Subduction zones are among the most complex tectonic systems due to their intricate geodynamics and the multiple interacting faults and structures affecting the subducting plate, the megathrust, and the upper plate and volcanic arc (Fig. S1 in the Supplement). They are also the locus of the great M∼8+ megathrust earthquakes and their disastrous effects. Because of their extraordinary nature, these megathrust earthquakes are the main focus of subduction seismicity and seismic hazard studies (e.g., Graham et al., 2021; Wirth et al., 2022). Yet, the seismic hazard in subduction zones results from several different sources, whose contributions depend strongly on the considered locations and hazard spectral period (e.g., Frankel et al., 2014), thus requiring specific hazard modeling techniques (Pagani et al., 2020a). The Lesser Antilles are a good example of the difficulties associated with subduction zone seismicity and seismic hazard studies, with additional issues due to the slow deformation rate of the arc, the limited land instrumentation coverage, and the multi-country situation. The characteristics of large damaging earthquakes, such as in 1839 (M = 7.5–8) and 1843 (M = 8–8.5), remain debated in terms of locations, depths, mechanism, or magnitudes (Bernard and Lambert, 1988; Feuillet et al., 2011a; Hough, 2013; van Rijsingen et al., 2021). Seismic hazard models are few (Bozzoni et al., 2011), generally at the scale of the whole Caribbean region with a small focus on the Lesser Antilles (Pagani et al., 2020b; Zimmerman et al., 2022). Yet, recent studies provide important new constraints on the Lesser Antilles seismotectonics and highlight key issues for seismic hazard in subduction zones, such as the very low present-day interseismic coupling of the megathrust (van Rijsingen et al., 2021), in disagreement with the Philibosian et al. (2022) hypothesis, and its implications for seismic hazard.

Estimation of the probabilistic seismic hazard assessment (PSHA) is based on (i) identifying earthquake sources, (ii) characterizing their magnitude–frequency distributions, (iii) the distribution of source-to-site distances, and (iv) predicting ground motion intensity (Baker et al., 2021). In this study, we focus on the first step by determining earthquake area and fault sources of the Lesser Antilles arc (hereafter the “seismotectonic zoning model”). Previous PSHA of the French islands was conducted in 2001 (Martin and Combes, 2001). This seismotectonic zoning model and resulting PSHA calculation were, as in this study, carried out in response to a request from government authorities. They both serve as a basis for the national seismic zoning revision process. The Martin and Combes (2001) crustal zoning was primarily based on structural data (gravimetric, magnetic, geologic, seismic and topo-bathymetric data from 0 to 30 km depth). The subducting-plate zoning was based on plate interface dip variations induced by the presence of ridges and fractures. No specific zoning for the plate interface was proposed. Previous seismotectonic zoning models of the Lesser Antilles did not consider mantle wedge seismicity and did not integrate specific zoning for the volcanic seismicity (Martin and Combes, 2001; Pagani et al., 2020b; Zimmerman et al., 2022). Due to its physical properties, the mantle wedge is generally not considered a site for seismic nucleation, with only weak and diffuse seismicity (Hasegawa et al., 2009). However, recent works indicate more sustained activity (Aotearoa / New Zealand, Davey and Ristau, 2011; Japan, Uchida et al., 2010; the Alps, Malusà et al., 2016; the Lesser Antilles, Laigle et al., 2013b). The Mw = 4.5 earthquake in Aotearoa / New Zealand (Davey and Ristau, 2011) and a possible 1974 M = 6.9–7.5 event in the Lesser Antilles (McCann et al., 1982) raise the question of the importance of considering this seismicity for seismic hazard assessment. Allocating this seismicity to the other seismogenic sources (i.e., upper-plate crust, lower plate, or the subduction interface) results in biased hypocenter distance hazard calculation. Moreover, working on PSHA in volcanic regions is a challenge regarding earthquake characteristics: low magnitude and high seismic wave attenuation. In the Lesser Antilles, the Nevis crisis of 1950–1951 caused damage to buildings, with a maximum magnitude of Mw = 4.3 (International Seismological Centre (ISC) catalog) and intensity VIII (Willmore, 1952). It is therefore important to be able to propose a way to consider it.

In this study, we present a new seismotectonic zoning model built for the Lesser Antilles, with seismogenic source characteristics provided for future seismic hazard assessment and subject to a detailed technical report (Foix et al., 2023a) requested from the government. The seismotectonic zoning model comprises the Lesser Antilles upper plate, subducting oceanic plate, subduction interface, mantle wedge, and volcanoes, based on a compilation and reanalysis of seismicity and fault catalogs, earthquake focal mechanisms, and geodetic data. Here, we present the main scientific points leading to the zoning model, as well as a focus on the Marie-Galante graben, and we estimate and compare the extension rates from seismic and geodetic data. The zoning model provides earthquake rates and maximum magnitudes, including uncertainties and multiple options associated with the current state of knowledge. We also provide recommendations on improvements necessary for future seismotectonic and seismic hazard studies.

The Lesser Antilles are the result of the subduction of the North American and South American plates beneath the Caribbean Plate at a present-day convergence rate of ∼ 20 mm yr⁻¹ (DeMets et al., 2010) (Fig. 1a, b). Due to the trench convex shape, the convergence direction is almost arc-perpendicular south of Guadeloupe and becomes more oblique to the north. The Lesser Antilles subduction is bounded at its northern end by the E–W en échelon strike-slip Anegada Passage system (Laurencin et al., 2017) and at its southern end by the strike-slip El Pilar Fault (Mann et al., 1991). Its western limit is marked by the Grenada Basin separating the active volcanic arc from the Aves Ridge.

The subducting-plate seafloor is marked by numerous fracture zones and ridges, affecting the accretionary wedge, the upper-plate tectonics, and the megathrust seismogenic behavior (Pichot et al., 2012; Ezenwaka et al., 2022, and references therein). The Barracuda Ridge and Tiburon Rise (Fig. 1b) mark the limit between the South American and North American plates, with convergence at the Barracuda Ridge (Patriat et al., 2011) and a N–S difference in the oceanic crust thickness (Kopp et al., 2011; Laurencin et al., 2018) and fracturing (Fig. 1b). The Wadati–Benioff seismicity is also impacted by the presence of the subducted fractures and ridges, as well as a possible slab tear (Harris et al., 2018).

The upper-plate tectonics are marked by a strong N–S asymmetry. South of ∼ 15° of latitude, the upper plate is characterized by the large Barbados accretionary wedge (e.g., Speed and Larue, 1982; Gomez et al., 2018; Deville, 2023), the Tobago forearc basin, and a single volcanic arc with few major active faults and structures (Fig. 1b). In contrast, the northern region is associated with a thin sedimentary wedge and long-term subduction erosion. The forearc basins are affected by trench-perpendicular normal faults, and the active volcanic arc (coupled with an old inactive arc in the forearc position) is affected by trench-parallel normal and strike-slip faults (Boucard et al., 2021; Feuillet et al., 2002, 2011b). The N–S asymmetry is also present in the arc crustal thickness, increasing from 20–25 km in the south to reach ∼ 35 km in the north beneath St Kitts (Schlaphorst et al., 2021). Across the arc, geodetic velocities indicate very small motions relative to the Caribbean Plate and < 1 mm yr⁻¹ of N–S intra-arc extension (Symithe et al., 2015; van Rijsingen et al., 2021), generally consistent with earthquake focal mechanisms and forearc basin structures (Allen et al., 2019; Lindner et al., 2023).

The overall N–S asymmetry of the Lesser Antilles subduction system also appears in the instrumental seismicity, with higher activity north of ∼ 15° of latitude compared to the southern region (McCann et al., 1984; Hayes et al., 2014; Fig. 1c). However, this pattern does not appear in the historical catalog (Lambert, 2009; Bertil et al., 2023; Lambert and Samarcq, 2024; Fig. 2a). This discrepancy highlights the difficulty and limits of the seismicity catalogs. Recent damaging earthquakes are associated with magnitudes M = 7–7.5 (e.g., 1953 Mw=7.3 south of Martinique, 1969 Mw=7.2 near Barbados, 1974 Mw=7.5 between Barbuda and Antigua, 2007 Mw=7.4 north of Martinique; Fig. 2c). All are attributed to normal faulting either within the subducting plate (1953, 1969, and 2007; Russo et al., 1992; Dorel, 1981; Régnier et al., 2013) or within the upper plate or mantle wedge (1974; McCann et al., 1982; Feuillet et al., 2002, 2011b). One of the outstanding characteristics of the Lesser Antilles subduction is the lack of instrumental large (M>6.5) thrust earthquakes associated with the subduction interface.

In seismic hazard analysis, seismotectonic zoning models are designed to fill knowledge gaps, and the Lesser Antilles arc and its active seismicity are poorly known. Area sources are built by crossing both qualitative and quantitative criteria with subjective interpretation. When knowledge is sufficient (data collected homogeneously over space and time), the zoning can be purely based on quantitative criteria. In this study, the most recent data have been collected and integrated. The seismotectonic zoning model comprises area sources and fault sources, with the former defining domains of uniform tectonic and seismicity characteristics. Area source boundaries are constructed following three principles, considering existing knowledge and uncertainties: (1) boundaries are defined in priority by (a) the seismicity distribution, (b) faults and local tectonics, (c) geodetic data, and (d) local geology. (2) A single large zone is preferred to several small zones, unless data clearly show different tectonic and seismicity characteristics that require zone divisions. (3) Area sources are chosen to prevent seismicity dilution within zones that are too large. Seismicity distribution, deformation style extracted from focal mechanisms, crustal fault locations, and tectonic features were superimposed using a geographic information system (GIS) tool as illustrated in Figs. 4, 7, 9, and 10. Boundaries depict a consensus between seismotectonic data and discussions with Lesser Antilles experts (Lesser Antilles Working Group). To address geological and tectonic complexities and specificities of the region, we have opted for a division into several zones instead of a simple north–south division. Specific detailed subdivisions are limited to cases where they are supported by the available data. This methodology better constrains each source and allows us to manage uncertainties more precisely by limiting the propagation of errors to the whole region.

The primary seismicity catalog is a homogenized Mw instrumental catalog (hereafter ISCU-cat) composed of seismicity extracted from the International Seismological Centre (ISC, 2023) catalog and regional catalogs from 1906 to 2021 (Bertil et al., 2022, 2023; Bertil, 2024). The Mw homogenization was done using reference magnitudes given by the Global Centroid Moment Tensor (GCMT; Dziewonski et al., 1981; Ekström et al., 2012) project and the National Earthquake Information Center (NEIC; Guy et al., 2015). This catalog is considered complete since 1964 for Mw≥4.3, and completeness for Mw≥4 is not reached until 1985 (Bertil et al., 2022, 2023; Bertil, 2024). Taking Mw≥4.0 for the whole catalog limits bias in regional completeness, even if few data exist before 1985 (6 %). No hypocenter relocations were conducted, but a first-order quality score is supplied based on the phase number (pn) from A (pn ≥ 1000) to E (pn ≤ 3) instead of location uncertainties. ISCU-cat comprises 0.2 % A, 5.3 % B, 18.6 % C, 75.7 % D, and 0.2 % E. The ISCU-cat spatial variations in seismicity are analyzed to determine area source boundaries and their associated Gutenberg–Richter distributions. The ISCU-cat needs for magnitude estimation improvements are discussed in Sect. 5.

ISCU-cat is completed for local analysis around Guadeloupe and Martinique with the instrumental catalogs IPGP (Institut de physique du globe de Paris; Saurel et al., 2022) and CDSA (seismological data center for the French Lesser Antilles; Massin et al., 2021) and regionally by the historical catalogs SisFrance Antilles (Vermeersch et al., 2002; Lambert, 2009; Bertil et al., 2022; Lambert and Samarcq, 2024) and SARA (South American Risk Assessment; Gómez-Capera et al., 2017) (Fig. 2a, b). SisFrance comprises 19 % of quality B (corresponding to ∼ 10 km location uncertainty) and 81 % of quality C (10–20 km uncertainty). For the SARA pre-1964 catalog, no uncertainties or quality is available. Historical seismicity allows us to discuss past activity of specific regions and to compare it with instrumental seismicity and geodetic data (Foix et al., 2023a). The exact location of some historical events is still debated, such as the 1839 and 1843 earthquakes. Bernard and Lambert (1988) and McCann et al. (1984) interpreted them as megathrust earthquakes, whereas van Rijsingen et al. (2021) proposed that the 1843 event had a smaller magnitude or different mechanism or location within the subducted slab and that the 1839 event could also be located in the subducted slab.

We construct a composite catalog of earthquake focal mechanisms comprising 572 events from the GCMT (Dziewonski et al., 1981; Ekström et al., 2012) and ISC (Lentas, 2018; Lentas et al., 2019) as well as Corbeau et al. (2019, 2021), González et al. (2017), and Ruiz et al. (2013), from 1977 to 2021, hereafter named FMAnt2021 (Focal Mechanisms Antilles 2021). From FMAnt2021, we compute average faulting types and P- and T-axis orientations on a regular grid using the Mazzotti et al. (2021) method (smoothing distance of 40 km, minimum of three mechanisms within a radius of 50 km). FMAnt2021 allows us to determine area source boundaries and deformation types.

The oceanic subducted plate surface is based on a unification of the slab models (hereafter Bie-Lau Slab) of Bie et al. (2019) and Laurencin et al. (2018) (Figs. 3b and S2 in the Supplement). We have georeferenced and digitized these interface geometries to combine them into one unique surface using a GIS tool every 10 km of depth. We then transformed it into a grid that can be extrapolated to get a surface and be used for earthquake sorting. Paulatto et al. (2017) slab and Moho geometries were not used in order to keep consistency along the arc. The slab top geometry may vary not only according to the publications, but also according to the interpretation of the presence or absence of a slab. Beneath the Lesser Antilles central area, at depths of around 170–200 km, Lindner et al. (2023) observed a seismic gap in the lithosphere of subducted American plates, whereas Braszus et al. (2021) observed a continuous slab, in agreement with the tomography of Bie et al. (2019). These differences would have an impact on earthquake sorting and the resulting statistics.

No unified database of crustal faults exists for the Lesser Antilles, and knowledge is heterogeneous along the arc. We use crustal faults from the Global Earthquake Model of the Caribbean region (Styron et al., 2020) completed with local studies of Boucard et al. (2021), Feuillet et al. (2001, 2002, 2004, 2011a, 2011b), Garrocq et al. (2021), Laurencin et al. (2017, 2019), and Leclerc et al. (2016) (Fig. 1b). Various unknowns and interpretations remain on fault activities. The Anegada Passage fault system (Fig. 3a, label 5) motion was interpreted from extensional faulting to sinistral or dextral transtension (Laurencin et al., 2017, and references therein). Fault activity is sometimes debated, such as for the V-shaped basin faults from Guadeloupe to Saint Kitts and Nevis (Feuillet et al., 2001, 2011a; Boucard et al., 2021). Moreover, structures still need to be imaged and understood south of Saint Lucia. Finally, we use the geodetic velocities from van Rijsingen et al. (2021) to calculate geodetic strain rates (Fig. 4) and, complemented by micro-atoll subsidence data (Philibosian et al., 2022), to test models of megathrust interseismic coupling on 2D cross-sections. We also use the geodetic velocities to calculate extension rate for some specific regions and determine an associated standard uncertainty (σ/√N, where σ is standard deviation and N denotes measurements).

Details of the area source limits and geometries are given in the Supplement (Figs. S1 and S2 and Table S1) and in a risk prevention department report (Foix et al., 2023a), with information on the Gutenberg–Richter distribution for each area source (Table S1 in the Supplement). The Supplement for this article also includes the minimum magnitude sensitivity analysis impact for the Marie-Galante graben (Fig. S4 in the Supplement) and a comparison between Martin and Combes (2001) and this study (Fig. S5 in the Supplement).

Recent studies proposed alternative methods to analyze seismotectonics. Mazzotti et al. (2011) incorporate geodetic data to derive the strain rate and seismic moment rates. Beauval et al. (2018) use geologic and geodetic slip rates to characterize a weakly instrumented zone. In this study, we propose an updated seismotectonic zoning model for seismic hazard assessment in the Lesser Antilles, enriched by numerous recent improvements in the understanding of the regional seismotectonic context. We mixed traditional seismicity-based methods with geological and geodetic data. The zoning comprises area and fault sources for the upper plate and area sources for the subduction interface, the subducting plate, the mantle wedge, and volcanic centers. Major updates compared to previous studies (Martin and Combes, 2001; Bozzoni et al., 2011; Pagani et al., 2020b; Zimmerman et al., 2022) consist of (i) better depth resolutions, owing to new slab and upper-plate geometries; (ii) a specific area source for the Marie-Galante graben, based on combined new tectonic, seismic, and geodetic data; (iii) new propositions of mantle wedge and volcanic zoning; and (iv) fully revised area sources for the subduction interface based on the integration of geodetic and coral data in a combined seismotectonic zoning model. Positioning area source boundaries may result in underestimating seismic hazard over short and long return periods, from 50 to 10⁵ years (Avital et al., 2018). To go further with PSHA calculations, we highly recommend adding an uncertainty in area source boundaries when it is possible. Alternative models may be used for PSHA estimations in regions with high levels of data and knowledge (e.g., California, Italy). In the Lesser Antilles, each subset of data is incomplete, which does not allow us to propose alternative zoning.

As well as the major updates listed above, we specifically compare our study to the seismotectonic zoning used for the French Lesser Antilles PSHA (Martin and Combes, 2001) and propose entirely revised upper-plate and intraslab zoning geometries (Fig. S5 in the Supplement). New knowledge on the geological context, Moho and slab geometries, seismicity catalogs and rates, and geodesy and seismic imaging allows us to propose a new zoning scheme. Only the trench and western volcanic arc zoning limits are consistent between the two models (Fig. S5 in the Supplement). The greatest change lies in the fact that we divide our model according to the subduction structure (upper-plate crust, interface, intraslab, mantle wedge, volcanoes), whereas Martin and Combes (2001) considered all the seismicity from 0 to 30 km depth to propose a “shallow” zoning. Recent seismotectonic zoning models also consider this subduction structure division (Zimmerman et al., 2022) as the seismicity of each domain presents specific characteristics, for instance in terms of maximum earthquake magnitude. Regarding outer-rise maximum-magnitude earthquake records around the world (e.g., Meng et al., 2012), we also decide to consider this region in our model in contrast to Martin and Combes (2001). We also exclude the division of the backarc region proposed by Martin and Combes (2001), as the sub-zones are not justified by a specific change in instrumental or historical seismic records or by specific fault data (activity, slip rate). Finally, our data set allows us to cover a larger area (Fig. S5 in the Supplement). This zoning model can also be used as wave attenuation laws in terms of seismic ground motion attenuation (e.g., Youngs et al., 1997).

Our study also highlights major remaining epistemic uncertainties and unknowns that can impact seismic hazard assessment, although the overall seismotectonics of the upper plate are relatively well defined in the few areas studied with dedicated surveys, given the very great structural heterogeneity observed. The specifics of most of the arc and forearc tectonics require further data collection and analyses. For example, an apparent velocity gradient of ∼ 0.6 ± 0.3 and 1.4 ± 0.4 mm yr⁻¹ for the eastern and northern components is observed in geodetic data between Saint Lucia and Saint Vincent and the Grenadines, without an associated pulse of earthquake activity or known active faults. This points to the need for a more complete active fault and structural database of the whole Lesser Antilles arc and forearc.

Similarly, homogeneous earthquake catalogs with improved locations and magnitudes are an important requirement for both instrumental and historical seismicity. The ISCU-cat (Bertil, 2024) empirical magnitude conversion laws have been made by converting from ML or Md of the Trinidad seismological network (TRN) or from the Fort-de-France seismic station data (FDF – IPGP network) to Mw (Bertil et al., 2023). Conversions are calibrated on data recorded between 1986 and 2014. Since 2014, local magnitude (MLv) calculations from IPGP observatories have differed (Massin et al., 2021), and TRN magnitude conversions have to be verified. Thus, the Gutenberg–Richter distributions proposed in Fig. S3 in the Supplement are more qualitative than quantitative. Changes in slope observed between magnitudes 3 and 4 could be a completeness effect or also the result of inappropriate Md-to-Mw conversions. For the Virgin Islands, no Md-to-Mw conversion is available and Md = Mw is used directly (Bertil et al., 2023). These Gutenberg–Richter distributions (Fig. S3 in the Supplement) clearly illustrate that improvements in the magnitude conversions for M<4.0 are still needed. We consequently describe this distribution in three levels of confidence: usable, questionable, and unusable (Table S1 in the Supplement).

One of the major sources of epistemic uncertainty concerns the subduction interface, with its lack of instrumental large-earthquake records and its overall very low geodetic interseismic coupling. We faced a lack of geodetic measurements in this area, which is mainly underwater, despite recent efforts. Worldwide, low interseismic coupling associated with limited large-earthquake activity is observed in the Greece, Calabria, South Sandwich, and Mariana subduction zones (e.g., Vanneste and Larter, 2002; Vernant et al., 2014; Carafa et al., 2018). Yet, focal mechanisms indicate that the Lesser Antilles megathrust can generate moderate M=5–6 earthquakes, at least in its northern half. Coral data are sometimes interpreted as indicative of deep (∼ 40–65 km), high interseismic coupling (Philibosian et al., 2022) or as coherent with geodetic vertical motions with observed subsidence (van Rijsingen et al., 2022). Issues of potential temporal variations in interseismic coupling, as in Mexico or Sumatra (Villafuerte et al., 2021; Philibosian et al., 2022), and the relationship between interseismic coupling and great-earthquake potential (e.g., Kaneko et al., 2010) remain to be addressed. Efforts on seafloor geodesy measurements would help to enhance our understanding of the Lesser Antilles intercoupling behavior of the plate interface, and we highly recommend considering them for future oceanographic missions.

Finally, the uncommon seismicity sources identified in the mantle wedge and the volcanic centers also require dedicated studies before they can be fully integrated into seismic hazard assessments. In both cases, issues such as the earthquake maximum magnitudes and mechanisms or appropriate ground motion attenuation laws demand dedicated global and, if possible, local studies. For the latter, the recognition that shallow, moderate volcano-tectonic earthquakes can constitute a significant source of hazard implies the need for further studies of the influence of eruptive phases on the triggering of volcano-tectonic earthquakes. The 10 km radius around each volcano should be revisited to be more specific to each volcano's characteristics. The creation of a specific volcano-tectonic earthquake catalog from instrumental and historical records is necessary to better encompass their seismic activities. Volcanic seismicity from the two French volcanoes La Soufrière and Mount Pelée can be downloaded from the IPGP observatory servers. ISCU-cat may contain events from volcanic activity, such as the 1950 Mw=4.3 earthquake from the St Kitts and Nevis seismo-volcanic crisis or the 2020 Mw=3.6 earthquake from the St Vincent seismo-volcanic crisis (Didier Bertil, personal communication, 2023; Joseph et al., 2022). Seismic records may not be sufficient to calculate magnitude–frequency distributions and their associated attenuation laws.