Content uploaded by Martin Jucker
Author content
All content in this area was uploaded by Martin Jucker on Aug 04, 2023
Content may be subject to copyright.
Generated using the official AMS L
A
T
E
X template v6.1 two-column layout. This work has been submitted for
publication. Copyright in this work may be transferred without further notice, and this version may no longer be
accessible.
Long-term surface impact of Hunga Tonga-Hunga Ha’apai-like stratospheric water vapor
injection
Martin Jucker,aChris Lucas,band Deepashree Dutta a
aClimate Change Research Centre and Centre of Excellence for Climate Extremes, University of New South Wales, Sydney, Australia
bBureau of Meteorology, Melbourne, Australia
ABSTRACT: The amount of water vapor injected into the stratosphere after the eruption of Hunga Tonga-Hunga Ha’apai (HTHH) was
unprecedented, and it is therefore unclear what it might mean for surface climate. We use climate model simulations to assess the long-term
surface impacts of stratospheric water vapor (SWV) anomalies caused by volcanic eruptions. The simulations show that the SWV anomalies
lead to strong and persistent warming of Northern Hemisphere landmasses in boreal winter, and austral winter cooling over Australia. Thus,
SWV forcing from volcanic eruptions like the one from Hunga Tonga-Hunga Ha’apai can have surface impacts on a decadal timescale.
We also emphasize that the surface response to SWV anomalies is more complex than simple warming due to greenhouse forcing and is
influenced by factors such as regional circulation patterns and cloud feedbacks. Further research is needed to fully understand the multi-year
effects of SWV anomalies and their relationship with climate phenomena like El Ni˜
no Southern Oscillation.
SIGNIFICANCE STATEMENT: Volcanic eruptions
typically cool the Earth’s surface by releasing aerosols
which reflect sunlight. However, a recent eruption released
a significant amount of water vapor — a strong greenhouse
gas — into the stratosphere with unknown consequences.
This study examines the aftermath of the eruption and re-
veals that surface temperatures across large regions of the
world increase by over 1.5°C for several years, although
some areas experience cooling close to 1°C. Additionally,
the research suggests a potential connection between the
eruption and sea surface temperatures in the tropical Pa-
cific, which warrants further investigation.
1. Introduction
Large volcanic eruptions can have significant and long-
lasting impacts on climate (Robock 2000), as demonstrated
by the El Chich´
on and Mt Pinatubo eruptions in the 1980s
and 1990s, respectively. These eruptions released massive
amounts of sulfur dioxide into the atmosphere, leading to
the formation of stratospheric sulfate aerosol and global
cooling of around 0.5-1.0 K. While the focus has largely
been on these larger eruptions, recent studies suggest that
even smaller volcanic events can have a measurable impact
on climate (Vernier et al. 2011; Santer et al. 2014).
On 15 January 2022, the submarine volcano Hunga
Tonga-Hunga Ha’apai (HTHH) erupted in the Southern
Hemisphere (SH) subtropical Pacific Ocean with unprece-
dented intensity, producing eruption plumes that reached
altitudes of 58 km in the upper stratosphere and lower
mesosphere (Carr et al. 2022; Proud et al. 2022). Despite
Corresponding author: Martin Jucker, publica-
tions@martinjucker.com
Deepashree Dutta’s current affiliation: Cambridge University, U.K.
releasing only 0.4-0.5 Tg of sulfur dioxide (Carn et al.
2022; Gupta et al. 2022), which is lower than high climate
impact events noted earlier, the eruption injected a large
amount of water vapor into the stratosphere, equivalent to
5-10% of the climatological amount (V¨
omel et al. 2022;
Mill´
an et al. 2022). This stratospheric water vapor (SWV)
can have both cooling and warming effects on the climate,
and the potential impacts of the HTHH SWV release on
the climate are still largely unknown.
Enhanced SWV can contribute to surface warming
(Dessler et al. 2013) and changes to the tropospheric and
stratospheric circulation patterns (Maycock et al. 2013).
It can also influence the formation of polar stratospheric
clouds (PSCs), and lead to increased hydroxyl radical con-
centrations, which both play a key role in stratospheric
ozone depletion (Solomon 1999; Tritscher et al. 2021).
The impacts of the HTHH SWV release on PSC formation
and ozone chemistry are uncertain and should be expected
to depend not only on chemical, but also dynamical evolu-
tion of the initial cloud.
Once it reaches the stratosphere, SWV is transported
in the zonal direction by the prevailing stratospheric
winds, and in the vertical and meridional direction by the
stratospheric overturning circulation, the so-called Brewer-
Dobson circulation (BDC, Brewer 1949; Dobson 1956;
Plumb 2002). As a result, the initially strongly confined
plume becomes a widely distributed SWV anomaly in the
zonal direction, but the meridional and vertical evolution is
much slower. In the zonal mean picture, SWV is advected
by the BDC on seasonal to multi-year timescales, but its
distribution in the stratosphere is also constrained by the
strength of zonal winds, such as the polar vortex or the
phase of the tropical Quasi-Biennial Oscillation (Holton
1
2
and Tan 1980; Baldwin et al. 2001), which can act as
transport barriers.
Therefore, further research is required to fully under-
stand the potential long-term effects of the HTHH eruption
on climate. Despite the lower amount of sulfur dioxide re-
leased, the injection of water vapor into the stratosphere
is a feature of the HTHH eruption that demands further
investigation to better understand its potential contribution
to climate variability.
2. Data and model setup
In this work, we examine the impact of an SWV dis-
turbance resembling that of the HTHH eruption and focus
on the long-term effect of SWV on surface climate. We
utilize data collected by NASA’s Aura Microwave Limb
Sounder (MLS) instrument to analyze the HTHH cloud’s
initial characteristics and evolution, and employ version
4 of the Whole Atmosphere Community Climate Model
(WACCM4) to address our research questions. Our re-
search focus is not on recreating the HTHH eruption in
every detail. Instead, we investigate the long-term climate
effects of a spontaneous and locally bounded perturbation
in stratospheric water vapor based on the scale and shape
of the perturbation caused by the HTHH eruption.
a. Initial water vapour perturbation
The details of data retrieval and processing for MLS
satellite data are given in Appendix A, and we contend
ourselves with summarizing the main results required for
setting up our model simulations.
We estimate the vertical profile and approximate shape
of the SWV anomaly in the initial states from three over-
passes from 16 January 2022, and compare SWV values
to those obtained from a 2005-2021 climatology. Our best
estimate from these data is an anomaly of about 100 Tg of
SWV, with a shape illustrated in Fig. 1.
Our estimate is between the 50 Tg derived from ra-
diosonde data (V¨
omel et al. 2022) and the 146 ±5 Tg
from Mill´
an et al. (2022), who utilized version 4 of the
MLS data. The key divergence between our estimate and
that of Mill´
an is that we identify a 47 Tg SWV anomaly
that is present before the 15 January 2022 eruption, and is
therefore not included in our estimates of the perturbation
associated with this specific eruption (Figs. A2, A3).
b. Model simulations
1) WACCM
We conduct our numerical experiments using the
Whole Atmosphere Community Climate Model version
4 (WACCM4)(Marsh et al. 2013). WACCM4 is a high-
top, fully interactive chemistry-climate model with a 1.875
×2.5°finite volume grid and 66 hybrid sigma levels ex-
tending from the surface to the lower thermosphere (140
km). WACCM4 includes all the physical parameteriza-
tions of the Community Atmosphere Model, Version 4
(Hurrell et al. 2013) and is coupled to the Community
Land Model Version 4 (Oleson et al. 2010) without dy-
namic vegetation or an active carbon-nitrogen cycle. The
fully interactive chemistry module in WACCM4 is based
on the Model for Ozone and Related Chemical Tracers
(MOZART) version 3 (Kinnison et al. 2007). The strato-
spheric temperature, chemical and radiative processes are
influenced by water vapor and volcanic aerosols (Zhu et al.
2022; Solomon 1999; Zhu et al. 2018; de F. Forster and
Shine 1999). In WACCM4, the temperature change due to
volcanic aerosols is calculated in a similar way to Tilmes
et al. (2009) following the Chemistry–Climate Model Val-
idation 2 (CCMVal2) protocols. In our experiments, the
climatological volcanic aerosol surface area density (SAD)
is based on observations for the year 2000 (Eyring et al.
2010). Similarly, the concentration of greenhouse gases
and halogens follow the seasonal cycle of year 2000. The
setup used in this study does not include the 11-year solar
cycle and the Quasi-Biennial Oscillation (QBO) package is
not activated, however our free running simulations include
an internally generated QBO.
The control experiment is forced with climatological
sea surface temperatures averaged over 1982–2001 from
the merged between Hadley-NOAA/Optimal Interpolation
SST analysesHurrell et al. (2008). The control experiment
is integrated for 43 years, and we use the results from last
the 38 years to form a 29-member ensemble (i.e., each 1st
January is treated as the start of an individual ensemble
member). The ensemble members are integrated for 10
years. To examine the response to a sudden increase in
stratospheric water vapor, we restart each of the 29 en-
semble members with a Gaussian water vapor anomaly
centered at 20°S and 181°E (Fig. 1). Specifically, a ma-
jority of the SWV mass (94%) is placed between 22 and
27 km altitude, with a smaller concentration near 37 km.
Horizontally, a Gaussian distribution is implemented with
a latitudinal width of 5°and a longitudinal width of 10°.
We inject 125Tg instead of the target 100 Tg of wa-
ter vapor in the initial conditions, because about 20% are
lost during the first few days due to the production of ice
clouds, which we attribute to several factors connected to
the idealized perturbation, such as the missing accompa-
nying temperature perturbation, and the strong localization
of the plume over only a few model grid points. The ice
cloud disappears within a few days, after which the to-
tal stratospheric water anomaly approaches the measured
100 Tg.
As the restarts are synchronized at the beginning of the
year, the anomalies are also added on January 1st for each
member. Branching a new perturbation member every
year also allows us to sample the effect of the QBO, with
14 members in the easterly (QBOE) and 15 members in the
westerly (QBOW) phase, as defined by the sign of the zonal
3
a) b)
0.00 0.05 0.10 0.15 0.20 0.25
fraction of total SWV anomaly []
1.0
10.0
100.0
pressure [hPa]
vertical SWV anomaly profile
Fig. 1. a) Artist’s view of the stratospheric water vapor cloud perturbation added at the start of each perturbation member simulation. b) Vertical
profile of the initial cloud above the eruption as derived from MLS.
mean zonal wind between 5S and 5N at 50hPa (Dunkerton
1990). We note here that during the HTHH eruption, the
real stratosphere was in a QBOE phase.
Despite using a full chemistry-climate model, this is
still an idealized experiment and there are significant limi-
tations on the interpretation of the response predicted here,
particularly around the use of a climatological SST. How-
ever, we focus on surface impact over land masses, and
the anomalous atmospheric circulations which are respon-
sible for those signals over land. We also chose to favor a
large number of ensemble members and longer simulations
over a fully coupled ocean, as that addition would severely
limit our capacity in terms of ensemble size and simulation
length.
2) MiMA
To alleviate some of the restrictions on atmosphere-
ocean interactions, we run additional ensemble simulations
with the Model of an idealized Moist Atmosphere (MiMA,
Jucker and Gerber 2017), an intermediate-complexity
moist general circulation model. The advantage of this
model is that it includes a mixed layer ocean, allowing for
an estimate of how SSTs might be influenced by the SWV
and ozone perturbations produced in WACCM.
MiMA is based on GFDL AM2.1 atmospheric model
with a spectral dynamical core, which we run at T42 reso-
lution ( 2.8 degrees) and 40 vertical levels up to 0.07hPa.
MiMA includes interactive water vapor including evapora-
tion and condensation, and a simplified convection scheme
following Frierson et al. (2007), but it does not include
any representation of cloud physics. Full radiative trans-
fer is computed with the Rapid Radiative Transfer Model
RRTMG (Mlawer et al. 1997), but here we fix ozone and
water vapor to the monthly values from our WACCM sim-
ulations (monthly means linearly interpolated to each radi-
ation time step). We include a homogeneous CO2concen-
tration of 390 ppm, the solar constant is set to 1370 Wm−2,
and surface albedo follows Garfinkel et al. (2020b) with a
value of 0.23 at low latitudes and 0.80 in polar regions.
We run MiMA in a configuration following the Southern
Hemisphere benchmark case of Garfinkel et al. (2020a),
which most notably includes a gravity wave scheme and
surface ocean heat fluxes mimicking the major ocean
currents plus a realistic Intertropical Convergence Zone
(ITCZ) and South Pacific Convergence Zone (SPCZ)
(Fig. B1). Land surface includes realistic topography, in-
creased surface roughness, restricted evaporation, higher
albedo (0.43) for the major deserts, and lower heat capac-
ity (10 million Jm−2K−1vs. 100-300 million Jm−2K−1for
water).
c. Statistical significance
Throughout the analysis, we compute anomalies as the
difference between control and perturbation for each en-
semble member, and then use a two-tailed t-test against
the null hypothesis that the ensemble mean anomalies are
zero. We have also used Kolmogorov-Smirnov and agree-
ment of sign tests, and the results are very similar. Rather
than hatching significant regions, we only plot anomalies
which are significant at the 90% confidence level. For
bar plots, we estimate the 90% confidence interval from a
1000-sample bootstrap method.
d. Labeling of time
We add SWV perturbations on 1 January, meaning that
calendar years are also a measure of years since eruption.
We follow the convention that the year of the eruption is
year 0, and 1 January one year after eruption is the start
of year 1. With this convention, the years correspond-
ing to the HTHH eruption are simply 2022+𝑛for year
𝑛. We will mostly discuss seasonal means, and attribute
4
December-January-February (DJF) to the year correspond-
ing to January-February. That is, DJF of year 1 is Decem-
ber of year 0 to February of year 1, e.g., one year after
eruption. DJF of year 0 does not include December, and
year 10 only includes December.
3. Stratospheric evolution
We first examine the evolution of the stratospheric per-
turbations induced by the SWV plume, and stratify the
results by initial QBO phase (Fig. 2). The total SWV mass
perturbation remains in the stratosphere for 7-8 years, and
the initial evolution is different based on initial QBO phase.
We note that during the eruption of HTHH in 2022, the
stratosphere was in a QBOE phase. In WACCM, QBOE
members retain more total water vapor mass than QBOW,
and by June 2023, the observed SWV anomaly from MLS
(Fig. 2, black) aligns closely with the QBOE ensemble
mean from WACCM (purple dashed). We attribute the
stronger peak in MLS during boreal winter of year 1 com-
pared to our simulations to the 2023 Sudden Stratospheric
Warming which the ensemble mean would not be able to
capture.
The differences in circulation between QBO phases re-
sult in distinct evolutions of the SWV anomaly during the
first 2-3 years (Fig. 3). The members in the easterly phase
(QBOE, those similar to HTHH) are subject to a weaker
BDC and easterly tropical winds during the initial phase,
and accumulate more SWV in the tropical stratosphere
during the first year. They then also transport more SWV
from the tropics into the NH during boreal winter one
year after eruption (also see Fig. 5c) . During the second
year, when the members which were initially in the QBOE
phase switch to QBOW and vice versa, the picture is in-
verted, and more SWV is transported towards Antarctica
to influence the ozone hole, while more SWV accumulates
in the tropical stratosphere for the (initially) QBOW mem-
bers. Most of the SWV differences between the two QBO
phases disappear by the end of year 3.
Besides differences in spatial distribution of SWV, the
differences in total SWV mass between QBOE and QBOW
members seen in Fig. 2 can be explained by the removal of
SWV via development of polar stratospheric clouds (PSCs,
Fig. 4). While QBOE members do not show any statis-
tically significant increase in PSCs in either hemisphere
(Fig. 4a), the QBOW members produce significantly more
PSCs in the SH autumn of years 1 and 2 (Fig. 4b). As
discussed above, QBOE members initially transport more
SWV into the NH, leaving less SWV in the SH to produce
PSCs. For QBOW members, the increased concentration
of SWV remaining in the SH means that more PSCs form,
and more SWV mass is lost to condensation. In agreement
with Fig. 3, the differences between QBOW and QBOE
phases disappears during year 3.
0123456789
time [years since eruption]
20
0
20
40
60
80
100
mass anomaly [Tg]
Total stratospheric water mass [Tg]
WACCM QBOE
WACCM QBOW
WACCM
MLS
Fig. 2. Total stratospheric water vapor mass anomaly [Tg] for (red,
solid) all WACCM members, (purple, dashed) QBOE members, (green,
dash-dotted) QBOW members, and (black, solid) MLS (until 11 July
2023). Red shading and purple/green thin lines show the ±1𝜎range
across members. The ensemble mean is solid where statistically signif-
icant, and dashed otherwise.
Even though there are more PSCs in autumn for QBOW
members, the ozone hole area is significantly larger dur-
ing Sep-Feb of the second spring/summer after eruption
for QBOE members (Fig. 5). This is because Antarctic
total column ozone reduction happens over an extended
period into summer for QBOE members, while it is lim-
ited to mid-winter and midlatitudes for QBOW members
(Fig. 5, bottom). We therefore expect a larger Antarctic
ozone hole during spring/summer of year 2, which trans-
lates to 2023/24 if applied to the HTHH eruption, with a
September-February mean increase of more than 2 million
km2(Fig. 5a). The reason for this is that there is not enough
time for the SWV to reach polar latitudes before the year
1 Antarctic polar vortex forms, and SWV can therefore
not penetrate the polar vortex during the first year. How-
ever, during the first DJF after eruption, SWV reaches high
southern latitudes, and an increased ozone hole therefore
forms during the following spring/summer (Fig. 5c).
4. Tropospheric and surface impact
For the remainder of this article, we will focus on the
seasonal impacts within the troposphere. The seasonal
timescale is where anomalies start to become significant,
while monthly data is too noisy for a meaningful analy-
sis, and 6- and 12-month averages start washing over the
seasonal signals (Fig. B2).
There is a well documented link between the strength
of the Antarctic polar vortex, the size and duration of
the ozone hole, and the phase of the Southern Annular
Mode (SAM) during the following summer (Baldwin and
Dunkerton 2001; Thompson et al. 2005; Lim et al. 2018).
5
1
10
100
pressure [hPa]
a) year 0 DJF b) year 0 MAM c) year 0 JJA d) year 0 SON
1
10
100
pressure [hPa]
e) year 1 DJF f) year 1 MAM g) year 1 JJA h) year 1 SON
1
10
100
pressure [hPa]
i) year 2 DJF j) year 2 MAM k) year 2 JJA l) year 2 SON
50 0 50
latitutde
1
10
100
pressure [hPa]
m) year 3 DJF
50 0 50
latitutde
n) year 3 MAM
50 0 50
latitutde
o) year 3 JJA
50 0 50
latitutde
p) year 3 SON
6
4
2
0
2
4
6
Q
1e 7
Fig. 3. Difference in (shading) water vapor and (contours) residual streamfunction between QBOW and QBOE ensemble members. Sign
convention is such that blue shading means QBOE members (as in 2022) show more SWV, and dashed contours in the SH mean weaker BDC for
QBOE members. Shading is only shown where statistically significant, contours are drawn in thick lines where significant. Contour interval is
108kg/s, and SWV is in kg/kg.
From this prior knowledge and the impact of the SWV
perturbation on the Antarctic ozone hole discussed above
(Fig. 5), we expect a positive SAM to develop two years
after eruption for QBOE members.
This is indeed the case, as seen two years after eruption
(2023/24 for HTHH) for QBOE members in the anomalous
geopotential height field at 300 hPa (Fig. 6a); although not
statistically significant throughout all longitudes, there is a
clear positive SAM signature with the characteristic dipole
structure between middle and high southern latitudes dur-
ing that year, and only for QBOE members.
As seen in Fig. 6, areas of statistically significant yearly
anomalies are quite small for most years, and we will now
focus on multi-year (years 3-7) mean anomalies, and justify
that choice after the fact. For now, suffice to say that
years 3-7 are the years with the most statistically significant
anomalies.
6
a)
0123456789
time [years since eruption]
3
2
1
0
1
2
3
mass anomaly [Tg]
Polar stratospheric cloud ice mass [Tg], QBOE
60-90S
60-90N
b)
0123456789
time [years since eruption]
3
2
1
0
1
2
3
mass anomaly [Tg]
Polar stratospheric cloud ice mass [Tg], QBOW
60-90S
60-90N
Fig. 4. Total stratospheric polar ice cloud mass anomalies for (green)
Antarctica and (purple) Arctica computed poleward of 60N/S for (a)
QBOE and (b) QBOW ensemble members. Lines denote ensemble
means, shading ±1𝜎, and lines are continuous if the ensemble mean is
statistically significant.
Averaged over those five years, there are substantial sur-
face temperature anomalies over the Northern Hemisphere
during winter and spring, reaching above 1.5°C over large
areas of North America in DJF and close to 1.5°C over
central Eurasia in MAM (Fig. 7a,b). There is also a cold
anomaly over Scandinavia in DJF, and the Arctic is anoma-
lously warm over most of the year, but most importantly
during SON (Fig. 7a-d). Note that our simulations do
not include interactive sea-ice, but these long-term Arc-
tic temperature anomalies should be expected to impact
sea-ice extent and concentrations, which would then also
positively feedback onto Arctic temperatures via albedo
feedback. In the Southern Hemisphere, the main surface
temperature anomalies are found over Australia, where the
local winters are almost 1°C cooler, and in the Amundsen
Sea, which is about 0.5°C warmer (Fig. 7c). In addition,
Western Australia also has slightly lower temperatures in
summer and autumn(Fig. 7a,b).
The main regions of significant rainfall anomalies are
located in the Pacific (most prominent in DJF) and in the
Indian Ocean in JJA (Fig. 7e-h). There is an indication
of a wave train emanating from the tropical Pacific north
and east towards North America in DJF (Fig. 7e), and a
similar wave train starting in the northern Indian Ocean
going south and east in JJA (Fig. 7g). In DJF and over
land, Europe and Western Australia receive slightly more
precipitation than usual, while the West Coast US is drier
than usual (Fig. 7e). In JJA, land anomalies include drier
summers over northern Eurasia, and wet anomalies along
China’s east coast and over northern Australia (Fig. 7g).
We now consider the year-by-year evolution of surface
anomalies within selected regions (rectangular boxes in
Fig. 7). Besides showing more temporal detail, this also
allows to justify our choice of year 3-7 averaging.
Fig. 8 confirms that the regional anomalies have a gen-
eral tendency to be largest during years 3-7. We also note
that by year 3, the differences between QBOE and QBOW
members are not significant (Fig. 3), and we use all mem-
bers from here on.
The North American surface temperature anomalies are
the highest, and gradually increase until they peak at around
1.8°C (area average) during year 4 (Dec 2025 - Feb 2026 for
HTHH) (Fig. 8a, green). The largest negative anomalies
are over Scandinavia and Australia, and they peak around
the same period (Fig. 8a, blue for Scandinavia, Fig. 8c,
blue for Australia). The Australian anomalies are also the
most persistent, with significant cooling from year 1 (JJA
2023) to 8 (JJA 2030).
For precipitation, there is a clear dipole developing be-
tween a drier northern tropical Central Pacific (Fig. 8e,
blue) and a wetter northern tropical Western Pacific
(Fig. 8e, orange) in DJF. In JJA, the Indian Ocean shows a
dipole between the northern and southern tropical regions
(Fig. 8f), with the northern edge around the Indian sub-
continent wetter and the tropical Indian Ocean just south
of the equator anomalously dry. These dipoles are similar
to the typical anomalies related to El Ni˜
no and the Indian
Ocean Dipole, even though our simulations do not include
interactive SSTs. Nevertheless, we will come back to the
global importance of these oceanic anomalies below.
5. Analysis
We now want to understand how those surface anoma-
lies develop, and focus on the two respective winter seasons
DJF and JJA, as those are the two seasons with the strongest
surface temperature and precipitation anomalies in the two
hemispheres. The simplest explanation for changes in sur-
face temperature forced by the SWV anomalies is a stronger
greenhouse effect causing surface heating. There is a clear
match between the regions of anomalous heat in the NH
7
a)
1234
time [years since eruption]
1
0
1
2
3
4
area [million km2]
mean ozone hole area [million km2], QBOE
hemi
SH Sep-Feb
b)
1234
time [years since eruption]
1
0
1
2
3
4
area [million km2]
mean ozone hole area [million km2], QBOW
hemi
SH Sep-Feb
c)
02468
time [years since eruption]
80
60
40
20
0
20
40
60
80
lat
Water vapor mass, QBOE
0
40
80
120
160
200
240
280
320
360
Q [mg/m2]
0
10
20
30
40
50
CLDICE [mg/m2]
d)
02468
time [years since eruption]
80
60
40
20
0
20
40
60
80
lat
Water vapor mass, QBOW
0
50
100
150
200
250
300
350
400
Q [mg/m2]
0
10
20
30
40
50
CLDICE [mg/m2]
Fig. 5. (top) Anomalous Antarctic ozone hole area, as defined by the area where total column ozone is below 220 DU for a) QBOE and b) QBOW
members. The anomalous area is averaged from September to February, as QBOE anomalies peak in December-February, while QBOW anomalies
peak around September. Error bars denote the 90% confidence interval. (bottom) Total anomalous stratospheric water vapor (red shading) and
cloud ice (blue shading) mass [mg/m2] above 100hPa, and total column ozone anomalies (black contours, -10 and -5 DU, solid where significant).
Shading is only shown where significant. MLS data until 11 July 2023 in cyan and a contour interval of 250 mg/m2. c) for QBOE members, d) for
QBOW members.
during DJF as shown in Fig. 7, and longwave forcing, as
those regions receive anomalous downwelling longwave
radiation at the surface during that season (Fig. 9a; boxes
as in Fig. 7). A similar conclusion can be drawn regarding
Arctic warming in JJA (Fig. 9b).
The strong regional character of the heating and cool-
ing patterns suggest that cloud and dynamical effects are
important in setting up the surface response to the SWV
perturbations. Cloud feedbacks play a dominant role in
setting up the cooling over Australia and warming over the
Amundsen Sea during years 3-7, as there are increases in
cloud fraction in these regions (Fig. 9cd). The cooling
over Australia can therefore be linked to shortwave cloud
forcing (Fig. 9gh) which matches the patterns of surface
cooling (Fig. 7), while the warming in the Amundsen Sea
is at least partically due to increased longwave cloud forc-
ing (Fig. 9f). Cloud forcing also matches the precipitation
response in the extratropical Pacific, where decreases in
precipitation are accompanied by fewer clouds, resulting
in positive cloud shortwave (more sunlight reaches the sur-
face) and negative cloud longwave forcing (less surface
longwave emission is being blocked; Fig. 9ceg). Cloud
anomalies are also consistent with the increase in precip-
itation over Europe in DJF (Fig. 7b), as the same region
shows increased cloud cover and cloud forcing (Fig. 9deg).
Interestingly, there are only minor cloud effects in the re-
gions of land surface warming in the NH described above.
Furthermore, cloud forcing even counteracts the green-
house heating of SWV over the arctic (Fig. 9h, green area
in the Arctic), and there is less cloud cover accompanied
by net positive cloud forcing anomalies over the east coast
of China and southern parts of North America ([Fig. 9ceg),
where the surface cools rather than warms. These obser-
vations suggest that dynamical effects are also important.
To assess the dynamical feedbacks to the imposed SWV
forcing, recall that there are strong precipitation anomalies
8
A)
a) U300, DJF, year 1 b) U300, DJF, year 2
c) U300, DJF, year 3 d) U300, DJF, year 4
4
3
2
1
0
1
2
3
4
B)
a) U300, DJF, year 1 b) U300, DJF, year 2
c) U300, DJF, year 3 d) U300, DJF, year 4
4
3
2
1
0
1
2
3
4
Fig. 6. 300 hPa zonal wind anomalies for (A) QBOE and (B) QBOW members [m/s].
in the Pacific in DJF and the Indian Ocean in JJA (Fig. 7).
The northern Pacific DJF anomalies suggest a wave train
originating in the tropical Central Pacific heading north
east towards North America (Fig. 7e). A clear wave pattern
originating over the Pacific can be seen in 350K potential
vorticity anomalies and wave activity flux (Fig. 10). A
similar pattern is also seen in other dynamical variables,
such as 300 hPa meridional and zonal wind or geopotential
height (Fig. B3). Thus, in DJF a stationary wave pattern
is established over the NH extratropics, with anticyclonic
PV anomalies over regions showing warming, and cyclonic
anomalies where the surface cools. Another consequence
of this global wave train is a strengthening of the zonal jet
over Europe, which accompanies the increased cloud cover
and rainfall described earlier (Fig. B3C). In contrast, there
is no clear circumglobal wave pattern in the SH, nor in JJA.
However, there is an increase in wave activity flux towards
the Amundsen Sea in JJA (Fig. 10b) which coincides with
meridional wind anomalies (Fig. B3B), increased cloud
cover (Fig. 9d) and surface warming (Fig. 7c), confirming
the importance of dynamical tropospheric feedbacks in
setting up surface anomalies globally.
9
a) TS anomalies, DJF, years 3-7 b) TS anomalies, MAM, years 3-7
c) TS anomalies, JJA, years 3-7 d) TS anomalies, SON, years 3-7
1.5
1.0
0.5
0.0
0.5
1.0
1.5
T2m [K]
e) P anomalies, DJF, years 3-7 f) P anomalies, MAM, years 3-7
g) P anomalies, JJA, years 3-7 h) P anomalies, SON, years 3-7
1.00
0.75
0.50
0.25
0.00
0.25
0.50
0.75
1.00
Q [mm/day]
Fig. 7. Year 3-7 seasonal mean anomalies for (top) 2m temperature and (bottom) precipitation, during (a,e) DJF, (b,f) MAM, (c,g) JJA, and (d,h)
SON. Only statistically significant anomalies are shown.
Idealised simulations
In order to assess possible impacts of the SWV anoma-
lies on SSTs, we ran additional simulations with the Model
of an idealized Moist Atmosphere (MiMA, Jucker and Ger-
ber 2017), as described in Section 2b. As a reminder, these
simulations are forced by WACCM SWV and ozone in the
radiative transfer calculations, and include an interactive
mixed layer ocean. Thus, we can estimate the radiative im-
pact of the SWV anomaly on surface temperatures around
the globe (excluding cloud feedbacks). We note that it was
noted by Jucker (2019) and Garfinkel et al. (2020b) that
due to various simplifications in the model, the seasonal
cycle at the surface is somewhat lagging compared to com-
prehensive models, and therefore some features appearing
in DJF in WACCM may leak into MAM in MiMA, and
similarly for JJA and SON.
10
a)
012345678910
time [years since eruption]
2.0
1.5
1.0
0.5
0.0
0.5
1.0
1.5
2.0
T2m [K]
TS anomalies, DJF
region
Scandinavia DJF
Eurasia DJF
NAmerica DJF
b)
012345678910
time [years since eruption]
2.0
1.5
1.0
0.5
0.0
0.5
1.0
1.5
2.0
T2m [K]
TS anomalies, MAM
region
Siberia MAM
Arctic MAM
c)
012345678910
time [years since eruption]
2.0
1.5
1.0
0.5
0.0
0.5
1.0
1.5
2.0
T2m [K]
TS anomalies, JJA
region
Australia JJA
Arctic JJA
Amundsen JJA
d)
012345678910
time [years since eruption]
2.0
1.5
1.0
0.5
0.0
0.5
1.0
1.5
2.0
T2m [K]
TS anomalies, SON
region
NAsia SON
Arctic SON
e)
012345678910
time [years since eruption]
1.5
1.0
0.5
0.0
0.5
1.0
1.5
Q [mm/day]
P anomalies, DJF
region
Pacific DJF
MC DJF
f)
012345678910
time [years since eruption]
1.5
1.0
0.5
0.0
0.5
1.0
1.5
Q [mm/day]
P anomalies, JJA
region
ION JJA
IOS JJA
Fig. 8. Year-by-year seasonal anomalies for the regions indicated by boxes in Fig. 7 for (a-d) surface temperature and (e,f) precipitation, and
(a,e) DJF, (b) MAM, (c,f) JJA, (d) SON. The red horizontal lines denote the averaging period for years 3-7. All members are used, and error bars
denote the 90% confidence intervals. Refer to legend for region and season. Colors are the same as the corresponding boxes in Fig. 7.
Surface temperature anomalies from these simulations
show a clear picture of warming winter hemisphere
landmasses, and cooling summer hemisphere landmasses
(Fig. 11). We note that additional simulations where only
ozone or only SWV anomalies where added suggest that
the surface temperature signal is dominated by SWV, and
ozone has only minor impact in these simulations (Figs. B5
and B6). The MiMA simulations confirm the robustness of
the SWV-induced wave structure in the Pacific (Fig. 11ef),
and they produce tropical surface temperature anomalies
consistent with an El Ni˜
no-like pattern (Fig. 11a-d). This
heating is produced by the zonally asymmetric distribution
of SWV in the tropics, and consistent with the increased
surface downward longwave flux over the tropical Pacific
in WACCM (Fig. 9a). Thus, it is possible that the SWV
forcing from the eruption would favor a positive phase of
ENSO on a multi-year timescale, but further work is re-
quired to confirm this, in particular with a model including
fully interactive ocean and cloud feedbacks.
On the other hand, these simulations confirm the im-
portance of cloud feedbacks, as the regions observed to
suffer cooling in winter in the WACCM simulations (Scan-
dinavia, Australia) consistently warm together with the rest
of the winter hemisphere in the cloud-less MiMA simula-
tions (Fig. 11a-d). The cooling of the summer hemisphere
land masses in MiMA is consistent with the widely absent
warming in WACCM in that season, and we attribute it to
increased shortwave absorption in the stratosphere due to
enhanced SWV levels.
11
a) DLS anomalies, DJF, years 3-7 b) DLS anomalies, JJA, years 3-7
4
2
0
2
4
DLS [W/m2]
c) CLDTOT anomalies, DJF, years 3-7 d) CLDTOT anomalies, JJA, years 3-7
0.04
0.02
0.00
0.02
0.04
CLD []
e) LWCF anomalies, DJF, years 3-7 f) LWCF anomalies, JJA, years 3-7
4
2
0
2
4
LWCF [W/m2]
g) SWCF anomalies, DJF, years 3-7 h) SWCF anomalies, JJA, years 3-7
4
2
0
2
4
SWCF [W/m2]
Fig. 9. Same as Fig. 7, but for DJF (left) and JJA (right) only, and for (a,b) downwelling longwave radiation at the surface (DLS), (c,d) total
cloud fraction, (e,f) longwave cloud forcing (LWCF), and (g,h) shortwave cloud forcing (SWCF). The same boxes as in Fig. 7a,d are overlaid for
easier comparison.
The MiMA simulations also show circumglobal wave
trains, although their structure is qualitatively different to
the wave trains discussed for the WACCM simulations
(Fig. B4). This is to be expected, as the warming and
cooling in each hemisphere is more zonally symmetric in
MiMA, and missing cloud feedbacks and simplified con-
vection can be expected to have an influence on storm track
behavior (Ceppi and Hartmann 2016; Fuchs et al. 2022).
Even so, the MiMA simulations provide evidence of the
robustness of the surface impacts from our SWV perturba-
tion simulations with WACCM discussed above.
12
7.999999999999999e-14
20
m2
s2
a) 350K PV & 200hPa WAF, DJF
7.999999999999999e-14
b) 350K PV & 200hPa WAF, JJA
0.20
0.16
0.12
0.08
0.04
0.02
0.06
0.10
0.14
0.18
Fig. 10. Potential vorticity at the 350K isentrope (shading, in PVU) and horizontal wave activity flux (arrows) for (a) DJF and (b) JJA. Only
statistically significant values are shown, and arrows within 15S-15N and outside 80S/N are masked for clarity. Both quantities are averaged from
year 3 to 7 as before.
6. Summary and Conclusions
Our analysis shows that volcanic eruptions like that of
Hunga-Tonga Hunga Ha’apai (HTHH) can have significant
impacts on the climate system. We find that the ozone
hole area is larger during the second spring/summer after
eruption, and a positive Southern Annular Mode develops
over the following austral summer. However, the impact
on the Antarctic ozone hole depends on the phase of the
Quasi-Biennial Oscillation (QBO) around the time of the
eruption, with the easterly phase (QBOE) showing more
impact than the westerly phase (QBOW). For QBOE (as
was the case during the HTHH eruption), we obtain an
ensemble mean, September to February mean ozone hole
area increase of more than 2 million km2. The effect of the
initial QBO phase vanishes by year 3.
Our WACCM simulations reveal significant surface tem-
perature and precipitation anomalies globally which peak
around years 3-7 after eruption, e.g., years 2025-2029 for
HTHH, but can already appear earlier. The Northern
Hemisphere experiences substantial surface temperature
anomalies during winter and spring, with strong warming
over North America and central Eurasia, and cooling over
Scandinavia. The Arctic also exhibits anomalous warmth
throughout the year, particularly during the September-
November period. In the Southern Hemisphere, cool
anomalies are observed over Australia, while the Amund-
sen Sea is warmer than usual.
Precipitation anomalies are found in the Pacific and In-
dian Oceans, with indications of wave trains originating
from the tropical Pacific and Indian Ocean. Europe and
Australia receive increased precipitation in winter, while
the West Coast US is drier than usual. In summer, north-
ern Eurasia experiences drier conditions, while China’s
east coast and western Australia see wet anomalies.
Through further analysis, we determine that cloud and
dynamical effects play important roles in setting up the sur-
face response to the SWV perturbations. Cloud feedbacks
contribute to the cooling over Australia and the warming
in the Amundsen Sea, while a circumglobal wave train
in northern winter midlatitudes contributes to temperature
anomalies of both signs across all longitudes, and the in-
creased rainfall over Europe.
Additional simulations with an idealized model with in-
teractive mixed layer ocean indicate that the Pacific anoma-
lies seen in WACCM might be accompanied by a positive
tendency of the El Ni˜
no Southern Oscillation, but further
work needs to be done to confirm this link.
13
a) TS anomalies, DJF, years 3-7 b) TS anomalies, MAM, years 3-7
c) TS anomalies, JJA, years 3-7 d) TS anomalies, SON, years 3-7
1.5
1.0
0.5
0.0
0.5
1.0
1.5
T2m [K]
e) P anomalies, DJF, years 3-7 f) P anomalies, MAM, years 3-7
g) P anomalies, JJA, years 3-7 h) P anomalies, SON, years 3-7
1.00
0.75
0.50
0.25
0.00
0.25
0.50
0.75
1.00
Q [mm/day]
Fig. 11. As Fig. 7, but for MiMA simulations: (a-d) Surface temperature and (e-h) precipitation anomalies.
14
Acknowledgments. MJ was supported by the ARC Cen-
tre of Excellence for Climate Extremes which is sup-
ported by the Australian Research Council via grant
CE170100023. This research was undertaken with the
assistance of resources from the National Computational
Infrastructure (NCI Australia), an NCRIS enabled capabil-
ity supported by the Australian Government. Additional
computing resources were provided through the ARC LIEF
grant LE200100040.
Data availability statement. All analysis scripts will
be made on github.com upon acceptance. Similarly, we
will make the post-processed simulation data as well any
configuration files for the models to duplicate our simula-
tions openly available via zenodo.com. Any direct access
to full simulation data can be arranged by contacting the
authors.
15
APPENDIX A
NASA Aura Microwave Limb Sounder (MLS) data
a. Data retrieval
The spatiotemporal evolution of the stratospheric water
vapor cloud is examined with data from the Microwave
Limb Sounder (MLS) instrument onboard the NASA Aura
satellite. Aura was launched in 2004 with a mission to
obtain measurements of ozone, aerosol and key gases
throughout the atmosphere. The satellite is in a sun-
synchronous, near polar orbit at 705 km above the surface
with the ascending node in daylight and an equator cross-
ing time at approximately 1345 LT. It constitutes part of
NASA’s ’A-Train’ constellation of satellites.
For this study, Version 5 (v5) Level 2 water vapor mea-
surements from MLS are used. For the calculations here,
the analysis is focused on the data acquired between 100
hPa and 0.1 hPa pressure levels. Monthly Level 3 data
from 2005 through 2021 are also used to compute the cli-
matology. Livesey et al. (2022) provide an assessment of
the characteristics and quality of the dataset. All MLS data
are acquired from the NASA Earthdata Search web portal
at https://search.earthdata.nasa.gov/search.
The MLS data users guide (Livesey et al. 2022) outlines
screening procedures to be used when examining the water
vapor data using the multiple quality control (QC) indica-
tors included in the L2 data files. These include criteria for
the precision (>0), quality (>0.7), status (even number)
and convergence fields (<2). Additionally, data at pres-
sures >316 hPa and <0.0001 hPa should not be used, and
values with low values of VMR (<0.1 ppmv) should be
screened out.
In this work, we apply two versions of the QC screen-
ing, one with full QC (QC-on) and one where only the
pressure and low value screenings are applied (QC-off).
The calculated values from these two versions only dif-
fer significantly from 15 Jan to 9 Feb 2022. While use
of the QC-off fields can introduce uncertainty and is not
recommended, it is required to effectively characterize the
water vapor plume during the initial weeks following the
eruption.
Processed L3 H2O files with precomputed monthly
zonal means from 2005 through 2021 are used to derive an
historical climatology for the MLS dataset. This monthly
climatology is interpolated to a daily time scale through
polynomial interpolation of the monthly values using the
built-in IDL function ’spline’ with a ’tension’ of 0.5. From
this procedure, a smooth, physically plausible interpolation
is produced.
Several different approaches are used to examine the
HTHH water vapor cloud. These are:
•Individual profiles and averages. Native measure-
ments in unitless volume mixing ratio versus pressure
for individual profiles is examined, as well as averages
of these profiles
•Zonal Means. Zonal means are computed with a 4°
latitude resolution by averaging all individual VMR
profiles in a given latitude bin. This is done at each
pressure level. This is consistent with the standard
resolution in the Level 3 files.
Two primary calculations are made with the MLS data:
1.) estimates of stratospheric integrated water vapor
(SIWV) and 2.) total stratospheric water vapor (TWV)
between the 100 hPa and 1 hPa levels and its anomaly.
These are calculated as follows:
•Stratospheric integrated water vapor: Water vapor is
vertically integrated between 100 and 1 hPa,
SIWV =−𝜖
𝑔∫1ℎ𝑃 𝑎
100 ℎ𝑃 𝑎
VMRd𝑝,
where 𝑉 𝑀 𝑅 is the (unitless) volume mixing ratio, 𝑝
the pressure, 𝑔the acceleration due to gravity and
𝜖=𝑀𝐻2𝑂/𝑀𝑎𝑖𝑟 ≈0.622 is the ration of the molec-
ular weights of water vapor and dry air. Units are
converted to g m−2, which is equivalent to microns
of ’precipitable water’. Typical values range between
2 and 3 g m−2. For reference, a uniform increase of
1 ppmv in this calculation is equivalent to an SIWV
increase of approximately 0.6 g m-2. This quantity
is reported in the zonal mean data and serves as the
basis of the total water vapor calculation below.
•Total water vapor and anomaly: Total stratospheric
water vapor (TWV) is computed as
TWV =𝐴· ⟨SIWV⟩cos 𝜙,
where angle brackets represent the zonal mean for
SIWV, 𝜙is the latitude, the overbar the average and
𝐴is the area of the globe between 80°N and 80°S
(5.02 x1014 m2). This is multiplied by 10−9to con-
vert into teragrams (Tg) of water vapor. This is done
for individual daily zonal means and for the monthly
climatological values (that are subsequently interpo-
lated into daily values TWV𝐶). Using these values,
daily anomalies of TWV (TWVA) are estimated as
TWVA = TWV - TWV𝐶.
Using the monthly climatological profiles perturbed by
normally distributed errors reflecting the accuracy and pre-
cision uncertainties noted by Livesey et al. (2022) with
a Monte Carlo approach, the uncertainty estimate in the
TWV calculation is ±4 Tg.
b. Determining the vertical profile of HTHH water vapor
The vertical profile of the water vapor anomaly in its
initial stages is estimated from the initial overpasses of
16
Fig. A1. Vertical Profiles of water vapor from three MLS overpasses.
See text for details.
the HTHH WV cloud. The aim here is to get the relative
heights and proportions of mass in the cloud for input into
the numerical model.
Three overpasses from 16 January 2022 are examined,
with start times of the water vapor cloud encounters at
16/023552 UTC, 16/041445 UTC and 16/150637 UTC,
22 to 36 hours after the eruption with a total of 28 profiles.
Mill´
an et al. (2022) confirmed through back-trajectory
analysis that these strong H2O signals originated from
HTHH. The vertical profiles from these overpasses are
presented in Fig. A1, with the profiles color-coded by over-
pass. The inset map shows the location of the overpasses
relative to the volcano and the geographic features of the
region.
These profiles all feature water vapor values higher than
the typical background values of 3-5 ppmv. Profiles fur-
ther west are higher in the atmosphere with green profiles
showing two peaks in water vapor VMR at 0.8 and 4 hPa
pressure levels, with peak values of VMR approaching 100
ppmv in the lower peak and 10-20 ppmv in the upper. Red
and blue profiles show broad peaks in the lower strato-
sphere, with peaks at approximately 20 hPa and 30 hPa,
respectively. Peak values for both sets of profiles approach
150 ppmv. The red and blue ones are generally rejected
by the QC process, producing too large values of ’conver-
gence’. For all three sets of profiles, the average profile is
plotted in Figure ST1 as the bold line of the appropriate
color. Mill´
an et al. (2022) examined of profiles using v4 of
the MLS data and reported values on 16 Jan in that paper
that were had considerably higher VMRs than shown here,
noting that v5 data does a poorer version of retrieving H2O
data in regions of ’extremely enhanced humidity’. This
is also consistent with a generally lower values of H2O
retrieved by v5 throughout the atmosphere as noted in the
data users guideLivesey et al. (2022).
V¨
omel et al. (2022) examined the HTHH H2O anomaly
using radiosonde data, showing strongly enhanced humid-
ity values between 20 and 30 km (approximately 90 to 10
hPa), with values at individual levels approaching 1000
ppmv in some instances. The heights where these anoma-
lies are found in the sonde data agree very well the heights
of the water vapor anomaly in the MLS data, although the
values of the anomaly are considerably lower in MLS. This
latter observation is not surprising, given the differences
in sampling between the measurements; MLS represents
a vertically smoothed observations while radiosondes are
high resolution, nearly ’instantaneous’ values. This favor-
able comparison with independent observations supports
the idea that these signals in the MLS are genuine, despite
not passing the QC and that they are suitable to use in
further analysis.
Using the average profiles from each of the three over-
passes, the mean vertical profile of H2O from the HTHH
eruption can be determined. This is done by merging the
three mean profiles; above 10 hPa, the mean of the green
profiles is used; below 10 hPa, an average of the mean red
and blue profiles is used. From this profile, the SIWV is
computed over each set of 3 non-overlapping levels and
divided by the total SIWV over all levels to obtain a frac-
tion of total mass with height, yielding a scalable profile
of the vertical distribution of H2O mass in the eruption.
This is shown in Fig. 1b. Two primary injection heights
are identified in the profile, one near 25 hPa and the other
near 4 hPa. A smaller third is also found near 0.8 hPa but
contains only a small fraction of the mass. Approximately
94% of the total mass is injected below 10 hPa, generally at
heights between 22 and 27 km. Most of the remaining 6%
of the mass above 10 hPa is found between approximately
36 and 38 km altitude.
c. Temporal evolution of total water vapor
Fig. A2 shows the evolution of TWV between 26 Nov
2021 and 31 Mar 2022. This includes the time before the
first eruption of HTHH on 19 December 2021 and the ini-
tial residence time of the H2O cloud in the stratosphere.
Time series for both QC-on and QC-off realizations are
shown, along with the MLS-based climatology. Prior to
15 January 2022 and after 9 February 2022, the two series
are effectively identical; as noted by Mill´
an et al. (2022)
the difference lies in the application of the QC during the
initial weeks of the eruption. The QC-off realization dis-
plays a sharp discontinuity at the time of the 3rd eruption,
although still does not reach its highest value until 2 weeks
past the eruption. The QC-on curve shows a much more
gradual increase after the eruption that is physically in-
consistent with the idea of a short, sharp eruption. RMS
uncertainty in the calculation of TWV is approximately ±4
Tg, corresponding to random errors of approximately 10%.
Other features of note in Fig. A2 are:
17
Fig. A2. Time series of QC-off (blue) and QC-on (red) time series
of v5 MLS H2O, v5 2005-2021 climatology shown as green line. Times
of eruption 1 (19 Dec 21) and eruption 3 (15 Jan 2022) highlighted.
•The substantial pre-existing anomaly in TWV of 40-
50 Tg that was present prior to the eruption. Analysis
of earlier data (not shown) indicates that anomalies
were near zero in early June 2019 and began to rise
more sharply in January 2020 and have slowly in-
creased since that time.
•An inflection point in the TWV time series that is
coincident with the first eruption of HTHH on 19
Dec 2021. Rather than a distinct increase in TWV, a
subtle shift in the rate of decline is noted, from more
quickly than climatology before to slower afterwards.
This is perhaps suggestive of a very small water vapor
emission from the first eruption or another (unknown)
change in the sink of stratospheric water vapor at that
time
d. Evolution of anomalies and estimated HTHH H2O emis-
sion
Fig. A3 depicts the TWV anomaly for the period from
25 November 2021 until 11 July 2023 using the QC-off
realization data. AS before, the sharp discontinuity at the
time of the eruption is readily apparent; peak value of the
anomaly is 150 Tg approximately two weeks after the
eruption. The anomaly field is particularly noisy around
the time of the peak. Also apparent are the pre-existing
anomaly and the inflection point coincident with the first
eruption. In the 5 days before the eruption, the average
anomaly is 46.8 Tg.
Because of potential errors introduced by the QC-off
processing and the slow evolution of peak anomalies, it is
difficult to directly estimate the anomaly associated with
HTHH. This is addressed here through curve fitting and
Fig. A3. Time series of anomalous global stratospheric water vapor.
Orange dash line represents the exponential fit. See text for details.
extrapolation. The water vapor anomaly is assumed to de-
cay exponentially following 𝐴=𝐴0𝑒𝑟 𝑡 , where 𝐴0is the
initial anomaly, 𝑟is the rate of change constant, presum-
ably negative, and 𝑡is the time in days. The rate of change
constant is estimated by calculating the regression coeffi-
cient of the logarithm of the anomaly data from 10 Feb-12
Nov 2022, the period over which the QC-on and QC-off
data are equivalent. The best estimate is 𝑟=-3.67x10−4
±2.10x10−5day−1(2-s uncertainty), a decay rate of ap-
proximately -1.1% per month. The initial value is then
determined by calculating the estimated 𝐴0value from all
the data in the regression and taking the mean. The pro-
vides an estimate of 𝐴0=144.6 ±2.0 Tg (1-sd uncertainty);
estimates range from 139.3 to 153.8 Tg. There is some sen-
sitivity to the choice of end time, and cutting off the data
in May or June gives slightly higher estimates of both vari-
ables (e.g. A0∼148 Tg, 𝑟= -2.1% month) but results in
a significantly worse fit for the later data. Note that the
slightly different values of 𝑟reported here compared to
those given in the main text arise from a fit of daily data
here versus monthly data in the main text.
Using these numbers, we can estimate the total H2O
anomaly from HTHH eruption as 144.6 – 46.8 = 97.8
±6 Tg, approximately 11% of the TWV for the season
as computed here. This estimate lies in the middle of
the radiosonde estimates of V¨
omel et al. (2022)V¨
omel
et al. (2022) (i.e. ’at least’ 50 Tg reported) and the MLS
estimates by Mill´
an et al. (2022)Mill´
an et al. (2022) (i.e.
146 Tg). We note that without considering the pre-existing
anomaly, our estimate of matches quite closely with the
latter result, as might be expected. It is not clear from the
text if that was considered in their result. Alternately, the
differences could be the result of using v5 versus v4 of the
MLS data.
From this estimate of 𝐴0and the model of exponential
decay, it is noted that estimates of H2O anomaly directly
18
from MLS remain biased quite low despite the use of QC-
off data. For example, on 16 January, the observed anomaly
value is 98.3 Tg, for an HTHH anomaly estimate of 51.5
Tg, approximately half of the ’true’ estimate. This bias has
at least two sources:
•The previously mentioned v4 vs v5 differences, in
particular those relating to the ’pointing error’(Mill´
an
et al. 2022). The earlier version of the algorithm
simply performs better in high humidity situations
than the current version.
•The TWV methodology here is reliant on the zonal
mean calculation which is strongly impacted by the
highly skewed zonal distribution of H2O before the
global dispersion of the H2O cloud. IN the first few
days in particular, H2O is concentrated in a small area
of the globe, while the rest has near-climatological
values. This has the effect of underestimating the
zonal mean, which feeds into the TWV calculation
and results in an underestimate of that value.
e. Anomaly Lifetime
The lifetime of the HTHH anomaly can be estimated
using the model of exponential decay. Using the estimated
value of 𝑟(-3.67 x10−4day−1), the estimated time to decay
back to the initial anomaly of 46.8 Tg is 3074 days; for
±2s, the values range from 2907-3260 days. Translating
to years, the range is about 8-9 years, meaning that the
water vapor anomaly should no longer be elevated (above
its 10-14 January 2022 value) at some point during 2030.
This calculation assumes that the rate from mid-February
to mid-November is the correct one and that no other sinks
of stratospheric H2O occur. As noted earlier, sensitivity to
the choice of endpoints is important.
19
APPENDIX B
Additional figures
q-ux, bench_SH
100
50
0
50
100
Fig. B1. Ocean surface heat fluxes [W/m2] for MiMA, which includes
the major ocean currents plus a realistic ITCZ and SPCZ.
20
A)
0000 0003 0006 0009
6
4
2
0
2
4
TREFHT
a) Scandinavia
0000 0003 0006 0009
7.5
5.0
2.5
0.0
2.5
5.0
TREFHT
b) Eurasia
0000 0003 0006 0009
time
6
4
2
0
2
4
TREFHT
c) NAmerica
0000 0003 0006 0009
time
7.5
5.0
2.5
0.0
2.5
5.0
TREFHT
d) Australia
T2m [K], 1-month rolling mean
B)
0000 0003 0006 0009
5.0
2.5
0.0
2.5
5.0
TREFHT
a) Scandinavia
0000 0003 0006 0009
4
2
0
2
4
TREFHT
b) Eurasia
0000 0003 0006 0009
time
4
2
0
2
4
TREFHT
c) NAmerica
0000 0003 0006 0009
time
6
4
2
0
2
4
TREFHT
d) Australia
T2m [K], 3-month rolling mean
C)
0000 0003 0006 0009
6
4
2
0
2
4
TREFHT
a) Scandinavia
0000 0003 0006 0009
4
2
0
2
4
6
TREFHT
b) Eurasia
0000 0003 0006 0009
time
4
2
0
2
4
TREFHT
c) NAmerica
0000 0003 0006 0009
time
5.0
2.5
0.0
2.5
5.0
TREFHT
d) Australia
T2m [K], 6-month rolling mean
D)
0003 0006 0009
6
4
2
0
2
4
TREFHT
a) Scandinavia
0003 0006 0009
2
0
2
4
TREFHT
b) Eurasia
0003 0006 0009
time
4
2
0
2
4
TREFHT
c) NAmerica
0003 0006 0009
time
5.0
2.5
0.0
2.5
5.0
TREFHT
d) Australia
T2m [K], 12-month rolling mean
Fig. B2. Spaghetti plots of regional rolling means of surface temperature anomalies over (blue) Scandinavia, (orange) Eurasia, (green) North
America, and (red) Australia from Fig. 7. Shown are all members, and the ensemble means with a thick line. A continuous thick line means the
ensemble mean is significantly different from zero. Rolling means are over (A) 1 month, (B) 3 months, (C) 6 months, and (D) 12 months.
21
A)
a)a) Z300, DJF, years 3-7 b)b) Z300, JJA, years 3-7
36
24
12
0
12
24
36
B)
a)a) V300, DJF, years 3-7 b)b) V300, JJA, years 3-7
2.6
1.8
1.0
0.2
0.6
1.4
2.2
3.0
C)
a)a) U300, DJF, years 3-7 b)b) U300, JJA, years 3-7
2.6
1.8
1.0
0.2
0.6
1.4
2.2
3.0
Fig. B3. Year 3-7 seasonal means of 300hPa (A) zonal wind [m/s], (B) meridional wind [m/s], and (C) geopotential height [m] anomalies. All
variables confirm the wave structure discussed in Fig. 10.
22
a)a)a)a) V300, DJF, years 3-7 b)b)b)b) V300, MAM, years 3-7
c)c)c)c) V300, JJA, years 3-7 d)d)d)d) V300, SON, years 3-7
2.6
1.8
1.0
0.2
0.6
1.4
2.2
3.0
Fig. B4. Meridional wind anomalies [m/s] similar to Fig. B3B, but for MiMA simulations.
23
a) TS anomalies, DJF, years 3-7 b) TS anomalies, MAM, years 3-7
c) TS anomalies, JJA, years 3-7 d) TS anomalies, SON, years 3-7
1.5
1.0
0.5
0.0
0.5
1.0
1.5
T2m [K]
e) P anomalies, DJF, years 3-7 f) P anomalies, MAM, years 3-7
g) P anomalies, JJA, years 3-7 h) P anomalies, SON, years 3-7
1.00
0.75
0.50
0.25
0.00
0.25
0.50
0.75
1.00
Q [mm/day]
Fig. B5. MiMA simulations similar to those discussed in Fig. 11, but with WACCM anomalies of SWV only.
24
a) TS anomalies, DJF, years 3-7 b) TS anomalies, MAM, years 3-7
c) TS anomalies, JJA, years 3-7 d) TS anomalies, SON, years 3-7
1.5
1.0
0.5
0.0
0.5
1.0
1.5
T2m [K]
e) P anomalies, DJF, years 3-7 f) P anomalies, MAM, years 3-7
g) P anomalies, JJA, years 3-7 h) P anomalies, SON, years 3-7
1.00
0.75
0.50
0.25
0.00
0.25
0.50
0.75
1.00
Q [mm/day]
Fig. B6. MiMA simulations similar to those discussed in Fig. 11, but with WACCM anomalies of ozone only.
25
References
Baldwin, M. P., and T. J. Dunkerton, 2001: Stratospheric harbingers of
anomalous weather regimes. Science (New York, N.Y.),294 (5542),
581–4, https://doi.org/10.1126/science.1063315, URL http://www.
ncbi.nlm.nih.gov/pubmed/11641495.
Baldwin, M. P., and Coauthors, 2001: The quasi-biennial os-
cillation. Reviews of Geophysics,39 (2), 179, https://doi.org/
10.1029/1999RG000073, URL http://doi.wiley.com/10.
1029/1999RG000073http://www.agu.org/pubs/crossref/2001/
1999RG000073.shtml.
Brewer, A., 1949: Evidence for a world circulation provided by
the measurements of helium and water vapour distribution in the
stratosphere. Quarterly Journal of the Royal Meteorological Soci-
ety,75 (326), 351–363, https://doi.org/10.1002/qj.49707532601,
URL http://onlinelibrary.wiley.com/doi/10.1002/qj.49707532603/
abstracthttp://doi.wiley.com/10.1002/qj.49707532601.
Carn, S. A., N. A. Krotkov, B. L. Fisher, and C. Li, 2022: Out of
the blue: Volcanic SO2 emissions during the 2021–2022 erup-
tions of Hunga Tonga—Hunga Ha’apai (Tonga). Frontiers in Earth
Science,10, https://doi.org/10.3389/feart.2022.976962, URL https:
//www.frontiersin.org/articles/10.3389/feart.2022.976962/full.
Carr, J. L., A. Horv´
ath, D. L. Wu, and M. D. Friberg, 2022: Stereo Plume
Height and Motion Retrievals for the Record-Setting Hunga Tonga-
Hunga Ha’apai Eruption of 15 January 2022. Geophysical Research
Letters,49 (9), https://doi.org/10.1029/2022GL098131, URL https:
//onlinelibrary.wiley.com/doi/10.1029/2022GL098131.
Ceppi, P., and D. L. Hartmann, 2016: Clouds and the At-
mospheric Circulation Response to Warming. Journal of Cli-
mate,29 (2), 783–799, https://doi.org/10.1175/JCLI-D-15-0394.1,
URL http://journals.ametsoc.org/doi/abs/10.1175/JCLI-D-15- 0394.
1http://journals.ametsoc.org/doi/10.1175/JCLI-D-15- 0394.1.
de F. Forster, P. M., and K. P. Shine, 1999: Stratospheric wa-
ter vapour changes as a possible contributor to observed strato-
spheric cooling. Geophysical Research Letters,26 (21), 3309–3312,
https://doi.org/10.1029/1999GL010487, URL http://doi.wiley.com/
10.1029/1999GL010487.
Dessler, A. E., M. R. Schoeberl, T. Wang, S. M. Davis, and K. H.
Rosenlof, 2013: Stratospheric water vapor feedback. Proceedings
of the National Academy of Sciences,110 (45), 18 087–18 091,
https://doi.org/10.1073/pnas.1310344110, URL https://pnas.org/doi/
full/10.1073/pnas.1310344110.
Dobson, G. M. B., 1956: Origin and Distribution of the Polyatomic
Molecules in the Atmosphere. Proceedings of the Royal Society
A: Mathematical, Physical and Engineering Sciences,236 (1205),
187–193, https://doi.org/10.1098/rspa.1956.0127, URL http://rspa.
royalsocietypublishing.org/cgi/doi/10.1098/rspa.1956.0127.
Dunkerton, T. J., 1990: Annual Variation of Deseasonalized Mean Flow
Acceleration in the Equatorial LowerStratosphere. Journal of the Me-
teorological Society of Japan. Ser. II,68 (4), 499–508, https://doi.org/
10.2151/jmsj1965.68.4{ \ }499, URL https://www.jstage.jst.go.jp/
article/jmsj1965/68/4/68 4 499/ article.
Eyring, V., T. G. Shepherd, and D. W. Waugh, 2010: SPARC
CCMVal Report on the Evaluation of Chemistry-Climate Models.
Tech. rep., SPARC, 426 pp. pp. URL http://www.sparc-climate.org/
publications/sparc-reports/.
Frierson, D. M. W., I. M. Held, and P. Zurita-Gotor, 2007: A
Gray-Radiation Aquaplanet Moist GCM. Part II: Energy Trans-
ports in Altered Climates. Journal of the Atmospheric Sciences,
64 (5), 1680–1693, https://doi.org/10.1175/JAS3913.1, URL http:
//journals.ametsoc.org/doi/abs/10.1175/JAS3913.1.
Fuchs, D., S. C. Sherwood, D. Waugh, V. Dixit, M. H. England, Y.-
L. Hwong, and O. Geoffroy, 2022: Midlatitude jet position spread
linked to atmospheric convective types. Journal of Climate, 1–44,
https://doi.org/10.1175/jcli-d- 21-0992.1.
Garfinkel, C. I., I. White, E. P. Gerber, and M. Jucker, 2020a: The Im-
pact of SST Biases in the Tropical East Pacific and Agulhas Current
Region on Atmospheric Stationary Waves in the Southern Hemi-
sphere. Journal of Climate,33 (21), 9351–9374, https://doi.org/
10.1175/JCLI-D- 20-0195.1, URL https://journals.ametsoc.org/doi/
10.1175/JCLI-D- 20-0195.1.
Garfinkel, C. I., I. White, E. P. Gerber, M. Jucker, and M. Erez, 2020b:
The building blocks of Northern Hemisphere wintertime station-
ary waves. Journal of Climate,33 (13), 5611–5633, https://doi.org/
10.1175/JCLI-D- 19-0181.1, URL http://journals.ametsoc.org/doi/
10.1175/JCLI-D- 19-0181.1.
Gupta, A. K., R. Bennartz, K. E. Fauria, and T. Mittal, 2022: Eruption
chronology of the December 2021 to January 2022 Hunga Tonga-
Hunga Ha’apai eruption sequence. Communications Earth & Envi-
ronment,3 (1), 314, https://doi.org/10.1038/s43247-022-00606-3,
URL https://www.nature.com/articles/s43247-022-00606-3.
Holton, J. R., and H.-C. Tan, 1980: The Influence of the Equatorial
Quasi-Biennial Oscillation on the Global Circulation at 50 mb. Jour-
nal of the Atmospheric Sciences,37 (10), 2200–2208, https://doi.org/
10.1175/1520-0469(1980)037 ⟨2200:TIOTEQ⟩2.0.CO;2, URL
http://journals.ametsoc.org/doi/abs/10.1175/1520-0469%281980%
29037%3C2200%3ATIOTEQ%3E2.0.CO%3B2.
Hurrell, J. W., J. J. Hack, D. Shea, J. M. Caron, and J. Rosinski, 2008: A
New Sea Surface Temperature and Sea Ice Boundar y Dataset for the
Community Atmosphere Model. Journal of Climate,21 (19), 5145–
5153, https://doi.org/10.1175/2008JCLI2292.1, URL http://journals.
ametsoc.org/doi/10.1175/2008JCLI2292.1.
Hurrell, J. W., and Coauthors, 2013: The Community Earth Sys-
tem Model: A Framework for Collaborative Research. Bulletin
of the American Meteorological Society,94 (9), 1339–1360,
https://doi.org/10.1175/BAMS-D-12- 00121.1, URL http://journals.
ametsoc.org/doi/10.1175/BAMS-D-12- 00121.1.
Jucker, M., 2019: The Surface of an Aquaplanet GCM.
URL https://research-iceberg.github.io/papers/M Jucker 201907/,
https://doi.org/10.5281/zenodo.3358284.
Jucker, M., and E. P. Gerber, 2017: Untangling the Annual Cy-
cle of the Tropical Tropopause Layer with an Idealized Moist
Model. Journal of Climate,30 (18), 7339–7358, https://doi.org/
10.1175/JCLI-D- 17-0127.1, URL http://journals.ametsoc.org/doi/
10.1175/JCLI-D- 17-0127.1.
Kinnison, D. E., and Coauthors, 2007: Sensitivity of chemical trac-
ers to meteorological parameters in the MOZART-3 chemical trans-
port model. Journal of Geophysical Research,112 (D20), D20 302,
https://doi.org/10.1029/2006JD007879, URL http://doi.wiley.com/
10.1029/2006JD007879.
Lim, E.-P., H. H. Hendon, and D. W. Thompson, 2018: Seasonal Evo-
lution of Stratosphere-Troposphere Coupling in the Southern Hemi-
sphere and Implications for the Predictability of Surface Climate.
Journal of Geophysical Research: Atmospheres,123 (21), 002–12,
https://doi.org/10.1029/2018JD029321, URL http://doi.wiley.com/
10.1029/2018JD029321.
26
Livesey, N., and Coauthors, 2022: Earth Observing System (EOS).
Aura Microwave Limb Sounder (MLS). Version 5.0x Level 2 and
3 data quality and description document. Tech. rep., Jet Propulsion
Laboratory, Pasendena, CA, 177 pp. URL https://mls.jpl.nasa.gov/
data/datadocs.php.
Marsh, D. R., M. J. Mills, D. E. Kinnison, J.-F. Lamarque, N. Calvo,
and L. M. Polvani, 2013: Climate Change from 1850 to 2005
Simulated in CESM1(WACCM). Journal of Climate,26 (19),
7372–7391, https://doi.org/10.1175/JCLI-D- 12-00558.1, URL http:
//journals.ametsoc.org/doi/10.1175/JCLI-D-12- 00558.1.
Maycock, A. C., M. M. Joshi, K. P. Shine, and A. A. Scaife, 2013: The
Circulation Response to Idealized Changes in Stratospheric Water
Vapor. Journal of Climate,26 (2), 545–561, https://doi.org/10.1175/
JCLI-D- 12-00155.1, URL http://journals.ametsoc.org/doi/10.1175/
JCLI-D- 12-00155.1.
Mill´
an, L., and Coauthors, 2022: The Hunga Tonga-Hunga Ha’apai
Hydration of the Stratosphere. Geophysical Research Letters,
49 (13), 1–10, https://doi.org/10.1029/2022GL099381, URL https:
//onlinelibrary.wiley.com/doi/10.1029/2022GL099381.
Mlawer, E. J., S. J. Taubman, P. D. Brown, M. J. Iacono, and S. A.
Clough, 1997: Radiative transfer for inhomogeneous atmospheres:
RRTM, a validated correlated-k model for the longwave. Journal of
Geophysical Research,102 (D14), 16663, https://doi.org/10.1029/
97JD00237, URL http://doi.wiley.com/10.1029/97JD00237.
Oleson, K. W., D. M. Lawrence, G. B. Bonan, and M. G. Flanner, 2010:
Technical Description of version 4.0 of the Community Land Model
(CLM). Tech. rep. https://doi.org/10.5065/D6FB50WZ.
Plumb, R. A., 2002: Stratospheric Transport. J. Meteor. Soc. Japan,
80 (1949), 793–809.
Proud, S. R., A. T. Prata, and S. Schmauß, 2022: The January
2022 eruption of Hunga Tonga-Hunga Ha’apai volcano reached
the mesosphere. Science,378 (6619), 554–557, https://doi.org/10.
1126/science.abo4076, URL https://www.science.org/doi/10.1126/
science.abo4076.
Robock, A., 2000: Volcanic eruptions and climate. Reviews of Geo-
physics,38 (2), 191–219, https://doi.org/10.1029/1998RG000054,
URL http://doi.wiley.com/10.1029/1998RG000054.
Santer, B. D., and Coauthors, 2014: Volcanic contribution to
decadal changes in tropospheric temperature. Nature Geoscience,
7 (3), 185–189, https://doi.org/10.1038/ngeo2098, URL http://www.
nature.com/articles/ngeo2098.
Solomon, S., 1999: Stratospheric ozone depletion: A review of
concepts and history. Reviews of Geophysics,37 (3), 275–316,
https://doi.org/10.1029/1999RG900008, URL http://doi.wiley.com/
10.1029/1999RG900008.
Thompson, D. W., M. P. Baldwin, and S. Solomon, 2005: Strato-
sphere–Troposphere Coupling in the Southern Hemisphere. Jour-
nal of the Atmospheric Sciences,62 (3), 708–715, https://doi.org/
10.1175/JAS-3321.1, URL http://journals.ametsoc.org/doi/abs/10.
1175/JAS-3321.1.
Tilmes, S., R. R. Garcia, D. E. Kinnison, A. Gettelman, and P. J. Rasch,
2009: Impact of geoengineered aerosols on the troposphere and
stratosphere. Journal of Geophysical Research,114 (D12), D12 305,
https://doi.org/10.1029/2008JD011420, URL http://doi.wiley.com/
10.1029/2008JD011420.
Tritscher, I., and Coauthors, 2021: Polar Stratospheric Clouds: Satel-
lite Observations, Processes, and Role in Ozone Depletion. Reviews
of Geophysics,59 (2), https://doi.org/10.1029/2020RG000702, URL
https://onlinelibrary.wiley.com/doi/10.1029/2020RG000702.
Vernier, J.-P., and Coauthors, 2011: Major influence of tropi-
cal volcanic eruptions on the stratospheric aerosol layer during
the last decade. Geophysical Research Letters,38 (12), L12 807,
https://doi.org/10.1029/2011GL047563, URL http://doi.wiley.com/
10.1029/2011GL047563.
V¨
omel, H., S. Evan, and M. Tully, 2022: Water vapor injection into the
stratosphere by Hunga Tonga-Hunga Ha’apai. Science,377 (6613),
1444–1447, https://doi.org/10.1126/science.abq2299, URL https://
www.science.org/doi/10.1126/science.abq2299.
Zhu, Y., and Coauthors, 2018: Stratospheric Aerosols, Polar Strato-
spheric Clouds, and Polar Ozone Depletion After the Mount Cal-
buco Eruption in 2015. Journal of Geophysical Research: At-
mospheres,123 (21), https://doi.org/10.1029/2018JD028974, URL
https://onlinelibrary.wiley.com/doi/10.1029/2018JD028974.
Zhu, Y.,and Coauthors, 2022: Perturbations in stratospheric aerosol evo-
lution due to the water-rich plume of the 2022 Hunga-Tonga eruption.
Communications Earth & Environment,3 (1), 248, https://doi.org/10.
1038/s43247-022- 00580-w, URL https://www.nature.com/articles/
s43247-022- 00580-w.