How great is the Great Glen Fault? Open Access

The striking geomorphological aspect of the Great Glen of Scotland is its ruler-straight NE–SW trend for 200 km defined by a relatively narrow, steep-sided valley with hundreds of metres of relief. In the middle of the 19th century, geologists identified the Great Glen as the location of an ancient fault line, detailing a normal sense of displacement of at least 1000 m of downthrow on its SE side (as reported by Kennedy 1946). Almost a century later, Kennedy (1946) compiled three lines of evidence to deduce that the fault also had a strike-slip component of 105 km of left-lateral offset (Fig. 2a). His evidence consisted of (1) connecting the Foyers and Strontian granites, (2) linking zones of ‘injection complexes’ (migmatites) and (3) inferring that the Moine Thrust was a linear structure that extended from the Northern Highlands through the Inner Hebrides to Islay. He also concluded that faulting had begun prior to the early Devonian but had ceased by Carboniferous time. Four decades later, Winchester (1973, 1974) constructed a map of metamorphic zones for the Scottish Highlands. The map pattern was complex but could be made more coherent by removing 160 km of left-lateral motion along the Great Glen Fault (Fig. 2b). Thus, two independent datasets had arrived at a similar conclusion: the Great Glen Fault was principally a strike-slip fault having 100–160 km of left-lateral offset.

However, the concept envisaged by Wilson (1962) that the Great Glen Fault was part of a fault system dismembered by continental drift, one that was comparable in magnitude with the San Andreas fault, seemed to be confirmed by palaeomagnetic data (van der Voo and Scotese 1981). Those data suggested that mid-Devonian North America and Europe were separated by at least 15° of latitude, thereby requiring 2000 km of sinistral motion along the Great Glen Fault for them to converge in the Carboniferous and form Pangaea. Later workers used geological reconstructions (Smith and Watson 1983; Soper and Hutton 1984; Soper et al. 1992; Stewart et al. 1999) and geochemical similarity of high-Sr and -Ba granites on opposite sides of the Great Glen Fault (Thirlwall 1989) to refute those data and the match-ups of metamorphic zones and Foyers–Strontian granites (e.g. Pankhurst 1979). This led to the conclusion that the Great Glen Fault recorded only a few hundred kilometres of sinistral offset (Fig. 2c and d). It should be noted that the disparate palaeo-latitudinal positions were not necessarily incorrect. What was incorrect was assigning the Great Glen Fault as Laurentia's plate boundary, now recognized as coincident with the Highland Boundary Fault (e.g. Briden et al. 1984; Chew and Strachan 2014): north of the Highland Boundary Fault is Laurentian crust (Scottish and Irish Highlands), to its south are Iapetus oceanic arcs and accretionary prisms (Scotland's Midland Valley and Southern Uplands, north–central Ireland) and, farther to the south and approximating the boundary between Scotland and England, the Iapetus suture along which are Avalonian crustal fragments (central and southern Ireland, England).

Ideas about the magnitude of offset along the Great Glen Fault changed again when it became apparent that rocks in the Northern Highlands north of the fault contain evidence for the 435–400 Ma Scandian event of the Caledonian Orogeny (e.g. Freeman et al. 1998; Dallmeyer et al. 2001; Kinny et al. 2003; Sherlock et al. 2003) whereas no deformational fabrics of that event have been documented in the Grampian Highlands that lie south of the fault (see synthesis by Chew and Strachan 2014). To explain this, Dewey and Strachan (2003) constructed a conceptual tectonic scheme that resurrected the idea of large-magnitude displacement along the Great Glen Fault (Fig. 2e and f). They postulated that the Northern and Grampian Highlands were separate tectonic blocks and that their juxtaposition resulted from 700 km or more of sinistral movement along the fault during a 40 Myr interval spanning 435–395 Ma. That view has dominated thinking for the past two decades but has been contested by the hypothesis of Searle (2021), within which the offset of the Great Glen Fault is thought to be significantly less, although how much less is not stated explicitly. Thus, the pendulum of displacement estimates for the Great Glen Fault has swung from invoking offsets of 100–160 km, to as much as 2000 km, a subsequent reduction to a few hundred kilometres, followed by reinstating displacements of 700 km (or more) and, most recently, a return to the view that the smaller-offset estimates were more probably correct.

Excepting Caledonian granites and related intrusions, the Scottish and Irish Highlands consist mostly of Meso- to Neoproterozoic (meta)sedimentary successions with lesser (meta)volcanic rocks (Torridonian, Moine and Dalradian supergroups, the first two now reclassified as the Wester Ross and Loch Ness supergroups; see Krabbendam et al. 2022). Establishing age constraints for these successions is difficult and researchers have relied on developing detrital zircon age spectra for ascertaining maximum depositional ages, identifying source areas and as a potential means of correlation (e.g. Rainbird et al. 2001; Cawood et al. 2003, 2007; Banks et al. 2007; Kinnaird et al. 2007; McAteer et al. 2010a, b; Krabbendam et al. 2017). When viewed as a composite (Fig. 3a), each succession shows multimodal peaks with varied amplitudes arrayed across 2 Gyr, and each peak can be confidently tied to Laurentian provenances ranging in age from Mesoarchean to Neoproterozoic.

Our interest in the evolution of the Great Glen Fault was sparked by an observation that certain units are marked by a distinctive detrital zircon age spectrum. That spectrum is dominated by modes of 1.8–1.6 Ga grains with or without a lesser mode of 1.5 Ga grains (Fig. 3b). Such a tight clustering is indicative of derivation from a terrane composed almost exclusively of late Paleoproterozoic rocks with minimal material attributable to older or younger rock units. This dominance of a singular source was noted for rocks of the Glenshirra, Grampian and Iona groups by previous workers, who suggested that they might be correlatives (Banks et al. 2007; McAteer et al. 2010a, b). Here we report, for the first time, that rocks belonging to the Tarskavaig nappe on the Sleat Peninsula of southern Skye also yield the same detrital zircon age spectrum (Fig. 3b; Supplementary material Table S1 Worksheet 1). As highlighted by previous researchers (Marcantonio et al. 1988; Daly et al. 1991, 1995, 2009; Daly 1996; Daly and Flowerdew 2005), the only rocks in the Scottish–Irish Caledonides from which abundant zircons with the appropriate ages can be sourced are the c. 1.8–1.6 Ga Rhinns Complex exposed in four places (Fig. 1): the Inner Hebridean islands of Islay and Colonsay (its northern tip) and, in Ireland, on Inishtrahul off the north coast of Donegal and the Mullet Gneiss of the Annagh Gneiss Complex in County Mayo (Supplementary material Fig. S1). Those rocks are part of a now dismembered Paleoproterozoic Makkovik–Ketilidian orogen of eastern Canada and southern Greenland (Supplementary material Fig. S2); the former also contains c. 1.5 Ga igneous rocks that define the Pinwarian event (Gower et al. 2008; Augland et al. 2015).

Our contribution is to emphasize how stratigraphically and geographically widespread that near-single-temporal-mode distribution is (Fig. 4b), being present in (meta)sedimentary successions from the NW Scottish Highlands to County Mayo, Ireland (Doonamo Formation; McAteer et al. 2010b). We stress that none of the units typified by this distribution have firm depositional age constraints at present: the only time bounds that exist are that those units must be younger than their youngest detrital zircons (typically c. 1.0 Ga) and older than Tonian (c. 1.0–0.7 Ga) or Ordovician–Silurian (c. 0.5–0.4 Ga) orogenic events during which they were metamorphosed and deformed. Hence, we are not implying that the near-single-mode-bearing units are all temporally correlative; rather, they share derivation from a provenance dominated by rocks spanning a restricted age range (mostly the Statherian Period of the Paleoproterozoic, c. 1.8–1.6 Ga).

Despite the lack of age control, what is particularly cogent for providing constraints on assessing original palaeogeographies is that the units dominated by 1.8–1.6 Ga detrital zircons are typified by poor sorting, range in composition from arkosic to feldspathic–quartzitic–lithic wackes and have clasts that are angular to subrounded across all grain sizes, and grain sizes change rapidly (on centimetre to metre scale) from silt–sand mixtures to pebble- and cobble-bearing beds (Figs 5 and 6); most of the units are also several hundred metres thick. Associated sedimentary structures are typically decimetre-thick sets and co-sets of trough and planar cross-bedding, metre-scale multistorey channels and, in places, climbing ripple bed-sets. Palaeocurrent data are broadly unimodal (transport directions mostly towards eastern and northern quadrants) signifying derivation from Laurentian provenances. These facies attributes have led to the conclusion that deposition occurred in alluvial fan-braided fluvial–nearshore marine settings and the textural and compositional immaturity indicates that distance of transport from source to sink was not large (see references and summaries given by Prave et al. 2024; Strachan et al. 2024). Two exceptions are the turbiditic rocks of the Corrieyairick Subgroup (lower Grampian Group) and the paragneiss of the Badenoch Group, in which original textural features have been obliterated by polyphase deformation and migmatization. Those exceptions aside, the facies attributes indicate that at the time of deposition Makkovik–Ketilidian–Rhinnian crust was exposed and must have been relatively nearby. This, however, poses a conundrum: the nearest exposures of that crust are 150–300 km distant in the Inner Hebrides and west of Ireland yet the coarsest facies, poorly sorted coarse-grained arkosic sandstone of the Tarskavaig nappe and Sleat Group (Fig. 5d–f) and pebbly sandstone and matrix-supported cobble conglomerate of the Glenshirra Group (Fig. 6b and c), are furthest from those exposures.

The basement to the rocks of the Northern Highlands is represented by the Archean Lewisian Gneiss Complex but what underlies the Grampian Highlands is less certain. Magmas that are in large part derived from crustal sources, particularly those forming granitoids, can retain distinctive elemental and isotopic compositions that are indicative of the nature of their source rocks (e.g. Chappell 1984; White et al. 2001). Although there are no exposures of Paleoproterozoic Rhinnian rocks on mainland Scotland, the isotopic and geochemical signatures of granites in the Grampian Highlands indicate that crust containing comparable components occurs at depth (Stephens and Halliday 1984; Halliday et al. 1993; Steinhoefel et al. 2008). To assess this further, we sampled the Strontian (Loch Sunart granodiorite, Glen Sanda granite), Foyers and Monadhliath plutons for inherited zircons and found that they do indeed contain 1.8–1.5 Ga inherited zircons (Fig. 3c; also Supplementary material Figs S3–S6 and Tables S1 and S2). This supports the inference based on the geochemical data that underlying the Grampian Highlands and Northern Highlands proximal to the Great Glen Fault there is a mid- to late Proterozoic protolith consistent with Rhinnian crust, and that it was prone to melting during Caledonian orogenesis (c. 435–405 Ma). Although the number of identified inherited cores is small (n = 73), the presence of older zircon cores (c. 2.8–2.7 Ga) in the Glen Sanda granite, younger zircons cores (c. 1.1–0.9 Ga) in all the sampled granites (particularly Monadhliath pluton) and detrital zircon age spectra for the granites (Supplementary material Fig. S4) similar to those of many of the Neoproterozoic metasedimentary units, together indicate variability in the make-up of the granite source rocks beneath the Grampian Highlands, possibly involving more than a single crustal unit. Analysis of additional plutons would be a good test and refinement of these inferences.

Given that every rock type (sedimentary, metasedimentary, igneous) across the Scottish–Irish Highlands is marked by a suite of inherited or detrital zircons derived from Archean to Proterozoic sources that define multimodal age spectra emphasizes that the essentially single-temporal-mode distribution of c. 1.8–1.6 Ga zircons is unique. This fact belies interpretations that it represents derivation from recycled or multicycle sedimentary inheritance processes, which would have contributed zircons of many different ages. Thus, we turned our thoughts to determining a palaeogeography that would satisfy the sedimentological constraints and the geochronological and geochemical data.

Offshore geophysical and coring surveys show that the onshore geological trends and distribution of the Archean Lewisian Complex and Paleoproterozoic Rhinns Complex extend for several hundred kilometres westward across the Scottish and Irish continental shelfs (Fig. 7a): Lewisian rocks underlie the Outer Hebridean shelf and George Bligh Bank whereas the Rockall Bank is underlain by gneissic rocks of the Rhinns Complex (Morton and Taylor 1991; Dickin 1992; Daly et al. 1995; Hitchen et al. 1997; Hitchen 2004; Holdsworth et al. 2019). The clearly defined boundary between those areas leaves little doubt that it is a fundamental crustal feature, being coincident with the present-day Anton Dohrn Transfer Zone (Dickin 1992; Doré et al. 1997; Corfield et al. 1999), that separates areas of differential syn- to post-Cretaceous exhumation across Britain and Ireland (Rateau and Chew 2020). Based on reconstructions of the North Atlantic prior to early Tertiary separation of North America and Europe (Barnett-Moore et al. 2016; Ady and Whittaker 2019; MacMahon et al. 2020), a Laurentian parentage for the Lewisian–Rhinnian crustal blocks is assured whereby the Scottish–Irish Highlands and their continental shelfs join to Archean and Makkovik–Ketilidian crust of eastern Canada and southernmost Greenland (Fig. 7b). There are no exposures or drill-core data to constrain the northeastern extent beyond the Scottish mainland of the postulated Makkovik–Ketilidian–Rhinns crustal fragment (i.e. Laurentian crust). However, geophysical modelling suggests that the Laurentia–Baltica crustal boundary occurs no more than a few tens of kilometres offshore of NE Scotland (Lyngsie and Thybo 2007, fig. 11b) thereby providing an approximate termination of the proposed Makkovik–Ketilidian–Rhinns crustal block (Fig. 7a).

Using the knowledge that Rhinnian crust occurs at depth beneath the Grampian Highlands, the reconstituted tectonic framework for the North Atlantic realm (Fig. 7b) shows a Makkovik–Ketilidian–Rhinns crustal fragment extending for more than 1500 km from eastern Canada to Ireland. However, at the location of the Great Glen Fault, the consistently east–west-trending boundary between Paleoproterozoic and Archean crust is disrupted by an abrupt change in trend and apparent sinistral offset. If the Grampian Highlands block (i.e. Makkovik–Ketilidian–Rhinns crustal block) is shifted SW by 250–300 km along the Great Glen Fault, it slots into a position that compensates that disruption and realigns the linear trend of the crustal boundary separating Archean from Paleoproterozoic rocks (Fig. 7c). That shift also resolves the sedimentological conundrum highlighted above. Marine geophysical surveys indicate that the Great Glen Fault projects between the Irish continental margin and Rockall Bank (Riddihough and Max 1976; Bailey et al. 1977; Max and Barker 1978; Dickin 1992). Using that projection as an offshore guide, placing the Grampian Highlands in the position that realigns the crustal blocks juxtaposes the texturally and compositionally immature sedimentary units dominated by 1.8–1.6 Ga detrital zircons adjacent to Makkovik–Ketilidian–Rhinns crust. This satisfies the lithofacies attributes (e.g. poor sorting, coarse-grained, angular clasts across all grain sizes) that suggest source areas must have been relatively nearby and, based on palaeocurrent data, located to the west and SW.

Our findings that explain the detrital zircon age spectra and sedimentology of the rocks yielding those spectra prompted another curiosity. Our proposed 250–300 km sinistral offset is comparable with the earlier estimates of Kennedy (1946) and Winchester (1973). Although the evidence for those estimates is now refuted, this nevertheless led us to think about the matching of other geological features that might inform on ways to assess or even test the magnitude of displacement along the Great Glen Fault. A wealth of data exists about the timing, geochemistry and composition of Caledonian plutons and related intrusive rocks that pepper the Scottish–Irish Highlands (Pitcher and Berger 1972; Wright and Bowes 1979; Hutton 1982; Halliday and Stephens 1984; Stephens and Halliday 1984; Chappell and Stephens 1988; Canning et al. 1996, 1998; Atherton and Ghani 2002; Ghani and Atherton 2006; Fowler et al. 2008; Neilson et al. 2009; Murphy et al. 2019; Murphy 2020; Archibald et al. 2021, 2022). From such data, Stephens and Halliday (1984) defined three major suites of Scottish granites: the Argyll and Cairngorm suites in the Grampian Highlands and the South of Scotland suite in the Southern Highlands, Midland Valley and Southern Uplands (plutons in the northern Highlands did not readily classify as suites). Of those, the Argyll suite and Donegal plutons share some key geochemical traits (Fig. 8). We are, however, circumspect in the use of such data. The Strontian and Foyers plutons also share Argyll-suite characteristics but differ in whole-rock rare earth element abundances (Pankhurst 1979), which invalidated the conjecture that they were the same pluton displaced by the Great Glen Fault. Nevertheless, it is striking what additional traits are shared between the Argyll suite plutons in the region centred on Glen Coe and the Donegal plutons in Ireland (Fig. 9): both have diverse petrological characteristics commonly with nested zoning patterns; both were emplaced between 430 and 405 Ma (summarized geochronology data were given by Archibald et al. 2022); both are associated with two of the most concentrated and well-developed occurrences anywhere in the Scottish–Irish Caledonides of Silurian appinite bodies and Tonian–Ediacaran metadolerite/gabbro intrusions (this aspect was highlighted as early as the 1960s; e.g. Pitcher and Shackleton 1966); and both intrude garnet-bearing upper greenschist–amphibolite-facies Dalradian rocks (Winchester 1973, 1974; Yardley 1976; Fettes 1979). Given these similarities and shared traits, the question we asked was: is it plausible that the Donegal–Argyll plutons and associated intrusive rocks were once more closely aligned?

The Leannan Fault in Donegal, Ireland (Fig. 1), like the Great Glen Fault, occupies a strikingly narrow, linear valley and has been postulated to be an extension of the left-lateral Great Glen Fault system (e.g. Dowling et al. 1954; Pitcher et al. 1964; Max and Barker 1978). Two observations have been used to deduce that its magnitude of displacement is limited to a few tens of kilometres. The first is based on presumed offsets of facies trends in Argyll Group strata (Pitcher et al. 1964; Alsop and Hutton 1990). This, however, is fraught with difficulties because facies changes in the Argyll Group are abrupt and trends cannot be followed for more than a few kilometres (e.g. Litherland 1980; Anderton 1988). Consequently, the utility of such features as pierce points for the Leannan Fault is equivocal. The second observation is a 30 km offset of the Lough Foyle synform (Fig. 9a), a structure that folds local D₁–D₃ features on the east side of the fault (Alsop 1992). In contrast, on the western side of the fault later structural overprinting and/or higher metamorphic grade, or later structural overprinting along the Main Donegal Granite Shear Zone, has masked those relationships (I. Alsop, pers. comm., 2024). This indicates that the fold and, hence, its offset formed during the later stages of strike-slip movement. Thus, that offset is a minimum, not a maximum, measure of displacement along the Leannan Fault. Consequently, we postulate that the Leannan and Great Glen faults are indeed a linked structure and that the commonalities shared by the Donegal and Argyll-suite plutons support the idea that they were once closer together.

Under that postulate, restoring 250–300 km of left-lateral offset along a combined Leannan–Great Glen fault juxtaposes both plutonic suites in a manner that is compatible with their allied features (Fig. 10): it brings into conjunction two of the largest, geochemically compatible and similar-aged suites of plutonic bodies in the Scottish–Irish Caledonides, it pairs the exceptional concentrations of appinite and metadolerite/gabbro intrusions and it matches similar metamorphic facies in the Dalradian rocks that encase the plutons. This proposed amount of restoration also corresponds to that required for aligning the trend of the boundary between Archean (Lewisian) and Paleoproterozoic (Rhinns–Makkovik–Ketilidian) crustal blocks and placement of compositionally and texturally immature sedimentary units against Rhinnian crust as a source of 1.8–1.6 Ga detrital zircons (Fig. 7c). One aspect that is different between the matched Argyll suite in Glen Coe and the plutons in Donegal is the presence of extensive volcanic rocks in the former and their absence in the latter. A plausible explanation is that the c. 1 km of downthrow on the SE side of the Great Glen Fault relative to its NW side has resulted in preserving extrusive units of the Argyll suite. It is noteworthy that the latest movement documented along the Great Glen Fault has been dextral transtensional and is as young as early Cretaceous (e.g. Tamas et al. 2023, and references therein). The amount of dextral reactivation, though, is limited to several tens of kilometres and, thus, does not severely affect our postulated reconstruction.

We have not included the Walls Boundary Fault in Shetland (Fig. 7a) in this reconstruction. Although often linked to the Great Glen Fault, it records early sinistral ductile fabrics that are overprinted by later brittle dextral movement (Watts et al. 2007). The 440 Ma Graven Complex (Lancaster et al. 2017) plausibly intrudes the earlier ductile fabrics and, if so, the left-lateral offset along the Walls Boundary fault predates the c. 430–400 Ma timing of sinistral motion along the Great Glen Fault. Given these uncertainties, how the Walls Boundary Fault fits within the sinistral Caledonian fault system(s) remains to be determined.

Ideas about the magnitude of displacement along the Great Glen Fault system and its role in the Caledonian Orogeny have oscillated between interpreting it as a sinistral transcurrent structural element in the tectonic framework of the Scottish–Irish Caledonides with displacements of 100–200 km (Kennedy 1946; Winchester 1973) to its role as a major tectonic boundary along which as much as 1000 km or more of displacement occurred (e.g. van der Voo and Scotese 1981). The geological evidence upon which those end-member models were constructed has been subsequently disproven, but debate continues, as evident by the concept of the Great Glen Fault having relatively small displacements being resurrected by Searle (2021) in his revised model of Caledonian orogenesis that contrasts with the 700 km or more of tectonic translation envisaged by Dewey and Strachan (2003).

Our contribution in this debate is based on integrating sedimentology with igneous geochemistry, and detrital and inherited zircon age data. The near-single-mode peak of 1.8–1.6 Ga zircons sourced from Rhinns–Makkovik–Ketilidian crust that characterizes compositionally and texturally immature sandstone-dominated units geographically widespread across the Scottish–Irish Highlands, and its presence as inheritance in the Caledonian granites, requires explanation. The Rhinns–Makkovik–Ketilidian orogen, when reconstructed into a pre-North-Atlantic-rift configuration, defines a 1500 km long linear belt along which the only disruption to that trend coincides with the Great Glen Fault. Restoring 250–300 km of sinistral displacement along a combined Great Glen and Leannan fault system re-establishes the linear trend of the orogen and, in so doing, explains sedimentological features indicative of deposition occurring proximal to source areas composed of Rhinns–Makkovik–Ketilidian crust. It also unites into a single coherent assemblage several otherwise widely separated yet comparable geological features: the compositionally and temporally compatible Donegal–Argyll granitic suites and the concentrated zones of Caledonian peri-batholitic appinite intrusions and Tonian–Ediacaran metadolerite/gabbro bodies.

Our postulated pairing of the Donegal and Argyll suite plutons offers a model that could be tested by additional detailed isotope and REE geochemistry coupled with high-precision and high-accuracy U–Pb geochronology. However, the striking feature that has been highlighted as central in discussions about the Great Glen Fault is the presence of Scandian-age (Silurian) deformational fabrics in the rocks of the Northern Highlands north of the fault and their absence in the rocks of the Grampian Highlands south of the fault. Although the 250–300 km of offset we propose for the Great Glen Fault falls short of the minimum of 400 km postulated by Dewey and Ryan (2022) to explain that relationship, we note that similarly scaled shear zones elsewhere, such as the Puros and Okahandja shear zones in Namibia as well as numerous others formed during the pan-African assembly of Gondwana (Schmitt et al. 2023), juxtapose disparate tectonothermal terranes not unlike the geological differences between the Northern and Grampian Highlands. Thus, given uncertainties in exact geometries of plate boundaries during the Scandian phase of the Caledonian Orogeny, it is possible that the magnitude of offset we suggest could account for the Scandian event in the Northern Highlands and the lack of deformational fabrics related to that event in the Grampian Highlands. In essence, our postulate offers a model of the Great Glen–Leannan fault system that can be tested. To do so requires forming collaborations that will stimulate joint research into Irish and Scottish geology to find shared ways to deepen knowledge about and further understanding of the Caledonian Orogeny.

We thank many colleagues over our many shared years of working in the Scottish and Irish Highlands. We also acknowledge the careful and thorough reviews of A. Bird, P. Cawood and B. Holdsworth that much improved the paper. Special thanks go to R. Strachan for freely sharing his ideas and deep knowledge of Scottish and Irish geology.

ARP: conceptualization (lead), writing – original draft (lead), writing – review & editing (lead); WES: conceptualization (supporting), formal analysis (supporting), writing – original draft (supporting), writing – review & editing (supporting); AEF: conceptualization (supporting), writing – original draft (supporting), writing – review & editing (supporting); IW: formal analysis (supporting), writing – original draft (supporting), writing – review & editing (supporting); KK: formal analysis (supporting), writing – original draft (supporting), writing – review & editing (supporting).

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

All data generated or analysed during this study are included in this published article and its supplementary information files.