In the context of the AEMET responsibilities, as a state agency, for the continuous monitoring of the meteorological and climatic conditions and of atmospheric composition, a specific instrumental deployment has been set up on La Palma. In particular, a new station for gas and particle monitoring was implemented at the San Antonio Volcano Visitor Center of Fuencaliente, at the southern tip of La Palma island, ∼ 15 km from the eruptive fissure of the Tajogaite volcano (Fig. 1). The FUE station included a wide range of instruments such as a Cimel sun–lunar CE318T photometer, contributing to the Aerosol Robotic Network (AERONET), for aerosol column measurements; a Lufft CHM15k ceilometer for aerosol and cloud vertical profiling; an all-sky camera for weather monitoring (Román et al., 2021); and a tephra trap.

A few days after the beginning of the eruption (on 25 September 2021), we deployed an EM27/SUN spectrometer (developed by the Karlsruhe Institute of Technology (KIT), in collaboration with Bruker Optics, Germany), which is the standard instrument of the Collaborative Carbon Column Observing Network (COCCON; Frey et al., 2019) dedicated to the measurement of greenhouse gases. This portable FTIR spectrometer, equipped with a Quartz beam splitter and two InGaAs photodetectors, provides low-spectral resolution (0.5 cm⁻¹) solar absorption spectra in the near-infrared (NIR) range (from 4000 to 11 000 cm⁻¹), allowing the analysis of COCCON standard species (O2, CO₂, CO, H₂O, and CH₄). It records double-sided forward–backward interferograms with a scanner velocity of 10 kHz and typically averages 10 scans, so that a spectrum is acquired approximately every minute. The spectral range of this instrument also allows one to obtain other gas species of interest for volcanology and air quality studies, such as halogen halides (HCl and HF) (Butz et al., 2017). From 10 October 2021 to 10 December 2021, following the approach of Butz et al. (2017), we combined the EM27/SUN with a UV–Vis DOAS spectrometer (model Avantes ULS2048). The DOAS instrument has a 50 µm wide slit entrance and allows one to record spectra in the 270–425 nm spectral range with a spectral resolution of 0.4 nm. We used a 200 µm wide quartz-made optical fiber. Both instruments shared the incident sun radiation from the EM27/SUN solar tracker to add simultaneous measurements of SO₂ with the same measurement configuration (Fig. 2).

The DOAS fiber was inserted and attached coaxially into the tube directing the light from the solar tracker toward the EM27/SUN spectrometer entrance (Fig. 2). Therefore, it allows one to collect the maximum light intensity with minimal disturbance to the solar beam transmitted to the EM27/SUN. The fiber was connected to the DOAS spectrometer, installed in a protective case sheltered from solar radiation near the EM27/SUN instrument. DOAS direct-sun absorption spectra were routinely recorded using the MobileDOAS software (unpublished acquisition program developed for mobile DOAS measurements by C. Fayt and A. Merlaud from the BIRA-IASB institute) with an integration time of about 30 s and, on average, 20 scans. The details of the DOAS and EM27/SUN spectral analysis and retrievals are given in Sect. 3.

The proximity of the island of Tenerife to La Palma and the location of the IZO station in the free troposphere (2373 m a.s.l.) resulted in this international reference observatory being affected several times by the Tajogaite volcanic plume. This allowed for a more in-depth study of various aspects of the volcanic eruption from a multi-instrumental perspective. Given its strategic location and its excellent atmospheric conditions, IZO indeed has a comprehensive, state-of-the-art program for atmospheric composition measurements. Uninterrupted meteorological and climatological observations started in 1916 and, since 1984, IZO has contributed to the GAW WMO (Global Atmosphere Watch of the World Meteorological Organization) program and to multiple international networks and databases (e.g., World Data Centre for Greenhouse Gases – WDCGG; World Ozone and Ultraviolet Radiation Data Centre – WOUDC; Network for the Detection of Atmospheric Composition Change – NDACC; Total Carbon Column Observing Network – TCCON; Collaborative Carbon Column Observing Network – COCCON; Aerosol Robotic Network – AERONET; Baseline Surface Radiation Network – BSRN; Micro-Pulse Lidar Network – MPLNET; EUMETNET EIG GNSS water VApour Program – E-GVAP; and National Oceanic and Atmospheric Administration (NOAA) Earth System Research Laboratory (ESRL) Global Monitoring Division (GMD) Carbon Cycle and Greenhouse Gases (CCGG) Group; Cuevas et al., 2024, and references therein). Within IZO's atmospheric research activities, the station is equipped with high-resolution (IFS-125HR) and low-resolution (EM27/SUN) FTIR spectrometers, which have been providing ongoing long-term solar absorption measurements since 1999 and 2018, respectively. The EM27/SUN spectrometer is the same instrument model as that implemented at the FUE station, allowing for the analysis of CO₂, CO, HF, and HCl species, as previously described.

The IZO FTIR spectrometers routinely contribute to NDACC (https://ndacc.larc.nasa.gov, last access: March 2025), TCCON (https://tccon-wiki.caltech.edu, last access: March 2025), and COCCON (https://www.imk-asf.kit.edu/english/COCCON.php, last access: March 2025) (Schneider et al., 2005; García et al., 2021). As part of NDACC activities, direct solar mid-infrared (MIR) absorption spectra are measured in the range from 700 to 4500 cm⁻¹, with a spectral resolution of 0.005 cm⁻¹. NDACC operations involve co-adding several scans to increase the signal-to-noise ratio, resulting in each spectrum acquisition taking several minutes. García et al. (2021) provide further details on the IZO FTIR program. The IZO IFS-125HR MIR solar spectra were used to analyze the SO₂ species alongside HCl and HF, which were also measured from the EM27/SUN spectra (unlike SO₂). This approach further allowed us to evaluate the uncertainties associated with our new retrieval methods for the HF and HCl species (see Sect. 3 and Appendix A). The details of the spectral analysis and retrievals are given in Sect. 3.

Moreover, as part of the GAW WMO program, continuous surface measurements of CO₂ (since 1984), SO₂ (since 2006), and CO (since 2008) are performed at IZO. Different in situ analyzers and measurement techniques have been used for measuring these gases: CO₂ with nondispersive infrared (NDIR) gas LI-COR analyzers, CO with gas chromatography (GC) Trace Analytical RGA-3 instruments, and SO₂ with ultraviolet (UV) fluorescence analyzers (Thermo 43C-Trace Level). Since 2015, CO₂ and CO have also been monitored using a cavity ring-down spectroscopy (CRDS)-based Picarro G2401 instrument. These observations are carried out following the strict GAW WMO measurement protocols, and their quality is periodically assessed by external audits by the World Calibration Center for surface O₃, CO, CH₄, and CO₂ (WCC-Empa). The bias for the CO₂ and CO measurements in the framework of the GAW WMO network is ±0.1 ppm and ±2 ppb, respectively (WMO, 2018). For SO₂, the uncertainties are expected to be around ±0.2 ppb (manufacturer specifications; see also Cuevas et al., 2024, and references therein). This continuous gas monitoring captured the Tajogaite plume composition on several occasions, when meteorological conditions allowed rapid and direct transport to the IZO station.

The SO₂ flux was retrieved by processing the images of the TROPOMI hyperspectral UV–SWIR (ultraviolet–shortwave infrared) sensor aboard the Sentinel-5P satellite. The images were processed by the traverse method, initially developed for the coarser-resolution TOMS satellite images by Bluth et al. (1994) and later adapted to more recent sensors such as OMI (Ozone Monitoring Instrument) and TROPOMI. The traverses are drawn across the plume semiautomatically, and the SO₂ flux is calculated using the following equation: 1F=ΣXi⋅Li⋅sin⁡(Θ)⋅v, where Xi is the SO₂ vertical column density (VCD), Li is the length of the pixel, Θ is the angle between the pixel row and the wind direction, and v is the plume transport speed. The SO₂ VCD was interpolated at plume height between the SO₂_1km and the SO₂_7km subproducts of version 3 of the TROPOMI SO₂ product, described in Theys et al. (2021). The plume speed was obtained from the Global Data Assimilation System model of NOAA, through the READY (Real-time Environmental Applications and Display sYstem) archived meteorology portal (https://www.ready.noaa.gov/index.php, last access: 19 October 2025). For the flux calculation, we used the average wind speed at the plume altitude over the analyzed plume portion. The plume altitude was estimated from visual observations, such as photographs, distal webcam images (from Roque de los Muchachos), and HYSPLIT (Hybrid Single-Particle Lagrangian Integrated Trajectory) trajectory simulations, picking the injection altitude that best reproduces the general plume direction observed in the TROPOMI image, and confirmed with the AEMET/IGN estimates for the coincident days. The SO₂ fluxes were finally estimated using the average of several traverses (usually a few tens and, on some occasions, up to 200 traverses, depending on the coherence of the plume and the wind field that transports it). The traverse method does not work in cases of plume stagnation in a low-wind environment or when the plume is split into several directions due to wind shear. These situations happened about 30 % of the time during the eruption, causing some gaps in the SO₂ flux time series. We also excluded images in which the plume was only partially captured.

During the eruptive period, between 28 September and 10 October 2021, mobile surface MultiGAS measurements (SO₂, CO₂, H₂O, and H₂S) were carried out in the volcanic plume, when meteorological conditions allowed it to be sampled at ground level at a high concentration. The instrument comprised an MSR145 datalogger, an Edinburgh Gascard NG for CO₂ (0–1000 ppm) with a pump, a City Technology T3ST/F electrochemical sensor for SO₂ (0–50 ppm), and a City Technology T3H electrochemical sensor for H₂S (0–20 ppm). SO₂ concentrations up to 7 ppm were measured at distances of about 2 km east and west of the vent (Fig. 1). Time series of concentrations of the different gas species were cross-correlated by adjusting the time lag (usually between 5 and 9 s) and smoothing parameter until the best R2 coefficient of determination was obtained. The measurements presented here have R2 values higher than 0.75.

Samples of aerosols, or particulate matter (PM), smaller than 10 µm (PM₁₀) were collected at two sites on La Palma, at El Paso and at Los Llanos de Aridane, and at IZO on Tenerife. We used high-volume samplers (30 m³ h⁻¹) and quartz microfiber filters (150 mm diameter). Sulfate concentrations were determined by ion chromatography (Metrohm™ 930 Compact IC FLEX), after a leaching extraction of the sample in deionized Milli-Q-grade water using the methods described in Rodríguez et al. (2012).

We report 14 new compositions of MIs hosted in olivine, clinopyroxene, and amphibole (kaersutite) crystals (Appendix B3; Supplement Table S1). We also report Cl, F, and S contents in tephra glasses that were measured alongside major elements during the analytical session described in González-García et al. (2023), although only the major elemental data were published in that study. The volatiles were analyzed using a Cameca SX-100 electron microprobe (EPMA) at the Department of Geosciences of the University of Bremen (Germany), with an accelerating voltage of 15 kV, beam current of 40 nA, and defocused beam of 10µm, following the methods described in González-García et al. (2023). The instrument was calibrated with a natural fluorite for F, pyrite for S, and Smithsonian scapolite for Cl. Counting times on peak were 120 s for F and 60 s for S and Cl. The analysis of F used the PHA (pulse height analysis) setting following Zhang et al. (2016); the interference of the FeLα line on the FKα peak was corrected using the overlay function of the Cameca software. The Smithsonian reference materials VG-2 glass, VG-A99 glass, and Kakanui hornblende (Jarosewich et al., 1980) were analyzed along with the samples for precision and accuracy control. Accuracy is better than 6 % for S and Cl and better than 20 % for F; reproducibility is typically better than 10 %. In addition, the composition of two sulfide droplets was semiquantitatively estimated by EDX (energy-dispersive X-ray) spectroscopy.

A scanning electron microscope (SEM) was used to obtain high-resolution backscattered-electron (BSE) images of two sulfide droplets found in the tephra sample LM-2309 (Las Manchas, 23 September). The BSE images were acquired using a JEOL JSM-7610F gun emission scanning electron microscope installed at the Institute of Earth System Sciences, Leibniz Universität Hannover, Germany, using an acceleration voltage of 15 kV and a working distance of 15 mm. Bruker ESPRIT software was used for image acquisition.

The HCl and HF retrieval strategy from the IFS-125HR spectra is based on the NDACC Infrared Working Group (IRWG) recommendations (IRWG, 2014) and on the adapted retrievals for volcanological applications reported in Taquet et al. (2019) and Stremme et al. (2023). However, they have been optimized here to properly capture tropospheric volcanic contributions up to 140 km from the eruptive fissure. Consistent with the NDACC approach, both species were retrieved using the nonlinear least-squares fitting algorithm PROFFIT (Profile Fit, Hase et al., 2004) and considering the specified spectral regions and interfering gases given in Table 3. The inversion procedure is solved using a first-order Tikhonov–Phillips regularization (L1; Rodgers, 2000) on a logarithmic scale, where the VMR a priori profiles for the interfering gases are taken from WACCM v.6 climatological profiles. The NCEP 12:00 UTC daily temperature and pressure profiles are employed for the radiative transfer simulations.

The most significant changes with respect to NDACC involved the a priori VMR profiles considered for the target gases, vertical L1 regularization, and the spectroscopic database. Similarly to the EM27/SUN analysis, we adopt modified HF and HCl a priori VMR profiles with high concentrations (1 × 10⁻⁴ ppm) up to the maximum plume altitude (∼ 6 km a.s.l.), which are completed for the IFS-125HR using WACCM v.6 information beyond this altitude. In addition, the 2020 HITRAN spectroscopic line lists were utilized for all gases. Finally, in contrast to the NDACC approach, where the lowermost and uppermost altitude levels are fixed to the a priori value to ensure stability in the retrieval, in this study, the first level is left unconstrained to provide flexibility in the retrieval process in the lower troposphere.

In the case of SO₂, a harmonized and standardized FTIR strategy is not available within NDACC. Therefore, in this work, we employ the strategy developed by García et al. (2022), which has been successfully applied to various NDACC FTIR sites affected by volcanic SO₂ emissions (Smale et al., 2023; García et al., 2025). This approach is based on the study by Taquet et al. (2019), which presents SO₂ total column amounts from the measured solar absorption spectra in the 2500 cm⁻¹ region using a scaling retrieval and the inversion code PROFFIT. Similarly to the HF and HCl volcanic products, the SO₂ a priori VMR profiles are adapted in the lower troposphere, while climatological WACCM v.6 profiles are considered for all interfering gases (Table 3).

Appendix A provides a summary of the comparison of the standard NDACC HCl and HF products with those developed in this study. It also includes the new IFS-125HR SO₂ retrievals as well as the comparison between all of the IFS-125HR and EM27/SUN products.

The temporal variability in the Tajogaite plume composition is examined using the time series of the ratios, some of them involving species with contrasting exsolution depths. Daily ΔCO₂ / SO₂, HCl / SO₂, HF / SO₂, HCl / ΔCO₂, HF / ΔCO₂, ΔCO / SO₂, and ΔCO / ΔCO₂ molecular ratios were estimated from the daily correlation plots of the total column time series, following the methodology detailed in Taquet et al. (2019, 2023), and are reported in Fig. 3. The same method used for column-averaged ratios was applied to calculate the surface concentration ratios from GAW and MultiGAS measurements (also presented in Fig. 3). The background contribution of atmospheric species (CO₂ and CO) to these measurements was removed using daily polynomial curves fitted from the surface measurements without the contribution of volcanic emissions (i.e., SO2<0.05 ppm). Additionally, in the same figure, we reported our MultiGAS ΔCO₂ / SO₂ measurements, obtained on 29 September and on 2 and 7 October from Las Manchas (∼ 500 m a.s.l., southwest of the eruptive fissure; Fig. 1) and from the El Jable viewpoint (2100 m a.s.l., east of the eruptive fissure; Fig. 1), ranging between 1.7 and 14.3. The scarcity of FTIR measurements from early November until the end of the eruption, across all measurement techniques, is mainly due to poor or unsuitable weather conditions.

Our column-averaged ΔCO₂ / SO₂ molecular ratios range between 9 ± 6 and 63 ± 28 (9–24 at IZO and 14–63 at FUE) during the eruption. These values are consistent with the surface measurements at IZO (ratios from 5.6 ± 0.1 to 18.3 ± 0.7) and with our MultiGAS measurements at La Palma (1.7–14.3). These values are also consistent with the proximal measurements reported in the literature including open-path FTIR (Burton et al., 2023; Asensio-Ramos et al., 2025) and MultiGAS (Burton et al., 2023; Ericksen et al., 2024) measurements, ranging between 2 and 52 (shaded area in Fig. 3). All the measured ΔCO₂ / SO₂ ratios define an increasing trend until at least 2 November 2021 and show more scatter after this date (Fig. 3).

HCl / SO₂ molecular ratios range between 0.02 ± 0.002 and 0.17 ± 0.01 (from 0.02 to 0.05 at IZO and from 0.02 to 0.17 at FUE) and show short-term variations around a nearly constant daily average of 0.05 ± 0.03 throughout the entire eruptive period. These ratios are consistent with the SO₂ / HCl values of 16.8 and 8 (HCl / SO2= 0.06 and 0.12, respectively) reported in Burton et al. (2023), which correspond to a lava-fountaining plume and spattering event (Fig. 3). It is also consistent with the more recently published ratios ranging between 0.04 and 0.2 (Asensio-Ramos et al., 2025; reported in Fig. 3 in shaded area). HF / SO₂ molecular ratios vary between 0.0012 ± 0.0002 and 0.081 ± 0.007 (from 0.001 ± 0.001 to 0.082 ± 0.007 at FUE and from 0.007 ± 0.002 to 0.037 ± 0.025 at IZO) and show a similar day-to-day variability to that observed for the HCl / SO₂ ratios throughout the eruptive period. HCl / ΔCO₂ molecular ratios exhibit values from (6 ± 1) × 10⁻⁴ to (4.1 ± 0.1) × 10⁻³ at FUE and from (2 ± 1) × 10⁻³ to (3 ± 1) × 10⁻³ at IZO, while the HF / ΔCO₂ ratios range from (0.5 ± 0.1) × 10⁻⁴ to (4.5 ± 0.1) × 10⁻³ at FUE and from (2.6 ± 0.3) × 10⁻⁴ to (2.7 ± 0.2) × 10⁻³ at IZO. Like HCl / SO₂ and HF / SO₂, the HCl / ΔCO₂ and HF / ΔCO₂ ratios exhibit similar day-to-day variability. Their fluctuations include short-term decreasing trends, as observed between 2 and 14 October 2021 and between 21 October and 4 November 2021. The ΔCO / SO₂ FTIR ratios span from 0.13 ± 0.01 to 0.66 ± 0.03 at FUE and from 0.02 ± 0.01 to 0.17 ± 0.07 at IZO, and they are relatively stable around an average of 0.24, with one extreme event observed between 1 and 4 November 2021.

During the initial phase of the eruption, prior to the eruptive pause on 27 September 2021, our ratios were comparable to those observed throughout the rest of the eruptive period, with ΔCO₂ / SO₂ ranging between 5.6 ± 0.1 and 9 ± 1.1, HCl / SO₂ between 0.031 ± 0.005 and 0.049 ± 0.007, and HF / SO₂ between 0.009 ± 0.003 and 0.022 ± 0.006. A significant and abrupt increase in all species-to-SO₂ ratios is observed on 2–3 November 2021, which also coincides with a minor peak in the HCl / ΔCO₂ and HF / ΔCO₂ ratios. This event represents a notable and enduring change in gas ratio variability involving CO₂ (i.e., ΔCO₂ / SO₂ and HCl / ΔCO₂) and coincides with a sudden decrease in the amplitude of seismic tremor (VLP and LP; Fig. 3; Bonadonna et al., 2022). Prior to this date, the variability in the ΔCO₂ / SO₂ ratio closely followed the increasing trend in VLP tremor amplitude, whereas it declined afterwards and exhibited a noticeable short-term variability until the end of the eruption. This noticeable change depicts two periods in our dataset (hereafter referred to as phase I and II), whose relationship with the previously described events and time frames of the eruption (Bonadonna et al., 2022; Ubide et al., 2023; Milford et al., 2023) will be discussed in Sect. 5. For HCl / ΔCO₂ and HF / ΔCO₂, the ratios are significantly lower during phase II (average of 0.0012 ± 0.0005 and 0.0007 ± 0.0004, respectively) than during phase I (average of 0.0027 ± 0.0009 and 0.0014 ± 0.001, respectively). For other species, only a brief spike is noted at this time, with ratios returning to phase-I levels at the onset of phase II.

Figure 4 presents the time series of ΔCO / ΔCO₂ ratios derived from FTIR solar absorption measurements at the FUE and IZO stations throughout the eruption, alongside in situ surface measurements at IZO (GAW data). The ΔCO / ΔCO₂ values observed at both sites and using both techniques are of the same order of magnitude and exceed the average atmospheric background ratio at IZO (∼ 0.0002) by more than 1 order of magnitude. At FUE, the FTIR-derived ratios show a progressive increase from 0.0016 to 0.016 during the first 30 d of the eruption, followed by a decrease to lower values before mid-November. The surface ΔCO / ΔCO₂ ratios at IZO fall within a similar range to those derived from FTIR at FUE, with some coinciding values in very good agreement. On average, the surface ratios at IZO are higher than the FTIR-derived ones at the same site. This discrepancy may be explained not only by the strong short-term variability in the ΔCO / ΔCO₂ ratios (only a few data points are coincident) but also by the fact that, although all of these points coincide with the presence of SO₂ (indicating the presence of volcanic plume), the correlation between ΔCO and SO₂ is relatively weak (R2<0.6), suggesting additional sources contributing to the CO enhancements. Furthermore, satellite imagery suggests that, on these days, the line of sight of the IZO FTIR instrument may have intersected aged volcanic plumes, potentially altering the retrieved ΔCO / ΔCO₂ ratios due to both geometric and compositional effects. The difference between the surface ΔCO / ΔCO₂ ratios observed at FUE and IZO and those reported by Asensio-Ramos et al. (2025) (shaded area) is discussed in Sect. 5.

SO₂ volcanic emission fluxes were estimated whenever the weather conditions made it possible, following the method described in Sect. 2.3 and reported in Fig. 5. The SO₂ volcanic emission fluxes retrieved during this eruption exhibited a remarkably strong correlation (R2=0.92) with the daily SO₂ masses (taken from the MOUNTS website; Valade et al., 2019). To fill the long-term gaps in our SO₂ flux time series, a less reliable mass-derived product was included, derived from the linear relation between the SO₂ volcanic emission fluxes and daily mass (empty circles in Fig. 5a). This was only applied to days with minimal accumulation. The SO₂ volcanic emission flux time series exhibit a decreasing exponential trend (red curve), with an equation of the form y=a×e-bx and a coefficient of determination of R2=0.63. Most mass-derived products were found to closely follow the overall trend (red curve in Fig. 5a), indicating that, despite inherent uncertainties, these estimates are likely robust enough to assess long-term variability in this case study. This also suggests that short-term variations in wind direction or partial plume coverage in satellite images (initial filtering criteria) may have a limited impact on the observed global trend.

Volcanic emission fluxes for the other species were estimated using daily species-to-SO₂ ratios and either (i) interpolating the exponentially decreasing fit of SO₂ fluxes or (ii) performing a linear interpolation of the SO₂ emission flux time series. The HCl, HF, CO₂, and CO volcanic emission fluxes are shown in Fig. 5b–d, concurrently with the time-averaged discharge rate (TADR, Fig. 5e) time series of Plank et al. (2023), multiplied by a factor of 2, as suggested by the authors to take the underestimation of the lava volume into account.

A significant observation is the long-term decrease in the volcanic emission fluxes of SO₂, HCl, HF, and CO, which aligns with the TADR trend throughout the eruption, in contrast to the nearly stable trend in CO₂. The similarity of the trends in the daily average SO₂ emission fluxes and the TADR is further supported by an excellent correlation (Pearson correlation coefficient = 0.94; see Fig. E1), defining a slope of 14.1 ± 1.2 kg of SO₂ per “thermal” cubic meter of discharged lava (lava volumes estimated using the radiant flux). This relationship includes 21 out of 27 of the available TADR–flux pairs) and is mainly valid from 7 October 2021 onwards (full circles in Fig. E1). The points corresponding to the onset of the eruption (outliers represented as hollow circles in Fig. E1) have either higher SO₂ fluxes for a given TADR until the 25 September or higher discharge rates after the 27 September eruptive break and until 30 September at least (the next pair is that of 7 October, belonging to the correlation).

Another important observation is that the SO₂ flux peak recorded during the first week of the eruption, accounting for approximately 20 % of the total SO₂ emissions, occurs during a period of apparently low TADR and around 10 d prior to the first peak with maximum values of TADR for the eruption. The relationship between the SO₂ volcanic emission fluxes and the TADR is examined in the light of the petrological data in Sect. 5.

Furthermore, the early-November peaks in the HF, HCl, and CO emission flux time series, which align with those observed in several ratio time series (Fig. 3), correspond to the inflection point in the overall flux decline, occurring near the end of phase I, as defined by Milford et al. (2023). As the CO₂ volcanic emission fluxes appear to be nearly constant throughout the entire eruptive period, we can interpret the lower HCl and HF / CO₂ ratios of phase II to be the result of globally lower fluxes during this period, in line with the pressure decrease in the reservoir (Charco et al., 2024).

Table 4 presents the average volcanic emission fluxes for each species over the entire eruption, distinguishing between the results from the two previously described methods. Total emissions were estimated by combining a Monte Carlo approach to account for uncertainties with trapezoidal integration to compute the area under the curve, and they are reported in Table 4. The average fluxes over the entire eruptive period and the estimated total emissions of SO₂, HCl, HF, and CO₂ (Table 4) provide insight into the scale of the emissions of this eruption with respect to other emission sources.

The total SO₂ emissions of 1.81 ± 0.18 Mt, derived from our exponentially decreasing fit, is similar to that reported in Milford et al. (2023) using the daily SO₂ mass derived from TROPOMI data (credit: ESA, MOUNTS). These total SO₂ emissions are comparable to the emissions of the submarine 2011 Tagoro eruption at El Hierro, which released between 1.8 and 2.9 Mt SO₂ into the ocean (estimated using the petrologic method; see Longpré et al., 2017).

During the Tajogaite eruption, the highest SO₂ emission fluxes occurred during the first 10 d of the eruption (median of 37 kt d⁻¹ during this period), whereafter a lower median of about 20 kt d⁻¹ was observed. These SO₂ emission rates are the same order of magnitude as recent basaltic eruptions such as, for instance, Piton de La Fournaise on Réunion Island in 2020 (average: 9 kt d⁻¹; max: 25 kt d⁻¹; Hayer et al., 2023) and Bárðarbunga in Iceland in 2014–2015 (average of 50 kt d⁻¹ over 6 months; Pfeffer et al., 2018), whereas they are lower than the emission rate of Kilauea in 2018 (average of 200 kt d⁻¹; Kern et al., 2020); however, the latter two events exhibit much higher eruptive TADR. For the Tajogaite eruption, the high SO₂ fluxes result from the high sulfur content of parental magma, as reflected by the average content of 3360 ppm in our MIs (Supplementary data, Table S1), similar to the value of 3500 ppm reported in Burton et al. (2023) and Dayton et al. (2024).

For CO₂, we obtained a steady average emission flux of 260 ± 24 kt d⁻¹ and total emissions of 19 ± 2 Mt over the course of the eruption. This result aligns closely with the estimates of 28 ± 14 Mt reported by Burton et al. (2023). These emissions represent 15 % of global subaerial volcanic and tectonic annual emissions (Fischer and Aiuppa, 2020) or the equivalent of the annual CO₂ budget of ocean island basalt (OIB) volcanism, as estimated by Lo Forte et al. (2024). The high CO₂ emissions with respect to the low extruded magma volume during the Tajogaite eruption, compared to other effusive eruptions, is explained by the extraordinarily carbon-rich magma, as it is reflected in both fluid and melt inclusions (up to 2 wt % CO₂ in MIs; Dayton et al., 2024). This is a characteristic of Macaronesian magmas and possibly of global OIB (Burton et al., 2023; Lo Forte et al., 2024; Van Gerve et al., 2024).

Daily total CO emissions measured during the eruption, averaging 2 kt d⁻¹, were exceptionally high, with a cumulative total of 0.12 ± 0.01 Mt. Only few volcanic CO emissions are reported in the literature, such as 0.15 kt d⁻¹ at Erebus volcano (Wardell et al., 2004), 0.007 kt d⁻¹ at Ol Doinyo Lengai (Oppenheimer et al., 2002), 0.16 to 0.27 kt d⁻¹ at Nyiragongo volcano (Sawyer et al., 2008a), and 0.0007 kt d⁻¹ at Erta Ale (Sawyer et al., 2008b), and they are more than 1 order of magnitude lower than our estimates during the Tajogaite eruption.

Finally, our estimated HCl and HF total emissions are about 50 ± 10 kt and 13 ± 2 kt, respectively, with corresponding averages of 604 ± 340 t d⁻¹ and 173 ± 86 t d⁻¹. These emissions are of the same order of magnitude as those observed for other basaltic volcanoes, such as Etna (300–1300 t d⁻¹ of HCl during the 2008–2009 eruption, as reported in La Spina et al., 2023, and 800 t d⁻¹ of HCl and 200 t d⁻¹ of HF in 1997, as reported in Oppenheimer et al., 1998), Bárðarbunga volcano (500 and 280 t d⁻¹ for HCl and HF, respectively, as reported in Galeczka et al., 2018). HCl and HF emissions from the Tajogaite eruption are more than an order of magnitude higher than those observed at Kilauea volcano, which were reported to be 12–22 t d⁻¹ of HCl and 6–9 t d⁻¹ of HF in 2008 and 2009 (Mather et al., 2012).

One of our key results is the remarkably strong consistency between the measured volcanic gas species-to-SO₂ ratios, regardless of the measurement site, the technique, or the instrument used (Fig. 3). The measurements conducted at the IZO station gave the excellent opportunity to assess the robustness of our estimated ratios, using both the EM27/SUN and IFS-125HR instruments, and their consistency with surface measurements. We found excellent agreement between the HCl and HF total columns (with volcanic plume contribution) derived from the IFS-125HR and EM27/SUN products (see Appendix A for details).

We found good comparability with respect to the available ΔCO₂ / SO₂ and ΔCO / SO₂ between surface and column measurements, reflecting an efficient vertical mixing. This also suggests that when the volcanic plume is detected by the surface measurements at the IZO station, the ground-level concentrations are representative of the average volcanic plume composition. As the IZO station is often located above the base height of the trade wind inversion (TWI) layer (Milford et al., 2023), volcanic plumes detected at IZO were typically transported rapidly through the low free troposphere. The progressive decrease in the plume injection height throughout the eruption, combined with seasonal changes in the vertical stratification of the atmosphere (TWI height), resulted in sparse detections of the plume at the IZO station after mid-November 2021 (Milford et al., 2023). This led to a reduction in the coincident surface and total column observations.

Moreover, the comparison of ratios at different distances from the eruptive vents (i) at IZO (140 km) and (ii) near the active vent measured by OP-FTIR or MultiGAS instruments (this work; Burton et al., 2023; Ericksen et al., 2024) allows qualitative assessment of the impact of in-plume reactions on our measurements. The ratios taken from Burton et al. (2023) were derived from either in situ ground-based or drone-borne MultiGAS measurements within the plume close to the volcanic vents or, after 2 October 2021, from open-path FTIR measurements pointing to the eruptive column and using the lava fountain as a source. Those reported by Ericksen et al. (2024) are ground-based MultiGAS measurements. In any case, the gas measured by these authors corresponds to the plume less than 1 km from the volcanic vents. Because CO₂ is a nonreactive species, a significant conversion of SO₂ into sulfate aerosols (H₂SO₄) during the transport between La Palma and IZO should increase the ΔCO₂ / SO₂ ratio. Hence, if significant conversion of SO₂ to sulfates occurred during transport, the IZO ratios should be higher than those measured closer to the volcano. To examine this aspect, we estimated the plume age for each recorded event using the HYSPLIT transport model, in both retro-trajectories and forward-simulation configuration mode. For meteorological data, we utilized 72 h extended files containing high-resolution meteorological information derived from the WRF-ARW (Advanced Research Weather Research and Forecasting) model as input. This model runs twice a day, using initial and boundary conditions from ECMWF's HRES-IFS (high-resolution Integrated Forecast System) data, with a resolution of 0.09° × 0.09° (for further details, refer to Appendix C). Table 5 shows the coinciding values of the ΔCO₂ / SO₂ ratios measured at less than 1 km from the eruptive fissure (Burton et al., 2023) and at IZO (this work) as well as an estimate of plume age for each event. Despite the limited number of coincident events at the two sites, no clear dependence of this ratio on distance was observed for plumes with an age of 12 h or less. A certain similarity was found, at least until the beginning of November, even in cases of relatively old plumes (∼ 12 h), suggesting a swift transport between La Palma and Tenerife and negligible in-plume reactions (or at least indistinguishable within the uncertainties in the ratios). In the troposphere, the SO₂ to SO4= oxidation rates vary significantly, from a few percent per hour by in-cloud droplet processes (driven by aqueous-phase oxidation, e.g., H₂O₂) to a few percent per day (in dry air, driven by OH radicals) (Seinfeld and Pandis, 1997). Our results suggest that this latter (slow dry oxidation) process may be the prevailing one during transport in the dry free-troposphere, from La Palma to IZO. This interpretation is supported by the sulfate aerosols measured in situ on La Palma (Rodríguez et al., 2025) and at IZO, when the volcanic plume reaches the station, plotted in Fig. 6. Figure 6a reports the statistical distribution of the ratio (in percent, %) of particulate sulfur (S(p), i.e., sulfate SO4=) over total sulfur (i.e., gas sulfur as sulfur dioxide (S(g)) plus S(p)) measured in the aerosols smaller than 10 µm (PM₁₀) at IZO and at La Palma during the eruption. Figure 6b shows the correlation plot of S(g) as a function of S(g)+S(p). We observe a higher maximum conversion rate at IZO (45 %) than on La Palma (20 %), as expected. However, 80 % of the dataset (Fig. 6a and b) presents a conversion rate of the sulfur content to SO₄ of below 15 % and 7 % at IZO and La Palma, respectively.

Furthermore, the time distribution of the S(p)/(S(p)+S(g) ratio (Fig. 6) suggests a higher conversion rate of SO₂ to sulfate during the first part of the eruption (until the beginning of November) compared to the second period. This trend appears closely tied to the volcanic plume's altitude relative to the trade wind inversion (TWI), as described by Milford et al. (2023). During the first period of the eruption (until early November), plumes from explosive activity and fountaining vents often rose above the TWI, and surface measurements at La Palma likely captured older, dilute, more oxidized emissions from effusive vents trapped in the TWI. Conversely, from the beginning of November, the entire plume, comprising both explosive and effusive components, was more frequently trapped below the TWI, leading to the detection of younger, more concentrated, and less oxidized emissions at ground level. In any case, the plumes reaching IZO are most likely dominated by explosive emissions which, despite substantial transport times, exhibit oxidation rates below 15 %. Such low conversion rates would not produce resolvable differences in our gas-to-SO₂ ratios. The last two events in Table 5 present some difference between both sites. On 16 October 2021, the FTIR and surface ratios at IZO are comparable, highlighting their robustness; however, these ratios are a factor of 2–3 lower than those reported by Burton et al. (2023). We remark that, for these days, the measurement target reported by these authors mention the base of the lava fountain instead of the spattering vents or passive degassing, as for the other three dates, implying different conditions and processes.

Finally, the ΔCO / ΔCO₂ ratios measured at FUE station (Fig. 4) and those recorded at IZO from surface measurements are on average higher and have higher variability than those recently reported in Asensio-Ramos et al. (2025) from open-path measurements (Fig. 4). This difference is likely due to the different measurement methods (solar absorption vs. open-path measurements using hot lava as source), implying different loci of measurements and gas contribution along their respective line of sight. Tajogaite volcano presented notable differences in eruptive behavior between the different vents along the volcanic fissure, with those at higher elevations being more explosive than lower vents. Recent studies suggest that eruptive dynamics may affect the abundance of redox-sensitive species (e.g., Oppenheimer et al., 2018; Moussallam et al., 2019). Furthermore, we note that most of the Asensio-Ramos et al. (2025) measurement sites until the beginning of November (i.e., when our highest ΔCO / ΔCO₂ ratios were recorded) were located to the north-northwest of the eruptive fissure. With winds dominantly blowing towards the south and southwest during this period, this configuration avoided a significant contribution of biomass- and building-burning plumes to their measurements. This was not the case for the FUE measurements, which are more likely to have been affected by this contribution (provoked by the advance of the lava flows). This hypothesis is also supported by the similarity of the ΔCO / ΔCO₂ time series at FUE with the time series of the areas covered daily by the advancing lava flows (Appendix D, Fig. D1), reflecting the extent of burnt vegetation. The typical values reported in the literature for the wildfires (Yokelson et al., 2007; Akagi et al., 2014; Vasileva et al., 2017; Álvarez et al., 2023) are generally higher than our values, by at least a factor of 5, which is likely explained by the different contributors of the measured plume, i.e., a mixing of volcanic plume and vegetation/infrastructure burning in the case of the 2021 La Palma eruption.

The ratios and emission flux time series and the total emission estimates presented here provide some information about the degassing processes during the Tajogaite eruption.

Our time series of SO₂ volcanic emission fluxes confirms the decreasing trend observed from the SO₂ daily mass time series from MOUNTS (http://www.mounts-project.com, last access: 25 October 2025; Valade et al., 2019) and reported in Milford et al. (2023). The concurrent decrease in SO₂ emissions with decreasing tephra accumulation rates and a decreasing plume height was suggested to reflect the decrease in the pressure in the plumbing system (Milford et al., 2023). This was confirmed by the co-eruptive deflation trend observed and inverted by Charco et al. (2024), possibly related to the pressure drop due to drainage of the reservoir. The relatively good fit of the SO₂ flux data obtained using an exponential function further supports this interpretation.

We found a good correlation between the SO₂ volcanic emission flux time series and the TADR (slope: 14.1 ± 1.2 kg of SO₂ per cubic meter of lava; R=0.94). A similar correlation between SO₂ emissions and effusive volumes was previously observed during the 2021 Fagradalsfjall eruption (Pfeffer et al., 2024). The few outliers to this correlation (empty circles in Fig. E2) occurred during three distinct periods: (1) the initial days of the eruption, coinciding with the peak in SO₂ emissions; (2) just after the 27 September eruptive pause, at the onset of a sharp increase in effusion rates; and (3) following the opening of the late-November vents, north of the main vent alignment. These outliers correspond to abrupt changes in the output rate, likely associated with transient perturbations in the surface thermal structure conditions, which are known to affect the reliability of TADR estimations based on radiant density models (Coppola et al., 2016). Interestingly, applying the TADR values derived from the Pleiades-based volume estimates of Belart and Pinel (2022), which are averaged over 6–7 d, would bring at least three of these outliers back in line with the main trend. This suggests that apparent short-term imbalances between SO₂ emissions and effusion rates may be rapidly compensated for, resulting in a coherent degassing–effusion relationship over multiday timescales. This is particularly evident at the beginning of the eruption, where the Belart and Pinel (2022) estimates yield significantly higher TADR values than those of Plank et al. (2023) (Fig. E2).

Beyond these transient deviations, the correlation between SO₂ flux and TADR remains remarkably consistent throughout the eruption, suggesting that the emitted SO₂ predominantly reflects syn-eruptive magma degassing. This coherence, maintained over nearly 3 months of activity, indicates that the degassing regime remained stable once the eruption was fully underway. The early deviation from this trend, characterized by an apparent excess of SO₂ emissions relative to effusion, may reflect the release of sulfur that had already exsolved in the shallow system prior to the eruption and its rapid release, followed, after the eruptive pause, by the evacuation of the partly degassed magma. While this interpretation is consistent with the observed trends (Fig. E2), it remains tentative, given the absence of composition data for the earliest days of the eruption.

This correlation confirms that the emitted SO₂ only proceeds from the ascending magma. We observed a similar behavior for HCl, HF, and CO emission fluxes, which contrasts with the almost constant CO₂ flux throughout the 85 d of the eruption. This observation is fully consistent with the degassing model proposed by Burton et al. (2023), who suggested a decoupling between CO₂ and SO₂ degassing processes. According to this model, a CO₂-rich volatile phase, already exsolved in the upper-mantle reservoir, could account for a large fraction of the emitted CO₂ (up to ∼ 80 % according to Dayton et al., 2024), sustaining nearly constant CO₂ fluxes through the system. This difference is partially reflected in the time series of the ΔCO₂ / SO₂ ratio that steadily increases from the beginning of the eruption to the end of phase I, mimicking the trend in the VLP tremor amplitude. Such co-evolution abruptly ends at the beginning of November, from which point the ratio becomes more variable. As the CO₂ volcanic emission fluxes are constant within uncertainties during the whole eruption and the SO₂ volcanic emission fluxes are mainly controlled by the magma discharge rate, the steady increase in the C/S ratio during the first part of the eruption thus reflects the progressive decrease in the proportion of the shallow (discharge) component relatively to the deep-reservoir CO₂-rich fluids. Within the framework of overall lower SO₂ fluxes due to waning activity, the variability in the ratios of phase II reflect the control of low SO₂ contents in the plume and short-term variability in the SO₂ emissions.

The early-November transition between phase I and phase II follows the appearance of new vents at the end of October (Muñoz et al., 2022), interpreted as further propagation/opening of the underlying dike intrusion. This transition shortly anticipates an abrupt and enduring drop in tremor amplitude (both VLP and LP frequency bands; Bonadonna et al., 2022), geochemical changes (Ubide et al., 2023; Dayton et al., 2024), and hydrologic and hydrochemical changes in the aquifer. The latter comprises, e.g., an influx of pure (most likely endogenous) CO₂ (Jiménez et al., 2024) that drastically increased the groundwater HCO3- content at several sampling points from 27 October 2021 (Amonte et al., 2022; García-Gil et al., 2023b) or the establishment of a direct relationship between the level in several groundwater wells and the tremor amplitude around 7 November 2021 (Garcia-Gil et al., 2023a). VLP tremor amplitudes are especially sensitive to variations in magma ascent dynamics and conduit geometry (D'Auria and Martini, 2009; Bonadonna et al., 2022). Similar drops in VLP tremor amplitude were observed at other volcanoes, such as at Piton de la Fournaise (Duputel et al., 2023), where they were interpreted in terms of a reduction in dike dimension, heralding the end of the eruption. All of these observations suggest that these events at the beginning of November constitute a turning point in the eruption, implying significant structural changes in the plumbing system.

This turning point is particularly evident from the split described in the time series in the Sr isotopic compositions of the matrix, and it is interpreted as the consequence of a deep-origin melt injection replenishing the feeder system (Ubide et al., 2023). This interpretation further relies on this compositional change occurring in close time relationship with an increase in the magnitude of seismicity, VLP tremor amplitude, and a short-term (5 d) rebound in the time series of daily SO₂ masses. We emphasize that the short-term increase in daily SO₂ masses observed between 28 October and 2 November 2021 should be interpreted with caution. First of all, at the depth of injection, with SO₂ being mostly soluble in magma until a few hundred meters depth (Burton et al., 2023), any increase in SO₂ emissions would be due to an increase in the lava discharge rate at the surface. Moreover, this apparent peak coincides with a period of low wind speeds and a reversal in wind direction at 700 hPa (ERA5 data), which likely caused plume stagnation and gas accumulation. These meteorological conditions can lead to an overestimation of SO₂ masses derived from satellite data. Therefore, we do not interpret this increase as a definitive sign of enhanced volcanic degassing. Furthermore, the deep-origin melt injection in this period is not supported by the absence of corresponding signals in the GPS baseline time series (Charco et al., 2024), TADR data (Plank et al., 2023), or our CO₂ fluxes or CO₂ / SO₂ ratios.

The observed multiparametric transition in the eruption dynamics at the beginning of November could be alternatively explained by a significant alteration of the magma pathway between the surface and the top of the magma chamber. As the eruption waned, the ascent rate decreased and the conduit became more unstable (Muñoz et al., 2022), with the opening of new vents from mid-November (González, 2022; Walter et al., 2023), resulting in interaction with the aquifer; changes in the tremor amplitude, mixing ratio, and/or composition of endmembers; and the return of radiogenic signatures.

Once released from the magma, volcanic gases suffer a number of processes such as oxidation, scavenging, and dissolution in aqueous fluids that can alter their original composition before their detection. Integrating petrological constraints helps to understand volcanic degassing processes, linking deep degassing to atmospheric observations and refining our understanding of element cycling and the environmental impact of volcanic plumes. Here, we report such an exercise, estimating expected emissions from petrological data and comparing them with our estimates derived from atmospheric measurements.

Volcanic emissions of greenhouse gases and reactive species represent critical inputs for climate models, as they contribute to baseline radiative forcing, perturb the oxidative capacity of both the troposphere and stratosphere, influence aerosol–cloud microphysical interactions, and play a significant role in the geochemical cycling of key elements (such as sulfur, carbon, and halogens) between the Earth's surface and atmosphere (Von Glasow et al., 2009, and references therein). Accurate quantification of these natural fluxes is essential for distinguishing anthropogenic signals from background variability in the atmospheric composition. The Tajogaite eruption provides a striking example of how a single volcanic event can temporarily dominate regional atmospheric budgets. Its SO₂ emissions were approximately 15 times greater than Spain's total anthropogenic SO₂ emissions for the year 2021 (123 kt; MITECO, 2023) and even exceeded the total EU anthropogenic SO₂ emissions for that year (1.4 Tg; EEA, 2023). Assuming a conservative 20 % conversion rate of S to sulfate aerosols (see Sect. 5.1 and Fig. 6), the eruption is estimated to have produced approximately 0.5 Mt of sulfate particles. However, as the plume remained below 8 km of altitude, well within the troposphere, these aerosols were likely short-lived and regionally confined, with a limited potential to affect atmospheric radiation budgets, and a negligible climatic forcing from aerosol loading. In terms of carbon emissions, the CO₂ released by the eruption amounted to approximately 10 % of Spain's anthropogenic CO₂ emissions for 2021 (https://www.miteco.gob.es/content/dam/miteco/es/calidad-y-evaluacion-ambiental/temas/sistema-espanol-de-inventario-sei-/es-nid-edicion-2025-.pdf, last access: 31 August 2025). Emissions of CO were also substantial, corresponding to about 7 % of the 2021 national anthropogenic CO inventory (1.64 Mt; MITECO, 2023). Halogen emissions were particularly notable. The total atmospheric HCl output was around 10 times higher than the annual UK emissions since 2017 (UK National Atmospheric Inventory) and represented roughly 20 % of total European anthropogenic HCl emissions in 2014 (220 kt; Zhang et al., 2022), which are primarily associated with the energy sector (38 %) and open waste burning (23 %). Similarly, the eruption's HF atmospheric emissions exceeded UK national totals for the same period by an order of magnitude. In contrast to the purely atmospheric pathway, a significant fraction of the halogens were likely scavenged from the plume by ash particles. This process, which accounts for an estimated 31 ± 8 kt of HCl and 39 ± 11 kt of HF, provides a distinct mechanism for their re-entry into the geosphere through ash deposition. Subsequently, these ash-bound halogens are remobilized by initial rainfall events (Medina et al., 2025), where they can enter and be transported through the natural elemental cycles of Cl and F in the soil, aquifer, and marine environments.

Volcanic emissions of chlorine are known to significantly influence tropospheric ozone (O₃), as these halogens participate in catalytic cycles that destroy O₃, particularly in the presence of sunlight and moisture (Gerlach, 2004). Studying these emissions allows for the assessment of the chemical forcing of volcanoes on the troposphere, the testing of atmospheric chemistry models using real events, and the improvement of our understanding of the climatic and chemical roles of volcanic eruptions, even moderate ones like Tajogaite. In this study, we looked for signs of such an impact in the local O₃ total column from FTIR spectroscopy but found no clear evidence. However, Fig. 7 displays the time series of the O₃ partial pressure up to 8000 m, retrieved from electrochemical concentration cell (ECC) ozonesonde measurements conducted by AEMET (García et al., 2021) from Puerto de la Cruz, Tenerife. These are shown together with cumulative SO₂, HF, and HCl emissions up to 18 November 2021, corresponding to the period with the most continuous and densely sampled flux measurements. A noticeable coincidence was observed between the two sharp increases in the cumulative HCl (and HF) emissions, occurring on the 18 October 2021 and 2 November 2021 and local ozone depletion (with O₃ values near zero) at plume altitudes. This coincidence is not observed for low HCl / SO₂ ratios.

Although geometric constraints and weather conditions affected the continuity of our cumulative flux estimates, this preliminary observation suggests that halogen-induced O₃ loss may occur locally, at least where the plume was present, even if only transiently. The observed O₃ loss appears to be short-lived, with concentrations recovering shortly after, arguing against a persistent or widespread effect. To better assess the intensity and duration of this impact, more continuous time series and refined flux retrievals are required. Nonetheless, this initial evidence from the Tajogaite eruption provides a valuable basis for future investigations.