Variations in stratospheric water vapour alter the radiative heat budget and the ozone mixing ratios . The processes which control the stratospheric water vapour budget, however, are poorly quantified . These processes are temperature-controlled dehydration, convective activity, methane oxidation and isentropic transport.

Due to their physical and chemical properties, water isotopologues have the potential to answer the open questions concerning the origin of stratospheric water vapour. The small mass difference between H2O and HDO (16O; the hydrogen isotope deuterium D denotes 2H) leads to different vapour pressures and zero-point energies. This causes equilibrium and kinetic fractionation effects during phase changes and chemical reactions. Each process which controls the stratospheric water vapour budget can be associated with certain fractionation effects and therefore leaves a specific isotopic signature in the water vapour compound see. This isotopic fingerprint allows us to comprehend the history of stratospheric water vapour and, in this way, can assist to explain the trends and variations in its budget.

In addition to in situ and remote-sensing measurements, the comprehensive simulation of the physical and chemical processes of water isotopologues on the global scale is needed to gain an improved understanding of the basic structure of the water isotope ratios in the stratosphere. Model studies of water isotopologues in the upper-troposphere–lower-stratosphere (UTLS) include approaches from conceptional to one-dimensional and two-dimensional models. applied the general circulation model (GCM) GISS-E, in order to study stratospheric entry values of the isotope ratios of water vapour. However, this model has a comparatively low resolution in the stratosphere and accounting for methane oxidation is prescribed with a fixed production rate.

In Part 1 of this article , an extension of the global climate chemistry model (CCM) EMAC (ECHAM MESSy Atmospheric Chemistry; MESSy stands for “Modular Earth Submodel System”) was presented and evaluated. This extension, namely the H2OISO (H2O ISOtopologues) submodel, comprises an additional hydrological cycle, including the water isotopologues H218O and HDO and their physical fractionation effects, based on previous studies by, e.g., and . Besides a vertical resolution resolving the tropical tropopause layer (TTL) and simulating the stratospheric dynamics explicitly, this expanded model system also includes the computation of the methane isotopologue CH3D and its chemical contribution to HDO through oxidation. Results of an EMAC simulation showed good agreement in stratospheric HDO and δD(H2O) with measurements from several satellite instruments: δD(H2O)=[HDO]/[H2O]RVSMOW-1⋅1000. The Vienna Standard Mean Ocean Water VSMOW; HDO standard is RVSMOW=155.76×10-6 see. Moreover, the results revealed a stratospheric tape recorder which ranges between the pronounced signal of MIPAS (Michelson Interferometer for Passive Atmospheric Sounding) observations see and the missing upward propagation of the seasonal signal in the ACE-FTS (Atmospheric Chemistry Experiment Fourier transform spectrometer) retrieval see.

The results of these simulations are now further analysed, with the aim of identifying the processes which determine the patterns of the isotopic signatures in stratospheric water vapour. The connection between the water vapour budget and its isotope ratio in the tropical stratosphere over the two simulated decades is presented in Sect. . The influence of isotope effects during methane oxidation on the δD(H2O) tape recorder signal is investigated in Sect. . In Sect.  the origin of the Northern Hemisphere (NH) summer signal of the δD(H2O) tape recorder is examined. Its generation exclusively in the NH is shown to be connected with in-mixing of extratropical air and ice lofting in association with clouds in monsoon systems. Furthermore, a possible pathway of isotopically enriched water vapour from the NH troposphere into the tropical stratosphere is presented. These analyses also reveal a possible underestimation of ice overshooting in the applied convection scheme which can have a significant effect on δD(H2O) in the lower stratosphere. This study constitutes the first application of the isotopic composition of water vapour in order to explore the reasons for changes in the stratospheric water vapour budget with global atmosphere chemistry–climate models.

The MESSy submodel H2OISO, now part of the EMAC model , comprises an additional hydrological cycle, separate from the actual cycle and including tracers for the water isotopologues H216O, HDO and H218O in the three phases vapour, liquid and ice, respectively. These tracers are treated identically to the standard state variables for water in the regular hydrological cycle of EMAC, with the addition of the physical fractionation effects for the isotopologues during phase transitions. The representation of these effects follows the water isotopologue-enabled ECHAM (ECMWF Hamburg) model see. Equilibrium and kinetic fractionation during the evaporation of water from oceans is described by the bulk formula of . Due to the limitations of the applied land surface scheme, we neglect any isotope fractionation from land surfaces for details, see. The implementation of the cloud and convection parameterisations (CLOUD and CONVECT) in EMAC follows the study of . For condensation within clouds and for the evaporation of cloud water, a closed system is assumed. An open system is used for the deposition of water vapour to ice. Due to the low diffusivities of the isotopologues in the ice phase, no exchange happens between ice and vapour. During the melting of ice and the freezing of water, as well as for the sedimentation of ice, autoconversion, accretion and aggregation, no fractionation is assumed. The representation of the fractionation during the reevaporation of raindrops follows the study by , who assume an isotopical equilibration of 45 % for large drops from convective rain and 95 % for small drops falling from stratiform clouds. Supplementary to this, an explicit accounting function for the contribution of CH3D oxidation to HDO, including a parameterisation of the deuterium storage in molecular hydrogen, has been developed in order to achieve realistic HDO mixing ratios and δD(H2O) values in the stratosphere. In Part 1 of this article , the model system and the implementation of HDO throughout the hydrological cycle, including its chemical representation, are presented in detail.

An EMAC simulation in the T42L90MA (∼ 2.8∘×2.8∘, 90 layers in the vertical up to 80 km (0.01 hPa), explicit middle atmospheric dynamics) resolution was carried out. The simulation was performed with specified dynamics (i.e. “nudged” towards ERA-Interim reanalysis data ECMWF; up to 1 hPa). The “Tiedtke–Nordeng” convection scheme was applied to the simulation. After starting from steady-state initial conditions in 1982, the simulation was evaluated during the 21 years from 1990 to the end of 2010. A detailed description of the simulation set-up and a description of the applied MESSy submodels are presented in Part 1 of this article . An evaluation of the model's hydrological cycle itself can be found in , who assess the EMAC base model ECHAM5. state that the modifications introduced by the MESSy system, as well as by the application of the T42L90MA resolution and nudging, produce a hydrological cycle similar to the results in and consistent with observations. An extensive evaluation of the isotopic composition of water vapour and of its chemical precursor CH3D in this EMAC simulation in the troposphere and the stratosphere is presented in Part 1 of the article . Overall, a reasonable representation of stratospheric HDO is reached, although with some systematic, but explainable, discrepancies.

Temporal variations in stratospheric water vapour during the last decades have been observed by various instruments. The reasons for these variations are much discussed see, e.g.,. Before analysing these changes with the EMAC model using the newly implemented HDO, it has to be ensured that the EMAC simulation features the main characteristics of the changes in stratospheric water vapour from 1990 to the end of 2010. Therefore, the equatorial water vapour mixing ratios at 30 km altitude of the simulation are compared with a combined HALOE (HALogen Occultation Experiment) and MIPAS data set in Fig. . A detailed description of the combination of the satellite data time series is given in the Supplement. A 2-year running mean was calculated for both time series in order to make the trends more visible by eliminating the signal of the quasi-biennial oscillation (QBO).

The combined HALOE–MIPAS data show an increase in stratospheric water vapour in the first half of the 1990s and a plateau hereafter until the year 2000. The water vapour mixing ratio drops by around 0.3 µmol mol-1 between 2000 and 2002 and stays at this lower level until the middle of the first decade of the 21st century. Hereafter, a slow increase can be observed until the end of the time series in 2010. This behaviour of stratospheric water vapour during the previous decades has also been reported and discussed, e.g. by , who analysed a combined HALOE and MLS (microwave limb sounder) data set, and is strongly connected to tropopause temperatures. The EMAC simulation generally reproduces these variations, although with a constant offset and a few differences. The general dry bias in EMAC has already been discussed by and in Part 1 of this article . Its main reasons are the slightly too cold hygropause in the nudging data see, e.g., and the coarse horizontal resolution of the model. In contrast to the satellite observations, in the EMAC simulation the drop around the year 2001 is preceded by an increase in water vapour. Moreover, the level of the water vapour mixing ratio after the drop does not fall below the level of the early 1990s. The Pearson's correlation coefficient between the observed and simulated time series is R2=0.50.

In order to estimate the correlation between the changes of water vapour and its isotopic composition, the monthly anomalies with regard to the 21-year monthly averages of the tropical water vapour mixing ratios and δD(H2O) are shown in Fig.  for the 21 years of the EMAC simulation at 18 km and at 30 km altitude. Again, the data were processed with a 2-year running mean, in order to obtain a better visibility of the trends. The anomaly of δD(H2O) is divided by 30 for better comparability.

At 18 km altitude, the Pearson's correlation coefficient between the two time series is R2=0.57, and this correlation decreases to R2=0.28 at 30 km. At 18 km altitude, both quantities are dominated by troposphere–stratosphere exchange processes. At 30 km altitude, the chemical effects, induced by CH4 oxidation for H2O and the different lifetimes of CH4 and CH3D for δD(H2O), become important. An interdependence of the two quantities can be observed at both altitudes, although, during certain periods, the development of the two time series is anticorrelated. The drop around the year 2001 can be seen in water vapour and in δD(H2O) at both altitudes. At 18 km, the more pronounced feature in δD(H2O), however, is the steep increase before the drop. The amplitude of this increase in δD(H2O) exceeds the amplitude of the drop almost by a factor of 2. Even though most of the variations in the two quantities are in phase, the signs of the anomalies are sometimes inverted. At 18 km altitude, δD(H2O) is generally at a lower level at the end of the 1990s compared to the early 2000s, after the drop. The short-term changes, in particular, seem to be different between the two quantities. This suggests that the processes that control stratospheric δD(H2O) are related but not equal to those that control the stratospheric water vapour budget. The tropopause temperatures, methane oxidation, convective activity and other processes determining water vapour in the stratosphere thus affect stratospheric H2O and δD(H2O) with different strengths. Knowledge of this behaviour can therefore help to connect the origin of certain variations and trends to changes in specific processes. The next sections thus aim to reveal the influence of individual processes on stratospheric δD(H2O), with a special focus on the tape recorder, since the strength of this phenomenon largely determines the intrusion of water vapour into the stratosphere.

In order to analyse the impact of the contribution of CH4 and CH3D oxidation on the δD(H2O) tape recorder signal, an additional EMAC simulation was conducted. The only difference in this simulation is a modified chemical tendency for HDO. The concept for this sensitivity simulation is an artificial deactivation of the chemical fractionation effects. In other words, δD(H2O) is not influenced by chemical isotope effects; CH3D oxidation alters always HDO in relation to CH4 oxidation, as if there was no isotope fractionation. A detailed description of this modification is given in the Supplement.

For the analysis of the impact of isotope effects during methane oxidation on the δD(H2O) tape recorder signal, the simulation with the modified HDO tendency is compared to the simulation with regular methane isotope chemistry. The set-up is the same for both simulations. Figure  shows the tropical tape recorder signals from 2004 to 2009 for the two simulations from 15 to 30 km.

Between 15 and 25 km, the δD(H2O) values are similar in both panels. In the tropical tropopause layer and the lower stratosphere, δD(H2O) is only weakly affected by methane oxidation. From 25 km upwards, increasingly higher δD(H2O) values can be observed in the simulation with regular methane isotope chemistry (upper panel). The effect of the chemistry on δD(H2O) increases with altitude in the stratosphere. This can be observed for the increased δD(H2O) values, which emerge during the NH summer, as well as for the low δD(H2O) values from the boreal-winter signal. The tape recorder signal in the simulation with the modified methane isotope chemistry (lower panel) extends further up. It is still present, although weak, at the top of this panel at around 30 km altitude. In the upper panel the δD(H2O) tape recorder signal above 25 km becomes increasingly overshadowed by high δD(H2O) values, which are generated by the different lifetimes of CH4 and CH3D, i.e. chemical isotope effects. The upward-propagating signatures fade out or, more specifically, mix with the high δD(H2O) values. These high δD(H2O) values show variations with a phase of around 2 years, which can be associated with the QBO.

For a better quantification of the differences between the two tape recorder signals, Fig.  shows the annually averaged difference in the δD(H2O) maximum and minimum as a function of altitude for the time period of Fig. . The black line denotes the simulation with, and the red line the simulation without, the methane effect.

The tape recorder amplitudes are equal below 20 km. As expected, further above, the amplitude of the simulation with chemistry effect on δD(H2O) decreases more quickly with altitude than the amplitude of the simulation without this effect. The high δD(H2O) values from the NH summer signal are not affected as strongly by methane oxidation as the low values from the NH winter signal are. To explain this, constant temperatures, and hence fractionation factors, and a constant background δD(CH4) are assumed, which is reasonable here. The isotope ratios of isotopically different reservoirs show different sensitivities to the addition of a compound with a certain isotope ratio. This means that the smaller the differences between the δ values are, the smaller is the modification. Since the high δD(H2O) values from the NH summer signal are closer to the δD(CH4) values (δD(CH4) is also based on VSMOW), which are around -50‰ here, compared to the low δD(H2O) values from NH winter, the summer signal is altered less. Additionally, the water vapour mixing ratios are also different here. The δD(H2O) values of the low water vapour mixing ratios from the NH winter signal are therefore again affected more strongly by the addition of (a similar amount of) isotopically enriched water vapour from methane oxidation. This leads to the conclusion that the production of H2O and HDO by the oxidation of CH4 and CH3D reduces the amplitude of the δD(H2O) tape recorder and overshadows the upward propagation of the signal.

Between 24 and 28 km, the amplitude of the δD(H2O) variations in the simulation with chemistry effect on δD(H2O) are similar, and above 28 km the amplitude in the simulation with the chemistry effect exceeds the amplitude of the simulation without. This, however, is not due to the tape recorder effect anymore but is caused by the QBO. The QBO also has an effect on the stratospheric water vapour budget and on the water vapour tape recorder see. As stated above, the cycle of the QBO can be seen in the high δD(H2O) values between 25 and 30 km in Fig. . Temperature and hence chemical fractionation factor variations and also dynamical differences between the QBO phases which mix in more or less strongly enriched water vapour lead to this cycle and hence to this increase in amplitude.

Both the water vapour mixing ratio and δD(H2O) exhibit enhanced values in the lower stratosphere during JJA (June, July, August). The underlying processes for this, however, may differ in some ways for the two quantities. In order to demonstrate this and to analyse the origin of the tape recorder signal, the water vapour mixing ratios and δD(H2O) in the UTLS for JJA are shown in Fig. .

Differences in the distribution of the enhanced values can be observed when comparing the two panels. In the left panel, enhanced H2O mixing ratios can be seen within almost the entire TTL, although decreasing with altitude and towards the southern latitudes. At the northern edge of the TTL, the high H2O mixing ratios exceed the tropopause and penetrate into the stratosphere. Some water vapour, however, also intrudes into the stratosphere in the central and the southern TTL. Isotopically enriched water vapour (see right panel) exclusively enters the stratosphere at the northern edge of the TTL. δD(H2O) values of above -650‰ can be observed, crossing the tropopause and entering the tropical pipe here. Note that the enhancement of δD(H2O) between 17 and 18 km is an artefact caused by the seasonal averaging. The signal originates from the northern edge of the TTL and remains in the tropical pipe during summer while being mixed with surrounding air masses comparatively quickly between 16 and 17 km. In the Supplement, zonal δD(H2O) across all latitudes is also presented for the other seasons (DJF – December, January, February; SON – September, October, November; MAM – March, April, May) from 10 to 30 km altitude. There, the high δD(H2O) values in the tropical pipe can still be seen during SON and DJF. In the central and southern parts of the TTL, the water vapour is isotopically strongly depleted, exhibiting values below -700‰. Low δD(H2O) values can be observed down to 14 km altitude in the central and southern TTL, while relatively high water vapour mixing ratios extend up to almost 16 km altitude in this region.

This shows that, in contrast to H2O, the enhanced isotope ratios in the tropical lower stratosphere during JJA originate exclusively from the NH. However, it is not clear if it originates from the intrusion of tropospheric water vapour into the stratosphere or from in-mixing of old stratospheric air from the extratropics. During DJF a similar signal of isotopically enriched water vapour at the edge of the TTL at around 40∘ S can be observed (see Supplement). In contrast to the situation in JJA, here this signal is at considerably lower altitudes and does not penetrate into the stratosphere.

In the lower stratosphere, air experiences rapid horizontal transport between the tropics and the midlatitudes above the subtropical jets . The region between the 380 and the 400 K isentrope is therefore crucial for the properties of stratospheric air. To provide an insight into the horizontal dynamics of this region, the average of δD(H2O) between the 380 and the 400 K isentrope is shown in a latitude–longitude representation for JJA in Fig. . Again the other seasons are presented in the Supplement.

In general, the image features a pattern with low δD(H2O) in the tropics and increasing values with higher latitudes. In the Northern Hemisphere, patterns can be observed which are associated with the Asian summer monsoon (ASM) and the North American monsoon (NAM). High δD(H2O) values can be seen over the entire North American continent. Over southern Asia, in contrast, very low values are dominant. Around this isotopically depleted centre of the ASM anticyclone, the water vapour is isotopically enriched. Over the West Pacific, at the outflow of the ASM anticyclone, the wind vectors indicate a considerable southward component, which drags isotopically enriched air from the extratropics towards the tropics and then westwards. This air may originate from in-mixing of old stratospheric air from the extratropics. show a similar pattern for ozone which, according to , can also result in a stratospheric-tape-recorder-like signal. However, state that this process is largely dependent on the species itself, or, more specifically, on its meridional gradient. The study shows that in-mixing plays a role in the annual ozone variations in the tropics but not for carbon monoxide, nitrous oxide or water vapour. Another possible explanation for these patterns is the intrusion of tropospheric air into the stratosphere. focus especially on slow ascent and dehydration through in situ cirrus formation, which can generate the δD(H2O) tape recorder signal, and point out the importance of ice overshooting convection on the pattern. In the remainder of this section we will examine some of these mechanisms in order to obtain a better understanding of their relative importance for the δD(H2O) tape recorder.

As a first application of the new H2OISO submodel within the EMAC model, stratospheric water vapour isotope ratios were investigated. The time series of water vapour in the nudged EMAC simulation analysed here reproduces the major variations in the recent decades. The time series of δD(H2O) shows similarities with H2O but differs mainly regarding short-term changes. This suggests that the processes controlling these two quantities coincide, but their effect on the respective value is of different quantity.

The impact of methane oxidation on the stratospheric δD(H2O) tape recorder signal was investigated by comparing the evaluated EMAC simulation with an additional simulation with a suppressed chemical isotope effect of methane oxidation on δD(H2O). Methane oxidation mainly affects water vapour and its isotopic signature above 25 km, where the δD(H2O) tape recorder signal fades out faster through this chemical effect. Additionally, the amplitude of the δD(H2O) tape recorder is reduced because methane oxidation influences the low δD(H2O) values and the low water vapour mixing ratios much more strongly than the higher ones. This result is not surprising, but it reveals the impact of the isotope chemistry on the tape recorder. also applied a correction for the methane effect on δD(H2O) to the ACE-FTS satellite retrieval. This led to the removal of the increase in δD(H2O) with altitude in the stratosphere as well. Moreover, it generated enhanced isotope ratios in the lower stratosphere during JJA and SON, compared to the simulation without the methane correction. However, the δD(H2O) tape recorder is still not clearly visible in the ACE-FTS satellite retrieval.

The determining processes for the generation of enhanced δD(H2O) during JJA in the tropical lower stratosphere were shown to take place exclusively in the NH. This is in contrast to water vapour itself, which is also influenced by direct transport through the TTL see, e.g.,. showed that for some species a tape-recorder-like signal can be generated through the in-mixing of old stratospheric air alone. However, and , for example, focus mostly on the slow ascent of tropospheric water vapour and convective ice overshooting when evaluating the reasons for the generation of the δD(H2O) tape recorder. In order to investigate the role of the individual processes, we first assessed the H2O to δD(H2O) ratio in the different hemispheres and seasons. A relationship between H2O and δD(H2O) that is much more spread out during JJA in the NH (compared to the SH and to DJF) suggested that both the intrusion of tropospheric water vapour and the in-mixing of old stratospheric air influence δD(H2O) here. In particular, due to extreme interhemispheric differences in the ice water content and its isotopic signature during the different seasons, the lofting of ice crystals is assumed to enrich water vapour in the NH upper troposphere during JJA. Confirmation is achieved through the results of sensitivity simulations with the modified HDO tendency for large-scale clouds and for convection. The averaged annual difference between the maximum and the minimum of δD(H2O) shows a clear reduction in the UTLS up to 23 km for the simulation without large-scale clouds affecting δD(H2O) and an enhancement for the simulation without convection affecting HDO. The isotopic enrichment during JJA is therefore not generated by convective events, which, in contrast, deplete the water vapour in the simulation, but by large-scale clouds in association with monsoon systems. However, this shows that the δD(H2O) tape recorder is strongly affected by tropospheric effects through clouds and convection and not only by the in-mixing of extratropical air masses.

Augmented ice lofting, especially in the ASM over the Himalayas, has been shown to isotopically strongly enrich the water vapour in the upper troposphere. In the outflow of the monsoonal anticyclones, this isotopically enriched water vapour is transported into the stratosphere on isentropic surfaces. Numerous studies, e.g. have shown that this sideways transport into the tropics in particular from the ASM also contributes significantly to the stratospheric water vapour budget. By analysing MIPAS satellite data, suggest that a slow dehydration through cirrus cloud formation plays a key role for the δD(H2O) tape recorder. However, the separation of this particular process within the individual parts of the model, i.e. large-scale and convective clouds, is not easily resolvable. In the model, convective clouds isotopically deplete stratospheric water vapour. However, the relative importance of this individual process for the annual signal has to be further investigated. present a somewhat different pattern of δD(H2O) in the UTLS by analysing ACE-FTS satellite data. In this retrieval, enriched δD(H2O) at 16.5 km altitude can mostly be found over America, and the patch of high δD(H2O) associated with the ASM, as seen in the EMAC data, is considerably weaker. It is still a matter of debate as to whether convective ice overshooting has a significant effect on the stratospheric water vapour budget see, e.g.,. According to and , however, it has a substantial effect on the stratospheric δD(H2O) signature. This ice overshooting occurs mostly in the inner tropics and has the potential to isotopically enrich the tropical lower stratosphere. However, the NAM is also associated with strong convective ice overshooting see, e.g.,. The direct intrusion of ice crystals into the stratosphere is known to be represented rather sparsely by the convection scheme applied here and taken from , and it has been shown to not affect stratospheric δD(H2O) in this simulation. Therefore, this discrepancy between model and observations may be due to the underrepresentation of convective ice overshooting in the applied convection scheme. The NAM region as well as the inner tropics could show considerably higher δD(ice) values in the UTLS. Furthermore, this is possibly also the cause for the too low δD(H2O) values in the lower tropical stratosphere in EMAC compared to satellite observations during NH summer, as shown in Part 1 of this article . A more detailed evaluation of this effect can be conducted through the implementation of water isotopologues into other convection schemes of EMAC. Future sensitivity studies can then also resolve the robustness of the patterns discovered here and possibly explain the differences between model results and observations more precisely.

The temporal variations in stratospheric δD(H2O) reveal connections to those in water vapour, but they show differences regarding the amplitudes. This provides additional information about the underlying processes of the changes and therefore can help to gain a better understanding of the reasons for the trends and variations in the stratospheric water vapour budget. First, however, this requires an understanding and quantification of the influence of the individual processes that are responsible for the patterns of δD(H2O) in the stratosphere. Isotope fractionation effects during methane oxidation blur the δD(H2O) tape recorder signal by damping its amplitude and overshadowing it at higher altitudes. This explains the weaker tape recorder signal in δD(H2O) compared to those in H2O and HDO. The in-mixing of old stratospheric water vapour with high isotope ratios from the extratropics alone does not suffice to describe the δD(H2O) tape recorder. Instead, the influence of the intrusion of tropospheric water vapour through clouds and convection contribute significantly to this pattern. The isotopic enrichment of upper-tropospheric water vapour through ice lofting in association with monsoon systems and further transport of these air masses into the tropical stratosphere in the outflow account for this influence. However, a quantification of the contributions of the respective processes and also of the individual monsoon systems is yet to be established and first requires further analyses of the discrepancies between the model results and satellite retrievals. These discrepancies indicate possible insufficiencies in the model, i.e. the underrepresentation of overshooting convection. This study has laid the foundations for further analyses in order to determine the connection between the patterns and changes in stratospheric H2O and δD(H2O). The additional information provided by the water isotope ratio is of significant support in unravelling the factors which contribute to trends and variations in the stratospheric water vapour budget.