In the wake of the Hayli Gubbi eruption Authors: Alexander Ukhov, Sateesh Masabathini, Marianthi Pateraki, Nikolaos Papagiannopoulos, Ibrahim Hoteit Published: 2026 ABSTRACT Explosive eruptions require rapid, observation-constrained estimates of ash and sulfur dioxide emissions for hazard assessment. We analyze the 23 November 2025 Hayli Gubbi eruption (Afar, Ethiopia) with WRF-Chem v4.8 and use backward trajectories seeded within TROPOMI aerosol index and sulfur dioxide (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document}) plumes to reconstruct time- and height-resolved source terms. The trajectory-derived ash source term, scaled to 0.8 Mt, captures the main ash transport corridor and the timing of the coarse-mode enhancement observed by AERONET at Nizwa University (Oman). For \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document}, a manually prescribed time-height scenario improves agreement with the TROPOMI plume relative to the first trajectory-based approximation. The simulation captures the large-scale \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} plume geometry and shows moderate agreement with aerosol optical depth and aerosol middle height. The corresponding \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} emission is ≈0.15 Mt. About 6% of the fine ash deposits within 28 h, including fallout to the southern Red Sea and Gulf of Aden. This study demonstrates a rapid, observation-constrained reconstruction for a single-eruption case study. Introduction The Hayli Gubbi volcano (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$13.50^\circ$$\end{document}N, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$40.72^\circ$$\end{document}E), located in the Afar Region of Ethiopia, is the southernmost volcano in the Mount Erta Ale range. The volcano is situated in a tectonically active sector of the Afar Depression, where extensional plate motion promotes magma ascent and sustained volcanism1. Before the recent event, the Smithsonian Institution’s Global Volcanism Program2 had no record of eruptions from this volcano since the beginning of the Holocene. However, on 23 November 2025, Hayli Gubbi experienced an explosive eruption. The initial blast was reported at around 08:30 UTC, and the ash plume spread eastward across the Red Sea; regional meteorological reports indicate transport over Yemen on 23 November and over Oman on 24 November1,3. In this study, we estimate the emitted amounts of ash and sulfur dioxide (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document}), reconstruct their time-varying vertical emission profiles, examine ash size distribution and fallout, and describe the early transport of the ash and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} plumes. We also explore whether modeled ash deposition over the sea coincides spatially with satellite-derived absolute Chl-a departures from climatology. The analysis is guided by satellite retrievals and ground-based observations and uses the online coupled chemistry-meteorology model WRF-Chem v4.84–6 as the primary modeling framework. Accurate representation of volcanic emissions is essential for simulating plume transport and assessing impacts on air quality and aviation. Although WRF-Chem v4.8 allows volcanic emissions to vary in time and height, such information is rarely available immediately after an eruption. Emission heights are therefore commonly estimated by comparing observations against simulations performed with different source configurations. More formal approaches include inverse modeling and data assimilation, which combine observations and transport models to constrain eruption source parameters more rigorously7–12. These methods provide a stronger framework for uncertainty treatment than heuristic source construction, but they are also more computationally and technically demanding. Here, we introduce a lighter trajectory-based source reconstruction approach. It should not be interpreted as formal inverse modeling or data assimilation; rather, it provides an observation-constrained estimate of time-dependent emission altitudes for rapid WRF-Chem simulations. To place the present workflow in the context of existing transport and source-attribution approaches, we briefly contrast it with widely used Lagrangian systems. HYSPLIT is a tool that supports backward trajectories, dispersion, and inverse applications, including volcanic ash forecasting and source-term estimation1,13,14. In its volcanic ash application, HYSPLIT represents ash using a prescribed particle-size distribution within a source layer extending from the volcano summit to the plume top15. More generally, the HYSPLIT modeling system supports multiple species, deposition, and generic conversion or decay terms16. However, we are not aware of a widely used backward-only tool that directly ingests \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} and AI plume (or other) 2D fields and converts them into WRF-Chem-ready time-height volcanic emission scenarios. The aim of the present work is not a full inter-model benchmark against HYSPLIT or other volcanic ash transport and deposition systems. Rather, the contribution of this paper is a WRF-Chem-specific source reconstruction workflow that links TROPOMI-based plume identification, backward-trajectory source attribution with Plume-traj, WRF-Chem-ready emission preprocessing with PrepEmisSources17, and forward WRF-Chem v4.8 simulation with 10 ash bins, time-varying emissions, refined \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} oxidation, sulfate settling, and ash-sulfate radiative feedback6. The methodological addition in this study is Plume-traj and its integration with PrepEmisSources and WRF-Chem to convert observed two-dimensional plume footprints into WRF-Chem-ready time-height source terms for a rapid case-study reconstruction. We do not present Plume-traj as a universal replacement for formal inversion, data-assimilation, or dispersion systems, but as a workflow tailored to observation-constrained source reconstruction for WRF-Chem-based volcanic plume simulations. Results Emission scenarios for ash and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} Fig. 1 shows the ash and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} emission scenarios used for the Hayli Gubbi simulation. The ash scenario was reconstructed with Plume-traj over 23 November 2025, 08:30 UTC to 24 November 2025, 09:30 UTC and contains 51 emission profiles updated every 30 min. It was scaled to the central 0.8 Mt ash case after ash-mass sensitivity runs with 0.7 and 0.9 Mt. The final \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} scenario started from the Plume-traj matrix and was adjusted manually against the TROPOMI \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} vertical column density at 11:00 UTC on 24 November 2025. It spans 23 November 2025, 08:30 UTC to 24 November 2025, 01:00 UTC, contains 34 profiles updated every 30 min, and releases 0.15 Mt of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document}. These profiles use Suzuki-type vertical source distributions18; implementation details are provided in the Code availability section. Ash emissions (Fig. 1a) are characterized by a dominant injection layer between approximately 7 and 15 km altitude, with peak emission centered near 7-10 km. The ash emission distribution reveals continuous ash release during the early and middle phases of the eruption, followed by a secondary, short-lived pulse at higher altitudes (≈11–14 km) toward the end of the eruption. The time-integrated ash emission profile confirms that the majority of the ash mass was injected between the mid-troposphere and upper troposphere. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} emissions (Fig. 1b) are represented by four discrete pulse groups in the manually adjusted scenario. Most of the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} mass is injected between ≈9 and 16 km, with the largest time-integrated contributions near 10-13 km and 15-16 km. The first pulse occurs shortly after 08:30 UTC on 23 November, a broader elevated pulse occurs from late morning to early afternoon, and two smaller pulses occur later on 23 November and near the transition to 24 November. This temporal structure should be interpreted as the selected observation-constrained \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} scenario, not as a uniquely constrained eruption source history.Fig. 1Reconstructed emission scenarios for ash and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document}. (a) Trajectory-derived ash source scaled to 0.8 Mt. (b) Manually adjusted \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} source totaling 0.15 Mt. Right panels show time-altitude emission rates (Mt \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {s}^{-1}$$\end{document}); left panels show time-integrated profiles (Mt). Dashed lines mark the approximate tropopause. The four \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} pulse groups are diagnosed in Supplementary Fig. S5. Reconstructed emission scenarios for ash and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document}. (a) Trajectory-derived ash source scaled to 0.8 Mt. (b) Manually adjusted \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} source totaling 0.15 Mt. Right panels show time-altitude emission rates (Mt \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {s}^{-1}$$\end{document}); left panels show time-integrated profiles (Mt). Dashed lines mark the approximate tropopause. The four \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} pulse groups are diagnosed in Supplementary Fig. S5. Plume evolution and transport Figure 2 illustrates the ash and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} plumes, represented as concentrations averaged between \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {10-25}^\circ$$\end{document}N, as simulated by WRF-Chem on 24 November 2025 at 12:00 UTC, shortly after the TROPOMI overpass. This snapshot shows the post-emission transport structure of the plume rather than the eruptive phase itself. The vertical profile in Fig. 2a is calculated by averaging concentrations over \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {10-25}^\circ$$\end{document}N at each model altitude level. The cross section reveals a well-defined, vertically structured volcanic plume extending from \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\approx 40^\circ$$\end{document}E to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$70^\circ$$\end{document}E (see Fig. 2b). Ash is concentrated mainly in a mid-tropospheric band, with the largest averaged concentrations near 9-10 km and detectable fine ash extending from the lower troposphere to ≈16 km. Here, the ash field represents the fine-ash fraction remaining aloft after transport and size-dependent settling, rather than the full emitted ash population. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} overlaps the ash plume but is shifted upward in the domain-mean profile, with pronounced structure between ≈10 and 18 km and additional lower-altitude \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} along parts of the transport corridor. Relative to the time-integrated injection profiles (Fig. 1a,b), the domain-mean ash maximum is displaced upward by roughly 1–2 km. This behavior may reflect the combined effects of plume evolution, preferential removal of larger particles, and possible radiative-dynamical adjustment of the ash cloud19. In contrast, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} remains concentrated within the upper-tropospheric range sampled by the prescribed \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} pulses.Fig. 2Averaged (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$10-25^\circ$$\end{document}N) vertical profiles (a) and longitude-altitude cross section (b) of ash (gray) and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} (blue) concentrations (μg \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {m}^{-3}$$\end{document}) on 24 November 2025 at 12:00 UTC. Concentration contours correspond to 1 μg \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {m}^{-3}$$\end{document}. Dashed lines indicate the tropopause, and terrain is shown in brown. Averaged (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$10-25^\circ$$\end{document}N) vertical profiles (a) and longitude-altitude cross section (b) of ash (gray) and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} (blue) concentrations (μg \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {m}^{-3}$$\end{document}) on 24 November 2025 at 12:00 UTC. Concentration contours correspond to 1 μg \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {m}^{-3}$$\end{document}. Dashed lines indicate the tropopause, and terrain is shown in brown. Ash aerosol volume size distribution Figure 3 compares AERONET aerosol volume size distributions (VSDs) at Nizwa University (Oman) with ash-bin VSDs calculated from the WRF output. AERONET observations were grouped into hourly slots near 08:00, 09:00, 10:00, and 12:00 UTC. For each slot, the background mean and standard deviation were computed over 7 November-9 December 2025 after excluding 23-25 November, and then compared with the AERONET retrieval on 24 November. The difference panels isolate the eruption-day enhancement relative to the background mean, with uncertainty propagated from the background variability and the 24 November proxy uncertainty. Starting near 09:00-10:00 UTC on 24 November, the AERONET VSD exhibits an enhanced coarse-mode contribution at radii of approximately 2-10 μm. WRF predicts a comparable emergence of coarse ash at the nearest AERONET times for the central 0.8 Mt ash case, while the 0.7 and 0.9 Mt ash-mass sensitivity runs indicate the sensitivity of the modeled ash-bin amplitudes. This comparison supports the simulated plume-arrival timing and the emergence of a coarse-mode aerosol signal, but it does not uniquely validate the emitted ash size distribution or total ash mass. Fig. 3AERONET aerosol volume size distributions (VSDs) and WRF ash-bin comparison at Nizwa University. Rows show UTC hour groups near 08:00, 09:00, 10:00, and 12:00. Columns show the AERONET background mean, 24 November retrievals, and their difference. Gray bars show the central 0.8 Mt WRF ash-bin VSD; whiskers show the 0.7 and 0.9 Mt ash-mass sensitivity bounds. Shaded envelopes indicate \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm 1\sigma$$\end{document} uncertainty, with difference uncertainty propagated in quadrature. AERONET aerosol volume size distributions (VSDs) and WRF ash-bin comparison at Nizwa University. Rows show UTC hour groups near 08:00, 09:00, 10:00, and 12:00. Columns show the AERONET background mean, 24 November retrievals, and their difference. Gray bars show the central 0.8 Mt WRF ash-bin VSD; whiskers show the 0.7 and 0.9 Mt ash-mass sensitivity bounds. Shaded envelopes indicate \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm 1\sigma$$\end{document} uncertainty, with difference uncertainty propagated in quadrature. Comparison with satellite retrievals Figure 4 compares the simulated \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} and ash plumes against TROPOMI observations on 24 November 2025 at 11:00 UTC. Figure 4a,b compare modeled and observed \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} vertical column density. Both fields show the main curved plume originating from the Hayli Gubbi volcano, advected northeastward across the Gulf of Aden and the southern part of the Arabian Peninsula, and subsequently transported toward the northern Arabian Sea. The dominant transport pathway follows the upper-tropospheric westerlies between \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\approx 10-25^\circ$$\end{document}N, producing a filamentary plume extending over 2000 km downwind of the source. However, WRF produces a narrower high-column \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} filament and does not reproduce the full broad low-column \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} area visible in TROPOMI, especially west of the volcano and over the western Gulf of Aden. We were also unable to identify a manually adjusted \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} pulse that transports the plume directly northward from the volcano, as seen in TROPOMI (Fig. 4b), suggesting that this near-source transport pathway is not represented in the driving meteorological fields or in the modeled local plume dynamics. The manually adjusted \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} scenario improves agreement with TROPOMI relative to the first trajectory-derived approximation shown in Supplementary Fig. S0. For the final \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} case, the spatial correlation is 0.435, RMSE is 4.861 DU, and mean bias is -0.736 DU. The first trajectory-derived \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} approximation reproduced the broad plume pathway but yielded weaker quantitative agreement with TROPOMI (R=0.270; RMSE=10.541 DU; mean bias=2.217 DU; Supplementary Fig. S0). To make the manually adjusted \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} source-to-plume mapping transparent, Supplementary Fig. S5 shows diagnostic WRF-Chem runs in which the four \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} pulse groups visible in Fig. 1b were emitted separately and their resulting column footprints were overlaid on the final \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} column field. This diagnostic identifies which source pulses contribute to each part of the modeled plume. The modeled ash cloud spatial footprint is assessed using TROPOMI AOD at 758 nm (Fig. 4c,d) and aerosol middle height (Fig. 4e,f). To account for background aerosol not represented by the volcanic WRF-Chem configuration, the mean MODIS AOD at 550 nm for 7 November-9 December 2025, excluding 23-25 November (Supplementary Fig. S4), was added to the WRF volcanic AOD before spectral adjustment to 758 nm. The resulting AOD field partly captures the main plume corridor from the source toward the northeastern Arabian Peninsula, but the local maxima and patchiness differ from TROPOMI, and the overall agreement remains modest (R=0.228; BIAS=0.046; RMSE=0.234). This background correction provides context for the optical comparison, but it does not remove uncertainty in the volcanic ash AOD. We therefore interpret the TROPOMI AOD comparison primarily as a plume-footprint consistency check rather than as strict validation of ash optical-depth magnitude. Animation (see Supplementary S3) provides qualitative support that the satellite-detected plume and the modeled ash transport have similar spatial patterns and downwind extent. The aerosol middle height comparison indicates that WRF captures the broad plume-height pattern (Fig. 4e,f), with moderate correlation (R=0.456), but with higher modeled plume mass-center heights overall (BIAS=2.086 km; RMSE=4.305 km). The largest discrepancy occurs over central Yemen, where TROPOMI indicates a low aerosol layer within the lowest ≈2 km that is not reproduced by WRF. This feature is also weak or absent in the AOD comparison, suggesting that it may reflect low-level background aerosol that is not represented in the simulation. Because TROPOMI products help constrain the source reconstruction, these comparisons should be interpreted as consistency checks rather than as fully independent validation. Fig. 4Comparison between WRF (left column) and TROPOMI (right column) at 11:00 UTC on 24 November 2025. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} vertical column density (DU) is shown in (a,b), background-adjusted aerosol optical depth at 758 nm in (c,d), and aerosol middle height (km) in (e,f). The WRF AOD in (c) includes the MODIS background-mean AOD used for the optical comparison. Insets in (a), (c), and (e) provide summary statistics. The red triangle marks the Hayli Gubbi volcano, and the blue circle in (c,d) marks the AERONET site at Nizwa University, Oman. Maps were generated by the authors using Matplotlib v3.7.1 (https://matplotlib.org/); coastlines and national borders were drawn from Natural Earth public-domain data (https://www.naturalearthdata.com/). Comparison between WRF (left column) and TROPOMI (right column) at 11:00 UTC on 24 November 2025. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} vertical column density (DU) is shown in (a,b), background-adjusted aerosol optical depth at 758 nm in (c,d), and aerosol middle height (km) in (e,f). The WRF AOD in (c) includes the MODIS background-mean AOD used for the optical comparison. Insets in (a), (c), and (e) provide summary statistics. The red triangle marks the Hayli Gubbi volcano, and the blue circle in (c,d) marks the AERONET site at Nizwa University, Oman. Maps were generated by the authors using Matplotlib v3.7.1 (https://matplotlib.org/); coastlines and national borders were drawn from Natural Earth public-domain data (https://www.naturalearthdata.com/). Ash deposition and exploratory chlorophyll departure comparison Figure 5a presents the simulated cumulative fine-ash deposition at 12:00 UTC on 24 November 2025, corresponding to the end of the WRF-Chem simulation. The ash fallout was calculated following the method described in20. In the central 0.8 Mt ash case, the total deposited ash mass over the model domain amounts to 0.0509 Mt, with an approximately ±12% spread across the 0.7 and 0.9 Mt ash-mass sensitivity runs. Fine ash deposition is concentrated over the southern Red Sea (0.0176 Mt, approximately ±12%), the northern part of the Gulf of Aden (0.0001 Mt, approximately ±12%), Ethiopia, Eritrea, and the south-western part of Yemen, reflecting the dominant northeastward transport of the plume during the first ≈28 h after eruption. Peak deposition (≳1 g \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {m}^{-2}$$\end{document}) occurs close to the source region and along the southern Red Sea, while lower but spatially extensive deposition (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$10^{-3}$$\end{document}-\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$10^{-1}$$\end{document} g \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {m}^{-2}$$\end{document}) spreads eastward across Yemen. Figure 5b shows an exploratory comparison with the normalized absolute Chl-a departure on 24 November 2025 relative to climatology (see Supplementary section S1.2 and Supplementary Fig. S1). Filamentary departures from climatology (up to 5σ) occur in the southern Red Sea near \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\approx 15^\circ$$\end{document}N, with a broader patch south of Yemen in the Gulf of Aden. The central, lower, and upper ash-fall contours overlap the main ocean-color perturbation south of Yemen, indicating that this spatial co-location is retained across the ash-mass sensitivity runs. Notably, the southern Red Sea does not exhibit departures as strong as those near Yemen in the Gulf of Aden, despite stronger modeled fine-ash deposition. One possible explanation is that ash deposited closer to the source contained a larger contribution from the faster-settling part of the fine-ash spectrum, whereas the more remote Gulf of Aden may have received a finer fraction with a longer near-surface influence on the ocean color signal. However, this explanation remains tentative because post-depositional processes are not resolved in the present study. We emphasize that satellite-derived Chl-a departures alone are not sufficient evidence of phytoplankton growth. We therefore use these maps only to document spatial coincidence and not to infer a confirmed biological response. Any biological response depends on factors not explicitly resolved here, including background nutrient limitation, the solubility and bioavailability of ash-derived nutrients, or optical complexity due to suspended ash that can be misinterpreted as a Chl-a departure21,22. A full attribution would require dedicated in situ observations and/or advanced remote sensing approaches23, which are not available in our study region and are beyond the scope of this work.Fig. 5Fine-ash fall and exploratory ocean-color comparison on 24 November 2025. (a) Fine-ash fall (g \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {m}^{-2}$$\end{document}) at 12:00 UTC for the 0.8 Mt case. (b) Normalized absolute daily Chl-a departure (σ units). Ash-fall \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$10^{-3}$$\end{document} g \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {m}^{-2}$$\end{document} contours show central (solid), 0.9 Mt (dotted), and 0.7 Mt (dashed) runs. The red triangle marks the volcano. Maps were generated by the authors using Matplotlib v3.7.1 (https://matplotlib.org/); coastlines and national borders were drawn from Natural Earth public-domain data (https://www.naturalearthdata.com/). Fine-ash fall and exploratory ocean-color comparison on 24 November 2025. (a) Fine-ash fall (g \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {m}^{-2}$$\end{document}) at 12:00 UTC for the 0.8 Mt case. (b) Normalized absolute daily Chl-a departure (σ units). Ash-fall \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$10^{-3}$$\end{document} g \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {m}^{-2}$$\end{document} contours show central (solid), 0.9 Mt (dotted), and 0.7 Mt (dashed) runs. The red triangle marks the volcano. Maps were generated by the authors using Matplotlib v3.7.1 (https://matplotlib.org/); coastlines and national borders were drawn from Natural Earth public-domain data (https://www.naturalearthdata.com/). Discussion and conclusion The 23 November 2025 eruption of Hayli Gubbi, the southernmost volcano of the Erta Ale range in Ethiopia, was an exceptional event for the Afar region: an explosive, high-column eruption from a volcano with no Holocene activity. The comparison with satellite and ground-based observations supports the plausibility of the source scenarios and simulated transport, but does not constitute fully independent validation of all source parameters. TROPOMI products provide important consistency checks on plume geometry and altitude, but \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} and AI fields are also used in the source-construction workflow. AERONET provides a more independent constraint at Nizwa University, mainly on plume-arrival timing and the emergence of a coarse-mode aerosol signal, but it does not uniquely determine the emitted ash size distribution or ash mass. In particular, the trajectory-derived ash scenario used here corresponds to a release of ≈0.8 Mt. The reconstructed ash matrix spans approximately 25 h, but most of the emitted ash mass is concentrated during the first approximately 10 h, with smaller later pulses. The manually adjusted \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} scenario releases ≈0.15 Mt. Ash was injected predominantly between ≈7 and 15 km, with a short late-stage contribution reaching approximately 11-14 km, whereas \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} exhibited a broader, higher emission layer at ≈9-16 km. On the following day, the simulated fine-ash plume showed a broader vertical distribution than \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document}, reflecting the combined effects of source-term structure, transport, and size-dependent settling. The \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} maximum remained near its injected level, and the plume remained predominantly tropospheric, implying limited direct stratospheric impact. Deposition calculations indicate that ≈6% of the central emitted ash mass was deposited within the first ≈28 hours, with hotspots near the source and along the southern Red Sea (0.0176 Mt), Yemen, and the Gulf of Aden (100 tons). Modeled fine ash fallout over the ocean coincides with pronounced normalized absolute Chl-a departures the day after the eruption, and the deposition contour overlap is preserved across the ash-mass sensitivity runs. The spatial coincidence between the deposition pattern and the ocean color signal is suggestive but does not demonstrate a biological response to the Hayli Gubbi eruption, and retrieval artifacts and other confounders cannot be excluded. The contrast between the weaker departure in the southern Red Sea and the stronger signal near Yemen in the Gulf of Aden may be influenced by differences in the fraction of deposited particles, post-depositional residence, near-surface ocean circulation, and background biogeochemical conditions. This caveat is particularly important in the context of the intense and recurrent volcanism in the southern Red Sea region24. We introduce and apply the Plume-traj workflow to reconstruct time- and height-resolved volcanic emissions for use in WRF-Chem v4.8. In this approach, parcel seeds are initialized within the observed plume and traced back to the volcano to derive an emission time-height matrix, which is then ingested into WRF-Chem through the PrepEmisSources preprocessor. The workflow provides a practical alternative to purely manual source tuning, but it does not constitute formal inverse modeling or data assimilation. In this case study, the trajectory-derived ash scenario was retained after ash-mass sensitivity testing, whereas the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} scenario was manually adjusted after the first trajectory-derived approximation showed limited agreement with the TROPOMI \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} field. Advantages, limitations, and operational scope The workflow provides a practical way to convert observed plume geometry into time-height source terms for WRF-Chem v4.8. However, this study demonstrates a case-study rapid-response workflow rather than a fully operational observatory tool. Several limitations should be emphasized. First, the ash mass range explored here is limited to ash-mass sensitivity runs with a fixed ash size distribution and does not constitute full uncertainty quantification. Second, the reconstruction depends on plume-mask thresholds, the chosen 20 km source cylinder around the volcano, and the meteorological wind fields used for the backward trajectories. Third, unlike formal data-assimilation approaches such as FALL3D+PDAF12, the present workflow does not provide posterior uncertainty estimates for the source term. Fourth, TROPOMI products are used both to constrain the source-construction workflow and to evaluate plume structure, so these comparisons are consistency checks rather than fully independent validation. Fifth, no independent lidar, radar, aircraft, or other column-height constraint was available here to validate the injection-height and transported-height diagnostics. Finally, the chlorophyll analysis is exploratory and documents spatial coincidence only; it does not establish a causal biological response to ash deposition. Methods Observations We used ground-based and satellite observations to evaluate the WRF simulation. In particular, we rely on the Tropospheric Monitoring Instrument (TROPOMI)25, Aerosol Robotic Network (AERONET)26 observations, and satellite observations of chlorophyll. Moreover, TROPOMI data were used to perform back-trajectory analysis with the Plume-traj tool introduced here. TROPOMI retrievals We used \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} vertical column density from TROPOMI, retrieved under the assumption of a 15 km \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} plume height27, since the eruption injected \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} into the upper troposphere. To facilitate comparison with TROPOMI retrievals, the modeled \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} column field was vertically weighted using the TROPOMI averaging kernels, which account for the altitude-dependent satellite sensitivity. Similar to11, we also employed the TROPOMI absorbing Aerosol Index (AI) to determine the spatial distribution of the volcanic ash plume. The AI is a qualitative indicator of elevated aerosol layers in the atmosphere. It can detect the occurrence of UV-absorbing aerosols, such as desert dust and volcanic ash. Positive AI values indicate the presence of UV-absorbing aerosol. To characterize background aerosol variability around the eruption, we computed the mean and standard deviation of MODIS AOD at 550 nm over 7 November-9 December 2025 after excluding 23-25 November (Supplementary Fig. S4). For comparison with the TROPOMI Aerosol Optical Depth (AOD) at 758 nm, the MODIS background mean at 550 nm was added to the WRF volcanic AOD and then spectrally adjusted to 758 nm, see Supplementary S1.1. TROPOMI also provides a retrieval of the aerosol middle height, which was used for comparison with the value calculated from WRF model output, following19. Backward trajectories initiated from AI-enhanced pixels and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} vertical column density fields were used to constrain ash and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} injection heights and times. Before applying the TROPOMI \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} data in the backward trajectory analysis, we first removed the contributions from anthropogenic sources located along the Iranian coastline, the Red Sea coast, and the Arabian Gulf, which represent the primary emission hotspots in the Middle East region28,29. AERONET AERONET retrieval algorithm provides column-integrated aerosol volume size distribution (VSD) dV/dlnr (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mu \text {m}^3/\mu \text {m}^2$$\end{document}) on 22 logarithmically equidistant discrete points in the range of radii between 0.05 and 15 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mu \text {m}}$$\end{document}30. The VSD provides key information on aerosol size, including mineral dust and volcanic ash. This product was used for comparison with VSDs computed from WRF output, following31. We use AERONET’s VSDs from Nizwa University in Oman (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$22.91^\circ$$\end{document}N, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$57.674^\circ$$\end{document}E). For each hourly slot used in Fig. 3, the 24 November uncertainty envelope uses the corresponding background-hour standard deviation as a proxy; the difference-panel uncertainty is propagated in quadrature from the background and 24 November envelopes. Satellite-derived Chlorophyll-a Volcanic ash plumes have been related to phytoplankton fertilization through nutrient input to the ocean euphotic zone. Satellite observations have documented chlorophyll-a (Chl-a) perturbations following several volcanic eruptions32,33. To explore whether a comparable signal followed the Hayli Gubbi eruption, this study utilizes ocean color data from the Sentinel-3 Ocean and Land Colour Instrument (OLCI) on Sentinel-3A/3B34. We use level 2 Water Full Resolution Chl-a products at 300 m resolution. Chl-a concentrations are retrieved using the Ocean Colour 4 for MERIS (OC4Me) algorithm35. To characterize baseline phytoplankton conditions, a climatological mean and standard deviation of Chl-a were computed for the analysis (see Supplementary S1.2). Model The WRF-Chem simulation (the terms WRF-Chem and WRF are used interchangeably later in the text) runs from 06:00 UTC on 23 November 2025 to 12:00 UTC on 24 November 2025, covering 30 hours and producing hourly output. The domain is centered at \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$23.5^\circ$$\end{document}N, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$47.2^\circ$$\end{document}E and contains 991×801 grid points in the zonal and meridional directions, respectively. The domain spans the Middle East at a horizontal resolution of 4 km. The vertical grid consists of 52 levels extending to 30 hPa (≈23.2 km), with the lowest four layers contained within the first 100 m above the surface. Model time step is 20 s. Meteorological initial and lateral boundary conditions are taken from the NCEP Global Forecast System (GFS) operational analyses and forecasts36, available every 3 hours at \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$0.25^\circ \times 0.25^\circ$$\end{document} resolution. The choice of the microphysics, radiation, planetary boundary layer, and land-surface parameterizations follows6, except that cumulus convection is not parameterized here because the 4 km horizontal grid spacing is within the convection-permitting regime. The simulation includes fully interactive, two-way coupling between volcanic aerosols and meteorology, allowing the radiative effects of volcanic ash and secondary sulfate to influence plume dynamics and the regional circulation.Table 1Ash mass fractions by particle-size bin.\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {Bin}$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r_{\min }\ (\mu \text {m})$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r_{\max }\ (\mu \text {m})$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r_{\textrm{eff}}\ (\mu \text {m})$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {Mass fraction (\%)}$$\end{document}150010005000.022505003750.03125250187.50.0462.512593.750.0531.2562.546.8750.5615.62531.2523.4357.377.812515.62511.7232.683.906257.81255.8642.291.9531253.906252.9315.8100.019551.9531250.9751.7Fractions are expressed as percentages of the total erupted ash mass. Each bin is defined by its lower and upper particle radii (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r_{\min }$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r_{\max }$$\end{document}) and by the representative particle radius (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r_{\textrm{eff}}$$\end{document}) used to calculate the gravitational settling velocity. All size limits in the table are given in terms of particle radius, where particle-size thresholds are discussed in the text in terms of diameter, d = 2r. Table 1Ash mass fractions by particle-size bin.\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {Bin}$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r_{\min }\ (\mu \text {m})$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r_{\max }\ (\mu \text {m})$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r_{\textrm{eff}}\ (\mu \text {m})$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {Mass fraction (\%)}$$\end{document}150010005000.022505003750.03125250187.50.0462.512593.750.0531.2562.546.8750.5615.62531.2523.4357.377.812515.62511.7232.683.906257.81255.8642.291.9531253.906252.9315.8100.019551.9531250.9751.7Fractions are expressed as percentages of the total erupted ash mass. Each bin is defined by its lower and upper particle radii (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r_{\min }$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r_{\max }$$\end{document}) and by the representative particle radius (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r_{\textrm{eff}}$$\end{document}) used to calculate the gravitational settling velocity. All size limits in the table are given in terms of particle radius, where particle-size thresholds are discussed in the text in terms of diameter, d = 2r. Ash mass fractions by particle-size bin. Fractions are expressed as percentages of the total erupted ash mass. Each bin is defined by its lower and upper particle radii (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r_{\min }$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r_{\max }$$\end{document}) and by the representative particle radius (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r_{\textrm{eff}}$$\end{document}) used to calculate the gravitational settling velocity. All size limits in the table are given in terms of particle radius, where particle-size thresholds are discussed in the text in terms of diameter, d = 2r. To determine the ash mass fractions for each ash bin, we follow37 and19, assuming that the ash particle size is log-normally distributed with a median radius of 2.4 μm and a geometric width of 1.8. The computed ash mass fractions and the size range of each ash bin are presented in Table 1. The adopted distribution implies that 99.5% of the erupted ash mass is contained in particles with a diameter less than 63 μm (ash bins from 6 to 10). Ash density is 2500 kg \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$m^{-3}$$\end{document} for all size bins. Simple aerosol chemistry is simulated using the chem_opt=402 configuration, which includes \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} and the 10 ash-size bins mentioned above. Emissions of sulfate and water vapor are omitted in this study. We account for major sinks: wet and dry deposition of ash, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document}, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} oxidation by OH and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {H}_2\ \text {O}_2$$\end{document}. The OH and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {H}_2\ \text {O}_2$$\end{document} fields are prescribed based on zonal and monthly mean OH and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {H}_2\ \text {O}_2$$\end{document} fields38 derived from the simulation of the Chemical Lagrangian Model of the Stratosphere (CLaMS)39 and Copernicus Atmosphere Monitoring Service (CAMS) reanalysis40, respectively. To reconstruct volcanic emissions consistent with the observed ash and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} plumes, we introduce the Plume-traj tool, which advects Lagrangian parcels backward or forward in time using the three-dimensional wind fields from the WRF simulation. In the current work, backward mode was used. As parcels, we considered ash represented by discrete particle-size bins (see Table 1) with corresponding gravitational settling velocities computed in6. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} was treated as an inert gas tracer, although a decay rate can be prescribed. In Plume-traj, parcel “seeds” were first initialized within the spatial extent of the observed plume on 24 November 2025 at 11:00 UTC, close to the TROPOMI observation time. In particular, ash seed locations were sampled within the TROPOMI AI footprint, while \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} parcel seeds were sampled within the retrieved TROPOMI \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} vertical column density. For the backward reconstruction, 90,000 parcels were initialized within each plume mask. Ash seeds were sampled from pixels with AI \ge 0.1, whereas \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} seeds were sampled from pixels with \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} vertical column density \ge 0.1 DU. Seed pixels were sampled uniformly within each thresholded two-dimensional mask, and the same seeding treatment was used for each ash bin. Then Plume-traj traced each parcel backward in time until it intersected a source cylinder with a 20 km radius centered at the Hayli Gubbi location and spanning the entire model atmospheric column, as shown in Supplementary Fig. S2. As a result, this backward-trajectory approach yields two diagnostics for each parcel hitting the cylinder: the time of arrival and the arrival height. For each successful backward trajectory, the reconstructed injection altitude is defined by the height at which the parcel intersects the source cylinder. Arrival time is discretized into uniform time intervals (30 minutes in our case), and arrival altitude is mapped to the WRF vertical grid. Thus, the resulting two-dimensional distribution of parcel counts as a function of height and time provides an emission time-height matrix. This information is used by the PrepEmisSources tool17 to construct emission scenarios normalized by the emitted mass and to write them to an auxinput13 NetCDF file readable by WRF when emiss_opt_vol=3 is selected. The trajectory-derived ash source term spans 23 November 2025, 08:30 UTC to 24 November 2025, 09:30 UTC and consists of 51 vertically resolved emission profiles updated every 30 min. The central ash simulation uses 0.8 Mt, while 0.7 and 0.9 Mt ash-mass sensitivity runs were used to assess the sensitivity of the AERONET ash-bin comparison and ash deposition footprint with the ash size distribution held fixed. The first trajectory-derived \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} source scenario was scaled to 0.3 Mt and captured the broad plume behavior as a first approximation (Supplementary Fig. S0). However, some backward \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} parcels did not return to the source region in the modeled wind field (Supplementary Fig. S2d,f), limiting the trajectory-derived \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} source reconstruction. This may reflect unresolved near-source plume dynamics during the explosive phase and/or uncertainties in the TROPOMI retrieval used to seed the parcels. The final \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} source term was therefore initialized from the Plume-traj matrix and adjusted manually in timing, altitude, and amplitude using visual plume-shape agreement and the TROPOMI \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} statistics at 11:00 UTC on 24 November 2025. It spans 23 November 2025, 08:30 UTC to 24 November 2025, 01:00 UTC, consists of 34 profiles updated every 30 min, and releases 0.15 Mt of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document}. The manually adjusted \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} profiles use Suzuki-type vertical source distributions18; implementation details are provided in the Code availability section. The 0.15 Mt \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} total is the selected scenario for this reconstruction and should not be interpreted as a posterior uncertainty-constrained emission estimate. For interpretability, four \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} pulse groups were defined from the final \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} source matrix and emitted in separate diagnostic WRF-Chem runs; the resulting \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {SO}_2$$\end{document} column footprints were evaluated at the TROPOMI comparison time and are shown in Supplementary Fig. S5. Supplementary Information Supplementary Information. Supplementary Information.