Aviation and the Global Atmosphere

	Other reports in this collection

3.5. Long-Term Changes in Observed Cloudiness and Cloud-Related Parameters

Contrails have long been considered possible modifiers of regional climate (Murcray, 1970; Changnon, 1981). Contrails may increase total cloud and cirrus cloud amounts, and consequently change the Earth's radiation balance. As a result, surface and upper tropospheric temperatures may change (Detwiler, 1983; Frankel et al., 1997). In the following sections, we examine changes in cloudiness and other climate parameters for their possible relationship to aircraft operations. Some data indicate changes in observed cirrus cloudiness. These observations are used to provide a first estimate of an upper bound on the increase in contrail-cirrus coverage since the beginning of the jet air traffic era. Limitations in attributing observed trends to aircraft are discussed in Section 3.5.3.

3.5.1. Changes in the Occurrence and Cover of Cirrus Clouds

3.5.1.1. Surface Observations

Although most studies reporting trends in cloud cover have considered total cloud cover (Henderson-Sellers, 1989, 1992; Angell, 1990; Plantico et al., 1990; Karl et al., 1993), we focus here on studies reporting trends in cirrus cloud cover, which is more relevant to the issue of aviation effects on cloudiness. Observations at Hohenpeissenberg, Germany, indicate that the frequency of high clouds during sunny hours increased from 45% in 1954 to 70% in 1995 (Vandersee, 1997; Winkler et al., 1997). Such large changes are not atypical of regional cloud climatologies (e.g., Rebetez and Beniston, 1998). Over the same period, global radiation during sunshine hours decreased by about 10%, indicating that the observed cloud trend is not an artifact. Similarly, cirrus frequency increased between 1964 and 1990 under cloudy-with-sun conditions by 27% over Hamburg and 15% over Hohenpeissenberg (in northern and southern Germany, respectively) (Liepert et al., 1994; Liepert, 1997). These changes are not directly attributable to aircraft, however; instead, they might be caused by an increase in tropopause altitude or higher relative humidity in the upper troposphere (Winkler et al., 1997). Cirrus cover over Salt Lake City and Denver has increased from about 12% annual mean cover to 20% in the period from the early 1960s (i.e., since the beginning of the jet aircraft era) to the 1980s, possibly because of increased air traffic over those cities. These trends are significantly lower in the 1990s, and similar observations over Chicago and San Francisco show no obvious trends (Liou et al., 1990; Frankel et al., 1997).

Global trends of observed total and cirrus cover can be deduced from ground-based cloud observations over land and ocean (Warren et al., 1986, 1988; Hahn et al., 1994, 1996) for 1971-91. For the period 1982-91, mean global trends in cirrus occurrence frequency were 1.7 and 6.2% per decade over land and ocean, respectively (Boucher, 1999). The decadal trend was 5.6% over North America and 13.3% over the heavy air traffic region located in the northeastern United States of America. The computed average change in cirrus occurrence (1987-91 relative to 1982-86) as a function of aviation fuel use at the observation locations is shown in Figure 3-18. The results indicate a statistically significant (97% confidence level) increase in cirrus occurrence in the North Atlantic flight corridor compared with the rest of the North Atlantic Ocean (Boucher, 1998, 1999).

From the same source of observations, cirrus cover was analyzed for the periods 1971-81 and 1982-91 and for a combined data set extending from 1971-92 (Minnis et al., 1998b). Data were averaged for gross air traffic regions (ATRs) and the rest of the globe. ATRs are rectangular areas on the globe that contain most air traffic routes and constitute 26% of the available regions with data. The ATR trends are larger for cirrus clouds than for total cloudiness and are most significant in the first period (1971-81), with the largest annual values found in the United States of America (3%), western Asia (1.6%), and the North Pacific (1.7%). No significant ATR trends were found over Europe. Changes for the rest of the globe were significant only over land but with values less than those over the United States of America and western Asia. Averaged over all ATRs, the cirrus cover increase between 1971 and 1981 amounts to 1.5% per decade, compared with 0.1% for the rest of the globe. From 1982-91, cirrus cover increased over all areas except over non-ATR land and North Atlantic regions, where it changed by -0.4 and 0%, respectively.

The combined 1971-92 cirrus cover data set shows somewhat different results from the two separate data sets. The mean trend for land was 0% per decade for ATRs, compared to -1.1% per decade for other land regions. Over the United States of America, however, the ATR trend is significant at 1.2% per decade. Over oceans, the mean ATR trend is 1.2% per decade and 0.6% per decade over the rest of the oceans. The Pacific ATR trend is strongest, at 1.5% per decade. Averaging of apparent trends from the two separate data sets gives results that differ from the mean trends in the combined data set. However, the relative difference in the mean trends between the ATR regions and the rest of the globe is roughly 1.1% in both cases, indicating that cirrus coverage in areas with significant air traffic is increasing relative to that over the remainder of the globe.

Using ground-based data sets, Figure 3-19 shows the seasonal trends in cirrus coverage over the United States of America and Europe. The average trend of the two separate data sets is compared with the seasonal cycle of contrail occurrence frequency over the United States of America and cover over Europe. Contrail occurrence over the United States of America was derived from 1 year of surface station observations (Minnis et al., 1997); contrail coverage over mid-Europe was computed from satellite data (Mannstein et al., 1999). Over the United States of America, the seasonal trends in cirrus coverage (statistically significant at least at the 75% confidence level) are roughly in phase with the seasonal cycle of contrail occurrence (see Figure 3-19a), suggesting that cirrus changes are related to contrail occurrence. In contrast, the seasonal cycle of cirrus cover trends for Europe does not resemble the seasonal cycle of contrail cover (see Figure 3-19b). The European data, which are not statistically significant, show that the trends observed for the United States of America are not so obvious for other air traffic regions.

3.5.1.2. ISCCP Observations

Data from the International Satellite Cloud Climatology Project (ISCCP, C2) (Rossow and Schiffer, 1991) between 50�S and 70�N from 1984 to 1990 have also been inspected for trends in total and high cloud cover over land and ocean regions (Minnis et al., 1998b). The trends are similar to those found in the 1982-91 surface observations, though details differ. Over all land areas, total cloudiness decreased by 0.5% per decade; high cloud cover grew at a rate of 1.2% per decade. Over the United States of America, total cloudiness decreased by 2.3% per decade, but high cloud cover increased by 5.5% per decade. The trends over Europe were of the same sign but half the magnitude. Over Asia, total cloud cover increased by 4.8% per decade, and high cloudiness grew by 2.7% per decade. Thus, the satellite data suggest a relatively stronger increase in cirrus over the United States of America than surface observations suggest. Over ocean, total cloudiness decreased by 1.2% per decade, whereas high cloud cover was enhanced by 3.7% per decade overall. Some of the discrepancies between the satellite and surface data may originate from the different spatial sampling patterns, calibration, and orbit drift issues peculiar to the satellite data (Klein and Hartmann, 1993; Brest et al., 1997). The satellite covers all regions almost equally, whereas surface observations are from fixed inhabited locations or from well-traveled ship routes.

Further analysis of the ISCCP satellite and the surface data sets was undertaken to isolate and quantify the effects of contrails (Minnis et al., 1998b). The apparent trends were calculated separately for regions having a mean value of contrail coverage less than and greater than 0.5% as computed by Sausen et al. (1998) for 1992 aircraft operations (see Section 3.4.3). The mean values of the computed contrail coverage in the two regions are 0.04 and 1.4%, respectively (Figure 3-16). Satellite data and surface observations show larger increases in cirrus cloudiness where contrails are expected to occur most frequently than in all other areas (Table 3-5). Satellite data show that, over land, the increase in high cloudiness in contrail regions was almost four times that in other areas. Over ocean, the difference, though less significant statistically, amounts to a differential increase of 1.6% cover per decade in contrail areas with respect to non-contrail areas. Cirrus trends over contrail regions as derived from surface observations are 0.8 to 2.3% per decade greater than those in the remainder of the globe for the two periods from 1971-92. The single 1971-92 surface data set shows a significant increase (1.6%) in cirrus over land contrail regions compared to remaining land areas. Over oceans, the relative difference is negative but insignificant because of the small number of samples and large variance. Although the differences in these trends are significant at confidence levels of 95% only for the land and global ISCCP data and the 1971-81 land surface data, they show consistent tendencies.

3.5.1.3. HIRS Observations

High-Resolution Infrared Radiation Sounder (HIRS) data from NOAA satellites have been analyzed for the period June 1989 to February 1997 to determine total and high cloud cover (Wylie and Menzel, 1999). During this period, high clouds observed by the NOAA-10 and -12 satellites increased by 4-5% over land and ocean in the Northern Hemisphere (from 23 to 65�N) but only by about 2% in the tropics. High clouds increased by about 3% over southern mid-latitude oceans. The trend values inferred from the NOAA-11 and -14 satellites are different and somewhat more uncertain because of orbit drift. Over oceans, they also indicate a larger high cloud increase in northern mid-latitudes (3.3%) than over the tropics (-0.4%). Further analysis of HIRS data is required to determine the extent of any contrail impact.

3.5.1.4. SAGE Observations

Data from the Stratospheric Aerosol and Gas Experiment (SAGE) II satellite instrument indicate that the frequency of subvisible cirrus clouds near 45�N is twice that at 45�S (Wang et al., 1996). Aviation, as well as hemispheric differences in atmospheric conditions and background aerosol (Chiou et al., 1997; Rosen et al., 1997), may contribute to such differences (Sausen et al., 1998).

3.5.1.5. Upper Bound for Aviation-Induced Changes in Cirrus Clouds

The line-shaped contrail cover and global extrapolation described in Section 3.4.3 (see Figure 3-16; Sausen et al., 1998) provide only a lower bound for aviation-induced changes in cirrus cloud cover because they are based on satellite observations that identify only contrails and additional cirrus clouds that are line-shaped. Although estimates of an upper bound of aviation-induced cirrus cover have not yet been established, evidence for a correlation of long-term increases in cirrus cloudiness with air traffic has been published (Liou et al., 1990; Frankel et al., 1997; Boucher, 1998, 1999). Here, observations of cloudiness changes described above are used to provide a preliminary estimate of this upper bound.

Differences in trends derived from observations indicate a stronger mean increase of cirrus amounts in regions with large computed contrail cover than in regions with low computed contrail cover, at least over land (see Table 3-5). The trend difference values vary and are of different statistical significances. The values are considered more meaningful over land because there is less air traffic over the oceans. Of regions with large computed contrail cover, only 14% occurs over oceans, and most of these regions occur near the coasts. In addition, less correlation is expected between computed and observable contrail cover over oceans because actual flight tracks often deviate significantly from idealized great-circle routes. Over land, surface-based observations for 1971-81 suggest a differential increase in cirrus cover of 2.3% per decade. The 1982-91 trend difference is smaller but still positive, and the combined 22-year trend difference is 1.6% per decade. The 7 years of ISCCP satellite data suggest even larger trend differences, both globally and over land. If a trend difference of 1.6% per decade is adopted as most representative of available data, and if that trend is assumed to have persisted for the 3 decades since the beginning of the jet aircraft era (the end of the 1960s), then the current increase in cirrus coverage from aircraft is 4.8% in areas with contrail cover greater than 0.5%. This value is about three times the currently computed linear-contrail cover of 1.4% in those areas and is equivalent to about 0.3% coverage of the Earth's surface, 0.2% more than for line-shaped contrails. If no counteracting process took place that reduced cloudiness over this same period, this value gives an upper bound for total aviation-induced cloudiness.

3.5.2. Changes in Other Climate Parameters

3.5.2.1. Sunshine Duration and Surface Radiation

Observations from 100 stations over the United States of America showed that the mean value for total cloud cover over the years 1970-88 increased by 2.0 � 1.3% relative to the years 1950-68 (Angell, 1990). Sunshine duration decreased by only 0.8 � 1.2%, presumably because cirrus clouds are often too thin to reduce measured sunshine duration. Sunshine duration decreased by an average of 3.7% per decade over Germany between 1953 and 1989 (Liepert et al., 1994). Contrails were estimated to be insufficient to account for the diminished sunshine. In an analysis of radiation trends in Germany from 1964 to 1990, solar radiation during cloudy periods with sun decreased strongly (by 20 to 80 W m^-2) (Liepert, 1997). The change might be attributable to aircraft, but the strong reduction in diffuse radiation revealed a more turbid atmosphere, which cannot be explained by increasing cirrus coverage alone.

3.5.2.2. Temperature

A substantial decrease in diurnal surface temperature range (DTR) has been observed on all continents (Karl et al., 1984; Rebetez and Beniston, 1998). The reasons for this decrease are not clear; it could be caused by a change in aerosol burden, cloud amount, cloud ceiling height (Hansen et al., 1995), soil moisture, or absorption of solar radiation by increased water vapor in the atmosphere (Roeckner et al., 1998). A weak correlation was found between contrail occurrence data and DTR over the United States of America (DeGrand et al., 1990; Travis and Changnon, 1997). The relationship was strongest during the summer and autumn and most significant over the southwestern United States of America. Rebetez and Beniston (1998) find large DTR trends in the Swiss Alps (a region with heavy air traffic) in correlation with low-level clouds but no reduction of DTR at higher elevation sites, contradicting a potential major impact on DTR from aviation-induced cirrus.

From data taken between 1901 and 1977 over the midwestern United States of America, increases of moderated temperatures (below average maximum and above average minimum), decreases in sunshine and clear days, and shifts in cloud cover, occurred after 1960; these changes were concentrated in the main east-west air corridor (Changnon, 1981). There are, however, many reasons for changes in global mean surface temperatures since 1880 (Halpert and Bell, 1997). The maximum increase in tropospheric temperatures occurred at northern mid-latitudes in the lower troposphere (Tett et al., 1996; Parker et al., 1997). However, radiative transfer models predict maximum heating rates right below contrails and cirrus in the upper troposphere (Liou, 1986; Section 3.6). Studies using a full-feedback global circulation model (GCM) indicate a nearly constant increase in temperature with altitude for an increase in thin cirrus cloudiness (Hansen et al., 1997). Thus, existing studies do not support the conclusion that aircraft-induced contrails and cloudiness caused changes in global tropospheric temperatures.

3.5.3. Limitations in Observed Climate Changes

The observations cited above suggest that some of the apparent trends in observed cirrus cloudiness and sunshine duration may be caused by air traffic. However, any observed change in cirrus cloudiness or climatic conditions may have other natural or anthropogenic causes (IPCC, 1996). Natural variability includes the El Ni�o Southern Oscillation (ENSO), the North Atlantic Oscillation (NAO), and the Pacific North America pattern (PNA). For example, the NAO index increased overall from the mid-1960s until 1995 (Hurrell, 1995; Halpert and Bell, 1997). However, there is no evident trend in the NAO index (Hurrell, 1995) and mean position of the Iceland low (Kapala et al., 1998; M�chel et al., 1998) over the period 1982-91 when significant increases in cirrus occurrence and cirrus amount were observed over the North Atlantic Ocean (Boucher, 1998, 1999). Changes in cirrus and subsequent changes in sunshine duration may reflect changes in cyclonic activity (Weber, 1990), ocean surface temperature and related atmospheric dynamics (Wallace et al., 1995), upper troposphere temperature and relative humidity, tropopause altitude (Hoinka, 1998; Steinbrecht et al., 1998), or volcanic effects (Minnis et al., 1993)-or they may be a result of global climate change caused by increased greenhouse gases. Changes in anthropogenic or natural emissions of aerosol or aerosol precursors at the ground might also influence cirrus cloud formation (Plantico et al., 1990; Hansen et al., 1996; Hasselmann, 1997). Cirrus trend values deduced from surface observations may also be influenced by a variety of issues. For example, reported cirrus frequencies depend in an unknown manner on how observers classify contrails as cirrus. Reliable cloud observations are obtained mainly during daytime periods (Hahn et al., 1995), which do not always correlate with the main air traffic periods. The significance of the statistical relationship between cloud changes and air traffic is limited because of the limited duration of satellite and surface cloud observations and the brief period of aircraft emissions in comparison to long-term climate variations. Thus, existing studies are not complete enough to conclude that aircraft-induced contrails and cloudiness have caused observable changes in surface and tropospheric temperatures or other climate parameters.