3.5 Long-term Variations

In this section we shall examine variations in water vapour on time scales longer than the quasi-biennial oscillation. We shall highlight the results from several instruments that have made long-term measurements on a nearly continuous basis and examine their results in some detail. Some of these instruments provide global coverage, and thus show that the evidence for an increase in northern mid-latitude stratospheric water vapour presented in Chapter 2 is not, at least in the 1990s, geographically confined to these regions. Long-term variations in water vapour in the upper troposphere will also be presented. Finally we will discuss some of the mechanisms that might cause the observed changes, and some of the important consequences of significant changes in atmospheric water vapour.

3.5.1 Stratospheric Measurements

Chapter 2 presented data from a large number of instruments that have measured water vapour at northern mid-latitudes since the 1950s. In this section we will highlight two specific data sets that have long records of regular measurements of stratospheric water vapour. The oldest data set that has provided quasi-continuous measurements of water vapour has been taken over Boulder, Colorado, since 1981 [Oltmans and Hofmann, 1995]. Since the early 1990s measurements from HALOE and from ground-based microwave instruments have also become available [Russell et al., 1993, Nedoluha et al., 1996]. The HALOE water vapour measurements are of particular value because they provide global coverage and because simultaneous measurements of methane are available.

The water vapour increase over Boulder, Colorado

Section 1.2.1 describes the operation of the balloon-borne, frost point hygrometer used to obtain vertical profiles of water vapour over Boulder, Colorado (40°N). This nearly 20-year record is unique in describing water vapour in the upper troposphere and lower stratosphere [Oltmans and Hofmann, 1995]. The relatively infrequent soundings (approximately monthly) do not provide adequate information to determine long-term changes in the troposphere because of the high variability of water vapour in this region. However, as has been shown in earlier chapters, the data are of suitable quality to provide information about changes in the lower stratosphere up to normal ceiling altitudes of the balloon of about 28 km.

The time series of water vapour mixing ratio in 2 km layers for three layers in the stratosphere is shown in Figure 3.30. The 16-18 km layer (Figure 3.30a) is the lowest level that is always in the stratosphere. This layer has a significant seasonal cycle and is the lowest altitude at which a linear trend fit to the data shows a significant increase in water vapour concentrations. Numerical increases are seen at altitudes even as low as 10 km but the large variability because of the mixture of tropospheric and stratospheric data makes the trend determination not statistically different from zero. At 20-22 km (Figure 3.30b) the seasonal cycle is small so that variability is a minimum in this region. The increase at 24-26 km (Figure 3.30c) is similar to that at all levels above 16 km (Figure 3.31). The increase of approximately 1%/year is highly significant at all levels above 16 km.

Figure 3.30. Time series of 2 km layer average water vapour mixing ratio in parts per million by volume (ppmv) over Boulder, Colorado, for (Top) 16-18 km which is the lowermost stratosphere over Boulder, (Middle) 20-22 km which is an altitude above the strong seasonal variation, and (Bottom) 24-26 km. The data are obtained from balloon-borne, chilled-mirror hygrometers that are launched approximately monthly. The solid line is a least squares fit to the data and m is the slope with the 95% confidence interval.

 

Figure 3.31. The linear trend of water vapour mixing ratio (percent per year) and 95% confidence interval (shaded area) as a function of altitude over Boulder. Significant increases of about 1% per year are found at all altitudes above 16 km.

The computation of seasonal changes does not yield significant linear trends in many instances but may yield some information on the pattern of changes. Figure 3.32 is similar to Figure 3.31, but it shows a linear trend as a function of height over Boulder for two different 4-month periods of the year. Above about 18 km, linear trends for December to March are stronger than linear trends for August to November, exceeding 1%/year. For the remainder of the annual cycle, linear trends still exceed 0.5% year-1. Linear trends for the April-July period (not shown) would fall between these lines. The stronger increase during December to March, which is when the water vapour minimum occurs over Boulder (Figure 3.10) has reduced the amplitude of the seasonal variation in the lowermost stratosphere over Boulder. This region of the stratosphere over Boulder has water vapour mixing ratios that can have only the equatorial lower stratosphere as their source. This suggests that an important component of the increase over Boulder is a change in the water vapour content of the source region or a weakening of the transport of air from this region to Boulder.

 

Figure 3.32. Four month averages of the linear trend of water vapour mixing ratio (in percent per year) and 95% confidence interval as a function of altitude over Boulder. Thick solid line and shaded confidence interval for December-March, thin solid line and diagonal hatched confidence interval for August-November. The April-July linear trend would fall between these two lines (not shown).

Increases in stratospheric water vapour in the 1990s

In addition to the multi-decadal increases in water vapour observed by balloon measurements at 40°N and by aircraft measurements over Southern England (section 2.5.4), a particularly rapid increase was observed globally during the years 1991-1996. This increase was documented with observations from the HALOE instrument, which first made measurements in October 1991 [Nedoluha et al., 1998a; Evans et al., 1998; Randel et al., 1999b]. In Figure 3.33 we show the global increases in water vapour (solid line B) and 2´ CH4+H2O (solid line A) as measured by HALOE from January 1992 to December 1996. The results in this figure are derived from time series of monthly, zonal, averages of HALOE version 19 H2O and 2´ CH4+H2O profiles starting in January 1992 and extending through December 1996. The HALOE H2O data are averaged vertically in 2 km increments over the altitude range 20- 70 km and in 10° latitude increments from 70°S to 70°N; i.e. 70° N to 60° N; 60° N to 50° N; and so on. The HALOE CH4 and H2O profiles are first combined on common altitudes to calculate individual 2´ CH4+H2O profiles and then these profiles are averaged similarly to H2O. Each of these monthly zonal averages are then averaged again with area weighting to produce a near-globally averaged time series. To investigate the annual average increase in H2O and 2´ CH4+H2O mixing ratios, multiple linear regression is used to fit the HALOE near-globally averaged time series to a model including a linear trend, an annual, semi-annual, and 27-month harmonic, and an autoregressive noise series of order 1.

Figure 3.33. Vertical profiles of near global (70°N-S) linear increases in HALOE 2´ CH4+H2O (profile A with no ticks on error bars) and H2O (profile B with ticks on error bars) derived from January 1992 through December 1996. Dashed line is the linear increase in HALOE H2O (profile B), but with a different fit to the quasi-biennial oscillation (see text). Linear increases are expressed in ppbv year-1 and the error bars on each curve represent the ±2 standard errors of the derived linear variation.

Since previous analyses of the linear variation in water vapour measured by HALOE during this period have indicated significant differences [WMO, 1999], we have included a second independently calculated linear increase in Figure 3.33. The primary difference between the two calculations is the replacement of the 27-month sinusoid (solid line B in Figure 3.33) with a time series based on zonal wind measurements at Singapore between 70 and 10 hPa (dashed line in Figure 3.33). Other analysis techniques also provide very similar results. We conclude that the details of the analysis technique should not result in significantly different linear variations for this period.

Figure 3.34 shows the change in water vapour detected by HALOE throughout the stratosphere and mesosphere from October 1991 to December 1996. The measurements show that the global increase shown in Figure 3.33 was present at all latitudes. Ground-based microwave instruments at 34.4°N and 45.0°S were therefore able to validate the increase detected by HALOE [Nedoluha et al., 1998a]. A comparison between the increase derived from the ground-based measurements and those from HALOE is given in section 2.5.5. Examination of the HALOE 2´ CH4+H2O measurements over a longer period of time (Figure 3.35) shows that the increase observed between 1992 and 1996 has not continued beyond 1996. Linear variations calculated for the period 1996-1999 (not shown) show no significant increases anywhere in the stratosphere. The strong change in low-frequency behaviour demonstrates that the global increases during 1992-1996 are not characteristic of monotonic decadal-scale trends, but rather suggestive of episodic change (possibly linked with the 1991 Mount Pinatubo volcanic eruption). The character of this time series suggests cautious interpretation of trends derived from short time series with arbitrary beginning and endpoints.

Figure 3.34. The linear increase calculated for the period October 1991 through December 1996 from HALOE H2O measurements that have been binned in 10° latitude increments. In units of ppmv year-1.

Figure 3.35. Time series of deseasonalised anomalies in 2´ CH4+H2O from HALOE measurements. The solid and open circles show anomalies averaged over the southern and northern hemispheres, respectively, and the smooth lines show the global means derived from a running gaussian filter intended to minimise the effects of the quasi-biennial oscillation (clearly seen in the time series at 10 hPa). Taken from Randel et al. [2000].

3.5.2 Tropospheric Measurements

In this section we shall investigate the long-term variations observed in the troposphere. In considering our ability to detect long-term changes in the troposphere, it is important to note that, with the exception of the MLS measurements, most of the measurements discussed here are taken by instruments that were designed to provide operational weather information. These measurements have therefore not generally put a high priority on the stability required for the study of long-term variations. Changes in instrumentation over time may therefore introduce time-varying biases into these measurements.

MLS observations of the upper troposphere

Figure 3.36 shows a 7-year time series of daily averaged data for humidity in specific and relative units, percent cloudiness and temperature in low latitudes (between 30° S and 30° N) derived from measurements made by MLS. MLS measures UTH from emission and reports vapour in relative humidity and mixing ratio units. The millimetre signals received by MLS can make observations in cirrus but the received signal will be enhanced by ice emission that is approximately half as strong on a per-mass-density content as vapour. Therefore ice has an impact on the data analysis and needs to be accounted for. Relative humidity values greater than 120% indicate likely presence of cloud and these were counted to derive a percent cloudiness. All measured humidity values exceeding 120% were then adjusted to 100% before analysing for humidity trends. The corresponding specific humidity values are set to ice saturation.

The data at 147 and 215 hPa are considered suitable for an analysis of long-term temporal variations. Daily averages of data taken between 30° S and 30° N were computed. To minimise sampling problems only days having at least 80% of the maximum possible science data are included in the analysis. An analysis of the global data set gives results similar to those found at low latitudes, but the magnitude of the long-term variations is reduced because the higher latitudes are much drier. The daily cloud counts are divided by the total number of observations for that day to give a percent cloudiness. The temperature data are exclusively from daily averaged NCEP reanalyses sampled at the MLS measurements points, and are entirely independent of the MLS measurements themselves. The plots show the daily average (grey dots) and a 12-month smoothed average (thick black line).

The relative humidity and cloudiness fields at both pressure levels show a clear minimum in 1994-1995, while the temperature field shows a maximum at this time. The anti-correlation in the variations of the temperature and relative humidity at 215 hPa produces a water vapour mixing ratio that has a shallow maximum at this level. Qualitatively the curves at 215 hPa seem reasonable. When the atmosphere cools the vapour carrying capacity of the atmosphere declines, hence relative humidity and cloudiness should increase. If cloudiness increases with declining temperature as shown, then more water is partitioned into ice and the vapour mixing ratio declines. At 147 hPa there is also a large decrease in relative humidity from 1991-1995, but this is accompanied by only a small increase in temperature during this period. This results in a clear minimum in water vapour mixing ratio during in 1994-1995. We note that there are several possible mechanisms that could cause variations on multi-year time scales that would not result in a linear trend over this time period, including the Pinatubo eruption that occurred 3 months prior to UARS launch, and ENSO.

Figure 3.36. Top panel to bottom: MLS volume mixing ratio, MLS relative humidity, MLS % cloudiness and NCEP temperature. Daily averaged data for 7 years between 30S and 30N. The columns represent different altitudes: a 3 km layer around 147 hPa (left column) and 215 hPa (right column). Dots are daily averages, and the black line is a 12-month running mean boxcar filter to remove the annual cycle.

HIRS observations of the upper troposphere

The HIRS observations provide information on relative humidity over ~200-500 hPa. Because this is a nadir-viewing instrument the vertical resolution is significantly coarser than that of MLS or SAGE. In addition, the level at which the emission is observed can also vary slightly. All of these differences result in decreased variability in the HIRS measurements relative to MLS and SAGE.

For this analysis, data from eight different HIRS instruments were used over a nearly 20-year time period of long-term variations. The intercalibration procedure uses the overlap periods of the different satellites to transfer the calibration from a baseline satellite to all other satellites [Bates et al., 1996]. The HIRS instruments used to derive UTH are filtered radiometers and slight differences in the filter response functions from one instrument to the next are measured in the laboratory before each instrument launch. Thus, the empirical in-orbit intercalibration can be compared with a physically based intercalibration using the known filter responses, a sample set of atmospheric temperature and moisture profiles, and a forward radiative transfer model [Bates et al., 2000]. In most cases, the agreement between the empirical and forward-modelled intersatellite biases is on the order of only a few tenths of a degree. This excellent agreement lends credibility to the assertion that most of the inter-satellite bias is attributable to known differences in the filter response functions and is removed by empirical inter-satellite calibration.

Although satellite remote sensing has the advantage of complete global coverage from a single instrument, the inversion of the radiometric observations into geophysical variables, such as the specific humidity profile or water vapour mixing ratio, is an ill-conditioned problem that requires the use of additional a priori data. Great care must be exercised to ensure that the use of such a priori data does not lead to systematic biases. The alternative, analysis of the radiances or their close equivalent UTH, reduces the dependence on a priori data, but does not provide information on the water vapour amount independent of temperature.

Assessing long-term variability of UTH for climate change studies is difficult because of the high variability due to the southern oscillation and the competing effects of water vapour and temperature changes on UTH. Figure 3.37 shows the time series for the annual averages of UTH for different regions and the results of a linear fit for each region. The subtropical regions have the lowest relative humidity (28.6% for 10° S-30° S and 33.1% for 10° N-30° N) and the mid-latitudes the highest relative humidity (44.3% for 30° N-60° N and 43.7% for 30° S-60° S). The relative humidity for the deep tropics (10° N-10° S) is 38.9%, and the average for the entire region (60° N-60° S) is 39.7%. None of the calculated linear variations are significant at the 99% level assuming that each year is independent. The linear changes for 60° N-60° S (0.04% year-1), 30° S-60° S (-0.1% year-1) and 10° N-10° S (0.1% year-1) are significant at the 95% level. These values should be viewed with caution because of the short length of the time series. In particular, in the deep tropics significant interannual variability and persistence, hamper the detection of long-term variations. The variability masks longer-term variations, and the persistence reduces the statistical significance of calculated linear changes because temporal autocorrelation reduces the number of independent observations.

The 30° S-30° N UTH variations measured by HIRS are also plotted in Figure 3.37 to provide a comparison with the variations observed by MLS (Figure 3.36). Of particular interest is the minimum in relative humidity that occurs in 1994. This minimum is contemporaneous with the minima observed in the MLS data at 147 hPa and 215 hPa, but is clearly much shallower. The shallower minimum seen in the HIRS data is qualitatively consistent with the shallower minimum seen with decreasing altitude in the MLS data, but care must be taken in comparing the two data sets because of the large differences in vertical resolution and because the HIRS measurements are not always sampling precisely the same pressure levels.

Figure 3.37. Annual mean upper tropospheric humidity over ice from HIRS for various latitude bands and linear fit statistics from 1980 to 1997. 60S-60N (solid line), 60S-30S (dotted line), 30S-10S (dashed line), 10S-10N (dot-dash line), 10N-30N (dash-3dots line) and 30N-60N (long dashed line).

Lower tropospheric measurements

Because the lower and upper troposphere are dynamically linked, and are not separable in the way the stratosphere and troposphere are, it is appropriate to examine long-term changes in lower tropospheric water vapour. Here we summarise long-term variations determined using in situ, meteorological observations and satellite observations, some not presented in the previous Chapters. Multi-decadal changes in water vapour over some Northern Hemisphere land areas have been deduced from surface and radiosonde observations. These observing systems are not research quality systems, nor were they designed for the purpose of monitoring multi-year or multi-decadal changes in water vapour. Changes in instrumentation and observing methods, often not documented, hinder analysis of long-term water vapour changes [Elliott and Gaffen 1991, Elliott 1995, Rind 1998]. Nevertheless, efforts to detect long-term variations of lower-tropospheric water vapour have been made because of the importance of lower-tropospheric moisture in the global energy and water budgets. Lower-tropospheric water vapour increases are expected to accompany increases in global surface temperatures, because of the temperature dependence of both evaporation from the surface and the capacity of the atmosphere to hold moisture. Therefore, monitoring tropospheric water vapour changes is part of the much broader global climate change detection effort.

The most comprehensive analysis of radiosonde-based water vapour variations focused on surface to 500 hPa precipitable water (column-integrated water vapour) at those Northern Hemisphere radiosonde stations with temporally homogeneous records for 1973-95. Ross and Elliott [1998] find increases over North America except for Northeast Canada, consistent with an earlier analysis of a shorter record [Ross and Elliott 1996]. Over Eurasia, only China and the Pacific islands showed coherent, statistically significant, regional increases. The remainder of Eurasia showed a mix of positive and negative changes with a tendency for decreases in water vapour over Eastern Europe and western Russia.

These results are broadly consistent with the Gaffen et al. [1992] analysis of a smaller global station network with shorter records. Other regional analyses are those of Zhai and Eskridge [1997], who found increases of about 1-3 % decade-1 in surface-to-200 hPa precipitable water over China for 1970-90, and Gutzler [1992, 1996], who found specific humidity increases of 3 to 9 % decade-1 at 1000, 700, and 300 hPa from four western tropical Pacific radiosonde stations.

In an attempt to analyse water vapour changes over a longer time period, Ross and Elliott [1999] found mostly positive annual variations in lower tropospheric specific humidity and dew point data at 25 Northern Hemisphere stations with temporally homogeneous records for 1961-1995. The increases were smaller than those for the 1973-1995 period and few were statistically significant.

The aforementioned studies are consistent in their findings of tropospheric water vapour increases at individual radiosonde stations over periods during which instrument changes are not believed to contaminate data records. In contrast, Peixoto and Oort [1996] reported decreases in zonal-mean relative humidity between 1974 and 1988. The decreases were more marked at 300 hPa, where they are more likely associated with instrument changes than at lower levels, and were more pronounced at higher latitudes than in the tropics.

One of the striking results of global studies of total-column water vapour (most of which resides in the lower troposphere) is the strong correlation with globally averaged tropospheric temperature. This result is found in comparisons between a merged radiosonde, HIRS and Special Sensor Microwave Imager (SSM/I) tropospheric water vapour data set and the global Microwave Sounding Unit (MSU) temperature data [Randel et al. 1996] for 1988-1994. A similar result is found in the SSM/I data alone, compared with MSU over oceans [Wentz and Schabel, 2000] for 1987-1998. Locally, both radiosonde data and satellite data show correlations between total-column water vapour and sea surface temperature or surface air temperature, although the correlations break down at very high temperatures in tropical locations [Stephens 1990; Gaffen et al., 1992; Bony et al, 1995].

An earlier analysis that used HIRS and MSU data to estimate precipitable water for 30° N-30° S, for the period 1979-1995, suggested a 3% average drying of the tropical belt between the two periods 1979-1987 and 1989-1995 [Schroeder and McGuirk 1998a, 1998b]. However, the analysis could be influenced by artificial drying signals associated with changes in radiosonde instrument types, because radiosonde data were used in the statistical regression of precipitable water from the HIRS data [Ross and Gaffen 1998]. Total-column water vapour data from (SSM/I) for July 1987 through October 1998 show a global (over ocean only) increase of 2.2% per decade [Wentz and Schabel, 2000].

In summary, radiosonde humidity measurements show some evidence for increases in lower-tropospheric water vapour over limited regions of the globe during the past one to four decades, at rates of several percent per decade. Satellite data yield better spatial coverage, but analysis of long-term changes has been limited both by shorter data records and problems merging the data from multiple satellites. A study of SSM/I data from 1987-1998 indicates increases comparable to those found in the radiosonde record.

3.5.3 Causes of Long-term Variations

While any long-term changes in upper tropospheric water vapour would be extremely important because of their radiative impact (see 3.5.4), it is not clear whether a significant change has indeed taken place. The long-term increase in water vapour in the stratosphere does, however, seem to be well documented. Several mechanisms that could be important in causing changes in stratospheric water vapour, both on a decadal time-scale and on longer time-scales, will be discussed in this section.

Only since 1991 is a continuous, near-global data set available for water vapour in the stratosphere. Since the quantity 2´ CH4+H2O is nearly conserved in the stratosphere (3.2.1) the ability of HALOE to measure both of these molecules is important in distinguishing different mechanisms that can cause an increase in water vapour. Figure 3.33 showed the increase in 2´ CH4+H2O. The only other significant reservoir of hydrogen in the stratosphere is H2 (~0.5 ppmv), and given the slow reaction rates associated with this species it is unlikely to contribute significantly to the observed changes. Of the three major hydrogen containing species in the stratosphere, only the entry level of H2O (3.2.1 and 2.5.4) can plausibly vary enough to provide the observed increase in 2´ CH4+H2O.

The increase in stratospheric water vapour seems to have stopped about 1996. This suggests that the large increase in the early 1990s might be related to the June 1991 eruption of Mt. Pinatubo [Nedoluha et al., 1998a], but there are certainly other possibilities. The 11-year solar cycle is near solar minimum at the same time that stratospheric water vapour peaks in ~1995-1996. While upper mesospheric water vapour is closely tied to the solar cycle, there is no known mechanism through which the solar cycle should affect stratospheric water vapour. There are many well known decadal scale oscillations which affect tropospheric parameters (e.g., the Arctic Oscillation, the Pacific Decadal Oscillation, etc.), and it is possible that some of the variations in stratospheric water vapour that have been observed are related to these oscillations. As an example, the Pacific Decadal Oscillation may itself help to prolong the period of the southern oscillation [Barnett et al., 1999], and it is possible that the variations in stratospheric water vapour are related to the enhanced El Niño activity of the early 1990s.

While the rapid global increase in water vapour in the early 1990s may be attributable to volcanic events or to decadal scale oscillations, the results shown in Chapter 2 (Figures 2.68 and 2.70) suggest that there is a much longer-term increase in water vapour in the lower stratosphere since the 1950s, across many instruments (Figures 2.70 and 2.72). Average water vapour mixing ratios as low as ~2 ppmv were reported for 140-170 hPa in the 1950s over Southern England by the British Meteorological Research Flight frost-point instrument (Figure 2.68e). This is significantly lower than the current entry-level water vapour mixing ratio. Such low mixing ratios are not currently found anywhere in the stratosphere, except in air that has been dehydrated in the extremely cold regions of the Antarctic vortex. Unless these older measurements are seriously in error, we must assume that these extremely low water vapour mixing ratios measured with the British Meteorological Research Flight frost-point instrument in the 1950’s and the Naval Research Laboratory frost-point in the middle 1960’s (see Figure 2.68d) point toward a very large change in global stratospheric water vapour over the last 50 years.

Changes in tropical tropopause temperature

Since the amount of water vapour in the stratosphere is very sensitive to the temperature at which air enters the stratosphere, long-term variations in the cold point tropopause temperatures may result in variations in the stratospheric water vapour budget. Even a small change in temperature could have significant effects, since the saturation vapour pressure of water vapour changes by ~15-20% for each 1 K change in temperature near the tropopause (at constant pressure).

The extent to which changes in tropical tropopause temperature affect stratospheric water vapour is dependent upon the saturation level of the air entering the stratosphere, and this is in turn dependent upon where tropospheric air enters the stratosphere. While the tropopause temperature calculated during the recent particularly cold years seems to suggest that the observed water vapour mixing ratios are consistent with tropical air crossing the tropopause uniformly throughout the tropics [Dessler, 1998], such a result is inconsistent with the stratospheric water vapour and tropopause temperature measurements from the 1980s and early 1990s [Vömel and Oltmans, 1999]. Regardless of whether or not air crosses the tropopause uniformly in the tropics, a long-term increase in the tropical tropopause temperature would probably result in an increase in water vapour.

Four studies of temperature trends at 100 hPa are summarised in WMO [1999]. These cover the years 1979-1994 and therefore do not include the particularly cold years of 1996 and 1997. The trends shown range from ~0.3° K decade-1 heating to ~0.5° K decade-1 cooling. Six studies of trends at 50 hPa do, however, all agree that from 1979-1994 there is a cooling trend in the tropical stratosphere, with numbers ranging from ~0.3 to 1.4° K decade-1.

Using analyses of the European Centre for Medium Range Weather Forecasting (ECMWF) from 1979-1998, Simmons et al. [1999] calculated a cooling trend of ~0.6° K/decade in the global means of the 100 hPa temperatures. They also noted that the tropical 100 hPa temperatures for 1996 and 1997 were particularly cold. While there was a change in the ECMWF analysis technique in January 1996, Simmons et al. consider it unlikely that this is the cause the very cold temperatures in the final 2 years of the study.

Zhou et al. [2000a] examined tropical tropopause temperatures using operational sounding data from 1973-1998. Instead of calculating the global trend at the 100 hPa level as did Simmons et al. [1999], they calculated the average temperature at the tropical cold-point tropopause. While the time scale, altitude, and latitude range of their study was different from that of Simmons et al., Zhou et al. found a cooling trend of 0.57° K/decade, in agreement with the trend calculated by Simmons et al..

Randel et al. [2000] conducted a study of the interannual variability of the tropical tropopause temperature from radiosonde data and from the NCEP reanalyses. They noted that, while the radiosonde data shows a cooling trend of 0.5 K/decade from 1979-1997, no such trend is apparent in the NCEP reanalyses. Furthermore, both the radiosonde and NCEP data showed that while the 1996 and 1997 years were relatively cold, they were similar to values in the period 1984-1987 (unlike the ECMWF results of Simmons et al [1999]). They also point out that there is widespread evidence of abrupt temperature changes due to changes in instrumentation and methods, and these changes may significantly impact any long-term analysis, especially for data extending back before the mid-1980s [Gaffen, 1994; Gaffen et al., 2000]. Reanalysis data also contain artificial abrupt changes [Randel et al., 2000], such as discontinuities introduced by satellite data assimilation begun in late 1978. These problems in both data sources hamper detection of long-term changes in tropopause temperature.

None of these studies of tropopause temperature extend back to the 1950s, but they are roughly coincident with the period of the measurements made in Boulder since 1981. Since observations show that stratospheric water vapour has been increasing over this period, especially during the early 1990s, we cannot make a direct link between long-term changes in tropical tropopause temperatures and increases in stratospheric water vapour. The lack of a relationship, if real, casts doubt on the hypothesis that the temperature of the tropical tropopause (or cold point tropopause) is the major governing process for stratospheric water vapour. As noted in section 3.3.3, other processes such as convection, cloud microphysics or circulation changes may be important.

Aircraft emissions

Another suggested source of increasing stratospheric water vapour is from aircraft emissions. The contribution of aircraft to atmospheric water vapour is directly from the H in the fuel. For a typical fuel hydrogen mass concentration of 13.8±0.02% by mass, aircraft emit with an emission index EI(H2O) of 1.23±0.02 kg of water vapour per kg of burned fuel [Schumann, 1996; IPCC, 1999]. Stratospheric emissions from the current world aircraft fleet are concentrated in Northern mid-latitudes as is expected from the air traffic pattern and the latitudinal variation of tropopause height. Depending on the definition of the tropopause, between 18 and 44% of current emissions of water vapour by aircraft are deposited in the stratosphere [Hoinka et al., 1993; Gettelman and Baughcum, 1999]. However, since flight routes are close to the tropopause and reach at most into the lowermost stratosphere, this effluent is rapidly returned to the troposphere with little expected accumulation [Holton et al., 1995].

For the IPCC assessment on the effects of aviation [IPCC, 1999], systematic tracer simulations were carried out with a suite of two-dimensional and three-dimensional atmospheric models to determine upper bounds for the accumulation of aviation aerosol in the atmosphere [Fahey and Schumann, 1999; Danilin et al., 1998]. These tracer simulations strongly suggest that aircraft emissions are not the source of observed decadal H2O changes at 40°N. The simulation results can be scaled by EI(H2O) to provide an upper bound (neglecting precipitation from the upper troposphere) for the accumulation of water vapour above the tropopause due to aircraft emissions. Taking the model that shows the largest tracer accumulation at 40°N, the computed H2O mixing ratio due to subsonic aircraft emissions reaches a maximum of 0.055 ppmv at 10 km altitude decreasing smoothly to 0.012 ppmv at 24 km. These values are small in comparison to current ambient values of 5.9 ppmv at 10 to 12 km and 4.2 ppmv at 22 to 24 km [Oltmans and Hofmann, 1995]. Assuming 5% yr-1 growth in fuel consumption, the change in aircraft-produced H2O ranges from 0.0034 ppmv yr-1 at 10 km to 0.0008 ppmv yr-1 at 24 km. These values represent a change of +0.006% yr-1 at 10 km and +0.018% yr-1 at 24 km. These increases in water vapour are more than a factor of 20 smaller than those found in long-term balloon observations (1-2 %/yr, Figure 3.31 and section 3.5.1).

The current contribution of aircraft emissions to stratospheric water vapour is therefore small, and unlikely to be a significant contributor to the observed increase. Nevertheless, far larger concentration changes and radiative forcing values would result from a large fleet of supersonic aircraft emitting water vapour into the stratosphere [IPCC, 1999].

Methane oxidation

The only well understood cause of long-term increases in stratospheric water vapour is the increase in anthropogenic methane emission. Increases in methane entering the stratosphere will lead to an increase in stratospheric water as the methane is oxidised. As can be seen in Figure 3.38, the increase in methane during the 1990s is less than ~0.01 ppmv yr-1 [Dlugokencky et al., 1994, Dlugokencky et al., 1998]. This should result in at most a ~0.02 ppmv yr-1 increase in stratospheric water vapour, far below what is observed in the HALOE data. For the period 1951 to 1981 Rinsland et al. [1985] derived column methane linear trends of 1.1±0.2% yr-1, while Zander et al. [1989] found 0.7±0.1% yr-1 for the period 1951 to 1986. A conservative upper limit for the total increase in methane since 1951 is ~0.5 ppmv. If all of the methane that entered the stratosphere were oxidised, then this would result in an increase of ~1.0 ppmv in water vapour since 1951. In regions, such as the lower stratosphere, where only a fraction of the methane entering the stratosphere has been oxidised this increase will be smaller. In particular, in the northern mid-latitude lower stratosphere regions where long-term measurements are available, ~1/2 of the methane will have been oxidised. The long-term increase in water vapour on these measurements due to increased methane entering the stratosphere would therefore be at most ~0.01 ppmv/year, which is below the rate of increase estimated from either the British Meteorological Research Flight data or the Boulder balloon data.

Figure 3.38. Top Panel: Globally averaged CH4 mole fractions (dots) based on NOAA Climate Monitoring and Diagnostics Laboratory cooperative air sampling network sites [see Dlugokencky et al., 1998]. A smooth curve and deseasonalised trend have been fitted to the global averages. Bottom Panel: The globally averaged instantaneous CH4 growth rate, calculated as the derivative with respect to time of the trend curve in the top panel.

The rate at which methane is entering the stratosphere is increasing slowly, and thus plays only a minor role during periods of rapid increase such as the early 1990s. Changes in the fraction of methane that is oxidised in a particular region can, however, cause rapid changes in water vapour at some altitudes. This is best seen in Figure 3.33, where the increase in H2O alone is, at some altitudes, significantly larger than the increase in 2´ CH4+H2O. Nedoluha et al. [1998b] found that there were no plausible changes in methane chemistry that could reproduce the observed changes in stratospheric methane in the early 1990s. Using a model, Nedoluha et al. [1998b] illustrated that a decrease in the tropical upward transport rate in the stratosphere could, however, cause the observed change in methane. Such a change in transport may, in turn, have been caused by the June 1991 eruption of Mount. Pinatubo.

Transport

Some of the long-term variations in stratospheric water vapour may be caused by changes in transport. As mentioned in section 3.5.1, the changes in water vapour observed at mid latitude locations (such as Boulder, Colorado or over southern England) may be due to changes in the transport of air from the tropical tropopause. The changes in transport need not be steady or uniform. Water vapour increases could be caused by less transport of dry tropical air, or by more transport of wet tropical air. Smith et al., [2000] found, using a seasonal analysis of the HALOE record, that the increases in HALOE water vapour between 1992 and 1999 may be driven by increases in the upper troposphere in boreal autumn, when water vapour concentrations are high in the tropics. Smith et al., [2000] attribute these changes to changes in the Asian monsoon. Similarly, Zhou et al., [2000a] have proposed that the increases in stratospheric water vapour might be a result of changes in the latitudinal extent of the upwelling in the tropics. If tropical upwelling extended to higher latitudes with warmer tropical tropopause temperatures, water vapour in the stratosphere would increase. Thus changes in the location or rate of transport in either time or space may affect the water vapour distribution in the stratosphere.

Summary of causes and expected future variations

The mechanisms described here do not, in any combination, offer an explanation of the long-term increase in stratospheric water vapour that is fully consistent with all of the observations. Since water vapour chemistry in the stratosphere is thought to be well understood, and since the increase in stratospheric water vapour is now very well documented, we conclude that there have probably been changes in the process of stratosphere-troposphere exchange especially at the tropical tropopause, a process that is still poorly understood (see section 3.3.3). The increase in water vapour entering the stratosphere is occurring despite an observed decrease in the average tropopause temperature. This suggests that the average tropical tropopause temperature does not fully explain the amount of water vapour entering the stratosphere, and that changes in transport into the stratosphere, or within the stratosphere are important. Such a conclusion is perhaps not surprising, since it has long been argued that the tropical tropopause temperature is too warm to explain the dryness of the stratosphere [Newell and Gould-Stewart, 1981].

As noted above, the increase in anthropogenic methane emission from the 1950s through the 1980s is responsible for perhaps one half (0.5 ppmv) of the increase in stratospheric water vapour between the British Meteorological Research Flight measurements and the present suite of measurements described in detail in Chapter 2 (Figure 2.68). IPCC reports from the early 1990s forecast continued increases in anthropogenic methane concentrations in the 21st century, which could result in continued increases in stratospheric water vapour. Figure 3.38 shows, however, that during the 1990s, there were significant changes in the growth rate of atmospheric methane concentration. If this change in the rate of methane increase is an indication that atmospheric methane is approaching a steady-state value, then any increase in stratospheric water vapour that is caused by increases in methane emission will soon cease. We note, however, that it is possible that recent changes in tropospheric chemistry are causing a decrease in the lifetime of methane, and that this is temporarily masking a continued increase in methane emissions [Dlugokencky, 1998].

Given the uncertainties in the future changes in methane, it is impossible to determine whether water vapour will continue to increase during the 21st century. It is certainly possible that if the positive trend in the amount of methane entering the stratosphere has come to a halt, then the increase in water vapour that has occurred over the last 50 years will also cease. Measurements since the mid-1990s do suggest that there has been no increase in water vapour since that time, and it is therefore possible that water vapour in the stratosphere has now stabilised.

We note, however, that a large fraction of the increase in stratospheric water vapour over the last 50 years cannot be attributed solely to increases in methane concentrations, but is instead probably the result of as yet unidentified changes in the exchange of water vapour from the troposphere. The amount of water vapour in the stratosphere will therefore only approach a steady-state value if the changes resulting from changes in methane concentrations, and the changes resulting from the as yet unidentified exchange mechanism, sum to zero. Since any changes in methane concentration that do occur in the future are likely to cause an increase in water vapour, and since the as yet unidentified mechanism has caused a large increase since the 1950s, it seems likely that some increase in stratospheric water vapour will continue in the 21st century. Based on figures such as 3.30 and 3.35, we have more confidence in saying that future increases (or decreases) in the water vapour content of the stratosphere may not be well described by a linear trend.

3.5.4 Consequences of long-term variations

Effect of water vapour changes in the stratosphere

Forster and Shine [1999] have shown that a prolonged increase in stratospheric water vapour would cause significant cooling of the stratosphere, and that this could be a significant contribution to the observed cooling of the stratosphere. They found that a fixed 0.7 ppmv increase in stratospheric water vapour mixing ratio would result in a global cooling effect, which is comparable to that caused by stratospheric ozone losses from 1979-1997. The cooling is particularly strong in the Arctic, where they calculate a temperature decrease of 3-7 K in the spring. Shindell et al. [1998], in modelling the effect of increasing greenhouse gas concentrations, found that a 5-7 K decrease in springtime Arctic temperatures resulted in ozone column depletions of ~50%. These results were obtained by assuming that chlorine activation takes place whenever temperatures fall below 195 K, an approximate threshold for the formation of polar stratospheric clouds. Kirk-Davidoff et al. [1999] combined the Shindell et al. [1998] results with the Forster and Shine [1999] results and pointed out that these implied that an increase in water vapour could have a significant effect on Arctic ozone depletion. In addition to its indirect effect on stratospheric temperatures, an increase in water vapour would also raise the saturation temperature required for the formation of polar stratospheric clouds. Kirk-Davidoff et al. [1999], however, found that the effect of increased saturation temperature on the formation of polar stratospheric clouds was about an order-of-magnitude smaller than the effect of the reduced Arctic temperature due to increased water vapour.

In addition to its effect on Arctic ozone depletion, a prolonged increase in water vapour at the rate observed in the early 1990s, would also have a significant impact on ozone recovery in the upper stratosphere. (Note, however, that the evidence quoted earlier does not necessarily support the idea of a prolonged, continuous increase in water vapour mixing ratios). Nevertheless, it appears that the changes observered over decadal time-scales are capable of exerting a significant effect on the radiative and chemical balance of the stratosphere. For example, Jucks and Salawitch [2000] find that a 2.0 ppmv increase in water vapour in this region would negate the recovery of ozone caused by a 15% reduction of odd chlorine, Cly. While this portion of the stratosphere does not significantly affect total column ozone amounts, this is the region where ozone recovery could first become apparent: this would be a sensible place to monitor for such evidence.

Furthermore, recent work has shown that stratospheric water vapour changes may contribute significantly to radiative forcing of surface temperature. Using an atmospheric model to simulate the last 20 years, and an increase of stratospheric water vapour consistent with the observations in Figure 3.30, Shindell [2000] illustrated that this simulation was consistent with observed changes in stratospheric temperature, and that the radiative forcing from the additional stratospheric water vapour contributed 0.2 W m-2 (~25 %) to climate forcing over the period. Thus continuing changes in stratospheric water vapour might significantly affect the evolution of climate forcing beyond greenhouse gas emissions. However, Shindell [2000] increased stratospheric water vapour concentrations by increasing the tropopause temperatures, which is not in accord with observations.

Effect of water vapour changes in the troposphere

The principal effects of variations of water vapour in the troposphere are expected to be: (i) the effect on the radiative balance, and (ii) the effect on cloudiness (which indirectly influences the radiative field). We will concentrate here on the effect on the radiation balance, since the effects of water vapour on clouds are beyond the scope of this assessment.

Water vapour throughout the troposphere exerts a strong influence over how the Earth loses radiative energy to space to balance the energy received and absorbed from the sun. The water vapour feedback (i.e., the feedback on global temperature caused by changes to water vapour resulting from increases in CO2 and other gases) is now almost universally accepted to be positive and strong. Lindzen [1990] hypothesised a negative feedback process occurring in the upper troposphere in regions of strong descent near tropical convection, and suggested that general circulation models cannot properly simulate this effect and furthermore that it could overwhelm the direct feedback in the boundary layer. Since that time, however, many observational and modelling studies [e.g., Rind et al., 1991, Allan et al., 1999] have indicated that general circulation models broadly (though with poor absolute accuracy at present) capture the essential processes governing upper tropospheric moisture. Hence, a positive water vapour feedback is expected except in quite localised regions of strong descent.

Doherty and Newell [1984] and Clough et al. [1992] demonstrated that the complex pattern of cooling rates, or energy loss to space, as a function of spectral frequency is considerably influenced by the far infrared pure rotation band of water vapour. Doherty and Newell [1984] introduced the use of a height-spectral wave number cross section, on which cooling rates are displayed. Figure 3.2 is an example of this cross section, taken from the work of Brindley and Harries [1998]. This shows, for two model atmospheres, tropical and sub-Arctic, that the pure rotation band of water vapour between about 100 and 500 cm-1 cools strongly to space from the upper and middle troposphere. Cooling rates up to 10-2 K day-1 (cm-1)-1 over some 200-300 cm-1 are calculated in the upper troposphere, that is an integrated cooling rate of about 3 K day-1. In addition, the continuum bands in the window region between about 800 and 1200 cm-1, and the n 2 vibration-rotation band between about 1200 and 1600 cm-1 also contribute to the cooling, but from deeper layers, near the surface and in the lower and middle troposphere. This is in addition, of course, to the strong cooling concentrated in a relatively narrow band around the 667 cm-1 CO2 band, extending through the whole depth of the troposphere and stratosphere, together with cooling due to other spectral features, such as the CH4 band centred at 1300 cm-1. With this description in mind, it is possible to discuss how humidity variations might affect the thermal radiation field.

Sinha and Harries [1995 and 1997] and Brindley and Harries [1998] demonstrated that the peak response to a change in the vertical water vapour mixing ratio distribution occurred at mid-troposphere levels within the atmospheric window for a warm, humid tropical case, and in the far infrared at upper tropospheric levels for a cold, dry sub-Arctic atmosphere. The quantitative effect of changes in upper tropospheric humidity are demonstrated in Figure 3.39 which shows the change in the greenhouse parameter (surface emission minus the outgoing longwave radiation) for +12% increases in water amounts for a tropical model atmosphere in the three layers (a) surface-800 hPa; (b) 800-500 hPa; and (c) 500-0 hPa. Comparing these results with Figure 3.1 shows that this 12% increase in the upper tropospheric humidity leads to a +2% increase, or about 0.5 W m-2, in the outgoing longwave radiation to space. The reason for the small % response of outgoing longwave radiation to a given % change in humidity arises because of the competing effects of an increase in humidity (and thus optical depth) which increases the emission, but which also raises the effective height of cooling to space to a higher, colder layer. Being colder, the enhanced emission due to the higher humidity is offset by the effect of the lower temperature [Harries, 1996, Slingo and Webb, 1997].

Slingo and Webb [1997] analysed the sensitivity of the spectrally resolved outgoing longwave radiation emerging from the top of the atmosphere, as the concentration of atmospheric components and temperature varied as predicted by a climate change calculation using the UK Hadley Centre Climate Model. The modelled stratospheric cooling caused a decrease in outgoing radiation in the CO2 and O3 bands centred at 667 and 1060 cm-1 and the predicted surface warming causing an increased emission to space in the window region. The response of the outgoing longwave radiation to the modelled increase in humidity was very complex, however, due to the near-cancellation between temperature and humidity effects mentioned above.

Because of this cancellation between increasing humidity and decreasing temperature with height, and because of the current uncertainty over the precise water vapour concentration in the upper troposphere, and even greater uncertainty over trends in upper tropospheric humidity and temperatures, it has not been possible to date to carry out a firm quantitative estimate of the upper tropospheric radiative consequences of long-term changes in upper tropospheric humidity. However, current knowledge indicates that the radiative response to long-term changes in upper tropospheric water vapour will almost certainly be significant.

Figure 3.39. Change in greenhouse parameter (surface emission minus outgoing long-wave radiation) for 12% increases in water vapour amounts for a tropical model atmosphere in three layers: Low, surface to 800 hPa (top); Middle, 800–500 hPa (middle) and high, 500–0 hPa (bottom).