Anticyclonic confinement of volcanic material
Already on 25 June, three days after the eruption, the SO2 plume started to cluster into distinct isolated structures clearly discernible in high-resolution TROPOMI SO2 maps (Fig. 1a, see also Fig. S1 for the complete SO2 map sequence). The two largest structures, denoted I and II, appear to merge on 28 June (Fig. 1b) then separate again on 1 June (Fig. 1c) but continue to move alongside each other during the first week of July (Fig. 1c,d). The structure II eventually moved towards the Canadian Arctic [9], whereas the structure I moved towards the Asian continent. Here we focus on the evolution of the structure I that showed remarkable stability and could be followed by various satellite sensors for months ahead [3,9,14].
Figure 1e shows the spatiotemporal evolution of this particular plume starting on 4 July, after it had traveled over northeast Siberia (Figure S1). The plume then made a u-turn over Alaska, crossed the Northern Pacific and returned to its source location by mid-July. During the second half of July, the SO2 plume is observed as an isolated inverted comma-shaped structure with a circular core moving along the Eastern flank of the Asian Summer Monsoon Anticyclone before entering the subtropical jet, which rapidly advected the structure all across Eurasia and North Africa in 10 days. Over time, the maximum column amount of SO2 in the plume gradually decreases by photochemical processes that convert SO2 into H2SO4 droplets with an e-folding timescale between 8-18 days [14] and 13-17 days [9]. Nevertheless, having preserved its compact shape, the plume could be followed in high-resolution TROPOMI SO2 total column measurements for eight weeks.
During its entire lifetime, the SO2 plume maintained a circular shape, with clear signs of continuous filamentation (erosion) at the edges as shown in Fig. 1e, and Fig. 2a,b,c. The compact circular shape is indicative of dynamical confinement, whereas the counterclockwise filamentation (or tailing) reveals the clockwise (anticyclonic) rotation of the plume. The filamentation at the edge of vortical structures is a common phenomenon in fluid dynamics associated with a vortex, especially in stably-stratified rotating fluids [26,27]. Indeed, it appears physically implausible that the compact shape of the plume could be preserved over such a long time in a highly dispersive environment [9] without a dynamical confinement mechanism, already demonstrated for the stratospheric smoke plumes [21,23,24,25]. The anticyclonic motion of the Raikoke volcanic plume is further corroborated by Aeolus satellite wind measurement and ECMWF ERA5 reanalysis. The phenomenon is hereinafter referred to as vorticized volcanic plume, VVP.
Vertical structure and composition of the VVP
The high-resolution transects through the VVP from late July to mid-August (Fig. 2 d,e,f) by the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) reveal a compact cloud of aerosol retaining a fairly compact shape over time and showing clear indication of the bottom-side elongation (tailing), characteristic of rotational motion of a plume [21]. The tail is clearly visible in the corresponding TROPOMI horizontal sections of the VVP (Fig. 2a,b,c) and although the CALIOP nighttime transects are more than ten hours out of phase with TROPOMI, complicating the cross-attribution of TROPOMI and CALIOP structures, the persistent anticyclonic filament observed by TROPOMI is most likely associated with the bottom-side elongation revealed by CALIOP.
According to CALIOP data, the VVP had a vertical extent of 1 - 1.5 km, whereas the horizontal extent of the associated SO2 structure increases from 200-300 km in early July to nearly 400 km by late July, that is when the VVP gets entrained by the subtropical jet. The horizontal extent of the VVP is thus two orders of magnitude larger than its vertical extent (aspect ratio or diameter over thickness of 133:1 to 400:1), or in other words, the VVP is a flat lense-like structure maintaining its shape for a long time. We note that the aspect ratio of the 6-wk old Raikoke VVP is about 3 times smaller compared to that of the largest smoke-charged vortices (SCV) produced by the Canadian wildfires in 2017 [25] and Australian bushfires in 2019/2020 [21].
Despite the very narrow vertical extent and the synoptic-scale horizontal extent of a few hundred kilometers, the primary VVP contained as much as 0.3±0.1 Tg or 24% of the total SO2 mass injected into the stratosphere, as derived from TROPOMI data (Fig. S2). Similar estimate of SO2 mass is obtained for the secondary VVP (0.39±0.12 Tg). Thus, around 54% of the Raikoke emission was contained by the long-lived vortices.
Aeolus observations of the VVP
While the persistent compact shape and circular filamentation of the structure are indicative of the vortical motion, direct observational evidence for the rotation of the plume is exclusively provided by ALADIN wind lidar onboard Aeolus satellite that sampled the VVP several times during its lifetime. Figure 3 displays two cases of Aeolus sampling across the 17-d and 40-d old VVP. Aeolus measures the horizontal wind component transverse to the orbital plane, i.e. a quasi-zonal component except at high latitudes, using both molecular (Rayleigh) and particulate (Mie) backscattering. Fig. 3a,c show the vertical cross section of the Mie wind measured inside the aerosol plume superimposed on the background flow derived by aggregating all Rayleigh winds within a 3-day/30° longitude window.
The compact Mie-wind feature between 16-18 km surmounting the meridionally more dispersed aerosol plume in Fig. 3a reveals a very particular horizontal gradient of wind speed and direction, changing from easterly at the southern edge to westerly at the northern edge, in a notable contrast with the near-zero background flow, thereby providing a clear indication of the anticyclonic motion of the plume. Another case presented in Fig. 3c reveals a very similar Mie wind pattern inside the vortex already at 21-22 km altitude (VVP self-lofting is discussed hereinafter). In both cases (and in all the other ones not shown) the absolute wind speed increases towards the structure’s edges, which is exactly what one would expect for a rotating vortex. The anomalous wind speed at the edges (i.e. with respect to the background flow) is estimated at 9 ± 4 m s-1, which translates into 38 hours turnover time, surprisingly similar to that of the largest Australian SCV with the 36-hour turnover time [21].
Additional evidence for the vortex was provided by an upper-air radiosounding at Hilo, Hawaii on 25 September, that is when a nearby lidar at Mauna Loa detected a strongly-scattering aerosol layer at 26 km altitude [3] coinciding with a notable anomaly in zonal wind velocity reaching 15 m s-1 (Fig. S3).
Diabatic lofting and circumglobal transport of the confined plume
The three-dimensional tracking of the rising plume from mid-July to late-September 2019 was provided by ref.[3,14]. Here we revisit this aspect and provide the VVP tracking since the eruption date (22 June) until mid-October from three satellite data sets: TROPOMI SO2 measurements of the circular plume during the first 8 weeks; CALIOP detections of the high scattering ratio anomaly as well as the Ozone Mapping and Profiler Suite (OMPS) Limb Profiler (LP) aerosol extinction profiles, enabling a robust tracking of the primary plume up to 15 weeks past the eruption.
During the first three weeks of its lifetime, the primary VVP has changed the heading direction three times in the meridional plane (Fig. 4a) and seven times in the zonal plane (Fig. 4b) before entering the steady flow within the subtropical jet, which was followed by a triple circumnavigation and a progressive equatorward shift of the plume. The plume could be reliably tracked by satellite sensors for more than 100 days, during which it has traveled the distance of 139,000 km, that is the longest travel distance for a coherent aerosol plume (cf. 66,000 for the largest Australian SCV [21]).
The VVP tracking in the vertical dimension using aerosol vertical profiling by OMPS-LP and CALIOP are in good agreement, enabling an accurate estimation of the diabatic lifting rate of the stratospheric aerosol cloud. Figure 4c shows a steady diabatic ascent from around 400 K (15 km) to nearly 700 K (27 km) in three months with an average climb rate of 2.7 K day-1 (110 m day-1). This is notably lower than the mean diabatic rise rates for the Canadian SCV (5.6 K/day)20 and Australian SCV (5.9 K day-1) [21]. The ascent of Raikoke VVP, although mostly linear in time, showed a sensible acceleration around D+60 time, when the diabatic rise rate increased to 4.5 K day-1 (Fig. S4). This is the period of the first VVP overpass above the Asian monsoon region characterised by frequent deep convection and extensive upper-level clouds during Boreal summer [28].
Figure 4d suggests that the diabatic ascent occurred in a stepwise manner, where the vertical steps correspond to the Asian monsoon overpass periods (see also Fig. S4). It is conceivable that the abundant highly-reflective convective clouds in the Asian monsoon region provided additional heating to the plume thereby accelerating its diabatic rise. Indeed, the multiple scattering of shortwave radiation over optically-thick convective clouds can result in significant warming of thin layers of absorbing aerosols [29,30,31].
Internal heating of the plume
In order to quantify the magnitude of internal heating, we used global navigation satellite system (GNSS) radio occultation (RO) temperature profiles collocated in space and time with VVP detections by OMPS-LP and CALIOP during the mid-July - early-August period when the VVP was transiting across Asia. Figure 5a displays a composited profile of temperature anomaly in the vertical coordinates relative to the plume centroid computed as deviation from the background temperature profile. The GNSS-RO measurements reveal a statistically-significant 2-km thick warm anomaly reaching 1 K near the plume’s centroid. Additional evidence for the aerosol plume heating was provided by an upper-air sounding in Israel on 5 August that sampled the VVP and revealed a 3 K warm anomaly (blue curve in Fig. 5a). The warm anomaly matches well with the stratospheric aerosol cloud detected by a nearby MicroPulse Lidar NETwork (MPLnet) lidar at Sede Boker, Israel (Fig. S5).
The horizontal extent of the warm anomaly was derived from ERA5 reanalysis which assimilates the GNSS-RO measurements. The longitude-altitude section of the temperature anomaly in ERA5 (Fig. 5b) agrees with the observed structure of the temperature anomaly and suggests its horizontal extent of 510 km, which is consistent with the diameter of the SO2 circular cloud estimated at 400 km.
The ERA5-derived warm anomaly reaching +0.7 K is readily comparable with the heating of +0.5 K reported for the Canadian SCV [25] yet considerably smaller than +4 K associated with the Australian SCV [21]. The heating of smoke plumes is mainly caused by the highly-absorptive black carbon particles. The volcanic plumes are primarily composed of weakly-absorbing sulfuric acid droplets produced by SO2 oxidation as well as the stronger-absorbing ash particles [2]. Estimates of the fraction of aerosol backscatter by ash particles inside the VVP, derived from CALIOP backscatter and depolarization (following ref.[32]) suggest a high ash backscatter fraction in the young VVP of up to 40% gradually decreasing to zero on a timescale of one month (Fig. S6), most likely due to progressive sulfate coating of ash particles [32,33]. A similar decay was observed for the other, non-rising plumes, however their ash backscatter ash fraction was a factor of two smaller compared to the rising VVP (Fig. S6).
While the diabatic self-lofting of the Raikoke confined plume was reported by ref.[3,14], here we provide the first observational evidence of the internal heating. The amount of internal heating caused by absorbing aerosols largely determines the diabatic lofting rate, however to maintain the heating at high degree for an extended time period, the absorbing material in the plume must remain at high concentration. This is conditioned by the dynamical confinement, which is a prerequisite for the VVP occurrence.
Potential vorticity of the VVP
The occurrence of long-lived anticyclonic structure such as the VVP suggests that some form of fluid dynamical conservation mechanism is at work here, of which potential vorticity (PV) is a likely candidate given the rotational nature of the VVP. The PV is to first order a conserved atmospheric quantity in the absence of diabatic processes. The VVP resides in the extremely stable summer stratosphere where the air masses with different PV tend not to mix. This way, an eruption-driven injection of large amounts of tropospheric air characterised by low PV directly into the high-PV stratospheric environment could trigger anticyclonic rotation as reported for the wildfire-generated SCVs that appear as low absolute PV kernels in the meteorological (re)analyses [24,25].
Note that the successful replication of such structures by operational meteorological models (that do not account for the volcanic or wildfire-induced vertical transport into the stratosphere), is linked with the assimilation of temperature and/or wind profiling from satellites and weather balloons [21].
In the case of Raikoke VVP, measuring 400×1.5 km at its maximum (i.e., 2-4 times smaller than the known SCVs [25]), the replication of such a small dynamical structure by meteorological analysis is more complicated due to limitations of the assimilated satellite observations. Nevertheless, the analysis of ERA5 data enabled identification of a low-PV anomaly associated with the VVP. Figure 6 reveals low-PV kernels (middle column) matching with the observed SO2 structures (left column). Similarly, the vertical PV cross sections (Fig. 6c,f) show a compact anomaly centered at the altitude of the satellite-derived aerosol plume. In Fig. 6a,b, one can see two distinct structures, of which the northern one is the primary VVP described in this study, whereas the southern one represents the secondary VVP (structure II in Fig. 1d) that was traveling side-by-side with the primary VVP during the first week of July before turning northeast. The low-PV structure associated with the primary VVP could be robustly followed in the reanalysis data until mid-July, becoming hardly discernible in ERA5 horizontal sections afterwards. Nonetheless, the composited vertical profile of PV anomaly derived by aggregating the reanalysis data from late-July to early August, when the VVP has been entrained by the subtropical jet, reveals a notable decrease of PV by 2 PVU symmetric to the aerosol plume centroid (Fig. 6i). A similar magnitude of PV anomaly (~3 PVU) was reported for the Canadian SCV [25].