From considerations of water vapor and helium distributions in the stratosphere, Brewer (1949) described a global circulation in which air enters the stratosphere at the equator where it is dried by condensation at low tropopause temperatures, travels poleward in the stratosphere, and sinks back into the troposphere at higher latitudes. With some added specifics, this general scheme is still accepted today, however, the mechanisms by which this circulation takes place have been an active area of debate. One of the major issues to be addressed is the fact that the stratosphere is so dry. Brewer (1949) held that air circulates by a slow mean motion into the stratosphere at the equator . As this air passed through the tropical tropopause, with it s very cold temperatures, a great deal of the water vapor in the air would condense and fall out, leaving a much lower water vapor mixing ratio. Brewer (1949) assumed that this was the only source for water vapor in the stratosphere and therefore explained the dryness of this area. There are 2 problems with this assumption: Weickmann et al. (1975) determined that the oxidation of CH4 provided another source of water vapor in the stratosphere and Kley et al. (1979) , using in situ techniques measured a minimum stratospheric water vapor content of 2.6 parts per million by volume (ppmv) over Brazil, which is too dry to be explained by Brewer s (1949) cold trap scenario. Local tropopause temperatures were not low enough to desiccate the air to this extent. This led to new attempts, to describe the mechanism responsible for the transport of air across the tropical tropopause, by researchers such as Newell and Gould-Stewart (1981) and Danielson (1982). The main competing theories will be discussed in sections 2 and 3. The controversy over the proposed mechanisms led to the development of The Stratosphere-Troposphere Exchange Project (STEP) (Russell and Pfister, 1993), conducted in January and February 1987in Darwin, Australia. The objective of the STEP Tropical Experiment was to use in-situ measurements from high flying aircraft to improve understanding of across tropopause exchange and dehydration processes in the tropics (Russell and Pfister, 1993) Some results from STEP will be discussed in section 4.
Brewer (1949) suggested a mechanism by which air is transported to the stratosphere across the cold tropical tropopause by a slow mean motion where it is freeze-dried. Kley et al. (1979) measured stratospheric water vapor mixing ratios of just under 3 ppmv over Brazil. One can calculate that in order to reach this concentration, the air must have been cooled to a temperature of -84oC , assuming saturation with respect to ice. Kley et al. (1979) determined that the local temperature of the tropical tropopause was only -80oC, leading to the suggestion that this air must have entered the stratosphere at some other location.
Newell and Gould-Stewart (1981) analyzed worldwide 100mb temperatures (as an approximation of the tropopause temperature) over a period of seven years, in order to try to identify the region of cross- tropopause transport. They identify areas where tropopause temperatures are sufficiently cold as to dehydrate the air. Using a criterion of a tropopause temperature of <-82.4oC, which corresponds to an ice saturation value of 3.5 ppmv. This is a slightly higher mixing ratio than the measurements made by Kley et al. (1979) and may therefore result in an overestimation of the entry area. Figure 1 shows a map from Newell and Gould-Stewart (1981) identifying the region where temperatures <-82.4oC were observed at 100 mb, for the month of January. This figure shows that the largest region meeting the temperature criterion was in the Western Pacific in the December - January season. This suggests that rather than the zonally constant transport at the equator proposed by Brewer (1949) there must be a favored region of cross- tropopause transport located over the tropical Western Pacific. Newell and Gould-Stewart (1981) termed this restricted region of transport into the stratosphere, the stratospheric fountain.
Further evidence supporting the idea of a restricted region of cross-tropopause transport are shown in Figure 2 (Danielson, 1992). This figure compares temperature profiles of Darwin, Australia (in the stratospheric fountain region) and Panama. The tropopause temperatures at Darwin are sufficiently cold to dehydrate the air, whereas at Panama, the minimum mixing ratio would only reach 8.2 ppmv (again assuming ice saturation).
There is a general agreement on where the bulk of the tropospheric air enters the stratosphere, but the mechanism by which this transport takes place is still an area of much debate. In this section, the two main proposals will be discussed.
Newell and Gould-Stewart (1981) define the stratospheric fountain area and then go on to describe the transport through this region by mean mass flux calculations. They take previous estimates of zonally averaged upwelling and confine the corresponding mass flux to the fountain region. These calculations yield an average vertical motion of 0.5x10-2 m/s through the entire region. They suggest that this mean large-scale ascent is responsible for the transport and subsequent dehydration of air entering the stratosphere.
The tropical Western Pacific is an area of very strong convection (Ramage, 1968). Danielson (1982) suggests that the strong convection this area must play a role in the cross-tropopause transport and therefore the dehydration mechanism. Consider a parcel of air rising moist adiabatically in a cumulonimbus turret. The turret will rise with positive buoyancy until it reaches the tropopause. At some height just above the tropopause, this parcel will reach its equilibrium height. However, because of its kinetic energy, the parcel will overshoot this equilibrium level. As the turret decelerates it will entrain warm, dry stratospheric air. This mixing with stratospheric air will alter the equilibrium level of the parcel to be in the lower stratosphere. The turret collapses and spreads horizontally, forming a large cirrus anvil composed of ice crystals. The anvil maintains static instability and buoyancy generated turbulence through radiative warming at anvil base due to upward radiation emitted by the warm surface, coupled with radiative cooling at cloud top. This instability will sustain mixing through the anvil layer resulting in an adiabatic temperature profile. Figure 3 shows a representation of this anvil formation.
The result of this radiative heating profile will be an upward turbulent flux of heat and water vapor. As water vapor and ice crystals are transported upward, cooling will produce supersaturation and thus, the larger ice crystals will grow until they are large enough to fall out. As larger crystals fall out of the volume, the smaller crystals will then have a chance to grow, until they are large enough to fall out. Thus, the downward flux of ice crystals act to remove water vapor from the air being transported upward in the anvil region. The result is that the air just above the anvil layer will be dried to the ice saturation values of the local temperatures. This could in principle continue until observed stratospheric water vapor contents are reached.
The Stratosphere -Troposphere Exchange Project Tropical Experiment (STEP Tropical) looked to address two main goals: (1) investigate the mechanisms and rates by which transports occur across the tropopause and (2) explain the dryness of the stratosphere (Russell and Pfister, 1993). In January-February 1987 from Darwin, Australia, in-situ measurements were made from the Environmental Research 2 aircraft with a group of advanced instruments. Measurements were made of temperature, pressure, winds, ozone, water vapor (Kelly et al., 1993), total water, radon (Kritz et al., 1993), condensation nuclei, cloud particles (Knollenberg et al., 1993), aerosols and radiation. Of particular importance was the high frequency of measurements of each quantity, necessary for the resolution of small-scale structures. Never before had such high precision measurements been made in this region. By taking these high resolution measurements the feasibility of the dehydration mechanisms discussed in the previous section could be investigated.
Figure 4, taken from Kelly et al. (1993), shows the mean water vapor profile of 10 aircraft flights over Darwin as a function of potential temperature. The most important feature to notice, is that just above cloud top (approx. q = 365K) there is a minimum water vapor mixing ratio between 1.8 to 3.2 ppmv. The fact that such low mixing ratios are measured supports the theory of the importance of the maritime continent region as a possible source of the dry stratospheric air measured over Panama. Figure 4 also shows the mean water vapor from 7 flights taken over Panama in September 1980. Note that the minimum in the Panama sounding is just under 4 ppmv and is achieved well above the tropopause , while just above the tropopause a mixing ratio of 6 ppmv is observed suggesting that in this region, penetrating cumulonimbus may have a hydrating effect rather than a dehydrating effect. This is exactly what Kley et al. (1982) concluded earlier from the Panama measurements. A comparison of the Darwin and Panama mean water vapor profiles suggests that air being transported across the tropopause in Darwin is being dried to it s minimum value locally, whereas for Panama, this is not the case (Kelly et al., 1993).
In order to hypothesize on the mechanism by which this cross-tropopause transport takes place, figure 5, also from Kelly et al. (1993) shows the relative humidity with respect to ice as a function of potential temperature for eight of the Darwin flights. From this plot, we get an idea of the location of clouds, or more specifically of where the anvil ice crystals from Danielson s dehydration mechanism are. The thickest clouds are observed just below 360K, which corresponds to being just below the average water vapor minimum. This is therefore consistent with the Danielson hypothesis.
Several properties of 226Ra (radon) make it an appropriate tracer of near-surface tropospheric air: (1) It s only source is at the earth s surface, (2) it has a relatively short half-life (3.8 days) and (3) because it is relatively insoluble in water, it is not noticeably removed by precipitation scavenging. So, not only does radon act to trace the recent source of the air, but due to its radioactive decay, its presence can also alert us to the time scales of transport. For example, if we measure quantities of radon in stratospheric air, we may conclude that this air must have been transported from near the surface, on a time scale of a couple days. Therefore, if the dry air in the stratosphere also contains radon, this may support the rapid transport of air across the tropopause by localized convection (Danielson, 1982) as opposed to the large scale mean ascent proposed by Newell and Gould-Stewart (1981).
Figure 6 (Kritz et al., 1993) shows the vertical profiles of temperature, total water, and ozone for flight 11 during STEP Tropical. This flight penetrated the cirrus shield of Cyclone Damien. Also noted in Figure 6 is the extent to which radon-rich air was detected. The minimum water vapor values are found at about a height of 17 km, while the layer of radon-rich air extends up to 18 km. These measurements clearly indicate the presence of appreciable radon levels in the dry air just above the cirrus anvil layer. This strongly suggests the idea that this air was recently transported from near the surface and therefore supports Danielson s hypothesis and the rapid transport by convective elements .
Danielson s dehydration scheme depends on the growth of ice crystals , as a sink of water vapor, which then become large enough to fall to lower levels without appreciable evaporation. During STEP Tropical Knollenberg et al. (1993) used imaging and light scattering instruments to measure ice crystal size distributions in tropical anvils. Figure 7 shows measurements of ice crystal mass and number density as a function of diameter for flight 11 during STEP (Knollenberg et al., 1993). These measurements show high number densities of the smallest crystals, and no crystals detected with diameters greater than 100 mm. Also notice that the cumulative mass is still dominated by the few larger crystals. It seems that the presence of many small ice crystals would be consistent with Danielson s dehydration mechanism, the larger crystals having fallen to lower parts of the anvil. However, the important question to address is whether the ice water mass is likely to be transferred to lower levels prior to sublimation, i.e. prior to hydration of the air. To address this question, Knollenberg et al. (1993) used the size distribution measurement from figure 7 along with a Stokes terminal velocity calculation to compute the average mass flux and sedimentation rate for the flight level. Figure 8 shows the results of this calculation. The sedimentation rate for the crystals observed in this flight were between 5 and 10 cm/s. Therefore, under typical anvil conditions, the larger ice crystals will be able to sediment to lower levels without sublimation (Knollenberg et al., 1993) and therefore the falling ice crystals will remove water from the region, and cannot contribute to the water vapor in the air.
Since these growing ice crystals are in an environment of ice saturation, the only way to add any water vapor to the air will be to raise the temperature of the parcel, through either adiabatic ascent , diabatic heating or mixing with warmer air. Adiabatic ascent is unlikely since the parcels are in the relatively stable stratosphere. Radiative heating rates depend on the size of the ice crystals in the anvil. Figure 9 (Knollenberg et al., 1993) shows the spectral variation of the absorption efficiency factor for ice spheres. Note that for smaller ice crystals, there is a strong absorption peak at about 3mm (near IR). Since these small crystals are also poor longwave emitters, we therefore expect to see heating of these crystals. The larger crystals are close to being blackbodies, and therefore we expect them to cool. This pattern of heating and cooling as a function of ice crystal size will tend to concentrate the ice crystals in the 5-100 mm size range, and therefore the sedimentation rates shown in figure 8 can be thought of as representative for anvil ice content in this region. Since there seems to be no reliable source of heating the parcels, we consider that the ice water content is removed from the anvil region through sedimentation, leaving behind air that is at ice saturation at the coldest cloud top temperatures (Knollenberg et al., 1993).
The question to be addressed here is, given all of these measurements from STEP Tropical, who is right about the cross-tropopause transport and dehydration mechanism, Newell and Gould-Stewart (1981) or Danielson (1982)? The answer is most likely both, with some reservations. When considering Newell and Gould-Stewart s (1981) hypothesis, Danielson (1982) points out that their criterion satisfies the necessary condition of cold temperature but does not assure a sufficient condition that the slow mean ascents (they propose as the cross-tropopause transport mechanism) are physically realizable. The area over Micronesia, identified as the stratospheric fountain region by Newell and Gould-Stewart (1981), is an area of the coldest tropical tropopause temperatures. It is also a region of maximum convection. In attributing the cross-tropopasue transports in this region to a slow mean ascent seems to neglect the predominant upward transport providers in this region. Danielson (1982), saw this discrepancy and suggested an alternative transport mechanism which could account for role of convection.
During STEP Tropical, an unprecedented set of measurements were made in the Micronesia region with the specific goal of addressing the dryness of the stratosphere. Figures 10a and 10b (Danielson, 1993) show profiles of temperature, water vapor and total water, taken in a stratospheric anvil during flight 7 of STEP Tropical. In figure 10b, the shaded region indicates where the total water is greater than the water vapor, indicating the presence of ice crystals. We see, a nearly adiabatic temperature profile in the cirrus cloud., with the coldest temperatures at cloud top. This temperature is about -89oC, and is consistent with the low water vapor mixing ratios that are observed here (<2.5ppmv). If we consider the sedimentation process to be taking place here, we could expect the ice water to have a downward flux in the cloud region. Leaving behind the air with low water vapor mixing ratios, with no source to rehydrate. Thus, Danielson s (1982) hypothesis seems to be consistent with these observations. However, if we look at figure 11 (Danielson , 1993), showing temperature profiles in the cirrus cloud shield of Cyclone Damien, from flight 11 during STEP Tropical, we see a slightly different mechanism happening. Here, ozone is used as a tracer of stratospheric air, and shows that tropospheric air extends up to the temperature inversion at about 90 mb, at the top of the cirrus cloud shield. This suggests that rather than penetrating the tropopause, that the convective cloud updrafts have lifted the tropopause (Danielson, 1993). Danielson (1993) suggests that this process comes closest to Newell and Gould-Stewart s (1981) slow mean ascent, although it is still on a much smaller scale. Figure 12 shows a schematic cross section of this process occurring in what Danielson (1993) refers to as an umbrella shaped cirrus cloud shield.
The conclusion is that both convective scale transport by overshooting cumulonimbus turrets and larger scale uplift associated with tropical cyclones and the subsequent formation of the cirrus cloud shield are the dominant mechanisms in transporting air to the stratosphere, and the subsequent dehydration of that air. It is however also necessary to note the importance of the interplay between the dynamical and microphysical processes in explaining the dehydration mechanism.
