Recent insights into the regulation of water vapour entering the stratosphere

Steven Sherwood, Yale University, New Haven, USA (Steven.Sherwood@yale.edu)

Introduction

Water vapour in the upper troposphere and lower stratosphere has come into increasing focus in recent years. Low-latitude water vapour at these levels is important not only for its greenhouse forcing, but also because the air here is advected into the stratosphere where water vapour plays an important role in regulating stratospheric chemistry and temperatures. The chemical role involves both the production of important OH hydroxyl radicals, and the formation of polar stratospheric clouds which subsequently help destroy ozone. The cooling role exacerbates ozone destruction as well.

As air passes the tropical tropopause it is dehydrated to a value approximately equal to the saturation vapour mixing ratio there, but the processes that decide exactly how much vapour stays in the air are not well understood. A recent SPARC assessment (SPARC, 2000) has thoroughly examined the observations of water vapour available in the stratosphere and concluded that previously reported, and unexplained, upward trends of about 1% per year are consistent over a number of available instrumental records spanning nearly four decades (Rosenlof et al., 2001). This SPARC effort has exposed both the inability of trends in tropopause temperature or chemical production to explain those of water vapour, and the acute need for more and better observations in the tropopause region. The failure of known influences on water vapour to account properly for these important trends reveals a genuine need to understand the stratospheric dehydration processes better than they are now.

Until recently, efforts to deduce further details of stratospheric dehydration have concentrated on precise quantitative comparisons of the mean mixing ratio found in the stratosphere to that expected from calculation of the saturation value at the tropical tropopause with respect to the ice phase. Comparison of these quantities based on data from the 1970's using seasonal tropopause climatologies led to the conclusion (Newell and Gould-Stewart, 1982) that dehydration had to be limited to the coldest regions of the tropopause (particularly over Indonesia and the western Pacific Ocean), since the observed stratospheric mixing ratios were lower than expected based on the wider tropics. They assumed further that air must be entering the stratosphere in these regions, dubbing this a "stratospheric fountain."

Recent work casts doubt on this entry conclusion, since observations show air to be sinking out of the stratosphere in the regions where the tropopause is coldest, as would be expected in a thermally direct baroclinic circulation (Sherwood, 2000), rather than ascending into it. Though contradicting the "fountain" image, this finding does not rule out localisation of dehydration in the indicated region since (contrary to common assumption) dehydration and upward motion need not have anything to do with one another, at least at large scales. More will be said about this important point below. Though not requiring a "fountain," the old argument in favour of regional dehydration has itself been found lacking since it rests on climatological tropopause values; if instantaneous water-vapour saturation minima from individual soundings are averaged rather than averaging temperature first on pressure surfaces, a much lower mean value results that is no longer inconsistent with stratospheric mixing ratios at least during the UARS era (Dessler, 1999). Thus the question of where air is dehydrated is not really constrained by the early observational methodologies; temporal fluctuations of temperature and moisture must be reckoned with as well as spatial ones.

Dehydration mechanisms

Any mechanism that can effectively dehydrate the stratosphere requires attainment of cold temperatures and removal of enough of the condensed water to reach stratospheric mixing ratios. Two mechanisms currently enjoy favour as candidates for doing this: first, dehydration within very cold convective overshoots as suggested by Johnston and Solomon (1979) and Danielsen (1982), and second, dehydration within large, stably-stratified air masses near the tropopause lifted by Kelvin waves or other transient motions (e.g., Potter and Holton, 1995). Both mechanisms involve temporary, local cooling of air below the climatological mean value. As it happens, either mechanism would probably be confined to more or less those cold-tropopause regions identified by Newell and Gould-Stewart. This is principally because both mechanisms are either directly or indirectly associated with intense convection, which is in turn highly correlated with cold tropopause temperatures within the tropics. Another factor is that dehydration would be easier in areas of lower mean temperature even if transient fluctuations were not exclusive to those areas.

The main objection to the convective mechanism is that convection rarely overshoots the tropical tropopause, while the minimum in water vapour is at least as high or higher than the mean tropopause thus requiring a dehydration mechanism active at that height. Most convective outflows occur between 150 and 200 hPa in the regions where intense convection occurs (outflows are even lower in places such as the ITCZ). However, some clouds do penetrate the tropopause. Quantifying the vertical extent of convective penetration and mixing in a way that is relevant to stratospheric dehydration (see Gettelman, this issue) is rather difficult.

The main hurdle faced by the second or "in-situ" mechanism is the difficulty of nucleating and removing ice crystals. Very few ice nuclei are present in the ambient air at these levels, so homogeneous nucleation is required to initiate freezing; this occurs only at supersaturations of ~60%. Once crystal formation begins, the water vapour mixing ratios quickly fall below the threshold for homogeneous nucleation, shutting off further crystal formation unless temperatures continue to fall. With the right ascent rates, however, reasonably effective dehydration can occur over time, but temperatures must fall farther than would be necessary with equilibrium thermodynamics (Jensen et al, 2001) and must remain low for at least a day or more. The cirrus clouds that would accompany this process are indeed widely observed, particularly over convective regions where temperatures are coldest and where wave activity is greatest, although they contain little ice. Current observations cannot yet tell us how much actual dehydration occurs in these clouds.

Recent support for, and clarification of, a role by the "overshoot" mechanism

Two recent studies (Sherwood and Dessler, 2000, 2001) have provided further support for the idea that a substantial contribution to the required dehydration comes directly from drying within convective cells. These studies reconsider the behaviour of convective overshoots in light of current understanding of the physics of convection and clouds, and use more recent data.

The earlier of these two studies used correlations found among water vapour, ozone, and other tracer data obtained by the ER-2 at low latitudes (see Figure 1) to argue that the youngest air crossing any given theta surface near the tropopause is also the driest, implying a rapid dehydration process followed by a slower ascent into the stratosphere. In interpreting the relationship, the authors assumed that (outside tropical convection) air can reach theta surfaces near the tropical tropopause only from below in the tropics or from the "overworld" (above 380 K) at mid-latitudes, in accord with prevailing views (Holton et al., 1995). In-situ drying of air during uniform ascent in the tropics should not produce such correlations because slow lifting can take many weeks to lift air from the peak outflow level of convection (13-14 km) up to the hygropause (16-18 km), which should allow plenty of time for horizontal mixing processes within the tropics to give each parcel the same dehydration opportunities regardless of its initial ozone or ascent rate. Nor can the relationship be caused by mixing with overworld air, because the latter's ratio of ozone to water is far too large. It would be worth examining these issues in global models with prognostic water vapour and ozone as they become available.

Probably the most important new point emphasised by the recent work relates to the time-scale argument above but is independent of the dehydration mechanism. This point is that the slowness of the residual, diabatic vertical motion means that dehydration can easily be localised in cold-tropopause regions regardless of residual circulation details. Both candidate dehydration mechanisms operate within an extended (at least a couple of km) vertical space. This means air is mainly processed through the dehydration regions horizontally rather than vertically. In other words, even air that crosses the tropopause in its warmer sectors will usually, at some point, have passed through the cold sectors where dehydration is occurring. Therefore, variations in how the ascent is horizontally distributed would not particularly be expected to cause variations in entry moisture, even if they meant that uplift occurred on average at warmer or colder tropopause temperatures. Instead, variations of (1) the temperature in the dehydration regions specifically, (2) the rate of horizontal processing of air through these regions, or (3) the efficiency of drying in these regions, would be required to change stratospheric moisture.

Earlier, Sherwood (2000) concluded that the observed sinking through the tropopause over Indonesia required a heat sink or diabatic cooling effect and suggested cooling by convective overshoots as the only apparently workable explanation for this. Building on the earlier studies, Sherwood and Dessler (2001) quantified the drying effect of overshoots by constraining a convective model to produce the correct cooling effect. Getting the right cooling required clouds overshooting the tropopause to cover about 0.5% of the tropics. With this number of clouds and CAPE (convective available potential energy) values consistent with observations, they were also able to obtain realistic profiles of water vapour (see Figure 2), ozone, and cloud ice including a water vapour minimum located just above the mean tropopause. The results were quite sensitive to assumptions about cloud mixing, and did not include a seasonal cycle, so they do not provide firm support for single-handed dehydration by overshoots. However, they go a long way toward deactivating criticisms of the possible importance of overshoots that are based on observed quantities of overshoots being too small, observed cloud-top heights being too low, or ozone below the tropopause being too great. In their simulations, dehydration in a limited region was easily able to control moisture entering the stratosphere throughout the tropics, through horizontal mixing.

It is important to emphasise what is different between the model of Sherwood and Dessler (2001) and common perceptions of overshooting behaviour. It is not physically possible for a convective updraft to shoot past its level of neutral buoyancy (LNB, which is always below the tropopause) and remain there without mixing with the environment. This point was recognised by earlier authors. But if such mixing does occur, it will cool the environment diabatically and thereby pull air gravitationally downward across theta surfaces until final detrainment occurs. Though this downward-moving air mass moves through less distance than did the overshooting updraft, it possesses greater mass since it is a mixture of overshoot plus environment. Thus, the net upward mass flux created by overshoots must be negative for some distance below the highest levels reached, if there is mixing. The sign of this flux at any given level (e.g. the tropopause) depends on overshooting and mixing details. But the only way to achieve permanent upward motion through the highest levels at which mixing occurs is by steady lofting outside of the overshoots, in balance with radiative heating.

Thus it is nonsense to ask what fraction of air "enters the stratosphere" by overshooting versus slow lofting, since the existence of overshooting/mixing only increases the necessary lofting rate at the key levels! The processes do not compete, they conspire. Our best estimates of the amount of ambient lofting through the tropopause (balanced by radiative heating) exceed the Brewer-Dobson mass flux in the lower stratosphere by a factor of several, which requires either flux out of the tropics just above the tropopause, the downward pulling process described above, or both.
What if overshoots occur but do not mix with their environment? If the convective elements were all of uniform LNB, the overshooting ones would sink back into the troposphere and would be much drier but otherwise no different from non-overshooting ones. Such behaviour would accomplish dehydration. But it would not explain the water-vapour/ ozone correlations, would not provide drying at a level higher than about 150 hPa, and would not account for the downward motion at the Indonesian tropopause. It would also not be consistent with the behaviour of other convective boundary layers including those in the laboratory, although quantitative comparison between these analogs and the tropopause region is not straightforward.
On the other hand, if non-mixing clouds had significant variations in LNB, this would produce the same net result as mixing except that the mixing effect would in this case be confined to levels at and below the maximum LNB rather than above, thus failing to reach the tropopause. In this case the H2O-ozone tracer correlations would still be explained, but not the sinking and drying effects inferred above the highest LNB. This option is possible if the height of the sinking and hygropause are exaggerated in the observations or if these features are caused by something other than convective processing of air.
The existence of a robust seasonal cycle in CO2 and other constituents is sometimes cited as evidence against significant overshooting, but these observations do not rule out an important impact of overshoots on constituents with shorter lifetimes. In either overshooting scenario described above, the overwhelming majority of tracer observations in the lower stratosphere (especially if they are collected far from the most vigorous convection) will inevitably be consistent with the existence of slow lofting of air since, at and above the tropopause, that is what over 99% of the air is doing. Discrepancies will be easy to find only when two species each having concentrations that can change significantly in a few weeks (e.g. water vapour and ozone) are examined together, in which case the lag between convective mixing and lofting can emerge (Figure 1). On a cautionary note, however, there must be a limit as to how deeply in altitude significant mixing or convective detrainment height variation can spread, beyond which damping of the seasonal cycles of all trace constituents entering the lower stratosphere would be implied that is inconsistent with observations. It is not clear yet what that limit is, especially since the diabatic ascent rate itself varies seasonally, but the matter should be considered carefully in future modelling work.

An issue with both mechanisms is the ability to remove frozen condensate. This issue is less problematic with overshoots, since these will be full of large particles on which vapour can condense and be removed in short order, but the issue must still be considered before the mechanism can be accepted as important. Danielsen's (1982) argument about anvil destabilisation is untenable in light of current understanding of radiative transfer, but may not be necessary. The question has already been explored to some degree for in-situ dehydration, but in this case progress is limited by our knowledge of statistics of the large-scale conditions.
Why do we care about the details of which of these processes are dominant? This question goes back to the observed long-term trend in water vapour. For this trend to be explained, some variable needs to be found that can change the effectiveness of an important dehydration process. From our arguments here, it does not appear that changes in the residual circulation would be very effective in doing this, although changes in the horizontal circulation could conceivably have an impact. If convective dehydration is important, then (as noted by Sherwood and Dessler, 2001) changing convective meteorology or ice microphysics should be able to change the moisture entering the stratosphere without any changes in temperature or circulation. Conversely, if in-situ removal were responsible, then one would look for changes in wave intensity or available ice nuclei near the tropopause to explain long-term changes in the drying effectiveness.

Conclusion

To summarise, two basic dehydration hypotheses (dehydration within energetic convective updrafts or by dynamical lifting at larger scales) have been discussed. Many conclusions can be drawn that would apply to either mechanism. Both mechanisms would act mainly in limited regions where energetic deep convection is prevalent and the tropopause is coldest. Though such localisation is not supported by the line of reasoning first used to argue it (based on mean mixing ratios in the stratosphere), it is supported both by the links between either mechanism and convection, and by time-scale arguments that tend to couple horizontal variations in dehydration to those of temperature rather than the residual circulation.

Either mechanism would operate over an extended vertical layer rather than on a surface such as the tropopause. Thus strict passage of air across some infinitesimal surface, regardless of how that surface is defined, should not be envisaged as a physically relevant event at least regarding stratospheric dehydration. Air passing into the stratosphere must run a gantlet of one or more dehydration mechanisms for at least a few weeks before it has safely ascended into the stratospheric overworld. During this time the air should have a chance to sample nearly all longitudes multiple times, which means that even a dehydration mechanism that is confined to a limited horizontal region can operate on nearly all air parcels that enter the stratosphere. Temperatures outside dehydration regions are irrelevant to stratospheric moisture.

The focus here has been on recent work concerning the first, or "overshooting," mechanism. This mechanism fell out of favour after data collected during the 1990's seemed to support only the alternatives, but recent work shows that the mechanism is still viable. Variables concerning this mechanism are how high overshoots reach, and to what extent they mix with their surroundings. Currently these variables are not well constrained. The results of Sherwood and Dessler (2000) are consistent with either variable energy or variable mixing of overshoots, but in any case argue in favour of rapid dehydration before air begins a lofting process of substantial duration, as opposed to dehydration during or after the lofting. The model of Sherwood and Dessler (2001) assumes overshoots with limited variability of LNB but robust mixing with the near-tropopause environment, and obtains a reasonable result, but this does not prove that mixing of overshoots is important or occurs as high as in their model.

Convective dehydration undoubtedly plays a dominant role in dehydrating tropospheric air from near-surface values of 30,000 ppmv to values of 10-20 ppmv at the main outflow levels. The most important question really amounts to whether convection finishes the job (down to ~4 ppmv) by dehydration in unusually strong updrafts, whether the job is finished by a post-convective "in-situ" removal process, or whether there is a combination "one-two punch" involving both. Following up on that question, we must then ask, what controls the effectiveness of these processes?

References

Danielsen, E. F., 1982, A dehydration mechanism for the stratosphere, Geophys. Res. Lett., 9, 605-608.

Dessler, A. E., 1998, A reexamination of the "stratospheric fountain" hypothesis, Geophys. Res. Lett., 25, 4165-4168.

Holton, J. R., P. H. Haynes, M. E. McIntyre, A. R. Douglass, R. B. Rood, and Pfister, L., 1995, Stratosphere-troposphere exchange, Rev. Geophys., 33, 403-409.

Jensen, E. J., L. Pfister, A. S. Ackerman, O. B. Toon, and A. Tabazadeh, 2001, A conceptual model of the freeze-drying of air due to optically thin, laminar cirrus rising slowly across the tropical tropopause, J. Geophys. Res., In Press.

Johnston, H. S. and S. Solomon, 1979, Thunderstorms as a possible micrometeorological sink for stratospheric water vapor, J. Geophys. Res., 84, 3155-3158.

Newell, R. E. and Gould-Stewart, S., 1982, A stratospheric fountain?, J. Atmos. Sci., 38, 2789-2796.

Potter, B. E. and J. R. Holton, 1995, The role of monsoon convection in the dehydration of the lower tropical stratosphere, J. Atmos. Sci., 52, 1034-1050.

Rosenlof, K. H. et al., 2001, Stratospheric water vapor increases over the past half-century. Geophys. Res. Lett., In Press.

Sherwood, S. C., 2000, A stratospheric "drain" over the maritime continent, Geophys. Res. Lett., 27, 677-680.

Sherwood, S. C. and Dessler, A. E., 2000, On the control of stratospheric humidity, Geophys. Res. Lett., 27, 2513-2516.

Sherwood, S. C. and Dessler, A. E., 2001, A model for transport across the tropical tropopause, J. Atmos. Sci., 58, 765-779.

SPARC, 2000, Assessment of upper tropospheric and stratospheric water vapour, WMO/TD No. 1043, SPARC Report #2, December, 2000.

 

Back to SPARC Newsletter 17 Homepage