Fig. 4. (a), (b), (c) and (d) are trends of globally zonally-averaged
SWV from 1984 to 2020. (e), (f), (g) and (h) are trends of temperature
from 1984 to 2020 at 70 hPa. (a) and (b) are from SWOOSH. (c), (d), (e)
and (f) are from ERA5. (g) and (h) are from JRA55. The left column is
the raw trend. The right column is trend removed IPWP by regression.
Dotted regions indicate significance at the 90% confidence level. A
five-year running average is used before calculating trends.
The importance of the warming IPWP for dehydration is further quantified
in Fig.5. What can be clearly seen in the bar plot is the dramatic
decline in the dehydration rate after regressing out the IPWP warming.
For SWOOSH, the drying trend in the tropical SWV entry reduces from
0.106 ppmv per decade to less than 0.06 ppmv per decade after linearly
removing the IPWP warming signal. The decadal decrease rate of SWV
associated with IPWP warming accounts for 43% of its tropical mean. In
ERA5, the trend of tropical SWV entry is -0.036 ppmv per decade and the
trend without IPWP is -0.014 ppmv per decade. IPWP warming contributes
61% of the decreasing trend of tropical SWV entry. The global mean of
both data sets is close to the tropical mean. It implies that global SWV
is mainly controlled by the tropical SWV entry. In sum, the linear
estimation based on SWOOSH for the contribution of IPWP warming to the
drying trend in the tropical SWV entry is about 43%, highlighting a
fundamental role of IPWP warming. However, as we have removed the
interannual signal from the analysis, the relationship between SWV entry
and IPWP may mainly be driven by secular trends. And there may be an
overestimation of the contribution of IPWP. In the next part, we further
present models to validate the effect of IPWP warming on SWV entry.