The composite “Grande Découverte–La Soufrière” volcanic complex, hereafter referred as “La Soufrière” was built during three main phases: Grande Découverte (0.25 Ma–42 ka), Carmichaël (42–11.5 ka), and La Soufrière phase that started 11.5 ka ago (Boudon et al., 1988, 1989). During the ongoing La Soufrière phase, the volcanic activity was roughly centered on the area where the current Soufrière lava dome is located. A series of ten flank collapses of low volume occurred during this phase, generating debris avalanches that flowed on the southwestern flank of the volcano, where the towns of Basse Terre and Saint Claude are now settled. Most of these debris avalanche reached the sea, 10 km away from the summit (Komorowski et al., 2005; Boudon et al., 2007; Legendre, 2012). Several dome-forming and small sub-Plinian eruptions occurred during this period (Legendre, 2012), including the last magmatic AD 1530 eruption at the origin of the Soufrière lava dome (see description below). Some deposits dated to AD 1657 have been interpreted as a small explosive Vulcanian eruption (Legendre, 2012), which would be the last magmatic event of La Soufrière. The historical period is characterized by six phreatic eruptions (AD 1690, 1797–98, 1812, 1836–37, 1956, and 1976–77) taking place from fractures and vents on La Soufrière lava dome (Barabé and Jolivet, 1958; Feuillard et al., 1983; Komorowski et al., 2005). The last phreatic eruption in 1976–1977 generated an important seismic crisis, 26 phreatic explosions, and forced the evacuation of about 73,000 persons for up to 6 months (Feuillard et al., 1983; Komorowski et al., 2005).

The AD 1530 eruption represents the last major magmatic event of La Soufrière and likely the only magmatic eruption during the last 1,400 years. It was first described by Vincent et al. (1979) and Semet et al. (1981), and revised by Boudon et al. (2008). Following a small phreatic phase, the magma reached the surface and generated a phreatomagmatic explosion followed by a flank collapse of the summit part of the volcano, the debris avalanche of which flowed on the southwestern flank of the volcano and reached the sea. Subsequently to the flank collapse, the volcanic conduit was decompressed down to the magmatic chamber, leading to an explosive sub-Plinian phase emitting pumice fallout and column-collapse pyroclastic density currents channeled into valleys. The column reached a height between 9 and 12 km, producing about 3.9 × 106 m³ DRE of juvenile products (Komorowski et al., 2008). The emitted products were first grey-white followed by white and black-and-white banded pumices. This event was followed by near-vent pyroclastic scoria fountains and ended with the growth of the present-day lava dome (Boudon et al., 2008).

The evolution of the magma reservoir prior to eruption was investigated by Pichavant et al. (2018), through a petrological and experimental study aiming at constraining both, the storage conditions of the resident magma and the characteristics of the intruding magma. These authors concluded that the main andesitic magma contained ∼46–60 vol% phenocrysts in equilibrium with a rhyolitic melt bearing 5.0 to 5.5 wt% H₂O, and was stored at a temperature (T) of 875 ± 25°C and pressure (P) of 175 ± 20 MPa (about 7 km deep, considering a rock density of 2.4 g/cm³). These pre-eruptive H₂O contents agree with the highest H₂O contents of ∼5.5 wt% measured in glass inclusions trapped in plagioclase and orthopyroxene phenocrysts. There is evidence from melt inclusions and residual glasses that T was locally as cold as 825 ± 25°C. The residual glasses are indicators of mixing between the resident and intruding melts, showing strong compositional heterogeneities from 64 to >76 wt% SiO₂.

Based on the wide compositional range of whole rocks and textural plus mineralogical evidence, Semet et al. (1981) first proposed that La Soufrière andesitic reservoir was remobilized by the intrusion of a deep mafic magma, with an about 40–60% proportion of arrival on resident magma. Pichavant et al. (2018) experimentally constrained the end-member magma composition to ∼50 wt% SiO₂ for the basaltic intrusion and 62 wt% SiO₂ for the resident andesite. Although long timescales of reservoir replenishment of about 35 ka have been proposed using U-Th-Ra disequilibria methods (Touboul et al., 2007), Pichavant et al. (2018) highlighted much shorter timescales of the order of tens of hours, based on diffusion modelling in orthopyroxenes (following the method given in Solaro et al., 2020).

In order to reconstruct the ascent-crystallization paths of the magmas in the AD 1530 sequence, one needs to infer the chemical composition of the melt from which the microlites crystallized. Ideally, this parental melt is the one in chemical equilibrium with the phenocryst rims in the storage region prior to magma ascent, and may be preserved as glass inclusions in these crystals. Yet, eruptions are often triggered by an intruding more mafic magma, whose melt mixes with the resident one, so that the melt from which microlites crystallize upon ascent may be more mafic than expected from the initial resident melt. Using the AD 1530 fused whole rocks as starting material (i.e., chemically including any late injection of mafic melt), the composition of the pre-eruptive interstitial melt was determined to be 71 ± 1 wt% SiO₂ from the phase-equilibrium experiments (Pichavant et al., 2018). Yet, the glass inclusions trapped in phenocrysts of plagioclase and orthopyroxene suggest melt SiO₂ contents of 74 ± 1 wt% (recalculated on anhydrous basis; Pichavant et al., 2018). Therefore, melts parental to microlites could span a SiO₂ range from ∼71 to ∼74 wt%, after and before mafic melt injection, respectively. As for the P-T conditions of ascent crystallization, it is commonly accepted that decompression-induced microlites crystallize nearly isothermally, that is, for the AD 1530 eruption, at pre-eruptive T of ∼875 or 825°C and P lower than the pre-eruptive P of 175 MPa.

The studied samples come from the lava dome (Figure 1B) and the deposit sequences of the Carbet Springs O1215 site and Galion River site (Boudon et al., 2008). Both deposit sequences have been combined to provide a typical sequence for the AD 1530 deposits (Figure 1C). The samples encompass: three white pumice clasts, including SG7B from the phreatomagmatic unit and SG5A and SG6A from the base of the sub-Plinian unit; two banded pumice samples from the upper part of the sub-Plinian unit: SG11A, in which SG11A_white and SG11A_dark refer to the white and dark layer, respectively, and O1215Eb that represents the white layer of a banded pumice; one dark scoria (SG3) from the Strombolian phase, and a dense block from the AD 1530 lava dome (1227A). The upper sub-Plinian O1215Eb pumice and the 1227A dome clast were sampled by Boudon et al. (2008). The Strombolian SG3 dark scoria, the lower sub-Plinian SG5A and SG6A pumice samples, the phreatomagmatic SG7B pumice, and the upper sub-Plinian SG11A pumice were sampled by Pichavant et al. (2018). These samples are basaltic andesites to andesites, with bulk-rock 56.1 to 61.7 wt% SiO₂, 2.7 to 4.2 wt% MgO, and 3.4 to 4.3 wt% alkalis (Boudon et al., 2008; Pichavant et al., 2018).

All samples, but the 1227A dome clast, show strong textures of bubble coalescence, with pore borders still showing outlines of initially individual bubbles (Figure 2). The sample vesicularities (Φ_b) are reported in Table 1 and shown in Figure 3A, as a function of the plagioclase content (Φ_Plag). The phreatomagmatic SG7B pumice (Figure 2A) shows Φ_b of 58 ± 4%. The lower sub-Plinian SG6A (Figure 2B) and SG5A (Figure 2C) pumice samples show Φ_b of 63 ± 4% and 67 ± 4%, respectively. The upper sub-Plinian O1215Eb (Figure 2D) pumice shows Φ_b of 76 ± 4%, representing the most vesiculated sample from the present study. Both parts of the upper sub-Plinian SG11A banded pumice (Figure 2E) show similar vesicularities, with Φ_b of 38 ± 9 and 39 ± 12% in SG11A_white (Figure 2F) and SG11A_dark (Figure 2G), respectively. The Strombolian SG3 dark scoria (Figure 2H) shows Φ_b of 60 ± 3%. The 1227A dome clast shows highly altered parts with up to millimetric highly hydrated and zoned silica crystals (not studied). The analyses were performed in the fresh-looking parts observed by SEM (Figure 2I), showing Φ_b of 7 ± 5%. Globally, Figure 3A shows that Φ_b decreases with increasing Φ_Plag.

The plagioclase microlite contents (Φ_Plag) and number density (N _A) are reported in Table 1 and displayed in Figure 3B. In the phreatomagmatic SG7B pumice, the microlites mostly consist of plagioclase and quartz, with some orthopyroxene, clinopyroxene, magnetite, and apatite (Figure 2A). Φ_Plag is 25 ± 7% and logN _A is 8.9 ± 0.1 m⁻². Quartz microlites represent 22 ± 5% of the total microlites. In the lower sub-Plinian SG6A pumice, the microlites mostly consist of plagioclase, coexisting with scarce orthopyroxene, magnetite, and apatite (Figure 2B). Φ_Plag is 7 ± 3% and logN _A is 8.3 ± 0.5 m⁻². In the lower sub-Plinian SG5A pumice, the microlites mostly consist of plagioclase, with numerous orthopyroxenes, clinopyroxenes, and magnetites (Figure 2C). Φ_Plag is 22 ± 4% and logN _A is 9.0 ± 0.1 m⁻². Pyroxene and magnetite represent 13 ± 3% of the total microlite content. In the upper sub-Plinian O1215Eb pumice, the microlites mostly consist of plagioclase, coexisting with scarce clinopyroxene, magnetite, and apatite (Figure 2D). Φ_Plag is 9 ± 3% and logN _A is 8.1 ± 0.2 m⁻². The upper sub-Plinian SG11A_white sample mostly contain microlites of plagioclase, with rare orthopyroxene and magnetite (Figure 2F). Φ_Plag is about 10% and logN _A is 9.1 m⁻². In the upper sub-Plinian SG11A_dark sample, microlites mostly consist of plagioclase, with numerous clinopyroxenes, rare orthopyroxenes, and magnetites (Figure 2G). Φ_Plag is 27 ± 8% and logN _A is 9.5 ± 0.4 m⁻². In the Strombolian SG3 dark scoria, the microlites mostly consist of plagioclase, with some orthopyroxene, clinopyroxene, and magnetite (Figure 2H). Φ_Plag is 20 ± 6% and logN _A is 9.1 ± 0.2 m⁻². The fresher-looking matrices of the 1227A dome clast show microlites mostly consisting of plagioclase, with numerous orthopyroxenes, clinopyroxenes, magnetites, and cristobalite (Figure 2I). Cristobalite appears as both elongated crystals and large “fish-scale” patches. The former habit suggests a primary magmatic origin (crystallization from the melt) whereas the latter reveals vapor-phase deposition (Schipper et al., 2015). The 1227A dome clast shows Φ_Plag of 38 ± 5% and logN _A of 9.6 ± 0.1 m⁻². Taken together, pyroxene and magnetite microlites represent about 10% of the total microlite content. Globally, Figure 3B shows that N _A and Φ_Plag positively correlate.

The major-element compositions of the analyzed microlites (plagioclase, magnetite, orthopyroxene, clinopyroxene, apatite, quartz, and cristobalite) are given in Supplementary Material and reported in Figure 4 for plagioclase, magnetite, orthopyroxene, and clinopyroxene. The phreatomagmatic SG7B pumice contains zoned plagioclase microlites, with cores up to An₇₃ and rims from An₂₉ to An₅₂ (Figure 2A). The less calcic microlites (∼An₃₀) have high orthoclase contents (Or, in mol%) up to Or_5-6, unlike microlites in the other samples that are < Or₁. Coexisting microlites are Mt₆₃ titanomagnetite, En_52-57 orthopyroxene, Wo_39-42En_35-39 clinopyroxene, and apatite (Mt, En, and Wo, are magnetite, enstatite, and wollastonite, contents in mol%, respectively). The lower sub-Plinian SG6A pumice contains unzoned An_54-66 plagioclase, coexisting with rare Mt₆₉ titanomagnetite, En_54-55 orthopyroxene and apatite (Figure 2B). The lower sub-Plinian SG5A pumice contains zoned plagioclases, with An_74-80 cores and An_50-67 rims, coexisting with Mt_64-65 titanomagnetite, En_66-70 orthopyroxene, and ∼Wo₄₅En₃₅ clinopyroxene (Figure 2C). The upper sub-Plinian O1215Eb pumice contains unzoned ∼ An₇₀ plagioclase coexisting with rare Mt_66-67 titanomagnetite and apatite (Figure 2D). In the upper sub-Plinian SG11A_white sample, zoned An₆₃ to An₇₇ plagioclase coexist with rare ∼ Mt₆₈ titanomagnetite and unzoned ∼ En₆₇ orthopyroxene (Figure 2F). In the upper sub-Plinian SG11A_dark sample, the plagioclases are zoned from ∼An₈₀ cores to ∼ An₆₂ rims (Figure 2G), a range similar to that of the SG11A_white microlites. Mt_69-70 titanomagnetite coexists with zoned clinopyroxene, from Wo₄₇En₄₁ (∼6 wt% Al₂O₃) core to Wo₄₄En₄₆ (∼3 wt% Al₂O₃) rim. The Strombolian SG3 dark scoria shows zoned plagioclases with An_73-85 cores and An_52-67 rims, coexisting with strongly zoned En_56-67 orthopyroxene and Wo_35-39En_38-47 clinopyroxene (Figure 2H). The 1227A dome sample shows strongly zoned plagioclases, with An_58-75Or_0-1 cores and An_43-45Or_2-3 rims. The coexisting microlites are Mt_54-58 titanomagnetite, zoned En_52-64 orthopyroxene, and Wo_35-38 En_38-39 clinopyroxene (Figure 2I).

The compositions of residual glasses are reported in Supplementary Material and displayed in Figure 5. All residual glasses are rhyodacitic to rhyolitic, spanning a SiO₂ range from 67.5 to 80.2 wt% and alkali contents from 3.5 to 7.5 wt%. The phreatomagmatic SG7B pumice and the 1227A dome sample have overlapping glass compositions, showing 79–80 wt% SiO₂, 9–11 wt% Al₂O₃, <1 wt% CaO, and 4–5 wt% K₂O. The upper sub-Plinian O1215Eb and lower sub-Plinian SG5A pumice clasts have overlapping glass compositions, showing narrow ranges of 74–76 wt% SiO₂, 12–14 wt% Al₂O₃, 2–3 wt% CaO, and ∼2 wt% K₂O. The lower sub-Plinian SG6A pumice contains 76.5 ± 0.3 wt% SiO₂, 11.8 ± 0.2 wt% Al₂O₃, 2.2 wt% CaO, and 2.3 wt% K₂O. The upper sub-Plinian SG11A banded pumice spreads a large range of glass compositions, with a continuum between the upper sub-Plinian SG11A_dark rhyodacitic and SG11A_white rhyolitic glasses, respectively showing 67–74 and 74–76 wt% SiO₂, 13–15 and 12–13 wt% Al₂O₃, 3–5 and 2–3 wt% CaO, and 1.5–2.2 and 2.1–2.5 wt% K₂O. In details, however, the glasses from both layers show small-scale chemical heterogeneities, with felsic areas (74–76 wt% SiO₂) alternating with more mafic ones (68–72 wt% SiO₂) (Pichavant et al., 2018; see also Figure 2F). The Strombolian SG3 dark scoria has rhyolitic glasses, with 73.7 ± 0.6 wt% SiO₂, 12.7 ± 0.1 wt% Al₂O₃, 3.3 wt% CaO, and 2.0 wt% K₂O. Globally, these samples show glass SiO₂ contents that do not correlate with the microlite contents and do not follow a crystallization linked to melt degassing (Figure 3C).

The H₂O contents of the residual glasses determined by EMP and SIMS are reported in Table 1 and displayed in Figure 6. The EMP and SIMS values are in good agreement within respective analytical errors, i.e., not differing by more than 0.5 wt%. One exception is the upper sub-Plinian SG11A_white sample, for which the SIMS data are about 1.3 wt% higher than the EMP values, for a reason that remains unclear. However, the SIMS values in SG11A_white (2.6 ± 0.3 wt% H₂O) do not compare to those in SG11A_dark (1.6 ± 0.2 wt% H₂O), although coming from the same pumice, and do not compare to the EMP values determined in both layers (1.1 ± 0.7 and 1.3 ± 0.6 wt% H₂O), so that the SIMS values in SG11A_white will be further discarded. The H₂O contents of the residual glasses are ∼2.5 wt% in the phreatomagmatic SG7B pumice, ∼1.5 wt% in the upper sub-Plinian O1215Eb and SG11A pumice, the lower sub-Plinian SG6A pumice, and the Strombolian SG3 scoria, and 0.4–0.7 wt% H₂O in the 1227A dome clast and lower sub-Plinian SG5A pumice.

The quench P calculated using the H₂O solubility model of Newman and Lowenstern (2002) are reported in Figure 7; note that calculating P at 825 and 875°C does not differ by more than ∼5 MPa for H₂O contents <3 wt%. The quench P, that can be viewed as the magma fragmentation P in the case of pumice, are ∼50 ± 20 MPa for the phreatomagmatic SG7B pumice, ∼20 ± 15 MPa for the lower sub-Plinian SG6A and upper sub-Plinian O1215Eb pumice, ∼16 ± 15 MPa for the upper sub-Plinian SG11A pumice, and ∼5 ± 2 MPa for the lower sub-Plinian SG5A pumice. For the Strombolian SG3 dark scoria, the calculated quench P is ∼28 ± 13 MPa and <5 MPa for the 1227A dome sample.

Back-calculating quench P (from Eq. 1 in Supplementary Material) by replacing α by Φ_b gives ∼10–30 MPa for the phreatomagmatic SG7B pumice, the lower sub-Plinian SG6A and SG5A pumice, the upper sub-Plinian O1215Eb pumice, and the Strombolian SG3 scoria, ∼50 MPa for both bands of the upper sub-Plinian SG11A pumice, and ∼130 MPa for the 1227A dome sample (Figure 7). Yet for the 1227A dome sample, the degassing obviously did not occur in closed-system, so that the quench P of 130 MPa can be discarded.

The calculated equilibrium P (using Eqs 2–6 in Supplementary Material) are shown in Figure 7 for p ≤ 200 MPa. p > 200 MPa, i.e., higher than the pre-eruptive P, were calculated for all samples, and mostly concern An _65-80 microlites that likely grew from melt compositions other than rhyolitic. The phreatomagmatic SG7B pumice shows crystallization P calculated from microlite composition from 5 to 115 MPa (71 wt% SiO₂ melt) and from 15 to 160 MPa (74 wt% SiO₂ melt). The crystallization P calculated from microlite crystallinity (i.e., plagioclase content) is 25 ± 5 MPa. If the residual glass SiO₂ equations are used, P are <10 and <65 MPa, respectively with the 71 and 74 wt% SiO₂ melts. In the lower sub-Plinian SG6 pumice, the P calculated from plagioclase composition range from ∼100 to >175 MPa for a 71 wt% SiO₂ melt and is > 135 MPa for a 74 wt% SiO₂ melt. From microlite crystallinity, P is 130 ± 25 MPa. From glass SiO₂, P is ∼55–80 MPa and ∼110–130 MPa when calculated from a 71 and 74 wt% SiO₂ melt, respectively. In the lower sub-Plinian SG5A pumice, most of the P calculated from microlite composition are >125 MPa, but rare P of ∼75 MPa are also obtained. P calculated from crystallinity are 50 ± 10 MPa. P deduced from the residual glass SiO₂ are ∼100 and ∼160 MPa, when calculated from a 71 and 74 wt% SiO₂ melt, respectively. In the upper sub-Plinian O1215Eb, SG11A_white, and SG11A_dark pumice samples, plagioclase microlite compositions are too calcic and no p < 175 MPa are calculated. In the O1215Eb pumice, the P is 140 ± 5 MPa from microlite crystallinity. From residual glass SiO₂, P are of 70–120 MPa and 125–180 MPa, respectively with the 71 and 74 wt% SiO₂ starting melts. In the Strombolian SG3 dark scoria, most of the P calculated from microlite composition are >125 MPa, but rare P of ∼75 MPa are also obtained. P calculated from crystallinity are 60 ± 22 MPa. P deduced from the residual glass SiO₂ are 100–135 MPa and 150–200 MPa, for a 71 and 74 wt% SiO₂ melt, respectively. In the 1227A dome sample, P calculated from microlite compositions range from 25 to >200 MPa, is of 20 ± 5 MPa from crystallinity, and of ∼50 MPa and <5 MPa, when calculated from residual glass SiO₂ starting from a 71 and 74 wt% SiO₂ melt, respectively.

To better fix the crystallization conditions of the microlites, it is useful to look for further constraints on the melt composition (i.e., either 71 or 74 wt% SiO₂). The phreatomagmatic SG7B matrix contains quartz, which requires SiO₂-saturated melts, i.e., SiO₂ contents > ∼79 wt% at ∼50 MPa (Martel et al., 2021). From the H₂O-SiO₂ contents of the residual glasses, quartz crystallization in SG7B would not be possible at 875°C starting from a 71 wt% SiO₂ melt, whereas it may be possible at 825°C if the starting melt contains 74 wt% SiO₂ (Figure 6). In contrast, the cristobalite-bearing 1227A dome sample is consistent with an origin by equilibrium crystallization from a 71 wt% SiO₂ starting melt at 875°C (Figure 6). In the upper sub-Plinian SG11A banded pumice, the residual glass compositions differ between the white and dark bands (Figure 5). Yet, crystallization from a melt with 74 wt% SiO₂ in SG11A_white and 71 wt% SiO₂ in SG11A_dark yields overlapping P of residual glass formation in the 130–175 MPa range (Figure 7), validating crystallization from a 74 wt% SiO₂ melt in the white bands whereas from a 71 wt% SiO₂ melt in the dark bands. The lower sub-Plinian SG6A pumice and the upper sub-Plinian O1215Eb pumice show residual glasses that are slightly more silicic than expected from equilibrium crystallization from a 71 wt% SiO₂ melt (Figure 3C). In addition, both samples contain apatite microlites, which were also found in the phreatomagmatic SG7B pumice for which the initial melt has been determined to contain 74 wt% SiO₂. Therefore, the microlites in the SG6A and O1215Eb pumice samples likely crystallized from a 74 wt% SiO₂ melt. For the lower sub-Plinian SG5A pumice and the Strombolian SG3 dark scoria, the choice of the starting melt composition is less clear. Yet, the SG3 sample does not contain apatite, is dark in colour and has a bulk-rock composition (∼58 wt% SiO₂ and ∼3.8 wt% MgO) close to that of the lava dome, comes from a Strombolian event typically involving melts more mafic than for sub-Plinian events, and chronologically just preceded the terminal dome growth, so that it is reasonable to assume a pre-eruptive melt composition similar to that of the dome, i.e., with 71 wt% SiO₂. As for the SG5A pumice, both SiO₂ contents, 71 and 74 wt%, will be considered.

Comparing the vesicularities measured in the samples to the calculated gas fraction for a closed-system degassing (from Eq. 1 in Supplementary Material) provides information on the magma degassing regime. Figure 8 shows that the phreatomagmatic SG7B pumice, the lower sub-Plinian SG6A pumice, the upper sub-Plinian O1215Eb pumice, and the Strombolian SG3 dark scoria have measured vesicularities close to the theoretical ones, suggesting that they probably degassed in closed system. The large variability in the calculated vesicularities for the phreatomagmatic SG7B pumice result from the high variability in the measured H₂O contents of the residual glasses (Table 1). In particular, locally high H₂O contents (∼3.0 wt%) lead to theoretical vesicularities (∼20 vol%) much lower than actually measured (∼60 vol%), which possibly result from rehydration of the residual glasses in the external H₂O-rich context that characterizes phreatomagmatism. Contrary to these samples, the lower sub-Plinian SG5A homogeneous pumice and the upper sub-Plinian SG11A banded pumice seem to have partially lost bubbles, which could result from local permeability development and partial outgassing in an open-system degassing regime. For the SG5A pumice, the H₂O contents of the residual glasses are lower than for the other pumice (0.7 vs >1.1 wt%; Table 1), whereas the measured vesicularities compare (i.e., ∼74 vol%), so that this could alternatively reflect post-fragmentation H₂O loss from the glasses (with a resulting overestimated theoretical vesicularity). The 1227A lava dome clast shows a much lower vesicularity (<10 vol%) than calculated (>90 vol%) from its low residual glass H₂O content (0.4 wt%), highlighting a strong outgassing, as expected from the slow ascent rates commonly prevailing during lava dome emplacements.

Martel and Iacono-Marziano (2015) demonstrated that the timescales of degassing, i.e. of bubble nucleation, growth, coalescence, and outgassing, primarily depend on the viscosity of the vesiculated magma under very low shear. The formulation of the bubble-melt viscosity used by these authors was set for crystal-free melts, so that the degassing timescales have to be taken with caution when dealing with crystal-bearing magmas. The bubble-melt viscosity was calculated from the steady simple-shear flow equation of Llewellin et al. (2002), calculating the viscosity of the hydrated residual glass at 875 and 825°C after Hess and Dingwell (1996) and using a capillary number (Ca) from 1.0 to 2.5 (characterizing little deformed bubbles). The calculations give bulk viscosities of 10^5.1–5.3 Pa s at 875°C and 10^5.6–5.9 Pa s at 825°C, and do not significantly differ for Ca varying from 1.0 to 2.5. Since all pumice samples show textural evidence of bubble coalescence, the ascent durations must have been longer than 1 h (Martel and Iacono-Marziano, 2015; their Figure 6). The magmas of the phreatomagmatic SG7B pumice, the lower sub-Plinian SG6A pumice, the upper sub-Plinian O1215Eb pumice, and the Strombolian SG3 dark scoria, did not outgas significantly, so that maximum ascent durations could be 10–100 h (0.5–4 days). The partially-outgassed lower sub-Plinian SG5A pumice and upper sub-Plinian SG11A banded pumice have calculated bulk viscosity from 10^5.2 Pa s for T of 875°C to 10^5.9 Pa s at 825°C, which suggest ascent durations >100 h. The 1227A lava dome sample that nearly fully outgassed has a calculated bulk viscosity of 10^5.2 and 10^5.8 Pa s at 875 and 825°C, respectively, suggesting emplacement duration >1,000 h (>40 days).

Ascent crystallization of microlites can be characterized by two main criteria. Firstly, microlite compositions should be compatible with crystallization from the pre-eruptive melt composition, i.e., with compositions more evolved than coexisting phenocryst rims (assuming an ideal case of phenocrysts no more growing upon ascent). Secondly, the proportion of microlites grown during ascent cannot be higher than expected for decompression rates permitting equilibrium crystallization; higher microlite contents would reflect inherited crystals. Based on these criteria, only two samples (the phreatomagmatic SG7B pumice and the 1227A dome clast) show major ascent-crystallization of microlites, three of them (the sub-Plinian O1215Eb, SG6A, and SG11A pumice samples) appear devoid of decompression-induced crystallization, and two others (the sub-Plinian SG5A pumice and the Strombolian SG3 scoria) show scarce ascent-crystallization, as detailed below.

The phreatomagmatic SG7B pumice fulfills the criteria for ascent-crystallization: the microlite compositions are more evolved than those crystallizing under pre-eruptive conditions (Figure 4) and the microlite content agrees with equilibrium crystallization at 825°C from a 74 wt% SiO₂ starting melt (Figure 3C). The continuous range of plagioclase microlite compositions suggests steady-state crystallization from reservoir to P of ∼15 MPa (Figure 7); note that ascent-crystallization also took place as overgrowths on phenocrysts, as evidenced by some ∼ An_28-40 plagioclase and En_53-54 orthopyroxene phenocryst rims (Figures 4A–C). The good agreement between crystallization P deduced from compositions and contents of plagioclase microlites and those calculated from the residual glass compositions (Figure 7) suggests a near-equilibrium crystallization process. Moreover, these crystallization P agree with the degassing P calculated from vesicularity and H₂O contents of the residual glasses (Figure 7), so that similar conditions are established for degassing and crystallization in the conduit.

For the 1227A dome sample, the P calculated from the compositions of the plagioclase microlites suggest that microlites continuously crystallized during ascent (Figure 7). Orthopyroxene and clinopyroxene microlite compositions (Figures 4C,D) confirm continuous crystallization during ascent. Degassing P retrieved from residual glass H₂O contents are <30 MPa, reaching p < 5 MPa using residual glass compositions, highlighting a crystallization process close to equilibrium, although likely partly hampered by the slow crystallization kinetics prevailing at low P. Yet, lava domes commonly cool upon emplacement, thus driving plagioclase to more albitic compositions than expected from isothermal decompression, so that the crystallization P calculated for this sample have to be taken as minimum values. The Mt_55-58 titanomagnetite microlites (Figure 4B) show about 10 mol% Mt less than both the microlites in the other samples and the dome phenocrysts (Supplementary Material). This confirms late-emplacement cooling of the AD 1530 lava dome, rather than evidencing more reduced conditions that appear unlikely since the dome cooled in air.

The lower sub-Plinian SG6A pumice and the upper sub-Plinian O1215Eb, SG11A_white, and SG11A_dark pumice, all show microlite compositions similar or less evolved than the pre-eruptive ones (Figure 4), with calculated p > 130 MPa (Figure 7). Therefore, most of microlites must have crystallized at reservoir depth prior to eruption, without significant decompression-induced crystallization. The strong contrast between these crystallization P and the degassing P (≤50 MPa) deduced from vesicularity and residual glass H₂O contents (Figure 7) shows that crystallization and degassing were largely decoupled. Since degassing is a process much faster than crystallization (e.g., Martel and Schmidt, 2003), strong kinetic crystallization delays are implied. In both bands of the SG11A pumice, the orthopyroxene microlites are more magnesian than the phenocryst rims (Figure 4C), likely reflecting crystallization from a more mafic melt. The SG11A_dark sample additionally contains Wo_44-47En_41-46 clinopyroxene microlites, which cover the compositions of the phenocrysts, from common Wo₄₂ (∼1 wt% Al₂O₃) augites to Wo₄₆ (∼6 wt% Al₂O₃) that can be considered as xenocrysts inherited from more mafic compositions (Pichavant et al., 2018). Also, SG11A_dark has a strongly higher microlite content than expected from equilibrium crystallization (Figure 3C). Overall, these observations argue in favor of mostly inherited microlites that crystallized from a more mafic melt intruding the reservoir, as described by Martel et al. (2006). Considering that SG6, O1215Eb, and SG11A_white pumice did not crystallize microlites during ascent, i.e., all microlites are inherited, suggests about 5–10 vol% inherited plagioclase microlites (Φ_plag values in Table 1) for the magmas coming from the coldest part of the reservoir.

The lower sub-Plinian SG5A pumice and the Strombolian SG3 dark scoria mostly crystallized plagioclase microlites at P of ∼125–200 MPa, i.e., similar to the pre-eruptive P, and some rare microlites at P of 50–75 MPa (Figure 7). The P deduced from the residual melt compositions are similar to the pre-eruptive P when calculated starting with a 74 wt% SiO₂ melt (SG5A) and ∼100–130 MPa when calculated with a 71 wt% SiO₂ melt (SG5A and SG3). Therefore, the amount of plagioclase microlites grown at 50–75 MPa was likely too low to drive the residual glass compositions to those expected at 50–70 MPa. This may be illustrated by the results of the phase-equilibrium experiments, showing that increasing the plagioclase content by 5 vol% from 75 to 50 MPa (Supplementary Material) does not increase the glass SiO₂ contents, which is capped at 79–80 wt% SiO₂ (Supplementary Material). The sub-Plinian SG5A sample contains orthopyroxene and clinopyroxene microlites that are compositionally less evolved than expected from decompression-induced crystallization at p < 175 MPa, either from 71 or 74 wt% SiO₂ melts (Figure 4). The Strombolian SG3 dark scoria contains orthopyroxene, clinopyroxene, and titanomagnetite microlites, whose compositions are on both sides, either mafic or felsic, of the pre-eruptive compositions (Figure 4). Additionally, the microlite contents are higher than expected from equilibrium crystallization starting from a 71 wt% SiO₂ melt (Figure 3C), suggesting a significant amount of inherited microlites. Therefore, most of the microlites in these two samples are inherited from mixing processes at reservoir level, with only a very few fraction crystallizing during magma ascent. Neglecting ascent-crystallization in the cases of the Strombolian SG3 scoria, the upper sub-Plinian SG11A_dark pumice, and the lower sub-Plinian SG5A pumice, may also suggest about 20–27 vol% inherited microlites from the Φ_plag values in Table 1.

Bubble number density has been demonstrated to correlate with decompression rate in rhyolitic melts (Toramaru, 2006). Yet, the AD 1530 magmas obviously underwent too much bubble coalescence (e.g., Figure 2) to make it possible to restore the original bubble number densities from the pumice samples. Alternatively, the decompression rate can be estimated from the microlite number density, in the case of a single event of crystal nucleation driven by melt H₂O exsolution (Toramaru et al., 2008). Among the samples that nucleated plagioclase microlites during ascent, the presence of inherited microlites (predominant in SG5A), around which a part of the decompression-induced crystallization occurred (zoned microlites), means that ascent-crystallization partly occurred as growth around preexisting crystals instead of nucleation of new crystals. This disagrees with the requirements of the decompression-rate meter of Toramaru et al. (2008), implying that the calculated data must be treated with caution. The calculations are reported in Figure 9, and converted below into ascent durations assuming a magma travel from the ∼7 km deep reservoir to subsurface. The phreatomagmatic SG7B pumice could have ascended at rates of ∼400 Pa/s (ascent velocities of ∼0.002 m/s; duration of ∼30 days), the lower sub-Plinian SG5A pumice at rates from 40 to 200 Pa/s (velocities between ∼0.003 and 0.015 m/s and duration from 30 to 2 days, calculated for 74 and 71 wt% SiO₂ melts, respectively), and the Strombolian SG3 dark scoria at rates of ∼300 Pa/s (velocity of ∼0.015 m/s; duration of ∼2 days). The decompression rate of 900 Pa/s (velocity of ∼0.05 m/s; duration of 1.5 days) calculated for the 1227A dome sample is unrealistically fast, likely biased by continuous microlite nucleation upon ascent and by non-isothermal crystallization (late microlites formed upon cooling), so that it will not be further considered.

In comparison with the timescales of degassing deduced from the viscosity of a vesiculated magma (from Martel and Iacono-Marziano, 2015), the timescales of magma ascent calculated from the microlite content (from Toramaru et al., 2008) are about the same order of magnitude for the Strombolian SG3 dark scoria and the sub-Plinian SG5A pumice when calculated using a 71 wt% SiO₂ melt (∼2 days versus maximum 0.5–4 days) and much longer for the phreatomagmatic SG7B pumice and the sub-Plinian SG5A pumice when calculated using a 74 wt% SiO₂ melt (∼30 days versus maximum 0.5–4 days). Concerning the sub-Plinian SG5A pumice, this could argue for a starting interstitial melt closer to 71 than 74 wt% SiO₂, i.e., tapping the main reservoir. As for the phreatomagmatic SG7B pumice, we conclude that the microlite number density could actually reflect a magma ascent duration of ∼30 days, whereas the shorter duration of 0.5–4 days attributed to degassing could reflect outgassing difficulties, as explained below.

The AD 1530 eruptive sequence is revisited below using temporal constraints from microlite crystallization and degassing, as illustrated in Figure 10.

Explosive eruptions, including Plinian, sub-Plinian, and Vulcanian events, and involving dacitic to rhyolitic magmas, commonly show syneruptive decompression rates from 100 to ∼3,000 Pa/s (ascent rates of 0.01–0.1 m/s), based on petrological, experimental, and monitoring data (Cassidy et al., 2018). These rates of decompression agree with the decompression rates of 270–3,500 Pa/s that we propose for the AD 1530 sub-Plinian event of La Soufrière. Nevertheless, refinements can be proposed by restricting the comparison to sub-Plinian events for which magma ascent rates are known, such as the 1996 explosion of Soufrière Hills, Montserrat, and the 2010 eruption of Merapi, Indonesia.

The September 17th, 1996, short-lived sub-Plinian event (VEI 3) of Soufrière Hills volcano (Robertson et al., 1998) produced homogeneous pumices. From the absence of reaction rims around the amphibole phenocrysts in the pumice samples, Devine et al. (1998) estimated magma ascent durations >0.01 m/s, i.e., < 4 days. These ascent rates are comparable to the ascent durations <4 days proposed for the AD 1530 sub-Plinian pumice samples of La Soufrière.

The 2010 eruption of Merapi volcano started on October 26th by a phreatomagmatic explosion, followed by 7 days of dome growth and collapse, a sub-Plinian event (VEI 4) on November 5th, and terminated by a new dome growth (Surono et al., 2012; Komorowski et al., 2013). This eruptive sequence compares to the AD 1530 sequence of La Soufrière. The 2010 Merapi sub-Plinian event produced white pumice and grey scoria. The matrices of the white pumice have some similarities with those of the AD 1530 pumice of Soufrière: they show limited outgassing and tabular feldspar microlites with number densities of 10^9.8 m⁻² and contents of 16–32 vol% (Preece et al., 2016) (Figure 3). Although the 2010 Merapi magma is more mafic (bulk-rock of ∼55–56 wt% SiO₂ and residual glasses of ∼68 wt% SiO₂ for the white pumice; Preece et al., 2014) and was stored at higher temperature (925–950°C; Erdmann et al., 2016) than for the AD 1530 La Soufrière magma, such textural similarities indicate comparable ascent conditions. The Merapi sub-Plinian event itself lasted 2 h and the preceding dome emplaced during 7 days, thus bracketing the ascent timescales of Merapi magma between 2 h for the pumice to 7 days for the denser clasts. These rates are in the same order of magnitude than those proposed for the AD 1530 sub-Plinian event (0.5–4 days; Figure 10).

Volcanic eruptions often start with a phreatic event that uncorks the conduit, and whether a magmatic eruption follows, it is still not predictable. The trigger of a magmatic eruption, and the possible transitions between effusive and explosive events, relies on interrelated and competing magmatic properties and processes, such as magma viscosity, degassing, conduit geometry and permeability, which ultimately control magma ascent rate (Cassidy et al., 2018). However, as only slight variations in the involved parameters may force a change in the eruptive style, we need to understand the factors controlling the conduit processes.

The study of the phreatomagmatic SG7B pumice from the AD 1530 deposit suggests that the early magma emplaced in a month or so at depth of ∼2 km, likely with little surface outgassing since bubbles did not escape the melt at depth. Such an intrusion could have been similar to that inferred during the 1976–77 phreatic eruption (Feuillard et al., 1983; Villemant et al., 2014). At the present day, a magma intrusion at depth could be detected via geophysical (seismic and ground deformation) methods, depending on the magma volume involved. Such geophysical signals of magma intrusion combined with an absence of surface degassing signals may be interpreted as gas pressurization at depth and potential precursors for an impending explosion. In AD 1530, the phreatomagmatic explosion likely destabilized the summit area of the volcano, generating a flank collapse that depressurized the underlying magma, leading to a sub-Plinian eruption. Such a scenario is nowadays all the more probable considering that the highly fractured and hydrothermalized summit lava dome is unstable (Rosas-Carbajal et al., 2016). In AD 1530, the collapse of the eruptive column during the sub-Plinian phase generated pyroclastic density currents channeled through two or three valleys, but the growth of the lava dome was not accompanied by destruction into pyroclastic density currents, as frequently occurred during the 1995–2010 Soufrière Hills eruption (Sparks and Young, 2002). Yet, superficial laterally directed explosions can also occur, particularly at the beginning of lava dome growth (e.g., historical eruptions at Montagne Pelée; Boudon and Balcone-Boissard, 2021). Such lava dome-related eruptions, which commonly depend on magma ascent rate (Martel, 2012) and lava dome size (Taisne and Jaupart, 2008), have to be considered as likely hazards in future eruptive scenarios of La Soufrière.

Therefore, geophysical and geochemical signals of shallow intrusion of new magma are of prime importance at La Soufrière volcano, because the produced heat could enhance the alteration of the lava-dome rocks eventually generating cascade processes including lava-dome and/or flank collapse and magmatic eruptions. Such a dramatic eruptive scenario threatens up to 70,000 people in southern Basse-Terre island of the region.