Notwithstanding the low abundance of water vapour in the stratosphere, with mixing ratios of the order of parts per million above the cold tropical tropopause , stratospheric water vapour variations play a crucial role for climate variability. Variations in stratospheric water vapour have been shown to account for up to 30 % of decadal variations in the radiative budget . In a future climate with rising greenhouse gas levels, climate models predict significant stratospheric water vapour increases, strongest in the lowermost stratosphere directly above the extratropical tropopause . The processes causing this increase are under debate . Recently, indications for an increase in stratospheric water vapour have been found from satellite observations , which could be partly related to decadal variability . As stratospheric water vapour acts as an impactful greenhouse gas, an increase in its abundance will cause a positive climate feedback, ranging between about 0.1–0.3 W m-2K-1 , and substantial effects on atmospheric temperatures and dynamics , with the largest fraction of these effects caused by water vapour increases in the lowermost stratosphere.

However, it is particularly the lowermost stratosphere where current global climate models show the largest differences between simulations and observations in stratospheric water vapour . These moist biases in the simulations appear to be related to numerical diffusion in model transport schemes; they can be significantly reduced when employing less diffusive Lagrangian schemes . showed that lower-stratospheric water vapour, in particular the model moist bias in that region, exerts a first-order effect on the zonal mean atmospheric circulation, with water vapour increases causing a strengthening of the stratospheric circulation, upward and poleward shifts of the subtropical jets, and a poleward shift of the tropospheric eddy-driven jet. Hence, improving the representation of stratospheric water vapour in climate models appears to be essential for the simulation of zonal mean circulation in climate models and future projections.

A particularly strong moisture source for the lower stratosphere is the anticyclonic Asian monsoon circulation during boreal summer , which causes the ascending moist air masses to partly bypass the coldest regions of the tropical tropopause and to retain relatively high moisture . Hence, the Asian monsoon circulation features the largest positive water vapour anomaly in the upper troposphere and lower stratosphere (UTLS) (here, “anomaly” is used for the deviation from zonal mean, while “bias” is used for the difference between model and observations). Although the air masses are, to some degree, confined within the monsoon anticyclone , Rossby-wave breaking along the anticyclone edge may cause substantial transport of moisture into the extratropical lowermost stratosphere . Analysis of the distribution of water vapour in the lower stratosphere simulated by a specific climate model has shown a substantial model bias in the Asian monsoon region . In the model, the maximum monsoon moisture anomaly was found to be displaced far into the Pacific region compared to satellite observations which show the moisture maximum in the Asian monsoon circulation. further showed that this model bias was likely a transport issue and could be reduced by increasing the model resolution.

Here, we combine a dedicated climate model experiment with the ECHAM MESSy Atmospheric Chemistry model (EMAC), statistical analysis of a suite of historical climate model simulations from the Coupled Model Intercomparison Project (CMIP6) and satellite observations by the Microwave Limb Sounder (MLS) to investigate regional water vapour anomalies in the UTLS and related effects on regional circulation systems. The EMAC model experiment employs either the standard “control” configuration or a new “modified-Lagrangian” transport scheme with the Chemical Lagrangian Model of the Stratosphere (CLaMS) coupled to EMAC (see Methods section in Appendix A), allowing the effects of model transport on stratospheric water vapour to be isolated. These EMAC model simulations are the same as those recently published by , but here we extend their zonal mean analysis to a regional level to address the following questions. (i) Where are the strongest biases in UTLS water vapour in current climate models compared to satellite observations? (ii) How do regional anomalies in UTLS water vapour affect regional atmospheric circulation systems, in particular in the Asian monsoon and in the Pacific region?

The strongest regional water vapour anomalies in the UTLS, as compared to the zonal mean distribution, occur in the Asian and North American monsoon circulations , with the relative strength between Asian and American anomalies under debate . At 100 hPa, a common level for studying UTLS water vapour, satellite observations by the MLS (see Methods section) show almost 2 ppmv (around 40 %) higher water vapour mixing ratios above Asia than above the Pacific during boreal summer (Fig. a). Current global climate models from the CMIP6 model intercomparison project show largest regional water vapour biases in that region (Fig. b) and a substantial displacement of the moisture anomaly towards the Pacific when compared to satellite observations. This moisture bias has already been reported by and has been related to transport effects, but only for a single model. Here, we show that in the Pacific UTLS the moisture bias is a common feature of the CMIP6 models, clearly visible in the multi-model mean (Fig. b) and reaching even higher values for specific models (Supplement Fig. S3).

Consistent with the CMIP6 models, the EMAC climate model in its control set-up (see Methods section) shows a strong moisture bias in the summertime UTLS above the Pacific (Fig. d). In EMAC, the moist bias extends even further across the Pacific than for the CMIP6 multi-model mean but overall lies within the spread of CMIP6 models (Supplement Fig. S3). After switching to a Lagrangian transport scheme, the modified-Lagrangian EMAC-CLaMS simulation (see Methods section) shows remarkably better agreement with MLS satellite observations in terms of UTLS water vapour (Fig. c). Hence, Lagrangian transport appears to significantly reduce the strong regional moisture bias in the summertime Pacific lower stratosphere in current global climate models. Note that the global mean water vapour mixing ratio in the modified-Lagrangian simulation is too low compared to MLS observations, due to a known cold bias in temperatures in the tropical tropopause layer (TTL) in temperatures in EMAC, but the focus of the present paper is on regional anomalies.

Water vapour is a radiatively active gas, thus we expect the additional moisture above the Pacific in the control EMAC simulation to affect atmospheric temperatures and circulation. The additional moisture causes enhanced long-wave cooling, which decreases temperatures in the UTLS above the Pacific. These decreased temperatures imply increased westerly zonal wind and decreased (more equatorward) meridional wind above the Pacific, consistent with thermal wind balance (see Methods section). Indeed, the westerlies in the Pacific UTLS close to the Equator strengthen in response to the additional moisture in the Pacific UTLS (Fig. d; see also the Discussion section).

Latitude–altitude cross-sections in the Pacific UTLS show that the additional moisture in the control EMAC simulation extends well above the tropopause, causing enhanced long-wave cooling and decreased temperatures particularly in the lowermost stratosphere (Fig. a–c). In the lowermost stratosphere, potential vorticity (PV) increases in response to the additional moisture, and the isentropic PV gradient strengthens, especially around the tropopause (Fig. d–f). Notably, the water vapour contours in the control EMAC simulation are not following the PV structure, indicating a decoupling of the simulated water vapour distribution from transport associated with large-scale dynamics. Satellite observations, on the other hand, show a clear anticorrelation between water vapour mixing ratios and PV in the UTLS (correlation coefficient -0.62, slope -0.53 ppmv per PVU, where PVU represents potential vorticity units), which is opposite in the control EMAC simulation (positive correlation 0.78 and slope 1.24 ppmv per PVU; see Fig. g, h). Again, the modified-Lagrangian simulation avoids this transport bias and results in much better agreement with the observations (correlation coefficient -0.86, slope -0.53 ppmv per PVU, Fig. i).

The unphysical correlation between water vapour mixing ratios and PV is an indication that the moisture bias in the control simulation is caused by small-scale, unresolved processes, namely numerical diffusion in the advective transport scheme seeand Methods section. The less diffusive Lagrangian transport scheme in the modified-Lagrangian EMAC-CLaMS model results in a more efficient sampling of cold tropopause regions and in less moisture transport into the lower stratosphere. Indeed, it is just the change in the model transport scheme and not the induced circulation changes which reduces the moist bias, as a sensitivity simulation without including the dynamical feedback of the water vapour changes provided very similar results to the modified-Lagrangian simulation with dynamical feedback included (see Supplement Fig. S1).

Moreover, the increased Pacific moisture modifies the Asian monsoon circulation. Differences in meridional velocity v between the control and the modified-Lagrangian simulations (Fig. a) show a strengthening of the equatorward flow on the eastern edge of the monsoon anticyclone related to the water-vapour-induced long-wave cooling, decreased temperatures, and modified zonal temperature gradient dT/dx above the Pacific (see Methods section). Thus, the increased water vapour above the Pacific in control EMAC causes a strengthened, eastward-shifted equatorward flow at the eastern edge of the monsoon anticyclone, associated with a zonal broadening of the monsoon circulation.

Eastward of the Asian monsoon circulation is the region of the Pacific westerly ducts , where westerly winds may extend to and across the Equator (Fig. e, f, white contours). Note that the term “westerly ducts” usually refers to the boreal winter circulation; here we refer to similar characteristics of the wind structure during summer. Differences between the control simulation and the modified-Lagrangian simulation of zonal u wind in the subtropics (Fig. b) show that the westerlies above the Pacific (east of 140∘ E) strengthen in response to enhanced Pacific moisture, consistent with the water-vapour-induced cooling and modified meridional temperature gradient dT/dy (Fig. b, and Methods section). Thus, the moist bias in the Pacific UTLS is associated with strengthened westerly ducts. Slight displacements between the zonal wind and water vapour changes indicate that, likely, the response is not entirely local (Figs.  and ).

As the Pacific UTLS moisture bias is similar between control EMAC and CMIP6 models, effects on regional circulation systems are also expected to be similar. In the CMIP6 models, the anticyclonic Asian monsoon circulation appears displaced towards the east when compared to ERA5 geopotential height (Fig. a, b), in particular when considering zonal geopotential height anomaly profiles across the anticyclone centre (Fig. c, and Methods section). All CMIP6 models show monsoon-associated geopotential height anomalies that are shifted eastward with respect to those in ERA5, with related differences larger than year-to-year variability (Fig. c, red error bars).

Based on similarity between CMIP6 inter-model correlations with the effects of the Pacific moist bias in the EMAC model experiment, we hypothesize in the following that CMIP6 model differences could partly be related to the moisture content in the Pacific UTLS. The CMIP6 inter-model correlation between local meridional v-wind velocity in the monsoon anticyclone and a water vapour index measuring the strength of the Pacific moisture anomaly (see Methods section) shows a significant anticorrelation above about 100 hPa and eastward of 120∘ E (Fig. c, with the mean correlation for winds averaged over a broader region shown in Fig. e). In other words, those models with a moister lower stratosphere above the Pacific simulate an anticyclonic monsoon circulation which is displaced further east. Above the tropopause in the lower stratosphere, the pattern of negative/positive CMIP6 inter-model correlation resembles the pattern of negative/positive differences in meridional flow in the EMAC model experiment (Fig. a). Hence, the enhanced equatorward flow at the eastern monsoon anticyclone edge appears related to the strength of the water vapour anomaly above the Pacific, in both the EMAC model experiment and in CMIP6 models. Those models with a moister Pacific UTLS simulate an eastward-displaced monsoon anticyclone.

Prevailing westerlies in the Pacific UTLS, the westerly ducts, allow Rossby waves to propagate further towards and even across the Equator and to cause inter-hemispheric exchange . Therefore, a realistic representation of the westerly ducts in atmospheric models is important for correctly simulating the global distribution of trace gases and pollutants, especially for species with strong emissions in the Northern Hemisphere. Zonal profiles of zonal wind in the northern tropics show that most CMIP6 models simulate too strong westerlies above the Pacific (Fig. d) compared to ERA5. Furthermore, CMIP6 inter-model correlations between zonal wind and Pacific UTLS moisture show a significant correlation in the Pacific subtropics around the tropopause (Fig. d, f), although weaker than for meridional wind at the eastern monsoon edge (Fig. c). The correlation for zonal wind implies that those models with increased Pacific moisture simulate stronger westerly ducts, consistent with the effect of enhanced Pacific moisture in the EMAC model experiment (Fig. b). Hence, the overestimated westerly ducts in CMIP6 models appear to be related, at least partly, to the model moist bias above the Pacific.

Differences in the simulated moisture bias are likely related to differences in diffusive model transport which are caused by differences in resolution or in the transport scheme. Across the CMIP6 ensemble, the strength of the bias is not related in a simple manner to horizontal or vertical model resolution. Although the smallest biases occur in those models with relatively high vertical resolution, different models with relatively high number of levels or horizontal grid points simulate very different biases (Fig. e–f). Exploring these issues in more detail would require dedicated model experiments with changes in the resolution in the UTLS.

Differences in model transport could also affect other trace gas species with large gradients, like ozone. However, as water vapour shows particularly steep gradients in the UTLS the associated effects on circulation are also expected to be comparably strong.

We show that a distinct moist bias above the Pacific in the summertime UTLS is a common feature of current climate models. This regional moist bias is not related to large-scale transport characteristics, and the expected anticorrelation between UTLS water vapour and PV breaks down. The bias can be significantly reduced by applying a Lagrangian, numerically less-diffusive transport scheme, as realized in the modified-Lagrangian EMAC-CLaMS simulation. The analysis of a dedicated climate model experiment and CMIP6 models further shows that the Pacific moist bias affects regional circulation patterns by enhancing local long-wave cooling and modifying temperature gradients. In particular, our results indicate that the water vapour increase in the Pacific UTLS strengthens the Pacific westerly ducts and causes zonal displacement of the Asian monsoon upper-level anticyclonic flow. Hence, improving the representation of water vapour in the Pacific UTLS region in models, e.g. by employing a less diffusive (Lagrangian) transport scheme, will affect regional circulation systems and could be a promising way to improve model simulations of current and future climate conditions, e.g. in the Asian monsoon and Pacific regions.