3.4. Contrail Occurrence and Persistence and Impact of Aircraft Exhaust on Cirrus

Aircraft cause visible changes in the atmosphere by forming contrails that represent artificially induced cirrus clouds. The conditions under which contrails form are discussed in Section 3.2.4. This section describes the formation, occurrence, and properties of persistent contrails and how they compare with natural cirrus.

3.4.1. Cirrus and Contrails

Cirrus clouds (Liou, 1986; Pruppacher and Klett, 1997) contain mainly ice crystals (Weickmann, 1945). The distinctive properties of cirrus and contrails derive from the physics of ice formation. Ice particle nucleation occurs either through homogeneous nucleation (when pure water droplets or liquid aerosol particles freeze) or through heterogeneous nucleation, when freezing of the liquid is triggered by a solid particle or surface that is in contact with the liquid or suspended within the liquid. Both processes depend strongly on temperature and relative humidity (Heymsfield and Miloshevich, 1995; see Section 3.2.4.2).

Comparisons of model simulations with observations suggest that we may understand homogeneous nucleation (Ström et al., 1997; Jensen et al., 1998b) and are making headway in understanding the potential contribution of heterogeneous nucleation (DeMott et al., 1998; Rogers et al., 1998). Each of these freezing mechanisms requires that the atmosphere be highly supersaturated with respect to the vapor pressure of ice before crystals can form. For instance, supersaturations in excess of 40-50% with respect to ice are needed for sulfuric acid particles to freeze homogeneously (Tabazadeh et al., 1997) at temperatures above 200 K. Observations of relative humidity with respect to ice at the leading edges of wave clouds are consistent with the requirement of large ice supersaturations for nucleation of ice on the bulk of the atmospheric aerosol (Heymsfield et al., 1998a; Jensen et al., 1998c). Hence, there is a large supersaturation range in which heterogeneous nuclei could lead to cirrus formation before the bulk of the atmospheric particles freeze (DeMott et al., 1997). This potential for heterogeneous nuclei to cause ice formation at ice supersaturations that are relatively low compared to those needed to freeze sulfate particles leads to concern about the role of aircraft exhaust in modifying ambient clouds (Jensen and Toon, 1997).

Ice crystal number densities are limited by competition between increasing saturation as a result of cooling in vertical updrafts and decreasing saturation as a result of growth of ice crystals. The depletion of vapor as a result of growth of the first few ice crystals nucleated prevents further ice nucleation on the remaining particles (Jensen and Toon, 1994). Once ice crystals form and take up available water vapor, supersaturation declines and further nucleation of ice ceases. This selectivity causes ice crystals in cirrus to be larger relative to droplets in liquid water-containing clouds-apart from differences in saturation vapor pressures over ice compared to water, which causes more water vapor to be available for deposition on ice particles than on water droplets. Large ice particles may precipitate rapidly. Many cirrus clouds have a "fuzzy" appearance because rapid precipitation causes optically thin edges of clouds to be diffuse, and precipitation allows particles to spread in the wind, forming long tails of cloud. Ice crystal nucleation also depends on available aerosol in the upper troposphere, the properties of which are only poorly known (Ström and Heintzenberg, 1994; Podzimek et al., 1995; Sassen et al., 1995; Schröder and Ström, 1997). In some locations, upper tropospheric particles are dominated by sulfates (Yamato and Ono, 1989; Sheridan et al., 1994). However, more recent data show that minerals, organic compounds, metals, and other substances may often be present in significant quantities (Chen et al., 1998; Talbot et al., 1998).

Cirrus clouds occur mainly in the upper troposphere. The mean tropopause altitude is about 16 km in the tropics and 10 km (250 hPa) north of 45°N latitude (Hoinka, 1998). The tropopause temperature at northern mid-latitudes varies typically between -40 and -65°C; it may reach below -80°C in the tropics. At mid-latitudes, the upper troposphere often is humid enough for cirrus and persistent contrails to form. The stratosphere is commonly so dry that cirrus (PSCs) and persistent contrails form only in polar winter (for T > -60°C and p < 250 hPa, a 10-ppmv H2O mixing ratio corresponds to less than 15% relative humidity). Ground-based observers report a mean cirrus cover of 13% over oceans and 23% over land (Warren et al., 1986, 1988). Satellite data (Wang et al., 1996; Wylie and Menzel, 1999) identify larger cloud cover (~40%) by subvisible (optical depth t at 0.55 mm below ~0.03) and semi-transparent (0.1 < t < 0.6) cirrus clouds. Cirrus clouds are typically 1.5 (0.1 to 4) km thick; are centered at 9 (4 to 18) km altitude; have a crystal concentration of 30 (10^-4 to 10⁴) L-1, with crystal lengths of 250 (1 to 8000) µm (see Dowling and Radke, 1990); and have an optical depth at 0.55 µm of about 0.3 (0.01 to 30) (Wylie and Menzel, 1999). These typical values of cirrus crystal concentration and mean particle size may be biased by early studies that failed to make adequate measurements of ice crystals smaller than about 50 µm. More recent studies suggest that crystal concentrations in cirrus are often on the order of 1 cm^-3 (Ström et al., 1997; Schröder et al., 1998b), although crystal concentrations in wave clouds can exceed 10 cm^-3. PSCs (between the tropopause and ~25-km altitude) and noctilucent clouds (~80-km altitude) contain ice and might therefore be included as very high altitude forms of cirrus clouds. Systems of cirrus clouds have a lifetime that may reach several hours or even days (Ludlam, 1980), but the lifetimes of particles within clouds is much shorter.

Persistent contrail formation requires air that is ice-supersaturated (Brewer, 1946). Ice-supersaturated air is often free of visible clouds (Sassen, 1997) because the supersaturation is too small for ice particle nucleation to occur (Heymsfield et al., 1998b). Supersaturated regions are expected to be quite common in the upper troposphere (Ludlam, 1980). The presence of persistent contrails demonstrates that the upper troposphere contains air that is ice-supersaturated but will not form clouds unless initiated by aircraft exhaust (Jensen et al., 1998a). Aircraft initiate contrail formation by increasing the humidity within their exhaust trails, whereas local atmospheric conditions govern the subsequent evolution of contrail cirrus clouds. Indeed, the ice mass in long-lasting contrails originates almost completely from ambient water vapor (Knollenberg, 1972).

Ice-supersaturated air masses are often formed when ice-saturated air masses are lifted by ambient air motions. While the air lifted, it may remain cloud-free until it is cooled adiabatically to near-liquid saturation (Ludlam, 1980). Other evidence for large supersaturation occurring in the upper troposphere is provided by cirrus fallstreaks that grow while falling through supersaturated air layers (Ludlam, 1980) and by a few localized humidity measurements (Brewer, 1946; Murphy et al., 1990; Ovarlez et al., 1997; Heymsfield et al., 1998b). Recent humidity measurements by commercial aircraft show that-in flights between Europe, North and South America, Africa, and Asia-14% of flight time was in air masses that were ice-supersaturated with a mean value of 15% (Helten et al., 1998; Gierens et al., 1999).

3.4.2 Cirrus and Contrail Models

Small-scale, regional, and global models have been used to study cirrus clouds and contrails. Small-scale models simulate the details of cloud formation in a single parcel of air along some defined trajectory (e.g., Jensen et al., 1994b). High-resolution 2-D models have simulated the dynamics and microphysics of cirrus (Starr and Cox, 1985) and contrails (Gierens, 1996; Chlond, 1998). Regional and global models describe the large-scale dynamics of clouds with simpler representations of microphysical cloud processes (Sundqvist, 1993; Lohmann and Roeckner, 1995; Fowler et al., 1996; Westphal et al., 1996). Regional-scale (typically 10-km horizontal grid-scale) and global-scale models are not able to resolve the vertical motions and small-scale temperature and humidity fluctuations (Gierens et al., 1997) that drive supersaturations involved in cirrus nucleation. However, global models with prescribed or parameterized contrail cover have been useful for understanding the radiative impact of a change in cloud cover as a result of aircraft emissions (Ponater et al., 1996).

3.4.3. Contrail Occurrence

At plume ages between 1 min and 1 h, contrails grow much faster horizontally (to several km width) than vertically (200 to 400 m), especially in highly sheared environments (Freudenthaler et al., 1995, 1996; Sassen, 1997). Young contrails spread as a result of turbulence created by aircraft vortices (Lewellen and Lewellen, 1996; Gerz et al., 1998; Jensen et al., 1998a,b,c), shear in the ambient wind field (Freudenthaler et al., 1995; Schumann et al., 1995; Dürbeck and Gerz, 1996; Gierens, 1996), and possibly radiatively driven mixing (Jensen et al., 1998d).

Contrails often become wide and thick enough to induce radiative disturbances that are sufficient to be detectable in multispectral satellite observations. They have been observed at 1-km spatial resolution with instruments such as the Advanced Very High Resolution Radiometer (AVHRR) on board National Oceanic and Atmospheric Administration (NOAA) polar-orbiting satellites (e.g., Lee, 1989) and at 4-km resolution in the infrared with the Geostationary Operational Environmental Satellite (GOES) (Minnis et al., 1998a). The AVHRR channels in the 11- to 12-µm range (4 and 5) are particularly suited to detect thin ice clouds because of the different emissivity of ice particles in this spectral range (King et al., 1992; Minnis et al., 1998c). Figure 3-12 shows, for example, a mid-European scene derived from AVHRR data in these channels, combined with the visible channel to represent the surface, for a day when many line-shaped contrails were formed by heavy air traffic (Mannstein, 1997). The figure shows that aircraft trigger contrail cirrus that evolves into cirrus clouds that are much more extensive in scale than the initial contrails. Such spread and deformed contrail cirrus can no longer be distinguished from naturally occurring cirrus. In Figure 3-12, contrails that still have a line-shaped appearance cover about 5% of the scene.

Aged contrails often cannot be distinguished from cirrus, which poses an observational problem in determining the frequency and area of coverage by contrails. An important example of the persistence of contrails and their evolution into more extensive cirrus is shown in Figure 3-13. An initial oval contrail observed in GOES-8 satellite images diffused as it was advected over California until it no longer resembled its initial shape 3 h later (Minnis et al., 1998a). The exhaust from this single aircraft flying for less than 1 h in a moist atmosphere caused a cirrus cloud that eventually covered up to 4,000 km² and lasted for more than 6 h. Other contrails and contrail clusters were observed to develop over periods of 7 to 17 h, spreading to cover areas of 12,000 to 35,000 km². Such dispersed contrails are usually indistinguishable from natural cirrus; hence, satellite detection algorithms based on the linear structures of young contrails will not detect these dispersed contrails.

At present, observations of actual frequency or areal coverage of Earth by contrails are limited to a few selected regions. Contrail frequency refers to the probability that a contrail will be observed somewhere within the scene being viewed; area of coverage refers to the fraction of the area of the scene in which contrails are observed. Estimates of contrail coverage or occurrence have been made directly or indirectly from surface (Detwiler and Pratt, 1984; Minnis et al., 1997) and satellite observations (Joseph et al., 1975; Carleton and Lamb, 1986; Lee, 1989; DeGrand et al., 1990; Schumann and Wendling, 1990; Betancor-Gothe and Grassl, 1993; Bakan et al., 1994; Mannstein et al., 1999).

Data for the frequency of contrails are available from surface-based observations at 19 locations across the continental United States of America for every hour during 1993-94 (Minnis et al., 1997). The data indicate that contrail frequency peaks around February/March and is at a minimum during July. Annual mean persistent contrail frequency (not the cover) for the 19 sites was 12%. When related to fuel use and extended to the remainder of the country, mean annual contrail frequency for the United States of America is estimated at 9% (Minnis et al., 1997). The relationship between fuel consumption and contrail frequency from this data set is shown in Figure 3-14. The correlation implies that contrail coverage is limited mainly by the number of aircraft flights, not by atmospheric conditions at cruise altitudes. Pilots flying over the former Soviet Union have reported that contrails occur most frequently in winter and spring and less often during summer (Mazin, 1996). Sky photographs taken from 1986 to 1996 over Salt Lake City, Utah, reveal a seasonal cycle in contrail frequency, with a maximum in fall and winter and a minimum in July (Sassen, 1997). These data are similar to observations from a site 64 km north (Minnis et al., 1997) of Salt Lake City. Sassen (1997) and Minnis et al. (1997) found that contrails occur with cirrus in approximately 80% of observations. The coincidence of cirrus and contrails suggests that cloud-free supersaturated regions are usually interspersed with areas in which natural clouds have actually formed.

From an analysis of AVHRR infrared images taken over the northeast Atlantic and Europe, a mean contrail cover of 0.5% was derived for 1979-81 and 1989-92 (Bakan et al., 1994). A distinct seasonal cycle was found, with a southward displacement of the maximum cover during winter. Maximum coverage (~2%) occurred during summer, centered along the North Atlantic air routes. Using AVHRR channel 4 and 5 differences and a pattern-recognition algorithm to differentiate line-shaped clouds from fuzzy cirrus clouds, the contrail cover-defined as line-shaped clouds over mid-Europe-was systematically evaluated for all of 1996 using nearly all noon passages of the NOAA-14 satellite (see Figure 3-15) (Mannstein et al., 1999). Contrails are not uniformly distributed; instead, they lie along air traffic corridors and accumulate near upper air route crossings. As with observations over the United States and the former Soviet Union, maximum coverage occurred during winter and spring. At noon on average over the year, line-shaped contrails cover about 0.5% of the area of the mid-European region shown in this figure. This coverage represents the lower bound for the actual contrail cover because the algorithm cannot identify non-line-shaped contrails. Contrail coverage at night over Europe is one-third of the noon-time contrail cover (Mannstein et al., 1999).

Contrails often occur in clusters within regions that are cold and humid enough to allow persistent contrails to form. Contrail clusters observed in satellite data indicate that these air masses cover 10 to 20% of the area over mid-Europe (Mannstein et al., 1999) and parts of the United States of America (Carleton and Lamb, 1986; Travis and Changnon, 1997), consistent with the fraction of air masses expected to be ice-supersaturated at cruise altitudes. Hence, as air traffic increases in these regions, persistent contrail coverage is also expected to increase, possibly up to a limit of 10-20%.

Estimates of global coverage by air masses that are sufficiently cold and humid for persistent contrails can be obtained using meteorological analysis data of temperature and humidity, a model to estimate the frequency of ice-saturation in each grid cell as a function of the analyzed relative humidity, and the Schmidt-Appleman criterion (depending on h; see Section 3.2.4.1) for contrail formation conditions. Using 11 years of meteorological data from the European Center for Medium Range Weather Forecast (ECMWF), the cover of suitable air masses is found to be largest in the upper troposphere, especially in the tropics (global mean value of 16%) (Sausen et al., 1998). This coverage is similar in size to the area covered by clusters of contrails in satellite data and the frequency of ice-supersaturated air masses observed along commercial aircraft routes (Gierens et al., 1999). The expected contrail cover is then computed from the product of the air mass coverage and the fuel consumption rate in the same region, with the latter chosen as one of several possible measures of air traffic. The product is scaled to give a 0.5% mean contrail cover in the European/Atlantic region (30°W to 30°E, 35 to 75°N) as considered by Bakan et al. (1994). The resulting contrail cover (see Figure 3-16) for 1992 fuel emissions and h = 0.3 is 0.087% (about 0.1%) in the global mean, with a local maximum of 5% over the eastern United States of America. The computed contrail coverage over southeastern Asia is only slightly smaller than that over Europe and North America. Although air traffic is much less extensive over southeastern Asia, this high contrail coverage may result because this region more often has ice-supersaturated air. However, the predicted contrail coverage over North America and southeastern Asia has not yet been verified by observations. Predicted regional differences are sensitive to the representation of the number of aircraft in operation. For example, in some regions, a large fuel consumption density is caused by a few large aircraft (Gierens et al., 1998). Studies have not been performed to evaluate the accuracy of humidity values provided by meteorological data. The results depend linearly on scaling by the cover observed in the reference region, as provided here by Bakan et al. (1994). Other data imply global cover values between 0.02 and 0.1% (Gierens et al., 1998). A value larger than 0.1% (possibly 0.2%) cannot be excluded because the analysis uses only thermodynamic contrail formation conditions and the scaling is based solely on observed line-shaped contrail cover.

3.4.4. Contrail Properties

The relatively small particles present in newly formed contrails serve to distinguish contrail radiative properties from those of most natural cirrus (Grassl, 1970; Ackerman et al., 1998; see also Section 3.6). Measurements of contrail particles with impactors and optical probes (see also Section 3.2.4 and Table 3-1) reveal a wide variety of size, shape, and spectral size distributions. The results depend on plume age, ambient humidity, ambient aerosol, and other parameters. At a plume age of 30 to 70 s, ice particles have been found to form a single-mode log-normal size distribution with a volume-equivalent radius in the range of 0.02 to 10 µm, a mean radius of about 2 µm, and maximum dimension of 22 µm. Particle shapes are mainly hexagonal plates, along with columns and triangles. The axial ratios of the columns were found to be less than 2, and the shapes of the crystals were already established for particles of about 1-µm radius (Goodman et al., 1998). Contrail particle sizes increase with time in humid air. Ice particles observed in 2-min-old contrails typically have radii of 2 to 5 µm with shapes that are almost spherical, indicating frozen solution droplets (Schröder et al., 1998b). On the other hand, some contrail particles strongly polarize light, indicating non-spherical shapes (Freudenthaler et al., 1996; Sassen, 1997). After 10 min to 1 h, contrail particle size distributions may range from 32 to 100 µm or even 75 µm to 2 mm (Knollenberg, 1972; Strauss et al., 1997).

Particle properties within a contrail core differ from those at the edges of the contrail. In the center of contrails, insufficient water vapor is available to allow the large number of ice particles to grow to large crystals. At the edges, ice supersaturation is greater, thereby sometimes allowing ice crystals to form up to 300 µm in diameter (Heymsfield et al., 1998a). The larger particles have various shapes, with the largest being bullet rosettes (Lawson et al., 1998). Such large particles are within the natural variability of cirrus particle sizes. As a consequence, old dispersed contrails appear to have particle sizes similar to those in surrounding cirrus (Duda et al., 1998; Minnis et al., 1998a).

Aircraft create a large number of new ice crystals that grow from ambient water vapor. As a contrail spreads, some evidence suggests that ice crystals are neither lost nor created in significant numbers. In one remote-sensing study, the total number of particles (about 2.6 x 10⁹ ice crystals per cm of contrail length) in a contrail was found to remain constant as the contrail dispersed and the particle concentration (particles per unit volume) dropped (Spinhirne et al., 1998). In situ measurements (Schröder et al., 1998b) reveal 10⁸ to 10⁹ particles larger than 4.5-mm diameter formed per cm of contrail length. In contrast, changes in particle number will occur when some particles eventually sediment out of the contrail and new particles are nucleated as a result of air motions induced by the contrail. The ice water content of new contrails increases with time to exceed the amount of water emitted by the aircraft by more than two orders of magnitude. However, the water content per unit volume remained approximately constant in the case observed by Spinhirne et al. (1998). The optical depth of contrails remains nearly constant during the first hour despite horizontal spreading (Jäger et al., 1998).

3.4.5. Impact of Aircraft Exhaust on Cirrus Clouds and Related Properties

Aircraft may perturb natural cirrus through the addition of water vapor, soot, and sulfate particles and by inducing vertical motions and turbulent mixing (Gierens and Ström, 1998). Observations of cirrus coverage in certain regions have found perturbations from anthropogenic aerosol (Ström et al., 1997). Persistent contrails are often associated with or embedded in natural cirrus (Minnis et al., 1997; Sassen, 1997). Such in-cloud contrails may be formed slightly above the Schmidt-Appleman temperature threshold because ambient ice particles that enter the engine inlet increase the humidity in the exhaust plume (Jensen et al., 1998a; Kärcher et al., 1998a). Some evidence associates in-cloud contrails with regions of enhanced absorbing material and enhanced ice crystal number densities (Ström and Ohlsson, 1998). The presence of HNO3 may increase the hygroscopic growth of supercooled cloud droplets (Laaksonen et al., 1997), and HNO3 dissolved in sulfate solution droplets may change their freezing behavior. The importance of such perturbations has not been quantified.

Soot particles originating from aircraft exhaust may act as freezing nuclei. In an atmosphere with few freezing nuclei, this perturbation could lead to an expansion of cirrus cover, a change in average particle size, and related changes in cloud surface area and optical depth (Jensen and Toon, 1997)-hence have consequences for radiative forcing (see Section 3.6.5). For aircraft to alter cirrus properties, exhaust particles would need to be more efficient or more abundant than ice nuclei currently existing in the atmosphere. If the predominant particle that freezes to form ice contains sulfate, aircraft soot could serve as important ice nuclei because sulfate particles require a relatively high supersaturation before freezing occurs. Over the central United States of America, however, as many as 0.1 to 0.2 cm^-3 freezing nuclei (effective at -35°C) have been found in the upper troposphere (DeMott et al., 1998), indicating that the number of freezing nuclei is not always small. Some observations show that aircraft exhaust does not contain large numbers of ice nuclei active at temperatures above -35°C (Rogers et al., 1998). Soot and metals were found to be significant, but not dominant, components of ice nuclei in contrails (Chen et al., 1998; Petzold et al., 1998; Twohy and Gandrud, 1998).

Aircraft measurements in and near clouds have indicated the presence of light-absorbing material contained inside ice crystals. The distribution pattern and the amount of measured absorbers suggest that the material is related to aircraft soot (Ström and Ohlsson, 1998) (Figure 3-17). For the same abundance of aerosol particles, clouds perturbed by absorbing material contained 1.6 to 2.8 times more ice crystals than unperturbed portions of clouds. These observations suggest that aircraft-produced particles enhance cloud ice particle concentrations, although they have not revealed the physical mechanism involved. Specifically, exhaust soot particles may have been involved in ice crystal formation within the cirrus or formed contrail ice particles within the exhaust plume before being incorporated into the cloud. These observations are roughly consistent with calculations of a cirrus cloud forming in a region of recent exhaust trails (Jensen and Toon, 1997).

Cirrus clouds may also be perturbed by enhanced sulfate aerosol. Small sulfate particles (e.g., 10-nm radius) are unlikely to be ice nuclei or cloud condensation nuclei. Larger particles would be more efficient. Such large particles may originate as ambient sulfate particles enlarged by growth from sulfur gases emitted by aircraft or by processing of liquid exhaust particles in short-lived contrails (see Section 3.2). Because ice nucleation processes tend to be self-limiting (Jensen and Toon, 1994), cirrus cloud properties would change only slightly if the number of sulfate particles increased while the size distribution and composition of ambient sulfate particles remained constant. On the other hand, large sulfate particles-such as those induced by the Mt. Pinatubo volcanic cloud-could lead to a significant increase in the number of ice crystals, optical depth, and radiative forcing of cirrus clouds (Jensen and Toon, 1992). Cloud modifications would require a large increase in the number of larger sulfate particles; such modifications would occur primarily in cirrus that were very cold (T < -50°C) and weakly forced by slow updrafts. Lidar observations of a cirrus cloud embedded in aerosol from the Mt. Pinatubo eruption show significantly higher than normal ice crystal concentrations (Sassen, 1992; Sassen et al., 1995). Satellite data analyzed after the Mt. Pinatubo eruption suggest that the volcanic aerosol reduced the occurrence of cirrus clouds with high extinction coefficients, increased the occurrence of clouds with low extinction coefficients, and increased extinction in optically thick clouds (Minnis et al., 1993; Wang et al., 1995). Hence, some available models and data suggest that aircraft exhaust could play a significant role in modifying the properties of clouds. However, the magnitude and nature of these modifications are not well understood.

Sedimentation of large particles in persistent contrails may remove water vapor from the upper troposphere, possibly reducing radiative heating by water vapor, and cause seeding of lower level clouds (Murcray, 1970; Knollenberg, 1972). Sedimentation of ice crystals, which has been observed occasionally (Konrad and Howard, 1974; Schumann, 1994; Heymsfield et al., 1998a), becomes important only in strongly supersaturated air (Hauf and Alheit, 1997) when large ice crystals form and have the potential to fall through lower-lying saturated air in which they will not evaporate readily. Because no attempts have been made to quantify the precipitation rate from contrails, the significance of this precipitation has not been assessed.

IPCC Homepage