Introduction

On January 13 and 15 of 2022, the Hunga Tonga-Hunga Ha’apai (HTHH) submarine volcano (21°S, 175°W) erupted and injected volcanic material into the stratosphere up to an altitude of 58 km1,2,3. Observations from several satellites showed enhanced levels of stratospheric sulfur dioxide (SO2) (by the TROPOspheric Monitoring Instrument, TROPOMI and the Ozone Mapping and Profiling Suite Nadir-Mapper, OMPS-NM), water vapor (by the Microwave Limb Sounder, MLS), aerosols (by OMPS Limb Profiler, OMPS LP and Cloud-Aerosol Lidar with Orthogonal Polarization, CALIOP), and possibly ice (by CALIOP) as we will discuss in this work. Unlike some land-based volcanos, HTHH injected more than 100 Tg of water into the stratosphere4,5, which normally contains about 1500 Tg of water globally defined for a tropopause pressure of 100 hPa. The MLS onboard NASA’s Earth Observing System (EOS) Aura satellite6 observed enhanced water vapor that persisted for more than 3 months. The CALIOP on the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) satellite7 observed two distinct aerosol types on January 16 after the eruption: one with weak 532 nm depolarization indicating spherical particles (such as sulfate aerosol) above 20 km, and layers at lower altitudes with large depolarization indicating particles with nonspherical shapes (such as volcanic ash and ice) (Supplementary Fig. 1). It is typical to see ice and ash immediately after an eruption (i.e., ref. 8). However, it is surprising to see abundant sulfate aerosol with large extinction/backscatter coefficients so quickly because SO2 usually takes more time to be oxidized and for sulfuric acid (H2SO4) to nucleate into sulfate aerosol9.

Stratospheric water vapor and volcanic aerosols are important to both chemistry10,11,12,13,14,15 and radiative balance16,17,18,19. Here, we demonstrate that enhanced H2O promotes faster sulfate aerosol formation in the stratosphere leading to larger particles and larger aerosol optical depth. The water vapor injection and a small amount of evaporating ice deposits substantial water vapor in the stratosphere. Large water vapor abundances result in the formation of hydroxide (OH) which acts to speed up the conversion of SO2 to H2SO4. Because of the faster formation and coagulation of sulfate aerosol, particles form with larger sizes and hence larger extinction than the ones without water injection. Finally, we predict the water vapor and aerosols that get transported toward the South Pole will linger near the polar vortex until at least October 2022.

Results

Both water injection and evaporating ice determine the water vapor remaining in the stratosphere

Here we use the Whole Atmosphere Community Climate Model (WACCM) to simulate the persistent water vapor enhancement from 10 to 50 hPa observed by MLS version 4 (Fig. 1a) from February to April. Zonal averages of simulated water vapor on March 1 shows good agreement with MLS observations (Supplementary Fig. 2). Figure 1a shows both the model (solid line) and MLS (dashed line) display a positive water vapor anomaly of 6–8 ppmv peaking at 30 hPa from February to April. The anomaly is averaged from 30°S to 0° and calculated using February, March, and April profiles minus the profile on January 5. The simulation has wider vertical extent and slightly smaller peak anomalies in March and April. Both the model and MLS (Fig. 1a) show that the positive water vapor anomaly slowly ascends from February 10 to April 1, which is largely related to the ascending branch of Brewer-Dobson circulation20 (Supplementary Fig. 2b).

Fig. 1: The H2O (water) profile and evolutions after the Hunga Tonga-Hunga Ha’apai eruption.
figure 1

a The zonal average H2O anomaly profiles between 30°S–0° on February 10, March 1, and April 1, 2022. The solid lines are simulations and the dashed lines are the MLS observations. The simulation and MLS anomaly curves are calculated using February, March, and April profiles minus the profile on January 5. b Simulated stratospheric H2O anomaly evolution in the first week. It is calculated using a case with H2O injection minus a case without H2O injection. The blue line is the ice, the orange line is the water vapor, and the black line is the total water. The green dashed lines indicate the water injection period.

Full size image

To be consistent with the MLS water anomalies5, we need to inject ~150 Tg of water in an area of ~6 × 105 km2, which is about three times larger than the anvil size observed by NOAA’s Geostationary Operational Environmental Satellite 17 on January 15, 2022. This large area is needed because the residual amount of water is determined by the ice vapor pressure curve as shown in Supplementary Fig. 3. If we inject the water into a smaller area, the model forms too much ice and cannot retain enough water at 30 hPa as observed. It is not unreasonable to inject in a larger area. MLS observations on Jan 16 (a day after the eruption) show the latitudinal spreading of water vapor anomaly is more than 10 degrees5. The simulated plume does not spread as fast as observed because it is hard for the global model to reproduce the vertical wind shear as observed due to the limited spatial resolution, especially for a tropical volcanic plume spreading in the first couple of days21. In the model, the injection is over 6 h on January 15, 2022 between 25.5 and 35 km with a majority of the water injected between 25.5 and 30 km (see Methods for details). The model simulation (Fig. 1b) shows that 10 Tg of ice falls out and 10 Tg of ice evaporates and contributes to the residual water vapor (~140 Tg) until April. In actuality, it is possible that more H2O was injected and formed ice particles, but these ice particles may have fallen out because of their large size, or because they aggregated with ash particles8.

The water injection significantly shortens the SO2 lifetime by providing abundant OH

An accurate volcanic SO2 lifetime is key for predicting the particle sizes in the volcanic cloud. With shorter lifetimes, the resulting aerosol concentration is higher, leading to more rapid coagulation and larger sulfate particles22. In the stratosphere, the volcanic SO2 lifetime is mainly determined by its reaction rate with OH, as well as the heterogeneous reaction on ash21. The depletion of OH in SO2-rich plumes slows the SO2/OH reaction9,21,23,24,25. However, during the HTHH eruption, the notable amount of water vapor injected rapidly increased OH15 (Supplementary Fig. 4) and shortened the SO2 lifetime as shown in Fig. 2. The simulated SO2 (blue dash line) is within the observational error bars (i.e., uncertainty of 35% and 30%26 for OMPS-NM and TROPOMI SO2 mass estimates, respectively). Clegg and Abbatt27 demonstrated that SO2 uptake on the ice surfaces is insignificant, so this mechanism is not included in the model. The SO2 lifetime (the e-folding time) of the SO2only case (red line) is 28 days, while the SO2 lifetime of the SO2_H2O case (blue solid line) is 12 days. Previous study21 showed that a considerable amount of SO2 might be undetectable because it falls below the detection limits of the instruments as it spreads through the atmosphere. However, the blue dashed line in Fig. 2 suggests that this effect is small for this eruption for the first 10 days.

Fig. 2: The time series of the stratospheric SO2 (sulfur dioxide) burden.
figure 2

The blue and red lines are two model cases with and without water injection. The dashed blue line is the SO2_H2O case excluding the SO2 below 0.2 DU (i.e., the approximate OMPS-NM (the Ozone Mapping and Profiling Suite Nadir-Mapper) SO2 detection limit). The circles and triangles are SO2 measurements by OMPS-NM and TROPOMI (the TROPOspheric Monitoring Instrument). The error bars imply an estimated uncertainty of 35% and 30%26 for OMPS-NM and TROPOMI SO2 mass estimates, respectively. The observational data are from https://so2.gsfc.nasa.gov/. The SO2only run has the SO2 injection; the SO2_H2O run has both the SO2 and H2O injections (detailed in Online methods).

Full size image

Enhanced water vapor impacts the stratospheric aerosol optical depth and radiation

We compare the stratospheric aerosol optical depth (sAOD) between OMPS-LP data and two model cases at near-infrared in Fig. 3, because the extinction coefficient retrieval near-infrared has the best accuracy among all the wavelength for the OMPS-LP instrument28. Here we present the SO2only and H2O_SO2 cases based on our best knowledge of the injection area and injection amount of SO2 and H2O. Additional model sensitivity tests with various injection amounts and latitude bands compared with several observational aerosol optical properties are detailed in Supplementary Note 1. Without the water injection, the SO2only case has a very small sAOD in the first 2 weeks after the eruption, because SO2 converts slowly to H2SO4 with a lifetime of ~1 month (Fig. 2). Both OMPS and the SO2_H2O case show almost immediate formation of a large amount of sulfate aerosol. OMPS LP sAOD retrieval during the first couple of days has large uncertainties when the volcanic clouds are optically thick and localized29. The backscatter coefficient at visible wavelengths for the SO2_H2O case and CALIOP shows a similar peak value of 0.005 to 0.01 km−1 sr−1 (Supplementary Fig. 5). In addition, compared to the SO2only case, sulfate particles in the SO2_H2O case doubled the sAOD in February because sulfate forms faster in the SO2_H2O case, and the more abundant particles coagulate to larger sizes. Also, sulfate particles swell to be a little larger because of the enhanced background water vapor (Supplementary Fig. 8). Generally, OMPS-LP observes a faster spreading of the plume to the northern hemisphere than is simulated (Fig. 3) and has higher optic values in the northern hemisphere in March (Supplementary Fig. 9). Sellitto et al.30 indicate the initial fast spreading of the HTHH aerosol plume is highly related to the strong cooling inside the plume. Our model is nudged to the observed meteorology, which does not capture this strong local cooling. Also, as we discuss in Result, the global model cannot reproduce the vertical wind shear as observed due to the limited spatial resolution. On March 1, the SO2_H2O case and the OMPS observation are consistent regarding the vertical extent of the plume from 20 to 30 km, and the extension to 40°S (Supplementary Fig. 9).

Fig. 3: The zonal mean column sAOD (stratospheric aerosol optical depth, background aerosol + volcanic aerosol).
figure 3

The top panel shows the OMPS LP (the Ozone Mapping and Profiling Suite Limb Profiler) at 997 nm. The middle and bottom panels show the simulated near-infrared sAOD (stratospheric aerosol optical depth) for cases without (the SO2only case) and with the water injection (the SO2_H2O case).

Full size image

Because of the increase in the burden of sulfate, the volcanic plume creates a negative radiative effect of about −1 to −2 W m−2 in the perturbed areas in January and February (Fig. 4). The global mean radiative effect at the top of the atmosphere in February is −0.13 W m−2 in the SO2only case and −0.21 W m−2 in the SO2_H2O case; the surface radiative effect is −0.16 W m−2 for the SO2only case and −0.21 W m−2 for the SO2_H2O case. These values are typical for middle-sized volcanic eruptions31. Even though the sAOD is approximately doubled in the SO2_H2O case, it only results in slightly more negative radiative forcing due to enhanced positive radiative forcing of the enhanced water vapor32,33 in the SO2_H2O case. Additional simulation with only H2O injection shows the radiative effect from water is positive in the Southern hemisphere but is much smaller than that of the aerosol (Supplementary Fig. 10).

Fig. 4: The zonal average radiative effect (W m−2) of the top of the atmosphere from January to March 2022.
figure 4

a The SO2only case minus The control case; b the SO2_H2O case minus the control case.

Full size image

Persistent volcanic water vapor and sulfate may impact stratospheric ozone chemistry

Here we focus our study on the transport of volcanic water vapor and sulfate in the Southern Hemisphere. We conduct three sets of 3-member ensemble simulations to explore the transport of volcanic water vapor and volcanic sulfate from January to October. The simulations are nudged to the observed meteorology until the end of March and then are free-running until October. Figure 5 shows that both H2O and sulfur in the condensed phase are slowly transported to the south from January to May. After mid-June, the majority of H2O and sulfur in the condensed phase reside between 30°S and 60°S. This is because the strong Antarctic vortex during June–July–August prevents volcanic materials from entering the polar cap. The black contours in Fig. 5 are the zonal wind showing the polar vortex starts to build up in April and remains through October between 50°S and 60°S. Figure 5c, d shows the percentage increase compared to the background levels of H2O and sulfur in the condensed phase. Even though the majority of volcanic material is outside the polar vortex, we see a ~10% increase of sulfur in the condensed phase at 90°S starting from April and the mass of the sulfur in the condensed phase almost doubles in October. The sulfate aerosol provides extra surface area density (SAD) for heterogeneous reactions affecting ozone chemistry. On the other hand, water increases by about 20% at the edge of the vortex (~60°S) after June. However, due to polar stratospheric cloud formation and dehydration, the positive water anomaly inside the polar vortex does not pass the Student’s t-test at a 90% significance level.

Fig. 5: The transport of volcanic H2O (water) and sulfur in the condensed phase from January to October, 2022.
figure 5

a, b The transport of volcanic column H2O and the mass density of volcanic column sulfur in the condensed phase towards the South Pole. “Column cond S” means the sulfur in the condensed phase. c, d Similar but plotted as the percentage increase relative to background H2O and sulfur. After April 1, white areas mean the anomaly is not significant using the t-test at a 90% confidence level based on three ensemble members after April 1, 2022. The contours are the zonal wind showing the polar vortex from April to October at around 60°S.

Full size image

The simulations predict an increase in water vapor and aerosol surface area density (SAD) near the vortex. These changes are expected to impact polar ozone because heterogeneous reactions on polar stratospheric clouds and volcanic particles convert inactive chlorine (ClONO2 and HCl) into photochemically active chlorine34,35. Figure 6a, b shows that water increases by about 2–3 ppmv and SAD increases by about 2 to 4.5 µm2 cm−3 near the edge of the vortex by the end of September. The SAD increases by about 1.5 µm2 cm−3 inside the vortex at 100 hPa. Water vapor increases between 50 and 10 hPa, while the aerosol increase is between 150 and 30 hPa. This difference in altitude occurs because of aerosol sedimentation during transport. Figure 6c shows the HCl+ClONO2 heterogeneous reaction rate36 as a function of temperature for different amounts of water vapor. The reaction probability increases about one order of magnitude as we increase water by 3 ppmv above 194 K. Volcanic aerosols are well known to be a factor to impact ozone depletion by providing additional surface area and suppressing the NOx cycle37. The additional H2O changes in HTHH can impacting the dynamics, HOx chemical cycles, heterogeneous reaction rate, and the Polar Stratospheric Cloud formation (PSCs). The water vapor mixing ratio determines the vapor pressure of HNO3 in the supercooled ternary solutions (Type Ib PSC) and impacts the ice PSC’s nucleation and growth. Both the direct impact and climate response caused by this eruption add complications to ozone assessment. Therefore, future studies will be conducted to explain how the volcanic H2O and sulfate impact the 2022 Antarctic ozone depletion.

Fig. 6: The simulated water vapor, the aerosol SAD (surface area density) anomaly, and ClONO2 + HCl (chlorine nitrate + hydrogen chloride) reaction probability.
figure 6

The simulated water vapor anomaly (a) and the aerosol SAD anomaly (b) in late September. These anomalies are calculated using the SO2_H2O case minus the control case. c The heterogeneous reaction rate as a function of water vapor assuming 0.1 µm particle size.

Full size image

Discussion

The HTHH eruption provides a natural testbed to show how increased water vapor in the stratosphere can impact the Earth system. This work demonstrates the importance of the water vapor injection on aerosol and aerosol-related chemistry using a state-of-the-art global climate model. The simulation shows that water vapor significantly shortens the SO2 lifetime and doubles aerosol extinction relative to the simulation with only an SO2 injection. The global radiative forcing of HHTH is on the order of −0.1 to −0.2 W m−2 though it is about −1 to −2 W m−2 from 30°S to 0° during the 2 months after the eruption. In the forecast simulation, volcanic water and sulfate persist near the polar vortex until at least October 2022. The model developed here provides opportunities for future studies on the climate impact, and effects on ozone, of HTHH eruption as it continues to evolve.

Methods

Model setup

We utilize the Community Earth System Model version 2 (CESM2) with the Whole Atmosphere Community Climate Model (WACCM)38 using 70 layers extending upward to 140 km. The vertical resolution is about 1 to 1.5 km in the stratosphere. The model is fully coupled to an interactive ocean, sea-ice, and land. The ocean and sea-ice are initialized on January 3 with output from a stand-alone ocean model forced by atmospheric state fields and fluxes from the Japanese 55-year Reanalysis39. Likewise, the land is initialized with output from a stand-alone land model forced with atmospheric data from the National Center for Environmental Prediction Climate Forecast System, version 2. These are the same initializations that were used in the subseasonal-to-seasonal project40. The atmosphere is initialized from a transient WACCM simulation38. Between January 3 and March 30, 2022, the atmospheric component is nudged to GEOS5 meteorological analysis41 with a 12-h relaxation using 3-h meteorology42. After April 1, three ensemble members with a fully interactive atmosphere and ocean are continued into the future until the end of October. The three ensemble members differ in the last date of nudging (namely March 29, 30, and 31). We conducted three sets of ensemble members: the control case without SO2 or H2O; the SO2only case with only SO2 injection; and the SO2_H2O case with both SO2 and H2O injection. We conduct a H2Oonly case with only H2O injection to investigate the radiative impact from H2O.

The SO2 injection amount is based on the TROPOMI and OMPS Nadir Mapper (NM) SO243 reported at https://so2.gsfc.nasa.gov/. We tune the SO2 injection vertical distributions based on comparisons between the simulated sulfate aerosol and OMPS LP aerosol extinction in March (i.e., Supplementary Fig. 9). We tune the H2O injection amount and vertical distribution mainly to retain enough H2O between 10 and 50 hPa as seen by MLS in Feb, March, and April (i.e., Fig. 1a). MLS also sees a small amount of water between 10 to 1 hPa in those days5. We inject 0.42 Tg of SO2 and 150 Tg of H2O from 4 UTC to 10 UTC on January 15, 2022 (ignoring the small eruption that occurred on January 13, 2022. Note that this small eruption also put ice into the lower stratosphere, up to 20 km). The vertical distribution of SO2 is 0.3 Tg from 20 to 22 km and 0.12 Tg from 22 to 28 km with constant number densities in these layers. The vertical distribution of H2O is 103 Tg from 25.5 to 27 km, 42 Tg from 27 to 30 km, and 5 Tg from 30 to 35 km with constant concentration. We inject the plume into an area of ~6 × 105 km2 between 22°S–14°S and 182°E–186°E, which is about three times larger than the anvil size observed by NOAA’s Geostationary Operational Environmental Satellite 17 on January 15, 2022. This is because the MLS observes H2O occupying ~10-degree latitude bands on January 155. Also, the global model cannot spread the plume as fast as observed as we discuss in Result. We have not considered ash injection since satellite observations (e.g., OMPS measured low UV Aerosol Index) do not show evidence of notable ash persisting for a long time. CALIOP sees that the majority of the high backscatter coefficient above 20 km correlates with low depolarization indicating the composition is mainly spherical sulfate particles. The Light Optical Aerosol Counter (LOAC) measurements at Reunion island (21.1°S, 55.3°E) on January 23 and January 26 mainly see submicron particles indicating no large volcanic ash or other large (>1 µm) volcanic particles present. Supplementary Note 1 presents several other model simulations with a larger SO2 injection amount (0.84 Tg) or a wider injection band (20-degree latitude band) for both SO2_H2O and SO2only cases as sensitivity tests.

For ice cloud formation, we modified the nucleation parameterization to allow both homogeneous freezing and heterogeneous nucleation to occur in the stratosphere when temperatures are cold (<−40 °C).

Observations

The MLS instrument onboard the EOS Aura satellite was launched into a near-polar sun-synchronous orbit in 2004 and measures atmospheric composition, temperature, humidity, and cloud ice. After the HTHH eruption, MLS detected enhanced water vapor. Here we use MLS version 4 quality screened data6 and do not use MLS ice water content as recommended by Millán et al.5.

The OMPS Limb Profiler (LP)44 data have been processed to create an aerosol extinction coefficient product from measurements of the limb scattered solar radiation at six wavelengths. The sensor employs three vertical slits separated horizontally. This work uses the version 2.0 OMPS aerosol extinction coefficient product at 997 nm wavelength28 with relative accuracies and precisions close to 15%. We use the center slit aerosol retrieval because it has the most accurate radiometric calibration and stray light corrections45.

SO2 mass estimates from OMPS-NM are based on the NASA standard SNPP/OMPS SO2 vertical column density (VCD) product, retrieved using a principal component analysis (PCA) spectral fitting algorithm43 assuming a fixed a priori profile centered at 18 km altitude. The estimated uncertainty of ~35% (error bar in Fig. 2) reflects potential errors that can be caused by the highly unusual injection height and the large amounts of aerosols and ice particles for the HTHH eruption. For OMPS data, the detection limit is based on the 1-sigma noise level of OMPS SO2 retrievals of ~0.1 DU over background, SO2-free areas. Any pixels with retrieved SO2 < 0.2 DU, we consider to be noise and for volcanic SO2 mass calculation those pixels are excluded (i.e., we calculate the total mass by summing up SO2 mass from all the OMPS pixels with SO2 >= 0.2 DU). In order to compare the modeled SO2 with observations as shown in Fig. 2, we apply a similar method as the observation by excluding modeled grid boxes with SO2 < 0.2 DU. This way, the model-based SO2 mass is calculated in a more consistent way with OMPS-based estimate. The comparisons between the modeled SO2 and OMPS-NM SO2 retrieval without applying detection limit (Supplementary Fig. 11) also agree within the error bars.