Highlights：The tropical lower stratosphere water vapour (SWV) experiences a drying process during 1984-2020 in both linear and nonlinear perspectives.The Indo-Pacific warm pool (IPWP) is the main factor in such drying of the tropical stratosphere.IPWP leads the coldest point region cooler modulating tropical SWV entry by enhancing equatorial waves.ABSTRACTA decreasing trend in the tropical (30°S~30°N) stratospheric water vapour (SWV) entry in recent four decades (from 1984 to 2020) is detected based on the Stratospheric Water and OzOne Satellite Homogenized (SWOOSH) measurements and the ERA5 reanalysis dataset using linear regression and Ensemble Empirical Mode Decomposition (EEMD) analysis. With the concurrent warming of the SST, the Indo-Pacific warm pool (IPWP) appears to be the most significant region among the tropical oceans based on correlation analysis. More than 43% of the decreasing tropical lower SWV trend is likely to be related to the IPWP sea surface temperature (SST) warming. To validate this relationship, two groups of idealized runs are carried out with version 4 of the Whole Atmosphere Community Climate Model (WACCM4) and version 5 of the Community Atmosphere Model (CAM5). Both simulations agree with the observational-based linkage. The IPWP-SST-warming forced simulations show that the temperature in the tropical tropopause has decreased at the rate of around 0.318 K per decade in the coldest point region, as the tropical convection over the IPWP has become more vigorous and excited stronger equatorial waves to produce adiabatic cooling around tropopause. This cooling tropical tropopause leads to a dehydrating tropical lower stratosphere at the rate of 0.025 ppmv per decade, as expected by the freeze-drying mechanism. These results imply the substantial warming trend of IPWP is an important factor for the long-term trend of the tropical SWV entry under climate change, and a better representation of this relationship in the model is critical for the SWV projection under future climate scenarios.Keywords: Stratospheric water vapour; Indo-Pacific warm pool; Trend; Tropopause; Coldest point regionIntroductionThe stratospheric water vapour (SWV) mainly originates from the troposphere: the moist air parcels at the bottom of the troposphere ascend, reaching the tropical tropopause layer (TTL) between 14–18.5 km, experiencing a severe dehydration process at the TTL (because the TTL has the coldest temperature in the lower atmosphere) (Gettelman and Forster 2002; Fueglistaler et al. 2009), then arriving in the stratosphere. The SWV is suggested to contribute significantly to global climate change by altering the infrared opacity of the atmosphere (e.g., Soden and Held 2006), providing a strong positive feedback at +0.3 W/(m2·K) to global warming (Dessler et al. 2013). When the SWV increases, it subsequently leads to warming in the troposphere and cooling in the stratosphere, and the warmer troposphere will, in turn, increase the SWV (Rind and Lonergan 1995; de F. Forster and Shine 1999; Solomon et al. 2010; Dessler et al. 2013; Fu et al. 2015), and by this positive feedback, the increase of the SWV will accelerate the rate of increase in global surface temperature and vice versa. Meanwhile, the SWV participates in stratospheric chemical processes as the primary source of stratospheric hydrogen oxide radicals. For example, it strongly affects heterogeneous chemistry on cold sulfate aerosol and the formation of polar stratospheric clouds, which promote chlorine activation and polar ozone loss (Evans et al. 1998; Shindell 2001; Stenke and Grewe 2005; Tian et al. 2009). So, it is critical to understand the decadal and long-term SWV variability and the relevant physical processes.Previous studies have already documented that the TTL temperature is very important for SWV variability. Because the TTL is the main area where air enters the stratosphere, changes in SWV are largely related to the tropical SWV entry and the TTL temperature largely determines the SWV entry values (Brewer 1949; Randel et al. 1998; Scaife et al. 2003; Fueglistaler et al. 2005; Rosenlof and Reid 2008; Schoeberl and Dessler 2011; Grise and Thompson 2012; Dessler et al. 2013). Therefore, the multi-timescale variations of the SWV ranging from daily to decadal timescales (e.g., Randel et al. 2004; Fueglistaler and Haynes 2005; Fujiwara et al. 2010; Hegglin et al. 2014) can be traced to TTL temperature variations (e.g., Randel et al. 2007; Rosenlof and Reid 2008; Randel 2010; Fueglistaler et al. 2013; Randel and Jensen 2013). As a layer between the stratosphere and troposphere at about 14–18.5 km (Fueglistaler et al. 2009), the TTL temperature is affected by both the stratospheric (top-down) and tropospheric (bottom-up) processes, including variability of the Brewer-Dobson circulation (BDC, a stratospheric mean meridional circulation), the quasi-biennial oscillation (QBO), and tropical convection (bottom-up). Because the TTL temperature and SWV entry values are generally the results of the interplay between the top-down and bottom-up processes, the SWV’s interpretation (e.g., Hegglin et al. 2014) and prediction (e.g., Gettelman et al. 2010) are complex. In the tropical troposphere, anomalous deep convection induces upward motion near the tropopause and thereby results in a cooling of the TTL (Highwood and Hoskins 1998). And deep convection is usually associated with the El Nino Southern Oscillation (ENSO), Asian Monsoon, and Madden-Julian Oscillation exert tropical planetary waves including the equatorial Rossby wave and Kelvin wave. In the stratosphere, the acceleration of the BDC causes the adiabatic cooling of the TTL through the enhanced large-scale vertical ascending motions, and vice versa (Holton et al. 1995; Thompson and Solomon 2005). Fu et al. (2010) also documented that the strength of the BDC is a main factor driving the seasonal variability of the TTL temperature. In short, the TTL temperature variability is driven by both the tropospheric (bottom-up) processes and the stratospheric (top-down) (Kumar et al. 2014).By using balloon-borne water vapour profiles above Washington DC and Boulder, Oltmans et al. (2000) observed an increase of lower SWV by about 1% per year during the 1960s and 1990s. This agrees with the Third Assessment Report of the IPCC, which reported water vapour in the lower stratosphere is likely to have increased by about 10% per decade since the beginning of the observational record. Dessler et al. (2013) showed observational evidence for stratospheric water vapour feedback—a warmer climate increases stratospheric water vapour, and because stratospheric water vapour is itself a greenhouse gas, this leads to further warming. Lin et al. (2017) found that the tropical tropopause layer will become warmer in response to carbon dioxide increase and surface warming. A few numerical studies reported that a moist stratosphere occurs under global warming scenario by using the forcing of quadrupling CO2 in different general circulation models (Zhang and Huang 2014; Banerjee et al. 2019; Li and Newman 2020; Wang and Huang 2020; Xia et al. 2021b). Keeble et al. (2021) suggested that CMIP6 multi-model mean SWV mixing ratios in the tropical lower stratosphere have increased by ∼0.5 ppmv from the pre-industrial to the present-day period and are projected to increase further by the end of the 21st century. Keeble et al. (2021) further pointed out that the largest SWV increases (∼2 ppmv) are simulated under the future scenarios with the highest assumed forcing pathway (e.g., SSP5-8.5).However, the increasing trend of the SWV seems stopped or becomes blurred after the 1990s. Although the SWV above Boulder is reported to increase during the periods of 1992 to 2002 in the balloon water vapour data (Randel et al. 2004). This increasing trend is not reproduced well by satellite data. Randel et al. (2004) also documented that the SWV near Boulder, Colorado (40°N) and the tropical mean (60°S-60°N) SWV have no significant trend in 1992-2002 based on the HALOE data. Besides, the Fifth Assessment Report of the IPCC (IPCC5, 2014) documented that the near-global satellite measurements of SWV show substantial variability but small net changes for 1992-2011, i.e., the satellite data show no clear trends for the SWV. Randel et al. (2006) even reported that the near-global SWV after 2001 decreased (or had persistent low values beginning in 2001), and this near-global SWV decrease is attributed to the enhanced tropical upwelling after 2001. Hurst et al. (2011) analyzed the balloon-borne SWV over Boulder, Colorado, then reported the multi-decadal variability of the SWV: the SWV increased by an average of 1.0 ± 0.2 ppmv (27 ± 6%) during 1980-2010, but in 2001-2005, it has an opposite trend to other periods. Recently, another strong drop is reported in the tropical SWV (10°S-10°N, similar to the SWV drop observed in the year ~2000) was observed in 2011-2012 (Urban et al. 2014). Hegglin et al. (2014) using observation data revised by transfer function found a negative trend in the lower and mid-stratosphere. Dessler et al. (2014) revealed that water vapour entering the stratosphere has no firm evidence of trend. Konopka et al. (2022) suggested that the stratosphere has become wetter after 2000. Tao et al. (2023) found that SWV has a robust multi-decadal variation and short-term trends in SWV are closely related to this multi-decadal variation. These imply that there are great uncertainties in the trends of SWV and the trends are sensitive to the period focused on.Many studies have suggested that tropical oceans have a crucial effect on the stratosphere (e.g., Hu et al. 2014; Hu and Guan 2018; Xie et al. 2020b; Xie et al. 2020a; Xia et al. 2021a). As an important component of the stratosphere, the SWV is no exception. Scaife et al. (2003) have pointed out that a positive trend in water vapour of around 0.1% per year and ENSO effects appear to explain no more than about one-tenth of the long-term trend by using model and observational data. Tropical SST variability, especially ENSO, has been known to be an important factor in determining the amount of water vapour being uplifted to the upper tropospheric region and the lower stratosphere by altering tropical convection (Su et al. 2006; Rosenlof and Reid 2008; Liang et al. 2011; Xie et al. 2012; Garfinkel et al. 2013a; Garfinkel et al. 2013b; Avery et al. 2017; Su et al. 2020). The tropical SST has been suggested to contribute to the drop of the lower stratospheric water vapour during 2000 (Brinkop et al. 2016; Ding and Fu 2018). The Indo-Pacific Warm Pool (IPWP) has been also documented as a vital region that affects the stratosphere and tropical lower SWV (Xie et al. 2014; Xie et al. 2018; Zhou et al. 2018). The warm phase of IPWP causes a drier lower stratosphere and vice versa. Zhou et al. (2021) further pointed out that such impact has seasonality and hemispheric differences. Almost the entire tropical ocean shows a warming long-term trend in SST over the last century (Deser et al. 2010). The tropical western Pacific is warm faster than the eastern Pacific in observations (Cane et al. 1997; Kanamitsu et al. 2002; Compo and Sardeshmukh 2010; Zhang et al. 2010; Li et al. 2017). There is widespread warming across the tropical Indian Ocean basin and SSTs have reached 28°C in the western Indian Ocean, which has expanded the Indo-Pacific warm pool region defined by the 28°C isotherm westward (Roxy et al. 2015). It is quite possible that the zonally asymmetric warming trend of tropical oceans will further influence the SWV entry and provide the principal source for its decadal trend. Previous studies have shown the impact of IPWP on interannual variability of tropical lower SWV and discussed its seasonal and hemispheric differences. However, the impact of IPWP on tropical SWV entry over longer time periods is unclear, especially when the IPWP is substantially warming in the past forty years. Therefore, in this study, we seek to understand the impact of IPWP continuous warming on tropical SWV entry in recent decades.In short, the decadal or long-term changes of the tropical SWV during the past decades seem not to strictly follow the presumed long-term increasing trend under global warming. So, we try to revisit and interpret the decadal or long-term change of the SWV with longer observations for the period 1984-2020 and explore the potential impact of IPWP warming on it. The remainder of the paper is arranged as follows. The data, model, and methods we used are presented in Section 2. In section 3, we revisit the long-term trend of SWV for the period 1984-2020, then focus on its link with the warming of IPWP. Section 4 is a summary of the principal findings.Data, model and methodsWater vapor dataStratospheric Water and OzOne Satellite Homogenized (SWOOSH) is a merged data set that ranges from 1984 to the present containing multiple satellite data and it not only offers values of SWV but also provides some ancillary information like standard deviation, number of data points and others (Davis et al. 2016). In addition, it also supplies a combined product which is a weighted mean value from the available satellite measurements mentioned above and it has a fabulous advantage that when one satellite measurement is missing, others will be filled in by using different algorithms. SWOOSH has been used in some studies (Hardiman et al. 2015; Gilford et al. 2016). Our study used this combined product with 31 pressure levels from 316 to 1 hPa and the version of SWOOSH is v2.7.In addition to SWOOSH, water vapour of ERA5 was also analyzed for comparison. ERA5 which covers the period from 1979 to the present, is the latest global atmospheric reanalysis product obtained from European Centre for Medium-Range Weather Forecasts (ECMWF). Based on the 4D-Var data assimilation scheme and Integrated Forecast System (IFS) CY41R2, it covers the earth with a 30 km horizontal grid and 137 hybrid sigma/pressure levels from the surface to 0.01 hPa (Hersbach et al. 2020). And the resolution we used is 0.25° × 0.25° in the horizontal and 37 levels from 1000 hPa to 1 hPa in the vertical. Because ERA5 has a cold bias in the lower stratosphere during 2000 and 2006, ERA5.1 is applied in this special time.Meteorological dataBesides ERA5, the major meteorological data we also used is Japanese 55-year Reanalysis (JRA-55). JRA-55 is conducted by Japan Meteorological Agency (JMA) and JRA5 is a comprehensive climate dataset with the applicant of 4D-Var. It covers the period from 1958, coinciding with the establishment of the global radiosonde observing system. The resolution we used is 1.25° × 1.25° in the horizontal and 37 levels from 1000 hPa to 1 hPa in the vertical.Another relevant meteorological data is sea surface temperature, and Hadley Centre Sea Ice and Sea Surface Temperature (HadISST) data set was applied in our study. It is mainly from Met Office Marine Data Bank (MDB), and it also covers data from Global Telecommunications System (GTS) since 1982. To enhance the coverage of data, when data from MDB is missing, the monthly median SSTs for 1871-1995 from the Comprehensive Ocean-Atmosphere Data Set (COADS) (now ICOADS) are included (Rayner et al. 2003). And its resolution is 1° × 1°.ModelThe Community Earth System Model version 1 (CESM1), developed by National Center for Atmosphere Research (NCAR), can simulate the state of climate from the past to the future. It consists of several relatively independent component models, including atmosphere, ocean, land, land ice, sea ice, and so on, and all of them have their own spatial resolutions. Besides, CESM1 has a central coupler that can exchange energy and information between different component models (Hurrell et al. 2013). In our study, we used version 5 of the Community Atmosphere Model (CAM5) to design experiments, which is one of the vital component models of CESM1, to simulate global atmosphere activity. CAM5 can not only run as one of the component models of CESM1 but also run as an independent model. It applies 30 vertical levels from the ground to 3.64 hPa. Version 4 of the Whole Atmosphere Community Climate Model (WACCM4), a comprehensive numerical model based on CAM, was also used in the experiment design. It can extend from the surface to the thermosphere, namely, from the surface to 5.1× 10-6 hPa (about 140 kilometers) with 66 vertical levels. In our study, the finite-volume dynamical core was applied in both CAM5 and WACCM4. We used the 1.9° × 2.5° medium resolution version which includes 96 longitude and 144 latitude points.CAM5 and WACCM4 are employed to investigate how and to what extent SWV entry changes in response to IPWP warming. WACCM4 can capture the negative relationship between the IPWP SST anomalies and the tropical SWV entry (Xie et al. 2018; Zhou et al. 2018). Two groups of experiments are carried out, which involve a control group forced by observed SST from 1955 to 2005 (E0, E1) and another forced by the observed SST in the IPWP region only (E2, E3). Details of transient experiments can be referred to in Table 1. Due to the limitation of WACCM4, E0 and E2 run only from 1955 to 2005. So, although CAM5 in E1 and E3 run from 1900 to 2005, we use the same period as WACCM4 in E0 and E2. We evaluate the two modes using control runs (E0, E1) by comparing them with the observational data and reanalysis data. Both can represent the structures in SWV and tropical tropopause temperature, with the pattern correlations exceeding 0.7 between observed and simulated climatology (Fig. S1 and Fig. S2).
