Fig. 1. (a) Time series of tropical (30°S-30°N) averaged SWV entry anomalies (solid lines) and their linear trends (dashed lines) based on SWOOSH (black lines, 82 hPa) and ERA5 (red lines, 70 hPa) from 1984 to 2020. The trends are shown in the dashed lines legend, and one asterisk denotes significance at the 90% confidence level and two asterisks denote significance at the 99% confidence level. The uncertainties are expressed by 2σ errors. The number in the top left-hand corner and top right-hand corner are the averages of anomalies over the first five years and the last five years. (b) The same as (a), but for the tropical tropopause temperature (30°S-30°N, 70 hPa) based on ERA5 (red lines) and JRA55 (black lines). (c) Vertical profiles of global SWV trend using SWOOSH (red line) and ERA5 (black line). (d) The same as (c), but for meridional variations of the global SWV (100-10 hPa) trend. Bold lines in (c) and (d) mean significance at the 90% confidence level.
The cooling trend in the lower stratosphere is captured by the tropical tropopause temperature (Fig. 1b). The linear cooling trends are 0.328 ± 0.083 K per decade in ERA5, while JRA55 has a relatively larger rate at 0.253 ± 0.08 K per decade. Coherently, the temperature average value also supports the SWV entry decrease but with a larger value in ERA5 than in JRA55. The free-drying regime works if it is assumed all air entering from the troposphere to the stratosphere passes through the extremely cold temperature. Thus, the decreased rate of SWV entry predicted by the Clausius-Clapeyron equation, for the linear cooling trend of the tropical tropopause at the value of 0.328 K per decade, is about 0.09 ppmv per decade, which is very close but slightly smaller than the observed value (0.106 ppmv per decade).
As the tropical SWV entry decreases, we can speculate that the global SWV may have the same change. So, the vertical and meridional resolved zonal-mean SWV trends are shown in Fig. 1c and 1d, respectively. The decreasing trends based on the two datasets are generally in agreement in the lower stratosphere, showing significant dehydration in the lower stratosphere for the period 1984-2020 (Fig. 1c). Above 30 hPa, however, the positive trend is only seen in SWOOSH. This increase in water vapour is possibly induced by the increase of methane (le Texier et al. 1988) and its production of water vapour via oxidization. The meridional variation shows that the drying trend in the southern hemisphere is stronger than that in the northern hemisphere. This hemispheric difference has not been clarified yet. In general, the linear trends of SWV based on observations show a consistent decreasing trend for the period 1984-2020. And it is similar to the tropical SWV entry.
However, the tropical SWV entry time series show large interannual variability, which can degrade the linear estimation of its long-term trend. So, we apply a nonlinear algorithm, EEMD, to confirm the drying trend. Fig. 2 shows the original tropical SWV entry time series in the tropics and its decomposed modes using EEMD analysis. In the original time series, the tropical SWV entry shows large interannual variability with the magnitudes up to about 1.5 ppmv (Fig. 2 (a) and (b)). The first three components are corresponding to the irregular oscillations, annual cycles, and interannual variations, with the peak-to-peak amplitudes at around 0.2 ppmv, 0.6 ppmv, and 0.2 ppmv, respectively (Fig. 2 (c), (d), (e), (f), (g) and (h)). With the number of EEMD modes increasing, the frequency of the oscillations becomes lower. Finally, the residual term has only one extremum which can be used to diagnose the nonlinear long-term change (Fig. 2 (i) and (j)). The residual component shows a general decreasing trend in the tropical SWV entry and the level around 2020 is still much lower than that around 1980. And, one thing that should be noted is that ERA5 exhibits strong non-linearity in its residual with a turnaround in 2005. This turnaround may be related to the preceding inconsistency between the amplitude of the drop and the estimated linear trend. And this turnaround makes decreasing trends much weaker in ERA5. We also applied EEMD on the tropopause temperature time series, and the results are in good agreement with a residual term showing a decades-long cooling trend (Fig. S3).