

The GIM model intercomparison with monthly mean values of observations demonstrates that the models capture some of the considerable variability within the observations. The range in observations may approach 20 ppb at 500 mb and up to 40 ppb at 300 mb, with the ranges in the models significantly greater. These ranges are significantly greater than the tropospheric ozone impacts of about 8 ppb anticipated from subsonic aircraft NOx emissions (Table 2-1).

2.3.1.3. Key Issues and Processes for Stratospheric Models

CFC and HSCT assessment activities have engaged 2-D (height and latitude) and, to some extent, 3-D (height, latitude, and longitude) models focused on the stratosphere over the past 10 years. These efforts have served to highlight a number of critical stratospheric model issues:

Aircraft plume processes

Stratospheric transport

Stratospheric gas-phase and heterogeneous chemistry

Sulfate aerosol

PSCs

Soot.

The issue of soot has been raised only to a small extent by the HSCT studies, although it has assumed a more prominent role in the subsonic aviation case (see Section 2.1.3). In the following paragraphs, we discuss these issues in the context of current model treatments of subsonic aviation impacts.

Aircraft emissions, whether supersonic or subsonic, are deposited primarily at northern mid-latitudes and over a limited vertical range. A key issue for models is how fast these emissions are dispersed to other regions of the atmosphere, such as the tropical stratosphere or the mid-latitude troposphere, where the response of ozone to the emissions will be substantially different. In addition, it is important to consider the chemistry occurring in the aircraft plume and wake before it has been expanded to the model grid scale. Initial attempts to combine near field, far field, and global models in series (Danilin et al., 1997) are the first global impact studies to be based directly on detailed microphysics and chemical kinetics occurring in the aircraft plume and wake. An increasingly robust plume and wake observational database is being collected to validate this approach (Kärcher, 1998; Kärcher et al., 1998b).

To date, most models used to assess the impact of aviation on the atmosphere have been 2-D, in which the time-consuming complexity of the real 3-D atmosphere is reduced to a manageable calculation by averaging around latitude circles. Because of this simplification, 2-D models do not adequately simulate all dynamic features of the atmosphere. Horizontal transport between mid-latitudes and the tropics (or polar vortex) is an inherently episodic, wave-driven process that is parameterized in 2-D models by eddy diffusion terms. Measurements of the NOY-to-ozone ratio in the LS have provided evidence for distinctly different airmass characteristics that are not well represented in 2-D models (Murphy et al., 1993; Minschwaner et al., 1996; Volk et al., 1996; Schoeberl et al., 1997). One method for improving the 2-D representation of tropical/extratropical air mass difference has been to reduce the horizontal eddy coefficient in the subtropical region. Efforts such as these have underscored the fact that an accurate model representation of tropical/mid-latitude air mass distinctions, including the extent of transport of tropical air into mid-latitudes, remains an important assessment uncertainty.

Model representation of bulk, global-scale vertical exchange between the stratosphere and the troposphere by diabatic circulation is likely adequate (Holton et al., 1995). However, most models do not adequately resolve tropopause-folding events or stratosphere-troposphere exchange along isentropic surfaces. To the extent that these processes are important, calculated aviation impacts will be sensitive to model horizontal and vertical resolution.

Model representation of gas-phase photochemical links between ozone and atmospheric trace species such as HOx and NOx may be the most mature area of model construction, although rate parameter uncertainties increase with decreasing temperature. This representation is facilitated by the existence of evaluated compilations of photochemical rate parameters (IUPAC, 1997a,b; JPL, 1997). Because of the sensitivity of reaction rates to temperature and photolysis rates to solar zenith angle, model treatments must account for temperature and solar flux changes as air parcels move around the globe and encounter day and nighttime conditions. Diurnal variations in calculated radical concentrations can be reproduced either by invoking an explicit time marching kinetic scheme or by applying a correction factor to concentrations calculated from averaged solar zenith angles.

The dependence of reaction rate coefficients on temperature, especially for PSC processes, can present a particular problem for 2-D models, which are constrained to zonal-mean temperature fields. One strategy to address zonal variations has been to describe the zonal mean temperature by a probability distribution (Considine et al., 1994). The applicability of this approach to PSC processes is an area of active investigation. Type II PSC particles, consisting of water-ice and uniformally formed at temperatures below 188 K, can be adequately captured in 2-D formulations. However, the temperature thresholds for PSC type I particle formation are highly variable because of the multitude of possible particle compositions, and they depend more heavily on the temperature histories of air parcels. Some of the type I PSCs considered in stratospheric models include solid nitric acid hydrates (e.g., trihydrate and dihydrate), mixed hydrates, and supercooled sulfate, nitrate, and water ternary solutions (Worsnop et al., 1993; Carslaw et al., 1994; Tabazadeh et al., 1994; Fox et al., 1995). Compositional details of modeled PSC type Is are important because they determine what the model will calculate for the size, density, and removal rates (by sedimentation) of the particles as well as the partitioning of NOy between gaseous and condensed phases.

Finally, stratospheric models must describe background sulfate and carbonaceous aerosol formation and evolution adequately to gauge perturbations from aircraft SOxO and soot emissions. In past studies, models have merely prescribed aerosol suface area distributions based on satellite observations. Recognition that aircraft exhaust may contain a large number of small-diameter sulfate particles has motivated development of aerosol microphysical schemes (Weisenstein et al., 1996).

2.3.1.4. Stratospheric Model Evaluation

The growing body of satellite, balloon, and aircraft chemical and meteorological data for the middle atmosphere has made it possible to devise tests of photochemistry and transport within stratospheric models. A number of 2-D and 3-D models have participated in two major intercomparison efforts, Models and Measurements (M&M) I and II (Prather and Remsberg, 1993; Park, 1999). These comparisons have focused on testing the ability of these models to estimate the atmospheric effects of a proposed fleet of supersonic aircraft that would operate near 20 km. As a direct result of the first M&M effort, a number of errors in the models were identified and corrected. Both M&M efforts have served to highlight important tests of model representations. Because of the supersonic aircraft focus, however, less analysis has been directed at model performance in the lowermost stratosphere, where subsonic aviation effects are expected. With the exception of ozone representation, rigorous tests of model representation of the dynamics and chemistry of the lowermost stratosphere and UT have not been performed to date. Poor agreement between model predictions and observations of ozone in this region of the atmosphere (typical errors greater than 50%) suggests that significant improvement will be required before stratospheric assessment models can be used to examine the impact of aviation (or, for that matter, any perturbation) on the lowermost stratosphere and UT. In the following paragraphs, we summarize comparison efforts for the following key issues (for altitudes above 15 km):

Photochemistry

Dynamics

Comparison of model data and observations of stratospheric ozone.

2.3.1.4.1. Photochemistry

The photochemical mechanisms employed by most of the models compare well with each other. Tests of the photochemical mechanisms were performed by comparing predicted concentrations of short-lived reactive chemicals from these models against a benchmark photo-stationary state model constrained by the distribution of precursors from each 2-D model. These comparisons provide a means of accounting for differences in the transport of long-lived species, such as NOy, and O3, within the models. The distribution of NOy versus altitude and the mixing ratio of N2O was markedly different among the various 2-D models. Most of the differences for calculated concentrations of hydrogen, nitrogen, and chlorine free radicals among the various 2-D models were shown to be caused by differences in NOy and to a lesser degree ozone. The benchmark model has been tested extensively against atmospheric observations and has been shown to generally reproduce observed concentrations of OH, HO2, NO, NO2, and ClO in the stratosphere to within ±30%, provided precursor fields and aerosol surface areas are accurately known.

However, no significant tests of the model photochemistry of the lowermost stratosphere were performed during the recent M&M workshop. The chemistry of this region is considerably different. For example, the relatively high ratio of CO to ozone implies that ozone production from the oxidation of CO is much more important in this region than at higher altitudes. Furthermore, at the tropopause and below, saturated conditions often exist; therefore, chemical processes occurring on ice particles may be important. In addition, because this air is influenced by mixing of reactive trace gases from the lower troposphere, these models must consider transport of a larger number of reactive species than they typically do.

2.3.1.4.2. Dynamics

Tests of the dynamics within the 2-D and 3-D models during both M&M I and II revealed a number of problems. In general, the mean age of air within the stratosphere is much older than predicted by these models. Measurements of CO2 (Boering et al., 1996) and sulfur hexafluoride (SF6) (Elkins et al., 1996), both of which are increasing rapidly, provide a means of dating stratospheric air. The models had a high dispersion in predicted conversion rate of N2O to NOy. It is unclear whether this dispersion reflects errors in dynamics or chemistry related to the high-altitude sink of NOy. This is a key point: If the assessment models are unable to accurately simulate observed concentrations of total NOy, their ability to predict the influence of additional NOy from aircraft on ozone will remain relatively uncertain. Transport in the lowermost stratosphere is considerably different, and in many ways even more difficult to represent in 2-D models, than transport at higher altitudes. This fact certainly does not bode well for the ability of current 2-D models to describe accurately the dynamic context within which the current subsonic fleet is operating.

2.3.1.4.3. Comparison of model data and observations of stratospheric ozone

As part of the M&M II effort, results from a group of stratospheric models were compared with a recently developed ozone climatology (WMO, 1998). The data used for the climatology are from sonde stations and from SAGE II, the latter data set having been evaluated by comparison with other satellite, lidar, sonde, and Umkehr data. Although agreement between models and between models and observations is relatively good above 25 km, differences between modeled and observed ozone are found to increase rapidly below 25 km and are largest between 20 km and the tropopause. The modeled ozone tends to be larger than observed ozone by up to a factor of 2 at these altitudes.

Overestimation of LS ozone in some models may be ascribed partly to the fact that they have tropopauses at mid-latitudes that are either invariant or do not vary correctly with season. However, based on chemistry and dynamics tests described in the preceding subsections, it is likely that differences between models and observations are caused in large part by deficencies in model transport representation.

2.3.1.5. Implications for Modeling Aviation Impacts

Global tropospheric 3-D CTMs are now the main modeling tools for climate-chemistry studies, including the role of subsonic aircraft NOx emissions. Although 3-D models with high temporal and spatial resolution have performed significantly better than 2-D or monthly averaged 3-D CTMs in the 222Rn, PhotoComp, NOx, and Ozone/GIM intercomparison exercises, key fundamental problems have been identified that are crucial to the representation of the impacts of subsonic NOx emissions from aircraft.

3-D CTM studies have provided only preliminary estimates of subsonic impacts, which exhibit significant scatter, as Table 2-1 shows. At present, we are unable to rationalize these real differences in results between studies because there is no one aspect of input data or process parameterization that can account for the spread in model results. Furthermore, the extent of model evaluation is highly variable, and no models have been evaluated comprehensively against all of the key issues detailed in Section 2.3.1.1.

These same difficulties apply to the subset of models adopted in Chapter 4 to examine the future impact of subsonic aircraft. There is no suggestion that these models have any distinguishing features that identify them as being inherently more or less reliable for assessment of the tropospheric impacts of subsonic aircraft NOx emissions. Furthermore, we have no concrete means of establishing a higher level of confidence in the models used in Chapter 4, compared with any of the similar 3-D models listed in Table 2-1.

Although the effects of present aviation on ozone are calculated to be much smaller in the stratosphere than in the troposphere -primarily because of the smaller fraction of exhaust released into the stratosphere-the performance of 2-D stratospheric models has not been extensively evaluated in the lowermost stratospheric region. Consequently, the results reported in Section 2.3.1.3 represent only preliminary estimates of subsonic aviation impacts on the stratosphere. The modeling situation is significantly better for evaluating the effects of future supersonic aircraft in that a number of intercomparisons have established the general quality of modeled middle stratosphere photochemistry. However, confident predictions of stratospheric effects of future aviation will require resolution of discrepancies between modeled and observed transport tracers.

2.3.2. Uncertainties in Observing Aviation Impacts

The data set resulting from ozonesondes is the only useful one for ozone trend analysis in the UT and LS. The error of an individual ozonesonde measurement has been evaluated to be ~5% in the LS, based on several intercomparison campaigns (WMO, 1998). The error is larger in the UT, where ozone densities, hence instrument signals, are substantially smaller. In addition, the background signals (i.e., dark current) of the sonde sensors have been checked relatively infrequently during the measurement period, giving rise to further measurement uncertainty. If the measurement error is random, one can improve the statistical significance of observed trends by increasing the observation frequency. Ozone densities vary greatly on time scales of days in the UT and LS, particularly in middle and high latitudes during the winter and spring. The cause of this variability is believed to be related to active dynamic transport associated with weather disturbances. Because the variability is largely random, it can be treated, to first order, as noise in the trend data. The variability is considered to be of the same order of magnitude (or larger) as noise from instrument measurement errors. The frequency of ozonesonde observation-once a week at most stations-is not enough to document these variations properly.

Long-term trends of external forcings other than aircraft greatly complicate analysis of ozone trends. The long-term variation of atmospheric chlorine loading is relatively well-documented, allowing for the construction of credible models to predict stratospheric ozone depletion. However, changes in gases important in UT photochemistry-such as NOx, oxygenated hydrocarbons, and water vapor-are much less well characterized. Feedbacks on tropospheric gases from climatic changes (e.g., greenhouse warming) may also have an impact on ozone in the UT and LS, but even the sign of this effect on ozone levels is uncertain.

In summary, because the database for ozone observations in the UT and LS is still relatively limited and because uncertainties in observational data, as well as model representations of non-aircraft ozone forcing phenomena, are quite large, it is presently impossible to associate a trend in ozone to aircraft operation with meaningful statistical significance.

2.4. Conclusions and Overall Assessment of Present Aviation Impacts on Ozone

Currently, there is no experimental evidence for a large geographical effect of aircraft emissions on ozone anywhere in the troposphere. Furthermore, the only evidence for an effect on NOx-the major ozone precursor in aircraft emissions anywhere outside the immediate vicinity (i.e., a few miles) of a jet engine's exhaust-has been obtained during a stagnant meterological condition when exhaust products built up over several days. Nevertheless, our understanding of UT/LS chemical and dynamical processes continues to improve and has progressed to a point where one can predict with some confidence the cruise-level effects of aviation.

Based on our current overall understanding of UT and LS processes, we are confident that NOx emissions from present subsonic aircraft lead to increased NOx and ozone concentrations at cruise altitudes, especially in air traffic corridors between and over Northern Hemisphere continents and at altitudes of 9-13 km. Based on the relatively large number of tropospheric model calculations, we are reasonably confident that tropospheric ozone increases from aircraft NOx have been on the order of 8 ppb, equivalent to 6% of the ozone density in the principal traffic areas.

Model studies, which have internal uncertainties associated with process parameterization and external uncertainties associated with the strengths of other very large NOx sources, have returned effects as low as 2% of the ozone density in high-traffic areas and as high as 14% in those areas. One of the major current limitations to the models' credibility in assessing aircraft emissions is the identification and quantification of background NOx levels and sources. Recent HOx measurements allow for a much better understanding of ozone production in the UT, and these measurements have shown that additional HOx sources are necessary to explain the observations. Moreover, these additional HOx sources cause the sensitivity of ozone production from NOx emissions to be higher than previously thought.

Much less confidence is attached to our understanding of the effects of NOx emissions in the lowermost stratosphere and aerosol emissions in the troposphere and stratosphere. The available data suggest that these effects are smaller than (and, in the case of aerosols, of opposite sign) those of NOx emissions in the UT (see Figure 2-8).

Based on our model predictions, the impact of present subsonic NOx emissions on ozone is well within the range of interannual variability of ozone concentrations in the UT as measured with ozonesondes. Furthermore, expressed as a decadal trend, the impact of subsonic NOx emissions on upper tropospheric and lower stratospheric ozone is smaller than or comparable to the span of confidence limits in the ozone trend analysis for mid-latitude stations. Finally, we note that aircraft NOx emissions should lead to decreased CH4 concentrations; however, any impact should be undetectable in the CH4 record.

IPCC Homepage