1 Introduction

Hunga Tonga-Hunga Ha'apai (HT-HH), a submarine volcano in the South Pacific (20.54°S, 175.38°W), reached its climactic eruption phase on 15 January 2022. The blast sent a volcanic plume into the mesosphere to altitudes of up to 57 km—a record in the satellite era (Carr et al., 2022; Proud et al., 2022). It also triggered tsunami alerts across the world (Carvajal et al., 2022; Ramirez-Herrera et al., 2022), waves that propagated globally (Wright et al., 2022), and ionospheric disturbances (Themens et al., 2022). Details about the HT-HH caldera complex, seismology, and volcanology are given by Kusky (2022) and Yuen et al. (2022).

In addition to particulate matter, volcanic eruptions can loft large quantities of gases into the stratosphere. Although around 80% of this gas volume can be magmatic H₂O (Coffey, 1996; Pinto et al., 1989), up to 90% of the volcanically emitted humidity is usually removed by condensation at the cold point tropopause (Glaze et al., 1997). Considerable amounts of CO₂ and SO₂ are also often found in volcanic plumes, along with HCl and other trace gases (e.g., Carn et al., 2016). SO₂ reacts with H₂O and OH to form submicron sulfate aerosols that reflect solar radiation, lowering surface temperature. For example, the radiative influence of the 1991 Mount Pinatubo eruption “put an end to several years of globally warm surface temperature” (McCormick et al., 1995), illustrating the capacity of volcanic eruptions to substantially alter global climate.

The composition of the HT-HH plume is unprecedented, as the eruption injected vast amounts of H₂O directly into the stratosphere. The high moisture content of the plume is perhaps not surprising since the HT-HH caldera was situated 150 m below sea level (Cronin et al., 2017), where water in contact with the erupting magma (at temperatures of ∼1100–1470 K) was superheated, resulting in explosive steam.

The Microwave Limb Sounder (MLS) onboard NASA's Aura satellite provides measurements of 15 trace gases, among them H₂O, HCl, and enhanced volcanic SO₂. MLS measures thermal emission from the Earth's limb, covering spectral regions near 118, 190, 240, and 640 GHz (Waters et al., 2006). MLS is well suited to observe volcanic plumes since microwave radiances are largely unaffected by sulfate aerosols. Moreover, the MLS two-dimensional retrieval exploits overlapping limb observations to better constrain trace gas gradients (Livesey et al., 2006), allowing the spatial heterogeneity of the plume to be captured.

Here, we use MLS version 4 (v4) data, instead of the most recent version (v5). Estimates of instrument pointing are required for atmospheric composition profile retrievals. V4 obtains pointing purely from O₂ signals, whereas v5 also uses the H₂O line. Poor fits to H₂O signals in regions with extremely enhanced humidity, such as those discussed here, lead to discrepancies in pointing as large as ∼2.5 km, degrading the accuracy of some v5 products.

2 Validity of MLS Measurements After the Eruption

Ten hours after the eruption on 15 January, MLS measured enhanced values of H₂O at altitudes up to 0.46 hPa (∼53 km), well above the stratopause (Figure 1c). Most of these measurements of enhanced H₂O did not pass the MLS quality screening (QS) criteria defined by Livesey et al. (2020), indicating that the retrieval achieved only a poor fit to the radiances. The poor performance of the standard data processing algorithms is unsurprising, as the largest H₂O values are more than an order of magnitude greater than any previously observed by MLS and more than 100 standard deviations above background levels. Here, data points with values greater than 7 standard deviations above the climatological January–February–March 2005–2021 average are identified as enhancements.

The eruption injected H₂O throughout a large vertical range encompassing most of the stratosphere, but on 15 January, MLS only measured the outer edge of the plume in the upper stratosphere, where strong winds advected the lofted H₂O to locations sampled by MLS. Near 80 hPa on this day, MLS also measured some enhanced H₂O injected by a previous, less violent, HT-HH eruption on 14 January.

For the next several days, most of the largest enhancements failed the QS. Figure 1d shows the profiles displaying the largest mixing ratios on 15, 16, 17, and 18 January. Back trajectories (as in Livesey et al. (2015); Santee et al. (2022)) indicate that these enhancements lie downwind from the HT-HH volcano (Figure 1b), and the measured spectral signature is well represented by radiance simulations (Figure 1e). Peaks centered on channels 5 and 22 on 16 and 17 January are SO₂ spectral lines; they indicate that these lower plumes contained more SO₂ than the high-altitude plume on 15 January.

As the plume dispersed, the daily number of profiles failing the QS increased, reaching a maximum on 19 January. Retrieval performance then returned to normal by 8 February, by which time the plume had dispersed sufficiently that maximum H₂O values had dropped to ∼50 ppmv, versus up to 350 ppmv immediately following the eruption (Figure 1).

Taken together, the back trajectories, radiance simulations, and return to typical retrieval quality confirm that the measured enhancements represent real volcanically enhanced H₂O values. However, the absolute magnitudes of the enhancements, especially for those failing the QS screening, are still in question because of the poor radiance fits. The MLS retrievals were not optimized to handle such strong H₂O enhancements. Thus, to fully quantify these injections and their uncertainties, we are developing a special retrieval for MLS measurements of the HT-HH plume. Preliminary results suggest that H₂O retrievals that better fit the radiances lie within 20% of current v4 estimates.

In addition, it is essential to account for the relatively coarse resolution of the MLS observations (∼3.2 × 230 km for H₂O at these altitudes, as quantified by the averaging kernels (Livesey et al., 2020)) in the presence of strong vertically confined plumes (Schwartz et al., 2013, 2020). Accordingly, mid-January maximum plume values of 1500 ppmv measured by radiosondes (Sellitto et al., 2022) are not necessarily inconsistent with observed MLS abundances, given the disparity in their respective resolutions.

Many chemical species measured by MLS show anomalous mixing ratios in the plume (Figure S1 in Supporting Information S1). However, only the H₂O, SO₂, and HCl spectral signatures can confidently be ascribed to real enhancements in those quantities; perturbations in other species are likely artifacts arising from SO₂ spectral interference. SO₂ is retrieved from a spectrometer that targets an O¹⁸O line but also covers many SO₂ lines, the strongest of which are located in channels 5, 11, and 20. The triple-peak structure in measured radiances within the volcanic plume (Figure 2b) can only be plausibly explained by an SO₂ enhancement.

HCl is currently measured by a spectrometer that targets an O₃ line but covers HCl lines in channels 3 and 25. The ∼5 K HCl radiance signature overlaps with an ∼180 K O₃ signal. The differences between the measurements and the simulations with and without accounting for contributions from HCl reveal its expected ∼5 K signature, suggesting that the observed enhancements represent real geophysical signals (Figure 2d). The HCl spectral signature is similar to that of the background because the HCl enhancements are not as dramatic as those of H₂O or SO₂.

MLS estimates of ice water content (IWC) are based on the differences between the measured radiances and the expected clear-sky radiances, with the residuals attributed to ice scattering and/or ice absorption. The clear-sky radiances are calculated using the retrieved atmospheric states; since most retrievals in the volcanic plume fail the QS in the days following the eruption, the derived IWC estimates are unreliable. In contrast, the quality of the MLS temperature, CO, and O₃ measurements is not affected by the plume.

3 Unprecedented Stratospheric H₂O Injection

Figure 3 compares the HT-HH HCl, SO₂, and H₂O stratospheric injections to other stratospheric injections (volcanic or otherwise) observed by MLS. Large injections are marked individually.

The HT-HH eruption did not inject vast amounts of either HCl or SO₂ into the stratosphere. The total injected mass of stratospheric SO₂ (calculated as described by Pumphrey et al. (2021)) was 0.41 ± 0.02 Tg, which pales in comparison to that from previous eruptions measured by MLS, such as the 2008 Kasatochi, the 2009 Sarychev, or the 2019 Raikoke eruptions, which each emitted ∼1 Tg (de Leeuw et al., 2021; Pumphrey et al., 2015). The mass of SO₂ injected by HT-HH is even less noteworthy in the context of the 17 Tg injected by the 1991 Pinatubo eruption (Read et al., 1993).

The only unusual aspect of the SO₂ plume is its injection height. SO₂ plumes are typically injected at altitudes no higher than 46 hPa (∼21 km) (Carn et al., 2016; Pumphrey et al., 2015). HT-HH is the only injection observed by MLS that produced maximum values of SO₂ at 14 hPa (∼29 km), with enhanced values detected up to 6.8 hPa (∼35 km)—outside the normally recommended pressure range for MLS SO₂. By 27 January, the SO₂ plume dropped below background levels (Figure S1 in Supporting Information S1).

The HCl injection was similarly unremarkable, with only 8 profiles during 16–18 January (barely) exceeding the threshold for enhancement (Figure 2c; Figure S1 in Supporting Information S1). As with SO₂, the only unusual aspect of the HCl plume is its injection height of 31.6 hPa (∼24 km), whereas previous eruptions reached no higher than 68 hPa (∼18.6 km).

In contrast, the magnitude of the HT-HH H₂O injection is unprecedented. Three natural pathways for direct injection of H₂O into the stratosphere exist: overshooting convection, pyrocumulonimbus (pyroCb) storms, and volcanic eruptions. The previous stratospheric H₂O record measured by MLS was 26.3 ppmv at 100 hPa associated with an overshooting convective event in August 2019 that spanned thousands of square kilometers and persisted for several hours (Werner et al., 2020). Two pyroCbs stand out in the MLS H₂O record: the 2017 Pacific Northwest (Pumphrey et al., 2021) and the 2019/2020 Australian New Year's (Kablick et al., 2020; Khaykin et al., 2020; Schwartz et al., 2020) events. Only the Australian pyroCbs injected enough H₂O to allow an accurate estimate of mass (19 ± 3 Tg).

The 2008 Kasatochi (Schwartz et al., 2013) and the 2015 Calbuco (Sioris et al., 2016) volcanic eruptions were the only others in the MLS record that injected appreciable amounts of H₂O into the stratosphere. Neither deposited H₂O at altitudes higher than 68 hPa (∼18.6 km), and both injections were too small for a reliable H₂O mass estimate.

The HT-HH eruption injected at least 146 ± 5 Tg of H₂O into the stratosphere, not only surpassing the magnitudes of all other injections in the MLS record, but also eclipsing a theoretical estimate of 37.5 Tg from Pinatubo (Pitari & Mancini, 2002). This stratospheric H₂O injection is unique in the satellite record (1979 to date). To put the HT-HH injection into perspective, the enhancement represents ∼10% of the estimated stratospheric H₂O burden of 1400 Tg (Glaze et al., 1997). Further, the H₂O plume injection height far exceeded that of any other injections in the MLS record (Figure 3).

4 Evolution of the H₂O Plume

To study the development of the H₂O plume, Figure 4 shows maps for selected days after the eruption and meridional and zonal mean anomalies based on all data points as well as only those that pass the QS criteria. On 15 January, the plume reached 0.46 hPa (∼53 km), with most of the MLS retrievals failing QS. On 16 January, two separate plumes are visible, one in the upper stratosphere (between 1 and 8 hPa) and the other in the lower stratosphere (between 10 and 80 hPa), where most of the H₂O volume was injected. On this day, the effects on the plume of strong wind shear between 1 and 8 hPa are already apparent.

By 22 January, the plume had almost entirely circled the globe at 2.1 hPa, while only traveling halfway around at 26 hPa. On average, through January and February, the plume moved ∼37° longitude per day at 2.1 hPa, but only ∼18° longitude per day from 31 to 6 hPa, consistent with winds from meteorological analyses (see Figure S2 in Supporting Information S1) interpolated to the MLS measurement times and locations as described by Manney et al. (2007). By 5 February, the plume covered all longitudes, with the largest enhancements from 38 to 21 hPa (∼22–26 km). By 31 March, the plume around 4.6 hPa had dropped to near background values.

Measurements from 31 March show the persistence of the H₂O plume in the lower and middle stratosphere. Concurrent with encircling the globe, the H₂O plume broadened slowly, spreading mostly northward around 26 hPa. This plume will require further monitoring as the eruption signal propagates into the upper stratosphere and toward the poles in the Brewer-Dobson Circulation (BDC).

5 Discussion and Summary

The importance of stratospheric H₂O is well established; it affects stratospheric chemistry and dynamics as well as atmospheric radiation. For example, excess stratospheric H₂O could lead to enhanced OH concentrations, slightly enhancing O₃ production through the CH₄ oxidation cycle but worsening O₃ depletion through the HO_x cycle, leading to a net decrease in O₃ (e.g., Dvortsov & Solomon, 2001; Stenke & Grewe, 2005). The enhanced OH concentrations could also increase the loss of CH₄, resulting in a decrease in its lifetime (e.g., Ko et al., 2013; Stevenson et al., 2020) and thus reducing its long-term effect on climate. In addition, if enhanced H₂O concentrations were to be entrained into the developing Antarctic vortex to an extent sufficient to raise the formation temperature of polar stratospheric clouds, then the earlier onset of heterogeneous processing would exacerbate cumulative chemical O₃ loss. In terms of transport, a study of the dynamical response to a uniform doubling of stratospheric H₂O concluded that such moistening could reduce stratospheric temperature and increase the strength of the BDC; it could also result in the tropospheric westerly jets becoming stronger and storm tracks shifting poleward (Maycock et al., 2013). Since the HT-HH injection is ∼10% of the stratospheric H₂O burden, a dynamical response of lesser magnitude than that found by Maycock et al. (2013) would be expected.

H₂O enters the stratosphere primarily in the tropics, where it freeze-dries at the cold point tropopause (Brewer, 1949). This mechanism gives rise to the “tape recorder,” whereby the annual cycle in tropopause temperatures is imprinted in alternating bands of dry and moist air rising in the tropical stratosphere (Mote et al., 1996). By short-circuiting the pathway through the cold point, HT-HH has disrupted this “heartbeat” signal (Figure 5a).

Consistent with the freeze-drying mechanism, unusually low tropopause temperatures around 2001 led to a sharp drop in the amount of H₂O entering the stratosphere (e.g., Randel et al., 2006; Rosenlof & Reid, 2008; Figure 5). This dry anomaly propagated via the BDC (Randel et al., 2006; Urban et al., 2014), slowly rising through the stratosphere and moving toward the poles. Using the propagation of the 2001 H₂O drop as described by Brinkop et al. (2016) as an analog for the transport of the HT-HH plume, we expect that ascent could carry volcanic H₂O to 10 hPa within ∼9 months. The excess H₂O could arrive in northern and southern midlatitudes in ∼18 and ∼24 months, respectively, over a broad domain in the upper stratosphere. Since part of the plume has entered the lower branch of the BDC, the elevated H₂O may reach lower stratospheric midlatitudes within a few months. The timescale for complete dissipation of the plume may be 5–10 years (Hall & Waugh, 1997).

Radiative calculations of the sudden drop in H₂O of ∼0.4 ppmv (at 100 hPa) in 2001 (Figure 5b) demonstrated that the radiative forcing from even small variations in lower stratospheric H₂O could induce decadal-scale changes in global-mean surface temperature (e.g., Solomon et al., 2010). The unprecedented HT-HH enhancement would correspond to ∼1.5 ppmv (at 31 hPa) if averaged over 60°S–60°N.

Previous studies of the radiative effects of stratospheric H₂O perturbations, including direct volcanic injection, have shown that they can cause surface warming (e.g., Joshi & Jones, 2009; Rind & Lonergan, 1995). As established in Section 3, the HT-HH eruption was unusual in that it injected extremely large amounts of H₂O. Preliminary climate model simulations (see Supporting Information S1 for details) suggest an effective radiative forcing (e.g., Forster et al., 2001; Myhre et al., 2013; Smith et al., 2020; Wang et al., 2017) at the tropopause of +0.15 Wm⁻² due to the stratospheric H₂O enhancement (Figure S3b in Supporting Information S1). For comparison, the radiative forcing increase due to the CO₂ growth from 1996 to 2005 was about +0.26 Wm⁻² (Solomon et al., 2010).

The HT-HH H₂O enhancement will exert a positive radiative forcing on the surface, offsetting the surface cooling caused by the aerosol radiative forcing (e.g., Sellitto et al., 2022; Zhang et al., 2022). Given the extraordinary magnitude of the HT-HH H₂O injection and the fact that its anticipated stratospheric residence time exceeds the typical 2–3 years timescale for sulfate aerosols to fall out of the stratosphere (Joshi & Jones, 2009), HT-HH may be the first volcanic eruption observed to impact climate not through surface cooling caused by volcanic sulfate aerosols, but rather through surface warming caused by excess H₂O radiative forcing.

In summary, MLS measurements indicate that an exceptional amount of H₂O was injected directly into the stratosphere by the HT-HH eruption. We estimate that the magnitude of the injection constituted at least 10% of the total stratospheric H₂O burden. On the day of the eruption, the H₂O plume reached ∼53 km altitude. The H₂O injection bypassed the cold point tropopause, disrupted the H₂O tape recorder signal, set a new record for H₂O injection height in the 17-year MLS record, and could alter stratospheric chemistry and dynamics as the long-lived H₂O plume propagates through the stratosphere in the BDC. Unlike previous strong eruptions in the satellite era, HT-HH could impact climate not through surface cooling due to sulfate aerosols, but rather through surface warming due to the excess stratospheric H₂O forcing. Given the potential high-impact consequences of the HT-HH H₂O injection, it is critical to continue monitoring volcanic gases from this eruption and future ones to better quantify their varying roles in climate.