3.1. Initial expansion of the last ice sheet

The extent of glacier ice in Scotland prior to expansion of the last ice sheet is unknown, but may be inferred from the flux of ice-rafted detritus (IRD) of British provenance in cores retrieved from the NE Atlantic. Investigation of the magnetic signature of IRD in core MD95-2006 from the distal northern part of the Barra–Donegal Fan (BDF) by Peters et al. (Reference Peters, Walden and Austin2008) indicates very limited contribution from Scotland prior to ∼38.5 ka, but an increased contribution thereafter, suggesting expansion of marine-terminating glaciers nourished in the Highlands. Core MD04-2882 from the Rockall Trough contains evidence of IRD derived from a nascent BIIS after ∼43 ka (Hibbert et al. Reference Hibbert, Austin, Leng and Gatliff2010) and suggests the intermittent presence of marine-terminating glaciers in western Scotland until ∼35 ka and the continuous presence of a tidewater ice margin thereafter. This evidence implies that an ice cap or icefield extended to sea level in the western Highlands several millennia before its expansion to cover low ground and the adjacent shelves.

Radiocarbon ages for organic material buried under till deposited by the SIS (Table 1) indicate that most low ground was ice free before ∼35 cal ¹⁴C ka and possibly as late as ∼32 cal ¹⁴C ka. Faunal and floral assemblages from the sites at Balglass Burn and Sourlie indicate cold tundra conditions and permafrost (Bos et al. Reference Bos, Dickson, Coope and Jardine2004; Brown et al. Reference Brown, Rose, Coope and Lowe2007), consistent with the coeval presence of glaciers in the Highlands. The Balglass site lies only ∼20 km from the Highland edge, yet the mean age of six samples from this site indicates that it was not over-run by Highland ice until after 36.0–34.9 cal ¹⁴C ka, and the youngest age suggests that it may not have been over-run until after ∼32 cal ¹⁴C ka. Of the uppermost four radiocarbon ages obtained by Whittington & Hall (Reference Whittington and Hall2002) for bulk samples from organic-rich sands and silts buried under till at Tolsta Head in eastern Lewis, the youngest (30.8–30.1 cal ¹⁴C ka) probably reflects contamination, as the underlying three samples yielded consistent ages averaging 33.4–32.8 cal ¹⁴C ka; there is also some uncertainty as to whether the overlying till was deposited by Outer Hebrides ice or mainland ice. Collectively, the limiting ages for ice expansion across low ground suggest that during the final millennia of MIS 3 glacier ice expanded from mountain source areas, over-running most low ground after ∼35 ka and possibly two or three millennia later.

Table 1 Radiocarbon ages relating to the expansion of the last Scottish Ice Sheet

Calibrated ages represent ±1σ range. Sources: ¹Brown et al. (Reference Brown, Rose, Coope and Lowe2007); ²Bos et al. (2004); ³Jacobi et al. (Reference Jacobi, Rose, MacLeod and Higham2009); ⁴Whittington & Hall (Reference Whittington and Hall2002); the uppermost age in the Tolsta Head sequence may reflect contamination and is excluded; ⁵FitzPatrick (Reference FitzPatrick1965); this date should be regarded as minimal (Sissons Reference Sissons1981; Hall et al. Reference Hall1995).

The timing of ice expansion across the NSB is broadly consistent with the ages obtained from terrestrial sites. Three radiocarbon ages for shells in marine sediments overlain by till within a core from the North Sea Plateau, ∼330 km NE of Shetland, yielded (in stratigraphic order) ages of 34.6–34.0, 34.5–33.8 and 34.1–33.3 cal ¹⁴C ka, implying that this area remained ice-free until at least ∼34 ka (Rise & Rokoengen Reference Rise and Rokoengen1984; Sejrup et al. Reference Sejrup, Haflidason, Aarseth, King, Forsberg, Long and Rokoengen1994). Similarly, samples of foraminifera and shell fragments in glacially deformed glacimarine sediments in core BGS BH 04/01 from the Witch Ground Basin ∼175 km NE of Rattray Head produced ages of 35.0–33.9, 39.1–30.9, 35.0–29.1 and 35.8–31.3 cal ¹⁴C ka (Graham et al. Reference Graham, Lonergan and Stoker2010). Although three of these ages have large uncertainties, the median ages associated with these samples (34.8–32.2 cal ¹⁴C ka) suggest that the Witch Ground Basin remained largely ice-free until ∼34–33 ka.

3.2. The overall pattern of ice-sheet expansion

Understanding of the evolution of the SIS has been transformed by remote mapping of landforms indicative of the direction of former ice movement (drumlins, streamlined drift ridges, mega-scale glacial lineations (MSGLs), ribbed moraine and meltwater channels) and the grouping of such landforms into discrete flowsets that represent different phases of ice movement. Overprinting of one flowset by another indicates the relative chronology of ice flow trajectories (Fig. 4), though it does not exclude the possibility that the oldest flowsets identified in this way represent pre-Late Devensian ice flow directions, and some supposed flowsets in areas of thin drift cover may be influenced by underlying rock structure (Hall & Riding Reference Hall and Riding2016; Merritt et al. Reference Merritt, Connell and Hall2017). Using this approach, and incorporating data on erratic transport and offshore landforms, Hughes et al. (Reference Hughes, Clark and Jordan2014) developed a first-order reconstruction of the stages of ice sheet expansion (Fig. 5). Stage 1 of their reconstruction envisages congruent ice caps over the Highlands and Outer Hebrides, and approximates conditions around 35–32 ka, when ice enveloped the Midland Valley. Stage 2 depicts the initial build-up of an ice divide over SW Scotland and the extension of Scottish ice deep into the Irish Midlands. By stages 3 and 4, thickening and expansion of the Irish Ice Sheet and Southern Uplands ice dome has resulted in the development of an ice divide between SW Scotland and Ireland; at this time the ice margin is inferred to have approached or reached the Atlantic Shelf edge. Stage 5 depicts the SIS terminating westward at the Atlantic Shelf edge and an eastward confluence with the FIS, forcing ice flowing eastward from Scotland to divert northwestward across Caithness and Orkney and southeastward across NE England and into the southern NSB. This stage approximates the maximum extent of glacier ice sourced in Scotland, which Hughes et al. (Reference Hughes, Clark and Jordan2014) suggested occurred between 28 ka and 26 ka. Southward ice expansion, however, is depicted as continuing during stage 6, by which time they depict the retreat of the northern ice margin, the development of an ice cap over Shetland and the partial decoupling of the SIS and FIS.

Figure 4 Flowsets of the northern and central parts of the last ice sheet in Great Britain as interpreted by Hughes et al. (Reference Hughes, Clark and Jordan2014). Red flowsets are interpreted to be isochronous and those in other colours are inferred to be time-transgressive. Reproduced from Hughes et al. (Reference Hughes, Clark and Jordan2014) Quaternary Science Reviews 89, 148–168, with permission from Elsevier. © 2014 Elsevier Ltd.

Figure 5 Stages in the growth of the last British Ice Sheet inferred by Hughes et al. (Reference Hughes, Clark and Jordan2014) from flowset and offshore evidence. Thick lines represent former ice divides or saddles, and thin lines represent flowlines inferred from flowsets and offshore landform data. Reproduced from Hughes et al. (Reference Hughes, Clark and Jordan2014) Quaternary Science Reviews 89, 148–168, with permission from Elsevier. © 2014 Elsevier Ltd.

Notable features of the pattern of ice-sheet build up proposed by Hughes et al. (Reference Hughes, Clark and Jordan2014) include the persistence of a migratory N–S ice divide extending southward from the NW Highlands, and the development of an ice divide between SW Scotland and NE Ireland, which limited the flow of Scottish ice into the ISB. They acknowledged that the timing of ice-sheet build-up is poorly constrained, and that their reconstruction (Fig. 5) may conflict with more detailed evidence in particular sectors. It nevertheless shows reasonable congruence with numerical models of the pattern of ice-sheet build-up (Boulton & Hagdorn Reference Boulton and Hagdorn2006; Hubbard et al. Reference Hubbard, Bradwell, Golledge, Hall, Patton, Sugden, Cooper and Stoker2009; Patton et al. Reference Patton, Hubbard, Andreassen, Winsbarrow and Stroeven2016) and represents a first approximation that forms a basis for future refinement.

3.3. The southern sector of the ice sheet

The evolution of the southern sector of the SIS has been reconstructed through identification of cross-cutting and overprinted flowsets that indicate migrating ice sheds and major changes in ice movement during ice-sheet expansion. This approach was pioneered in this sector by Salt & Evans (Reference Salt and Evans2004) and subsequently developed by Finlayson et al. (Reference Finlayson, Merritt, Browne, Merritt, McMillan and Whitbread2010, Reference Finlayson, Fabel, Bradwell and Sugden2014) for west-central Scotland, and Evans et al. (Reference Evans, Livingstone, Vieli and Ó Cofaigh2009) and Livingstone et al. (Reference Livingstone, Ó Cofaigh and Evans2008, Reference Livingstone, Ó Cofaigh and Evans2010b, Reference Livingstone, Evans, Ó Cofaigh, Davies, Merritt, Huddart, Mitchell, Roberts and Yorke2012, Reference Livingstone, Roberts, Davies, Evans, Ó Cofaigh and Gheorghiu2015) for northern England and southern Scotland. Palaeoflow directions reconstructed from landforms are complemented by evidence provided by striae, erratic transport and lithostratigraphy. The methodology employed is essentially similar to that employed by Hughes et al. (Reference Hughes, Clark and Jordan2014; Fig. 4), and it is noteworthy that the flowsets derived independently by three groups of researchers are broadly similar.

Finlayson et al. (Reference Finlayson, Merritt, Browne, Merritt, McMillan and Whitbread2010, Reference Finlayson, Fabel, Bradwell and Sugden2014) showed that initial expansion (after ∼35 ka) of ice nourished in the SW Highlands resulted in ice movement eastward across the Midland Valley and southward down the Firth of Clyde, eventually meeting the expanding Southern Uplands ice cap in south Ayrshire (Fig. 6a). Flowset evidence indicates that during this stage Scottish ice may have penetrated up to 200 km southwestward into the Irish Midlands (Greenwood & Clark Reference Greenwood and Clark2009), and the presence of Ailsa Craig microgranite erratics along the coasts of Wales and eastern Ireland (Sissons Reference Sissons1967; McCabe & Ó Cofaigh Reference McCabe and Ó Cofaigh1995) suggests that ice from the SW Highlands reached the ISB. Subsequent expansion and thickening of Southern Uplands ice, however, led to its confluence with the Irish Ice Sheet and the formation of a persistent but migratory ice divide across the North Channel between NE Ireland and SW Scotland, so that ice N of the divide was rerouted westward across the Malin Shelf (Finlayson et al. Reference Finlayson, Fabel, Bradwell and Sugden2014) and only ice nourished in the Galloway Hills flowed south into the ISB. A N–S-trending ice divide also developed across the Firth of Clyde from the SW Highlands to the western Southern Uplands, feeding ice movement both westward across the Malin Shelf and eastward across the Midland Valley (Fig. 6b). The flowset evidence mapped by Finlayson et al. (Reference Finlayson, Merritt, Browne, Merritt, McMillan and Whitbread2010) suggests that this divide subsequently migrated ∼60 km eastward (Fig. 6c), possibly in response to drawdown of ice streaming westward across the Malin Shelf.

Figure 6 Stages in the evolution of the last ice sheet in west central Scotland identified by Finlayson et al. (Reference Finlayson, Merritt, Browne, Merritt, McMillan and Whitbread2010): (a) advance of ice from the SW Highlands across the Firth of Clyde, Ayrshire and the Midland Valley; (b) establishment of an ice divide between the SW Highlands and western Southern Uplands results in eastward flow of ice across the Midland Valley (c) eastward migration of the ice divide results in dominantly westward ice movement, feeding the Hebrides Ice Stream. Reproduced from Finlayson et al. (Reference Finlayson, Merritt, Browne, Merritt, McMillan and Whitbread2010) Quaternary Science Reviews 29, 969–988, with permission from Elsevier. © 2009 NERC.

Using a similar approach, Livingstone et al. (Reference Livingstone, Ó Cofaigh and Evans2008, Reference Livingstone, Ó Cofaigh and Evans2010b, Reference Livingstone, Evans, Ó Cofaigh, Davies, Merritt, Huddart, Mitchell, Roberts and Yorke2012, Reference Livingstone, Roberts, Davies, Evans, Ó Cofaigh and Gheorghiu2015) and Evans et al. (Reference Evans, Livingstone, Vieli and Ó Cofaigh2009) reconstructed the evolution of the last ice sheet in the Southern Uplands and northern England. Ice from the Galloway Hills and Southern Uplands initially predominated, carrying erratics into the Eden Valley and possibly crossing the Stainmore Gap through the Pennines. Subsequent development of a NW–SE ice divide across the Solway Firth from SW Scotland to the Lake District (Fig. 7) represents the continuation of that identified by Finlayson et al. (Reference Finlayson, Merritt, Browne, Merritt, McMillan and Whitbread2010; Fig. 6b) farther north. This divide separated southward ice flow from Galloway into the ISB and eastward flow of ice from the Southern Uplands and Lake District into the Tyne valley. During the LLGM, high ground in the Southern Uplands and the Cheviots appears to have been occupied by cold-based ice domes (Mitchell Reference Mitchell2007, Reference Mitchell2008) that fed ice flowing eastward across the Midland Valley and southwards into the Tweed valley (Everest et al. Reference Everest, Bradwell and Golledge2005). Ice flowing eastward through the Midland Valley to the Firth of Forth was diverted SE near the coast, joining that from the Tweed and the Tyne and extending southwards to NE Yorkshire (Fig. 7). Livingstone et al. (Reference Livingstone, Evans, Ó Cofaigh, Davies, Merritt, Huddart, Mitchell, Roberts and Yorke2012) suggested that this general pattern of ice movement persisted for several millennia. Southward-flowing ice from Galloway continued to flow into the ISB, joining ice from the Lake District, Wales and Ireland to form the Irish Sea Ice Stream (ISIS). Scourse & Furze (Reference Scourse and Furze2001) placed the southernmost extent of the ISIS at a transition from subglacial till to glacimarine facies at ∼49.5°S, but Praeg et al. (Reference Praeg, McCarron, Dove, Ó Cofaigh, Scott, Monteys, Facchin, Romeo and Coxon2015) have presented lithostratigraphic and geophysical evidence suggesting that the ISIS extended >150 km farther south, potentially reaching the Celtic Sea shelf edge as a result of short-lived surging. Bayesian modelling based on published age data suggests that the ISIS achieved its maximum southern reach between ∼24.3 and 23.0 ka (McCarroll et al. Reference McCarroll, Stone, Ballantyne, Scourse, Hiemstra, Evans and Fifield2010; Chiverrell et al. Reference Chiverrell, Thrasher, Thomas, Lang, Scourse, McCarroll, Clark, Ó Cofaigh, Evans and Ballantyne2013), but TCN and luminescence dating evidence implies that it impinged on the Scilly Isles at ∼26–25 ka (Smedley et al. Reference Smedley, Chiverrell, Ballantyne, Burke, Clark, Duller, Fabel, McCarroll, Scourse, Small and Thomas2017a).

Figure 7 Ice flow in southern Scotland and northern England during the maximum expansion of the BIIS, as depicted by Livingstone et al. (Reference Livingstone, Evans, Ó Cofaigh, Davies, Merritt, Huddart, Mitchell, Roberts and Yorke2012). Thick dash-dotted lines represent ice divides, arrows indicate ice-flow vectors and dashed arrows in red indicate possible alternative ice-flow vectors. The inset shows location of this sector within the BIIS as depicted by Clark et al. (Reference Clark, Hughes, Greenwood, Jordan and Sejrup2012). Reproduced from Livingstone, S. J. et al. (Reference Livingstone, Evans, Ó Cofaigh, Davies, Merritt, Huddart, Mitchell, Roberts and Yorke2012) Earth-Science Reviews 111, 25–55, with permission from Elsevier. © 2011 Elsevier B.V.

3.4. The northeastern and eastern sector of the ice sheet

Although it is generally accepted that the SIS was confluent with the FIS in the NSB, there remains uncertainty regarding the pattern and sequence of ice movement during both ice-sheet expansion and retreat. Here we consider first the terrestrial evidence for key areas (Shetland, Orkney, Caithness and NE Scotland), then the changing interpretations of the pattern of ice movement in the NSB during the LLGM.

3.4.2. Caithness and Orkney

The last glaciation of Caithness has traditionally been interpreted as a two-stage event in which initial ice movement northeastward from Sutherland was succeeded by N to NW movement of Moray Firth ice, which deposited an extensive grey till containing marine shells (Peach & Horne Reference Peach and Horne1881; Hall & Whittington Reference Hall and Jarvis1989). Lithostratigraphic research by Hall et al. (Reference Hall, Auton, Michie, McL Pearson and Riding2011) and Hall & Riding (Reference Hall and Riding2016) indicates a more complex event stratigraphy in which the radial outflow of ice from ice centred over the hills of the Sutherland–Caithness border was succeeded in turn by (1) the retreat of the ice margin from the north coast; (2) an initial advance of Moray Firth ice northwestward across the Caithness plain; and (3) the retreat of the ice margin, represented by deposition of fan gravels at the north coast. This was followed by the main northwestward advance of Moray Firth ice across Caithness, which deflected the flow of Highland ice northwards and deposited the widespread shelly till (the Forse Till). Excluding outliers, six TCN ages obtained by Phillips et al. (Reference Phillips, Hall, Ballantyne, Binnie, Kubik and Freeman2008) for the deglaciation of sites in northern and SE Caithness range from 20.5±3.2 ka to 16.0±1.1 ka, indicating that the main advance of Moray Firth ice across Caithness occurred during the LLGM.

On Orkney, the evidence provided by striae, glacial lineations, erratic transport and till containing marine shell fragments and palynomorphs demonstrates a general ESE to WNW movement of glacier ice across the archipelago (Peach & Horne Reference Peach and Horne1880; Sutherland Reference Sutherland1984; Hall et al. Reference Hall, Riding and Brown2016b). Eight TCN ages for low ground have produced post-LGM ¹⁰Be exposure ages (Phillips et al. Reference Phillips, Hall, Ballantyne, Binnie, Kubik and Freeman2008), confirming that the last movement of glacier ice across the archipelago occurred during the Late Devensian; a single (recalibrated) TCN age of 20.6±1.2 ka for a boulder at 467 m altitude on Ward Hill indicates the complete over-running of all high ground. Unlike Shetland, there appears to be no evidence for a former independent ice cap on Orkney (Hall et al. Reference Hall, Riding and Brown2016b).

Hall et al. (Reference Hall, Riding and Brown2016b) identified three tills on Orkney, and related these to three phases of ice movement. The earliest till appears to record ice movement from a more southerly direction than the overlying widespread Scara Taing Till, the stratigraphic equivalent of the Forse Till of Caithness. This till was emplaced by ice moving SE–NW, and in northern Orkney contains rare far-travelled erratics that can be linked to sources in the Northern Highlands, the Inner Moray Firth, the Grampian Highlands, Fennoscandia and possibly eastern Scotland, implying a complex history of transport and reworking. The uppermost till was apparently deposited by a late readvance of ice across low ground. Noting that the palynomorph assemblages in the Orkney tills are dominated by material sourced from the inner Moray Firth and the shelf SE of Orkney, Hall et al. (Reference Hall, Riding and Brown2016b) argued that the rare Scandinavian erratics detected in Orkney till are reworked, and that the last FIS did not impinge on these islands.

Although Hall & Riding (Reference Hall and Riding2016) and Hall et al. (Reference Hall, Riding and Brown2016b) acknowledged that expansion of the FIS and Shetland Ice Cap could have been responsible for the deflection of Moray Firth ice NW across Caithness and Orkney, they advocated other scenarios for such deflection, such as eastwards shift of the former ice shed or, somewhat confusingly, “… ice congestion in the northern North Sea”. An alternative solution is that a SW–NE aligned ice divide (Sejrup et al. Reference Sejrup, Clark and Hjelstuen2016) or migrating ice divide (Merritt et al. Reference Merritt, Connell and Hall2017) developed between NE Scotland and SW Norway during the LLGM (Fig. 8) forcing ice from the Moray Firth to flow NW toward the Atlantic Shelf edge.

Figure 8 Sequential stages in the evolution of the eastern sector of the SIS as depicted by Merritt et al. (Reference Merritt, Connell and Hall2017): (a) southeastward diversion of Moray Firth ice from an ice divide linking Caithness and the Shetland Ice Cap; (b) northwestward re-routing of Moray Firth ice by a receding ice shed linking Buchan and Shetland (position 1) that subsequently retreated to position 2. Dashed lines represent flow directions. Thick lines with open diamonds denote ice divides; those with filled ticks denote a receding ice shed. The red hexagon represents the position of the Witch Ground Basin. From Merritt et al. (Reference Merritt, Connell and Hall2017) Journal of Quaternary Science 32, 276–294. © 2016 John Wiley & Sons, Ltd.

3.4.4. The North Sea Basin (NSB)

Confirmation that the central and northern NSB were completely covered by conjoined ice sheets during the LLGM comes from several sources. Radiocarbon ages of molluscs and foraminifera within undisturbed marine or glacimarine sediments overlying glacially-deformed marine sediments or till range from ∼26.5 to ∼17.2 cal ¹⁴C ka (Sejrup et al. Reference Sejrup, Haflidason, Aarseth, King, Forsberg, Long and Rokoengen1994, Reference Sejrup, Nygård, Hall and Haflidason2009, Reference Sejrup, Hjelstuen, Nygård, Haflidason and Mardal2015; Graham et al. Reference Graham, Lonergan and Stoker2010). Though the oldest ages may be stratigraphically unrepresentative (Sejrup et al. Reference Sejrup, Clark and Hjelstuen2016), these dates confirm that the underlying sediments were deformed or emplaced by Late Devensian (Late Weichselian) ice, and the younger ages indicate marine sedimentation following ice-sheet retreat prior to ∼18–17 ka. Microfabric analyses of key formations in the NSB indicate complete ice cover during the LLGM (Carr et al. Reference Carr, Holmes, van der Meer and Rose2006) and mapping of moraines and subglacial tunnel valleys from bathymetric and seismostratigraphic data (Lonergan et al. Reference Lonergan, Maidment and Collier2006; Bradwell et al. Reference Bradwell, Stoker, Golledge, Wilson, Merritt, Long, Everest, Hestvik, Stevenson, Hubbard, Finlayson and Mathers2008b; Clark et al. Reference Clark, Hughes, Greenwood, Jordan and Sejrup2012) also indicates complete ice cover across the northern NSB during the LLGM. Finally, research by Graham et al. (Reference Graham, Lonergan and Stoker2007, Reference Graham, Lonergan and Stoker2010) based on 3D reflection seismic data covering the Witch Ground Basin (Fig. 8) has revealed NW-trending buried mega-scale glacial lineations, 5–20 km long, interpreted as demonstrating the confluence of the SIS and FIS, and the development of a former ice stream that flowed NW towards the Orkney–Shetland channel or SE towards the central NSB. The timing of ice build-up in the NSB remains uncertain. Analysis of the available chronological data by Hughes et al. (Reference Hughes, Gyllencreutz, Lohne, Mangerud and Svendsen2016) suggested that parts of the NSB may have been ice free as late as 29–28 ka, but that the central and northern NSB were completely ice-covered by 27 ka.

There are conflicting views regarding the pattern of ice movement in the NSB during the LLGM. The traditional view was that the expanding FIS acted as a barrier that forced eastward moving ice from NE Scotland to turn northwestward across Caithness and Orkney, and ice from SE Scotland to turn southeastward then southward toward the English coast. A refinement of this interpretation is that the NW-trending MSGLs in the Witch Ground Basin identified by Graham et al. (Reference Graham, Lonergan and Stoker2007, Reference Graham, Lonergan and Stoker2010) represent a confluence zone between the FIS and SIS that extended NW across northern Orkney, implying that Shetland was over-run by ice from Norway (Bradwell et al. Reference Bradwell, Stoker, Golledge, Wilson, Merritt, Long, Everest, Hestvik, Stevenson, Hubbard, Finlayson and Mathers2008b; Chiverrell & Thomas Reference Chiverrell and Thomas2010; Hughes et al. Reference Hughes, Clark and Jordan2014; Fig. 9). Clark et al. (Reference Clark, Hughes, Greenwood, Jordan and Sejrup2012), however, depicted a broad ice divide extending from NE Scotland to SW Norway, an interpretation consistent with modelling experiments (Boulton & Hagdorn Reference Boulton and Hagdorn2006; Patton et al. Reference Patton, Hubbard, Andreassen, Winsbarrow and Stroeven2016). This view was developed by Sejrup et al. (Reference Sejrup, Clark and Hjelstuen2016), who argued that the MSGLs in the Witch Ground Basin were produced by ice streaming to the SE rather than NW. The migratory ice shed proposed by Merritt et al. (Reference Merritt, Connell and Hall2017; Fig. 8b) represents a variant of this model consistent with switching ice-flow directions on land. The concept of a SE-migrating divide spanning the northern NSB appears to accommodate the available terrestrial and offshore evidence better than the ‘confluence' model (Fig. 9), but remains to be confirmed.

Figure 9 Confluence of the SIS and FIS in the NSB during the LLGM as depicted by Bradwell et al. (Reference Bradwell, Stoker, Golledge, Wilson, Merritt, Long, Everest, Hestvik, Stevenson, Hubbard, Finlayson and Mathers2008b), with hypothesised flow lines. Dark shading depicts the inferred confluence zone (CZ) and arrows show the orientations of mega-scale glacial lineations. Hatching represents the approximate areas occupied by shelf-edge fans. Also depicted is diversion of ice from western Scotland into the Minch and Hebrides Ice Streams by the Outer Hebrides Ice Cap. From Bradwell et al. (Reference Bradwell, Stoker, Golledge, Wilson, Merritt, Long, Everest, Hestvik, Stevenson, Hubbard, Finlayson and Mathers2008) Earth-Science Reviews 88, 207–226. © 2008 NERC. Reproduced with permission from Elsevier B.V.

3.5. Western Scotland, the Hebrides and the Atlantic Shelf

The evidence provided by striae, erratic distribution and blockfields indicates that during ice-sheet build-up, the islands of Skye, Mull and Arran developed cold-based mountain ice caps that diverted the flow of mainland ice and probably remained centres of ice dispersal through the lifetime of the SIS, though all other islands in the Inner Hebrides were ultimately over-run by ice flowing westward from the Scottish Mainland (Bailey et al. Reference Bailey, Clough, Wright, Richey and Wilson1924; Tyrrell Reference Tyrrell1928; Gemmell Reference Gemmell1973; Dahl et al. Reference Dahl, Ballantyne, McCarroll and Nesje1996; Ballantyne & McCarroll Reference Ballantyne, McCarroll, Nesje and Dahl1997; Ballantyne Reference Ballantyne1999; Finlayson et al. Reference Finlayson, Fabel, Bradwell and Sugden2014; Ballantyne et al. Reference Ballantyne, Benn, Bradwell, Small, Ballantyne and Lowe2016). A larger independent ice cap formed over the Outer Hebrides. Mapping of striae and ice-moulded bedrock by Flinn (Reference Flinn1978b), Von Weymarn (Reference Von Weymarn1979) and Peacock (Reference Peacock1984, Reference Peacock, Ehlers, Gibbard and Rose1991) demonstrated former radial ice movement on Lewis and Harris, though occurrences of shelly tills and mainland erratics indicate that mainland ice impinged on eastern Lewis at some stage, and probably crossed northernmost Lewis (Peacock Reference Peacock1984). Striae, ice moulding and erratic transport imply that a N–S-aligned ice divide developed along the western margin of the Uists and Benbecula (Coward Reference Coward1977; Flinn Reference Flinn1978b; Peacock & Ross Reference Peacock and Ross1978; Selby Reference Selby1989). Preservation of blockfields and tors on mountain summits on Harris and South Uist suggests that these were occupied by cold-based ice throughout much or all of the Late Devensian (Ballantyne & McCarroll Reference Ballantyne and McCarroll1995; Ballantyne & Hallam Reference Ballantyne and Hallam2001). The Outer Hebrides Ice Cap appears to have acted as a centre of ice dispersal throughout the evolution of the SIS, feeding ice W across the Hebrides Shelf (Selby Reference Selby1989; Stoker et al. Reference Stoker, Hitchen and Graham1993; Ballantyne et al. Reference Ballantyne, Ó Cofaigh, Coxon, McCarron and Mitchell2017), NE into the Minch Ice Stream (Bradwell et al. Reference Bradwell, Stoker and Larter2007), and southward to feed the Hebrides Ice Stream (Howe et al. Reference Howe, Dove, Bradwell and Gafeira2012; Dove et al. Reference Dove, Arosio, Finlayson, Bradwell and Howe2015; Fig. 9).

The evidence for extension of the SIS to the Atlantic Shelf edge has been inferred from the stratigraphy of large submarine shelf-edge fans (trough-mouth fans), and the distribution and alignment of moraine banks at or near the shelf edge. Four fans abut the shelf edge: the Barra–Donegal Fan (BDF; ∼6300 km²); the Sula Sgeir Fan (SSF; ∼3750 km²); and the smaller Rona Wedge and Foula Wedge (Fig. 9). Deposition of glacigenic sediment on the SSF has been linked to a former ice stream, the Minch Ice Stream (Stoker & Bradwell Reference Stoker and Bradwell2005; Bradwell et al. Reference Bradwell, Stoker and Larter2007; Bradwell & Stoker Reference Bradwell and Stoker2015b) and the BDF acted as depocentre for sediment deposited by converging ice from western Scotland (the Hebrides Ice Stream) and Ireland (Dunlop et al. Reference Dunlop, Shannon, McCabe, Quinn and Doyle2010; Howe et al. Reference Howe, Dove, Bradwell and Gafeira2012; Ó Cofaigh et al. Reference Ó Cofaigh, Dunlop and Benetti2012; Finlayson et al. Reference Finlayson, Fabel, Bradwell and Sugden2014; Dove et al. Reference Dove, Arosio, Finlayson, Bradwell and Howe2015). The link between the Rona and Foula Wedges, submarine troughs and former ice streams is less clear, though the latter may have acted as a depocentre for ice flowing westward through the Orkney–Shetland channel (Fig. 9).

The advance of the ice margin toward the edge of the Malin Shelf is indicated by the lithostratigraphy and ice-rafted debris (IRD) flux recorded in core MD95–2006, a 30 m-long sediment core recovered from the distal northern part of the BDF. At 21.5 m depth within this core there is an abrupt change from hemipelagite and muddy contourite to glacimarine mud with sandy turbidites, which has been interpreted as indicating the proximity of the ice margin (Kroon et al. Reference Kroon, Shimmield, Austin, Derrick, Knutz and Shimmield2000; Knutz et al. Reference Knutz, Austin and Jones2001). This change is accompanied by a marked increase in basaltic IRD derived from the Cenozoic igneous provinces of western Scotland and NE Ireland. Radiocarbon dating of polar and subpolar foraminifera recovered from this core suggests that the onset of ice-proximal glacimarine conditions occurred shortly after ∼30 cal¹⁴ C ka, at the MIS 3/2 transition, though the precise timing of maximum ice extent is difficult to define (Wilson et al. Reference Wilson, Austin and Jansen2002; Bradwell et al. Reference Bradwell, Stoker, Golledge, Wilson, Merritt, Long, Everest, Hestvik, Stevenson, Hubbard, Finlayson and Mathers2008b). Based on this record, and associated geophysical data (Knutz et al. Reference Knutz, Jones, Austin and van Weering2002), it has been inferred that ice feeding the BDF reached its maximum extent at ∼27 ka, generating turbiditic flows on the fan (Wilson & Austin Reference Wilson and Austin2002; Scourse et al. Reference Scourse, Haapaniemi, Colmenero-Hidalgo, Peck, Hall, Austin, Knutz and Zahn2009). This proposition may be supported by a marked increase in IRD flux in deep-ocean core MD04-2822, from a location distal to the BDF, at ∼27.4 ka, though this could also represent partial ice-margin collapse or surging behaviour (Hibbert et al. Reference Hibbert, Austin, Leng and Gatliff2010). The SSF also comprises marine or glacimarine muds alternating with packages of glacigenic mass-flow deposits and sandy turbidites, the latter indicating the proximity of the former ice-sheet margin (Stoker & Holmes Reference Stoker and Holmes1991; Stoker Reference Stoker1995; Baltzer et al. Reference Baltzer, Holmes and Evans1998), and sediment sequences on the Rona and Foula Wedges also demonstrate ice-proximal deposition, probably during MIS 2 (Stoker et al. Reference Stoker, Hitchen and Graham1993, Reference Stoker, Leslie, Scott, Briden, Hine, Harland, Wilkinson, Evans and Ardus1994). Arrival of the ice margin at or near the West Shetland Shelf break may be indicated by increased IRD flux at ∼29 ka in deep-ocean core MD04-2829 from the Rosemary Bank (58.95°N, 09.57°W; Scourse et al. Reference Scourse, Haapaniemi, Colmenero-Hidalgo, Peck, Hall, Austin, Knutz and Zahn2009). Collectively, the available dating evidence appears to place the arrival of the Hebrides and Minch ice streams on the outer shelf within the period 30–27 ka. The latter date has been assumed by several authors, but it seems possible that these ice streams may have reached the outer shelf edge 2–3 ka earlier, depending on the interpretation of the lithostratigraphy and IRD flux in deep-ocean cores.

In this context, ¹⁰Be exposure ages obtained for samples from eight glacially-transported boulders on the island of North Rona on the Shetland shelf (59.1°N, 05.8°W) are interesting. These indicate the timing of retreat of the ice margin from the outer shelf, and yielded a reported mean age of 24.8±2.7 ka (Everest et al. Reference Everest, Bradwell, Stoker and Dewey2013). Recalibration produces individual ages of 34.0 ± 2.0 ka to 23.4±1.4 ka; excluding three outliers, the weighted mean age of the remaining five samples is 28.7±1.4 ka. This appears to imply both the extension of the ice margin in this sector and its subsequent retreat before ∼29–28 ka. This conclusion should be treated with caution, however, because of the large dating uncertainty, and in the light of research suggesting that ¹⁰Be exposure ages for sites at the periphery of former ice sheets may be compromised by subsurface production of ¹⁰Be by deeply-penetrating muons, yielding exposure ages that may be ‘too old', even when they exhibit good internal agreement (Briner et al. Reference Briner, Goehring, Mangerud and Svendsen2016; Smedley et al. Reference Smedley, Scourse, Small, Hiemstra, Duller, Bateman, Burke, Chiverrell, Clark, Davies, Fabel, Gheorghiu, McCarroll, Medialdia and Xu2017b).

Submarine moraine banks recording former ice margin positions on the Atlantic Shelf were first mapped by Selby (Reference Selby1989), Stoker & Holmes (Reference Stoker and Holmes1991) and Stoker et al. (Reference Stoker, Hitchen and Graham1993). Identification of the pattern of moraine banks on the Hebrides Shelf, the West Shetland Shelf and the northern NSB has been transformed by mapping based on the Olex bathymetric database (www.olex.no), which has been used to generate images of seabed relief (Fig. 10). Ridges evident from Olex-based imagery have been mapped by Bradwell et al. (Reference Bradwell, Stoker, Golledge, Wilson, Merritt, Long, Everest, Hestvik, Stevenson, Hubbard, Finlayson and Mathers2008b) and Clark et al. (Reference Clark, Hughes, Greenwood, Jordan and Sejrup2012), both of whom interpreted these features as ice-marginal moraines (or moraine banks) deposited by grounded ice. These ridges often exceed 1 km in width, suggesting that they represent the prolonged deposition and glacitectonic deformation of sediment at oscillating grounding-line margins. The maps produced by Bradwell et al. (Reference Bradwell, Stoker, Golledge, Wilson, Merritt, Long, Everest, Hestvik, Stevenson, Hubbard, Finlayson and Mathers2008b; Fig. 11) and Clark et al. (Reference Clark, Hughes, Greenwood, Jordan and Sejrup2012) differ only in detail.

Figure 10 Merged onshore-offshore (topographic-bathymetric) surface model depicting the relief of Scotland north of the Southern Uplands and the adjacent continental shelf. Offshore data are derived from the Olex database (www.olex.no) and onshore relief from the NEXTMap Britain digital surface model (Intermap Technologies). Reproduced from Bradwell et al. (Reference Bradwell, Stoker, Golledge, Wilson, Merritt, Long, Everest, Hestvik, Stevenson, Hubbard, Finlayson and Mathers2008b) Earth-Science Reviews 88, 207–226. © 2008 NERC. Reproduced with permission from Elsevier B.V.

The outermost ridges (‘Group 1 ridges' in Fig. 11) are broadly arcuate features at or near the shelf edge. These are typically 2–10 km wide and up to 60 km long, and were interpreted by both Bradwell et al. (Reference Bradwell, Stoker, Golledge, Wilson, Merritt, Long, Everest, Hestvik, Stevenson, Hubbard, Finlayson and Mathers2008b) and Clark et al. (Reference Clark, Hughes, Greenwood, Jordan and Sejrup2012) as demonstrating the extension of the SIS to the shelf break (Fig. 9). Farther south, the Olex database is incomplete, but using the multibeam swath bathymetric data of the Irish National Seabed Survey, Dunlop et al. (Reference Dunlop, Shannon, McCabe, Quinn and Doyle2010) mapped moraines deposited by westward-moving ice from Scotland on the Malin Shelf east of the BDF between 55.5°N and 56.3°N. Here, the outermost ridge runs approximately parallel to the shelf edge, which exhibits furrows attributed to iceberg scour, suggesting that the grounded ice margin became marine-based as a result of advance into deep water, or as a consequence of sea-level rise.

Figure 11 Seafloor landforms on the Atlantic shelf and northern North Sea Basin mapped by Bradwell et al. (Reference Bradwell, Stoker, Golledge, Wilson, Merritt, Long, Everest, Hestvik, Stevenson, Hubbard, Finlayson and Mathers2008b) from the bathymetric data in Figure 10. Solid lines: ridges (moraines or moraine banks). Dashed lines: channels, interpreted as tunnel valleys excavated by subglacial meltwater. Reproduced from Bradwell et al. (Reference Bradwell, Stoker, Golledge, Wilson, Merritt, Long, Everest, Hestvik, Stevenson, Hubbard, Finlayson and Mathers2008b) Earth-Science Reviews 88, 207–226. © 2008 NERC. Reproduced with permission from Elsevier B.V.

Although similar shelf-edge moraines W of Ireland have been shown to represent the limit of the last BIIS (Peters et al. Reference Peters, Benetti, Dunlop and Ó Cofaigh2015; Ballantyne & Ó Cofaigh Reference Ballantyne, Ó Cofaigh, Coxon, McCarron and Mitchell2017), there is persuasive evidence that those on the edge of the northern Hebrides Shelf were deposited by an earlier, pre-MIS3/2 ice sheet (Stoker et al. Reference Stoker, Hitchen and Graham1993, Reference Stoker, Leslie, Scott, Briden, Hine, Harland, Wilkinson, Evans and Ardus1994; Stoker Reference Stoker1995, Reference Stoker, Hitchen, Johnson and Gatliff2013). Seismostratigraphic research by Stoker & Holmes (Reference Stoker and Holmes1991) demonstrated that the shelf-edge moraines in this sector pre-date similar moraines on the West Shetland Shelf, being separated by an angular stratigraphic discordance. Though the time interval represented by this discordance is unknown, they postulated that that the two sets of moraines were deposited during two different glaciations. This conclusion has been supported by amino-acid diagenesis, which indicated that the moraines on the outer Hebrides Shelf pre-date MIS 3 (Stoker & Holmes Reference Stoker and Holmes1991; Stoker Reference Stoker, Hitchen, Johnson and Gatliff2013; Stoker & Bradwell Reference Stoker and Bradwell2005). Recent reappraisal of the pattern of moraine systems on the northern and central Hebrides Shelf by Bradwell & Stoker (Reference Bradwell and Stoker2015a) conforms to the stratigraphic evidence, and suggests that ice moving westward from the Outer Hebrides during the LLGM extended no farther than mid-shelf (Fig. 12). This interpretation is supported by research on the St Kilda archipelago, 65 km W of the Outer Hebrides and 40–60 km E of the shelf break, which has demonstrated that the last ice sheet failed to encroach on these islands (Sutherland et al. Reference Sutherland1984; Hiemstra et al. Reference Hiemstra, Shakesby and Vieli2015; Ballantyne et al. Reference Ballantyne, Fabel, Gheorghiu, Rodés, Shanks and Xu2017). Collectively, the evidence outlined above conflicts with the simple ‘shelf-edge' configuration of the last ice sheet promoted by Bradwell et al. (Reference Bradwell, Stoker, Golledge, Wilson, Merritt, Long, Everest, Hestvik, Stevenson, Hubbard, Finlayson and Mathers2008b) and Clark et al. (Reference Clark, Hughes, Greenwood, Jordan and Sejrup2012). The actual extent of the last ice sheet on the northern Hebrides shelf remains uncertain. It has been depicted as a broad arc extending west of the Outer Hebrides but terminating just east of St Kilda (Selby Reference Selby1989; Stoker et al. Reference Stoker, Hitchen and Graham1993). Bradwell & Stoker (Reference Bradwell and Stoker2015a, p. 317) suggested that the LLGM ice margin was “… situated close to the present-day coastline in NW Lewis…”, although this interpretation is based solely on the interpretation of submarine moraine configuration from bathymetric imagery. Conversely, the TCN ages obtained for North Rona (28.7±1.4 ka; Everest et al. Reference Everest, Bradwell, Stoker and Dewey2013) and a radiocarbon age of 22.5±0.3 ¹⁴C ka BP (27.2–25.9 cal ¹⁴C ka) reported by Peacock et al. (Reference Peacock, Austin, Selby, Graham, Harland and Wilkinson1992) for a bivalve from glacimarine sediments west of moraine banks south of St Kilda suggest the extension of Outer Hebrides ice to at least mid-shelf.

Figure 12 Reconstructed ice-margin positions around northern Scotland (coloured lines), interpreted from the alignment of submarine moraines by Bradwell & Stoker (Reference Bradwell and Stoker2015a). Stages 1 and 2 are inferred to represent pre-MIS 3/2 moraines, and Stage 3 moraines are interpreted as the outermost Late Devensian moraines. Their interpretation of subsequent ice-sheet retreat (stages 4–10) implies early deglaciation of the northern Outer Hebrides and persistence of an ice cap centred on Orkney and Shetland after retreat of the ice margin to the present coast of NW Scotland. The dates depicted are selected (unrecalibrated) TCN ages. Reproduced from Bradwell, T. & Stoker, M. S. (Reference Bradwell and Stoker2015) Earth and Environmental Science Transactions of the Royal Society of Edinburgh 105, 297–322 with permission.

The evidence outlined above contravenes the widespread belief that there is “…unequivocal evidence for glaciation to the continental shelf edge all the way from SW Ireland to the Shetland Isles” (Clark et al. Reference Clark, Hughes, Greenwood, Jordan and Sejrup2012, p. 141), a proposition widely accepted in most recent accounts (Hubbard et al. Reference Hubbard, Bradwell, Golledge, Hall, Patton, Sugden, Cooper and Stoker2009; Chiverrell & Thomas Reference Chiverrell and Thomas2010; Gibbard & Clark Reference Gibbard and Clark2011; Hughes et al. Reference Hughes, Clark and Jordan2014, Reference Hughes, Gyllencreutz, Lohne, Mangerud and Svendsen2016). It demonstrates that westward advance of the last ice sheet was not halted by increasing water depth in all sectors. On the West Shetland Shelf, however, the mapped shelf-edge moraines are associated with stratigraphic units of inferred Late Devensian age (Stoker & Holmes Reference Stoker and Holmes1991; Bradwell & Stoker Reference Bradwell and Stoker2015a; Fig. 12), implying that ice moving NW across the shelf in this area terminated at the shelf break.

3.6. Ice sheet expansion: synthesis

The evidence summarised above indicates that marine-terminating ice existed in Scotland for several millennia prior to the growth of the last SIS, and that ice expansion across low ground and the adjacent shelf commenced within the period 35–32 ka. To the south, the expanding ice sheet initially invaded Ireland, the ISB and northern England, before the establishment of an ice divide across the North Channel limited southward ice flow. A N–S ice divide developed between the SW Highlands and N England, producing dominantly westerly ice flow toward the Malin Shelf and easterly flow towards the NSB. Ice from eastern Scotland was confluent with the FIS in the NSB, but there is debate as to whether this ultimately resulted in a confluent flow of ice towards the NW or the establishment of a (migratory) ice shed between NE Scotland and Norway that resulted in a northwesterly flow of ice from the Moray Firth and the NSB across Caithness and Orkney. Both Shetland and the Outer Hebrides probably developed independent ice caps that persisted as centres of ice dispersal throughout the expansion of the SIS. Glacigenic sedimentation on shelf-edge fans, the seismostratigraphy of the outer shelf, and the alignment of submarine moraine banks collectively suggest that the last SIS probably extended to the Atlantic shelf edge in most sectors, but no farther than mid-shelf on the Hebrides Shelf.

Assumption of ice-sheet expansion from the Scottish mainland after ∼34 ka, as suggested by the terrestrial dating evidence, implies that growth of the SIS to its maximum extent on the Atlantic shelf occurred over 4000–7000 years. Irrespective of the exact timing of the LLGM in this sector, the northern parts of the SIS appear to have reached their maximum extent prior to the global LGM of ∼26.5–19 ka (P. U. Clark et al. Reference Clark, McCabe, Schnabel, Clark, Freeman, Maden and Xu2009), probably reflecting the relatively small size of the SIS and its position adjacent to the Atlantic Ocean, which was a source of moisture-bearing airmasses that fed rapid ice-sheet growth so that ice centres responded rapidly to climate deterioration (Hughes et al. Reference Hughes, Gyllencreutz, Lohne, Mangerud and Svendsen2016; Patton et al. Reference Patton, Hubbard, Andreassen, Winsbarrow and Stroeven2016).

5.1. Palaeo-ice streams of the last Scottish Ice Sheet

Ice streams are corridors of fast-flowing ice within ice sheets (Bentley Reference Bentley1987; Bennett Reference Bennett2003; Truffer & Echelmeyer Reference Truffer and Echelmeyer2003; Benn & Evans Reference Benn and Evans2010; Rignot et al. Reference Rignot, Mouginot and Scheuchl2011). Velocities reach several hundreds of metres per year, often over deforming sediments, and their margins are defined by shear zones. Palaeo-ice streams that developed within former ice sheets can be identified by a range of criteria, notably a convergent flow pattern, a distinct flow track with sharply defined lateral margins, markedly elongate bedforms and lineations, and terminal sediment accumulations, often in the form of trough-mouth fans (Stokes & Clark Reference Stokes and Clark1999, Reference Stokes and Clark2001).

On the basis of such evidence, four eastward-flowing ice streams that drained the SIS have been identified; all exhibit topographic control and extended into the NSB. The Tyne Ice Stream drained ice from the Southern Uplands, Cheviots, Lake District and northern Pennines (Fig. 7). During the LLGM, flow was W–E, but a subsequent shift to southeasterly flow reflects the increased dominance of Scottish ice (Livingstone et al. Reference Livingstone, Ó Cofaigh and Evans2008, Reference Livingstone, Ó Cofaigh and Evans2010b, Reference Livingstone, Evans, Ó Cofaigh, Davies, Merritt, Huddart, Mitchell, Roberts and Yorke2012, Reference Livingstone, Roberts, Davies, Evans, Ó Cofaigh and Gheorghiu2015). The ∼20 km-wide Tweed Ice Stream, reconstructed by Everest et al. (Reference Everest, Bradwell and Golledge2005) from the distribution and orientation of drumlins, megadrumlins and megaflutes, flowed NE to E between the Lammermuir and Cheviot Hills, draining an area of ∼3500 km² and extending ∼65 km to the present coastline. The Strathmore Ice Stream flowed eastward in a ∼45 km-wide corridor between the Highland edge and NE Fife, is defined by streamlined bedrock and bedforms, and exhibits evidence of a post-LLGM northeastward shift in flow (Golledge & Stoker Reference Golledge and Stoker2006). Farther north, there is evidence of eastward streaming of ice in the Moray Firth, although changes in the offshore configuration of the Moray Firth Ice Stream remain to be worked out in detail (Merritt et al. Reference Merritt, Auton and Firth1995, Reference Merritt, Auton, Connell, Hall and Peacock2003, Reference Merritt, Connell and Hall2017; Graham et al. Reference Graham, Lonergan and Stoker2009). In the NSB, the NW–SE-trending MSGLs identified by Graham et al. (Reference Graham, Lonergan and Stoker2007, Reference Graham, Lonergan and Stoker2010, Reference Graham, Stoker, Lonergan, Bradwell, Stewart, Ehlers, Gibbard and Hughes2011), crossing the Witch Ground Basin (see section 3.4.4), have been interpreted as representing development of a former ice stream, although there is debate as to whether this represents former ice streaming NW towards the Orkney–Shetland channel (Fig. 9) or SE away from an ice divide over the northern NSB (Sejrup et al. Reference Sejrup, Clark and Hjelstuen2016). Hall & Glasser (2003) used the term ‘ice stream' to describe the effects of wet-based sliding ice in Glen Avon (Cairngorms), but whether true ice streams developed in the Cairngorm troughs is questionable.

In the western NSB, the coalescence of ice emanating from the Forth, the Tweed and probably the eastern Grampians (Davies et al. Reference Davies, Roberts, Bridgland, Ó Cofaigh and Riding2011) fed the North Sea Lobe (NSL), a major ice stream that moved southwards along eastern England, reaching the Norfolk coast. Livingstone et al. (Reference Livingstone, Evans, Ó Cofaigh, Davies, Merritt, Huddart, Mitchell, Roberts and Yorke2012) suggested that the NSL was initially constrained by the FIS to the east, but persisted after a decoupling of the FIS and BIIS. Moss samples from beneath the lower of two tills (the Skipsea Till) at Dimlington on the Holderness coast of east Yorkshire produced radiocarbon ages of 22.9–21.9 cal ¹⁴C ka and 22.4–21.9 cal ¹⁴C ka (Penny et al. Reference Penny, Coope and Catt1969), and OSL dating indicates the deposition of the Skipsea Till at ∼21.6 ka (Bateman et al. Reference Bateman, Evans, Roberts, Medialdea, Ely and Clark2017), implying that the NSL reached its maximum extent several millennia after the LLGM on the Atlantic shelf. At Dimlington, laminated silts and cross-bedded sands separate the Skipsea Till from an upper till, the Withernsea Till. OSL and radiocarbon ages suggest that the latter was deposited by a readvance that culminated at ∼16.8 ka, although as Bateman et al. (Reference Bateman, Buckland, Whyte, Ashurst, Boulter and Panagiotakopulu2011, Reference Bateman, Evans, Roberts, Medialdea, Ely and Clark2017) acknowledged, this timing appears contrary to wider evidence for advanced deglaciation by this time.

To the south, ice from Galloway joined that from England, Wales and Ireland to feed the Irish Sea Ice Stream (ISIS), which drained >17 % of the BIIS. The southern limit of this ice stream has been placed near the edge of the Celtic Sea Shelf (Praeg et al. Reference Praeg, McCarron, Dove, Ó Cofaigh, Scott, Monteys, Facchin, Romeo and Coxon2015), implying a maximum travel distance exceeding 800 km. To the west, several authors have depicted a Hebrides Ice Stream that was confluent with a North Channel–Malin Shelf Ice Stream fed by ice from northern Ireland (Dunlop et al. Reference Dunlop, Shannon, McCabe, Quinn and Doyle2010) and SW Scotland (Finlayson et al. Reference Finlayson, Fabel, Bradwell and Sugden2014), and probably terminated at the shelf break, feeding sediment to the BDF (Fig. 9). Bedform evidence from the main flow track of this postulated ice stream is lacking, but Howe et al. (Reference Howe, Dove, Bradwell and Gafeira2012) and Dove et al. (Reference Dove, Arosio, Finlayson, Bradwell and Howe2015) have shown that elongate streamlined bedforms cut southwestward across structural trends on the seafloor of the Sea of the Hebrides. They interpreted these as representing the onset zone of the Hebrides Ice Stream, where ice accelerated from an area dominated by bedrock obstacles onto the sediment-dominated shelf. Streamlined bedforms in western Kintyre probably represent associated westward ice streaming from the Firth of Clyde, and Finlayson et al. (Reference Finlayson, Fabel, Bradwell and Sugden2014) have suggested that the resulting drawdown of ice resulted in a ∼60 km westward shift of the N–S ice divide across central Scotland. Whether the landform evidence in the Sea of the Hebrides relates to an ice stream at the LLGM or during overall retreat (as suggested by cross-cutting flow indicators) is uncertain (Dove et al. Reference Dove, Arosio, Finlayson, Bradwell and Howe2015).

The most intensively researched Scottish palaeo-ice stream is the Minch Ice Stream, first identified by Stoker & Bradwell (Reference Stoker and Bradwell2005). During the LLGM, it drained an area of 10,000–15,000 km² and was fed by ice from the NW Highlands, Skye and eastern Lewis, which converged in the North Minch and followed a ∼200 km-long, 40–50 km-wide trough that terminates at the SSF (Fig. 15). Bradwell et al. (Reference Bradwell, Stoker and Larter2007) identified at least nine land-based onset zones characterised by streamlined landforms. Spectacular bedrock megagrooves formed by subglacial abrasion and meltwater erosion occur in a 20 km-wide corridor north of Loch Broom (Bradwell et al. Reference Bradwell, Stoker and Krabbendam2008c), but probably reflect erosion over several glacial cycles. In the onset zone around Loch Laxford, Bradwell (Reference Bradwell2013) showed that the zone of accelerated ice flow can be identified from the distribution of bedrock landforms such as roches moutonnées, whalebacks and plucked rock faces. Within the North Minch, the signature of former ice streaming includes an assemblage of ‘hard bed' subglacial forms, including large submarine crag-and-tail features, bedrock megaflutes and megagrooves, as well as drumlinoid bedforms in areas of sediment cover (Bradwell & Stoker Reference Bradwell and Stoker2015b). Farther N, the ice stream followed the trough that terminates at the SSF, though it is uncertain whether grounded ice reached the shelf edge or terminated on mid-shelf (Bradwell & Stoker Reference Bradwell and Stoker2015a), possibly with a floating ice shelf extending to the shelf break. Fluctuations in the size and velocity of the ice stream may have resulted in the periodic drawdown of ice in source areas, causing shifts in the location of the main N–S ice divide across the NW Highlands (Bradwell et al. Reference Bradwell, Stoker and Larter2007). Eastward migration of this divide probably explains the eastwards and westwards transport of erratics across the Moine Thrust Zone (Lawson Reference Lawson1990) and from an augen-gneiss outcrop ∼15 km E of the present watershed (Sutherland Reference Sutherland1984). This interpretation is also consistent with the easterly migration of the ice-divide inferred by Hughes et al. (Reference Hughes, Clark and Jordan2014; Fig. 5).

Figure 15 Main palaeoglaciological features of the Minch Ice Stream. White lines indicate proposed ice-stream tributaries and flowlines. Thick lines are inferred terminus positions proposed by Bradwell & Stoker (Reference Bradwell and Stoker2015a), who considered that the outermost line probably represents a pre-MIS3/2 ice limit, but the inner lines represent Late Devensian ice margin positions. Hatching indicates an area of subglacial bedforms and iceberg scours, but lacking moraines. From Bradwell & Stoker (Reference Bradwell and Stoker2015b) Boreas 44, 255–276. © Boreas Collegium. Reproduced with permission from John Wiley & Sons Ltd.

5.2. Ice-sheet models

Although the evidence for major ice streams draining the SIS is compelling, the history of most ice streams is poorly constrained. In their reconstruction of ice-sheet evolution, Hughes et al. (Reference Hughes, Clark and Jordan2014) envisaged initial development of a thick integrated ice sheet, with ice-stream development contributing to subsequent thinning and ice-divide migration during and after the ice sheet achieved its maximum extent, but did not preclude ice-stream formation during ice-sheet expansion. The first numerical simulation experiments driven by proxy climate functions (Boulton & Hagdorn Reference Boulton and Hagdorn2006) indicated the early development of ice streams draining the predominantly cold-based upland core of the SIS, with the resultant drawdown of the ice-sheet surface. Their models also identified major ice discharge arteries at the locations of the Minch, Hebrides, Irish Sea and Moray Firth Ice Streams.

Later thermomechanically-coupled models of the BIIS driven by a scaled NGRIP oxygen isotope curve for the period 38–10 ka (Hubbard et al. Reference Hubbard, Bradwell, Golledge, Hall, Patton, Sugden, Cooper and Stoker2009) have indicated a more detailed scenario. A key finding of the optimal (most geologically-consistent) model is that it retrodicts major ‘binge and purge' cycles during the lifetime of the ice sheet, identifying prolonged phases of build-up of cold-based, high-viscosity ice succeeded by ‘purge' phases triggered by abrupt warming. Such ‘purge' phases are characterised by the widespread development of ice streams around the cold-based core of the modelled ice sheet. At such times, the ice sheet surface is drawn down but the ice margins advance. The optimal model generates transient but recurrent ice streaming at all the Scottish ice-stream locations described above, and also intervening areas of fast-flowing ice that may have captured ice flow from the known ice-stream locations. It retrodicts limited ice cover (mainly over the Highlands, with tidewater-terminating margins) prior to ∼33 ka, followed by extension of the ice margin to (or near to) the Atlantic Shelf edge at ∼29 ka during the first major ‘purge' event, immediately followed by the development of all the Scottish ice streams identified from the geomorphological evidence. It also suggests, however, that small ice streams may have developed during ice-sheet build-up. More generally, this model suggests that although the largest ice streams (the Hebrides and Irish Sea Ice Streams) were active throughout much of the last glacial cycle, smaller ice streams were only intermittently active over centennial timescales.

A notable feature of the optimum model of Hubbard et al. (Reference Hubbard, Bradwell, Golledge, Hall, Patton, Sugden, Cooper and Stoker2009) is that it suggests that the initial advance of the SIS to the Atlantic Shelf edge, deep into England and Ireland and far out into the NSB at ∼29 ka was followed first by shrinkage to a much smaller ice mass centred on the Scottish Highlands by ∼27 ka, then by climate-driven re-expansion to form an even more extensive ice sheet by ∼23.7 ka. There appears to be no evidence to support drastic shrinkage of the SIS between 29 ka and 27 ka (many authors have inferred that ∼27 ka approximates the timing of the SIS maximum) but this may be because of ‘erosion censoring': later, more extensive ice advances obscure or obliterate the geomorphological evidence produced by earlier less extensive events (Kirkbride & Winkler Reference Kirkbride and Winkler2012). The same caveat applies to moraine sequences on offshore shelves: nested ice-marginal moraines represent ice extent at particular times, but intervening events involving less extensive ice cover are likely to be absent from the landform record. There may therefore be a prolonged temporal hiatus between one set of moraines and the next, during which the ice margin may have retreated and readvanced on several occasions. Thus although ‘outer' moraines are older than ‘inner' moraines, the intervening time period may have involved complex changes in ice extent or flow direction.

Below we outline the results of attempts to reconstruct the deglaciation history of the SIS at ice-sheet scale, before considering the detailed evidence for particular sectors. Apart from the numerical modelling by Hubbard et al. (Reference Hubbard, Bradwell, Golledge, Hall, Patton, Sugden, Cooper and Stoker2009), the former are based on the pattern of offshore landforms, terrestrial flowsets and a patchy distribution of radiocarbon and TCN ages, some of questionable validity. Not all reconstructions are adequately integrated with the stratigraphic record. At the time of writing, a major initiative (the BRITICE-CHRONO project), designed to refine the chronology of BIIS retreat, is nearing completion, and promises to yield new insights into the retreat behaviour of the SIS (www.britice-chrono.group.shef.ac.uk).

7.1. The northern sector of the last Scottish Ice Sheet

Deglaciation of the northern Hebrides Shelf and West Shetland Shelf has been reconstructed by Bradwell & Stoker (Reference Bradwell and Stoker2015a), who employed the pattern of ice-marginal landforms mapped from high-resolution echo-sounder data and stratigraphic and sedimentological data (Stoker et al. Reference Stoker, Hitchen and Graham1993; Stoker & Varming Reference Stoker, Varming, Ritchie, Ziska, Johnson and Evans2011; Stoker Reference Stoker, Hitchen, Johnson and Gatliff2013) to identify stages in ice-margin retreat (Fig. 12). Intervening and nearshore sequences of small, sharp-crested ridges in various locations are interpreted as De Geer moraines, implying that the larger moraine banks represent grounded ice margins terminating in shallow water.

On stratigraphic grounds (see section 3.5 above), Bradwell & Stoker (Reference Bradwell and Stoker2015a) assigned the outermost moraines of their stages 1 and 2 (Fig. 12) to deposition by a pre-Late Devensian ice sheet, implying that the oldest moraines of Late Devensian age are represented by their stage 3 ice limit, depicted at the edge of the West Shetland Shelf but on mid-shelf farther SW. As North Rona, where ice retreat was originally dated to ∼25 ka (Everest et al. Reference Everest, Bradwell, Stoker and Dewey2013; recalibrated as ∼29 ka; section 3.5), lies outside this limit, this interpretation appears to imply that the last ice sheet locally extended beyond the mapped stage 3 limit but left no footprint, unless the stage 2 ice limit has been misattributed. Stage 4 implies subsequent recession of the ice margin along the entire sector and suggests early deglaciation of northernmost Lewis. If the suggested configuration of subsequent stages is correct, it implies: (1) very early deglaciation of Cape Wrath, possibly the first part of mainland Scotland to emerge from the ice sheet; (2) early deglaciation of the North Minch and probably parts of eastern Lewis; and (3) progressive decoupling of ice nourished in NW Scotland from an ice mass that covered NE Caithness, Orkney and Shetland. The configuration of this ice mass is less certain. Bradwell & Stoker (Reference Bradwell and Stoker2015a) suggested a correlation of their stage 9 limit W of Orkney with moraine banks SE of Orkney (Fig. 12), but this solution appears to imply southwards ice movement across the northern coast of Caithness and shrinkage of the ice margin towards a residual ice cap on Orkney. The terrestrial field evidence, however, suggests that the final ice movement across both Caithness and Orkney was SE–NW, with no reported evidence for a late-stage Orkney ice cap (Hall et al. 2011, 2016b; Hall & Riding 2016).

The published dating evidence for these events is slender. The inferred stage 3 ice limit must post-date the retreat of the ice margin from North Rona at 28.7±1.4 ka, and TCN dating of the Wester Ross Readvance places the ice margin across the fjords and peninsulas in the south of the area depicted in Figure 12 at 15.3±0.7 ka (see section 8.1 below). Samples from bedrock surfaces and boulders on low ground at sites on Orkney have, excluding outliers, produced five recalibrated ¹⁰Be exposure ages of 17.2±2.4 ka to 16.0±2.0 ka (Phillips et al. Reference Phillips, Hall, Ballantyne, Binnie, Kubik and Freeman2008), with a weighted mean age of 16.5±1.2 ka (Fig. 19), indicating that Orkney was largely ice-free by that time. Two samples from Dunnet Head on the N coast of Caithness have a mean age of 17.5±1.1 ka, suggesting deglaciation of the Pentland Firth prior to the final deglaciation of Orkney.

7.2. Shetland

As outlined earlier, there is persuasive (though contested) evidence that an independent ice cap persisted over Shetland throughout the last glacial cycle (Hall 2013). The pattern of offshore moraines is consistent with this interpretation (Figs 11, 16), indicating oscillatory retreat of an independent Shetland Ice Cap from the shelf edge and from near the edge of the Norwegian Channel. Bradwell et al. (Reference Bradwell, Stoker, Golledge, Wilson, Merritt, Long, Everest, Hestvik, Stevenson, Hubbard, Finlayson and Mathers2008b), Clark et al. (Reference Clark, Hughes, Greenwood, Jordan and Sejrup2012) and Hughes et al. (Reference Hughes, Clark and Jordan2014, Reference Hughes, Gyllencreutz, Lohne, Mangerud and Svendsen2016) all depict the ice margin progressively retreating towards Shetland and the eventual isolation of an ice cap over the archipelago. The TMC model of Hubbard et al. (Reference Hubbard, Bradwell, Golledge, Hall, Patton, Sugden, Cooper and Stoker2009) suggests separation of the Shetland Ice Cap before 17 ka. Radiocarbon dates from marine shells recovered from vibrocores indicate that Shetland ice retreated from the Viking Bank, ∼155 ENE of Shetland, before 18.3–17.3 cal ¹⁴C ka (Peacock Reference Peacock1995) and from a site ∼70 km E of Shetland before ∼15.1–14.1 cal ¹⁴C ka (Peacock & Long Reference Peacock and Long1994). Basal radiocarbon ages obtained from organic material at sites in southern Shetland indicate deglaciation prior to 15.4–15.2 cal ¹⁴C ka (Whittington et al. Reference Whittington, Buckland, Edwards, Greenwood, Hall and Robinson2003), 16.2–14.8 cal ¹⁴C ka (Birnie Reference Birnie2000) and 16.1–15.8 cal ¹⁴C ka (Hulme & Shirrifs 1994), although the last-mentioned inferred the onset of postglacial sedimentation as early as 16.9–16.6 ka. Collectively, the available dating evidence suggests that isolation of the Shetland Ice Cap occurred before 17 ka, and that much of Shetland was deglaciated between 17 ka and 16 ka.

7.3. The North Sea Basin (NSB)

Interpretation of the timing and nature of the decoupling of the SIS and FIS in the NSB has proved controversial. Forminifera and molluscs within stratified glacimarine or marine sediments overlying till (deformed marine sediments) in a core from the Witch Ground Basin (∼162 km NE of Rattray Head) have produced radiocarbon ages of 27.0–26.3, 25.3–24.2 to 23.8–23.0 cal ¹⁴C ka (Sejrup et al. Reference Sejrup, Haflidason, Aarseth, King, Forsberg, Long and Rokoengen1994). The oldest date suggests that the central NSB may have become deglaciated as early as ∼27 ka (Sejrup et al. Reference Sejrup, Haflidason, Aarseth, King, Forsberg, Long and Rokoengen1994, Reference Sejrup, Nygård, Hall and Haflidason2009, Reference Sejrup, Hjelstuen, Nygård, Haflidason and Mardal2015; Graham et al. Reference Graham, Lonergan and Stoker2010). Other samples in contexts indicating deglaciation of this area are, however, much younger (<22 cal ¹⁴C ka), suggesting a hiatus in the stratigraphic record (Sejrup et al. Reference Sejrup, Clark and Hjelstuen2016).

Bradwell et al. (Reference Bradwell, Stoker, Golledge, Wilson, Merritt, Long, Everest, Hestvik, Stevenson, Hubbard, Finlayson and Mathers2008b) suggested that the initial decoupling of the SIS and FIS followed a bathymetric depression that extends from the shelf edge N of Shetland southwards to the Witch Ground Basin. They argued that a calving margin propagated rapidly southward from the shelf edge, creating a marine embayment between the two ice sheets, an interpretation accepted by several authors (Graham et al. Reference Graham, Lonergan and Stoker2010; Sejrup et al. Reference Sejrup, Hjelstuen, Nygård, Haflidason and Mardal2015; Merritt et al. Reference Merritt, Connell and Hall2017). The reconstruction by Clark et al. (Reference Clark, Hughes, Greenwood, Jordan and Sejrup2012) offered two possible scenarios: (1) early (∼25 ka) and complete separation of the BIIS and FIS; and (2) partial decoupling of the SIS from the FIS by ∼23 ka along a marine embayment, a scenario similar to that proposed by Bradwell et al. (Reference Bradwell, Stoker, Golledge, Wilson, Merritt, Long, Everest, Hestvik, Stevenson, Hubbard, Finlayson and Mathers2008b). They noted, however, that the latter scenario requires persistence of a residual ice dome south of the proposed embayment, for which there is no evidence.

An alternative interpretation, proposed by Sejrup et al. (Reference Sejrup, Clark and Hjelstuen2016), is based on bathymetric landform evidence indicating that an initial westward extension of ice flowing NW along the Norwegian Channel (a broad bathymetric trough parallel to the coast of SW Norway) is overprinted by moraines indicating ice advance from a Shetland–Orkney ice centre. This evidence suggests that acceleration, thinning and retreat of the Norwegian Channel Ice Stream (NCIS) debuttressed the ice flowing from Shetland and Orkney, causing it to expand eastward (Fig. 17). Radiocarbon-dated records of glacigenic debris flows at the NCIS terminus demonstrate extremely high sediment flux at ∼20–19 ka (Nygård et al. Reference Nygård, Sejrup, Haflidason, Lekens, Clark and Bigg2007), which Sejrup et al. (Reference Sejrup, Clark and Hjelstuen2016) attributed to the onset of NCIS acceleration and thinning, and radiocarbon ages relating to its subsequent retreat suggest decoupling of the ice sheets in the eastern NSB at ∼18.5 ka, accompanied by drainage of a former ice-dammed lake in the southern NSB (Fig. 17). This interpretation places the line of disengagement of the BIIS and FIS farther east than previously depicted, suggesting that much of the ice occupying the NSB at the time of separation was of Scottish rather than Fennoscandian provenance. Sejrup et al. (Reference Sejrup, Clark and Hjelstuen2016) envisaged that decoupling of the two ice sheets led to a reorganisation of flow in the BIIS, accompanied by readvances, ice thinning and a rapid retreat of its eastern margins (Fig. 17). This interpretation is consistent with the persistence of a residual ice mass over Orkney and Shetland (Bradwell & Stoker Reference Bradwell and Stoker2015a) and does not require the existence of a remnant ice dome in the southern NSB, but conflicts with radiocarbon ages from the Witch Ground Basin and eastern Scotland that appear to imply much earlier deglaciation of these areas (McCabe et al. Reference McCabe, Clark, Smith and Dunlop2007; Sejrup et al. Reference Sejrup, Haflidason, Aarseth, King, Forsberg, Long and Rokoengen1994, Reference Sejrup, Nygård, Hall and Haflidason2009, Reference Sejrup, Hjelstuen, Nygård, Haflidason and Mardal2015).

Figure 17 Reconstructions of successive ice-sheet configurations in the North Sea Basin (NSB) proposed by Sejrup et al. (Reference Sejrup, Clark and Hjelstuen2016): (a) maximum configuration with an ice divide extending from NE Scotland to SW Norway and an ice-dammed lake in the southern NSB; (b) initiation of ice-sheet collapse: expansion and acceleration of the Norwegian Channel Ice Stream (NCIS) (dark blue lines) causes drawdown of ice and destroys the ice divide. Grounding line retreat in the Norwegian Channel progressively debuttresses ice on either side; (c) decoupling of the BIIS and FIS as the NCIS retreats, causing northward drainage of the ice-dammed lake. Red arrows indicate readvances of the BIIS; and black arrows indicate net westward retreat of the BIIS ice margin. Reproduced from Sejrup et al. (Reference Sejrup, Clark and Hjelstuen2016) Geology 44, 355–358, with permission from the Geological Society of America.

Evidence for readvances of Scottish ice into the Witch Ground Basin has been identified by Sejrup et al. (Reference Sejrup, Hjelstuen, Nygård, Haflidason and Mardal2015), who presented bathymetric, stratigraphic and chronological evidence for two such events, the Fladen 1 and Fladen 2 readvances. The location and alignment of the associated ice margins suggests that although they may have been fed by ice from the Moray Firth, they could reflect advance of ice from an Orkney–Shetland centre to the NW. The associated radiocarbon ages are not all in sequence, but Sejrup et al. (Reference Sejrup, Hjelstuen, Nygård, Haflidason and Mardal2015) argued that the Fladen 1 readvance probably occurred at ∼17.5 ka and the Fladen 2 event at ∼16.2 ka. For the outer Moray Firth, west of the Witch Ground Basin, Graham et al. (Reference Graham, Lonergan and Stoker2009) described evidence suggesting that the formation of W–E-aligned streamlined bedforms indicative of fast eastward flow of grounded ice was succeeded by a westward retreat of the ice margin, then the formation of broad N–S-aligned morainic banks (the Little Halibut Bank and Bosies Bank) and multiple composite ice-marginal ridges. Farther W, a series of transverse ridges indicate marine grounding-line positions or ice-push moraines that are succeeded westward by a zone of hummocky moraine. They interpreted this evidence as demonstrating that initial retreat of the ice margin in the outer Moray Firth was characterised by readvances, stillstands and possible ‘surge-like' activity; the hummocky moraine farther west they interpreted as stagnant ice topography, indicating rapid retreat.

7.4. The Irish Sea Basin (ISB)

As outlined in section 3.3, the development of an ice divide between Scotland and NE Ireland during the LLGM appears to have limited the Scottish contribution to the Irish Sea Ice Stream (ISIS) to ice nourished in SW Scotland (Livingstone et al. Reference Livingstone, Evans, Ó Cofaigh, Davies, Merritt, Huddart, Mitchell, Roberts and Yorke2012; Hughes et al. Reference Hughes, Clark and Jordan2014; Figs 5, 7). Initial retreat of the ISIS from the Celtic Sea Shelf appears to have been rapid, with the complete deglaciation of the southern ISB by 21.9–20.7 ka (Chiverrell et al. Reference Chiverrell, Thrasher, Thomas, Lang, Scourse, McCarroll, Clark, Ó Cofaigh, Evans and Ballantyne2013; Patton et al. Reference Patton, Hubbard, Bradwell, Glasser, Hambrey and Clark2013; Smedley et al. Reference Smedley, Chiverrell, Ballantyne, Burke, Clark, Duller, Fabel, McCarroll, Scourse, Small and Thomas2017a). Foraminifera in raised marine muds in NE Ireland have yielded ages of 20.4–19.9 cal ¹⁴C ka, implying ice-free conditions in the eastern sector of the northern ISB before ∼20 ka (McCabe & Clark Reference McCabe and Clark1998; P. U. Clark et al. Reference Clark, Evans, Khatwa, Bradwell, Jordan, Marsh, Mitchell and Bateman2004).

Interpretation of terrestrial lithostratigraphy and flowsets suggests that the subsequent deglaciation of the northern ISB was complex. Livingstone et al. (Reference Livingstone, Evans, Ó Cofaigh, Davies, Merritt, Huddart, Mitchell, Roberts and Yorke2012) identified evidence for a readvance of the ice margin, the Blackhall Wood Readvance, fed in part by ice from the Solway region and Galloway. This event they tentatively correlated with the Gosport Oscillation of Merritt & Auton (Reference Merritt and Auton2000), and they suggested that it may reflect widespread cooling of the NE Atlantic region due to a decrease in Atlantic meridional overturning circulation at ∼19 ka (P. U. Clark et al. Reference Clark, Evans, Khatwa, Bradwell, Jordan, Marsh, Mitchell and Bateman2004; Hall et al. 2006). Onshore stratigraphic evidence suggests subsequent deglaciation of the Solway lowlands and almost all of the ISB, followed by a second readvance, the ‘Scottish Readvance'. During this event, ice from southern Scotland reoccupied the Solway region and extended as far south as the Isle of Man (Livingstone et al. Reference Livingstone, Evans and Ó Cofaigh2010a, Reference Livingstone, Evans, Ó Cofaigh, Davies, Merritt, Huddart, Mitchell, Roberts and Yorke2012, Reference Livingstone, Roberts, Davies, Evans, Ó Cofaigh and Gheorghiu2015; Fig. 18). Various authors (McCabe et al. Reference McCabe and Clark1998; McCabe & Clark Reference McCabe, Knight and McCarron1998; Roberts et al. Reference Roberts, Dackombe and Thomas2007; Merritt & Auton Reference Merritt and Auton2000) have suggested that the Scottish Readvance correlates with the Killard Point Readvance in NE Ireland, dated to 17.3–16.6 cal ¹⁴C ka (Ballantyne & Ó Cofaigh Reference Ballantyne, Ó Cofaigh, Coxon, McCarron and Mitchell2017), but Livingstone et al. (Reference Livingstone, Evans and Ó Cofaigh2010a, Reference Livingstone, Evans, Ó Cofaigh, Davies, Merritt, Huddart, Mitchell, Roberts and Yorke2012) noted that it may have been a short-lived event that could reflect local reorganisation of ice flow rather than a regionally significant advance of the ice margin.

Figure 18 Extent of the Scottish Readvance as represented by Livingstone et al. (Reference Livingstone, Roberts, Davies, Evans, Ó Cofaigh and Gheorghiu2015). Only the ice limits south of the Solway Firth, along the Cumbrian coast and across the Isle of Man can be confidently identified. Purple areas represent proposed ice-dammed lakes. From Livingstone et al. (Reference Livingstone, Roberts, Davies, Evans, Ó Cofaigh and Gheorghiu2015) Journal of Quaternary Science 30, 790–804. © The authors. Reproduced with permission from John Wiley & Sons, Ltd.

Decoupling of the Scottish and Irish Ice Sheets over the North Channel probably occurred between ∼16.5 ka and ∼16.0 ka (Clark et al. Reference Clark, Hughes, Greenwood, Jordan and Sejrup2012; Finlayson et al. Reference Finlayson, Fabel, Bradwell and Sugden2014), marking the end of ice occupance in the ISB. Retreat of Irish ice from the North Channel was succeeded by a readvance of ice from west-central Scotland across NE Ireland, the East Antrim Coastal Readvance. McCabe & Williams (Reference McCabe and Williams2012) reconstructed the limits of this readvance across 1800 km² of NE Antrim, and argued that it occurred at ∼15.6–15.0 ka. This timing was challenged by Finlayson et al. (Reference Finlayson, Fabel, Bradwell and Sugden2014), who argued that it conflicts inter alia with older TCN exposure ages from the Isle of Arran. They envisaged that invasion of SW-flowing Scottish ice across NE Antrim represented ‘debuttressing' of the SW margin of the SIS following collapse of the Irish Ice Sheet around 16.5 ka, but both the timing and cause of this event remain uncertain.

7.5. The western sector of the last Scottish Ice Sheet

The western sector of the SIS comprised a northern zone, where ice from the Outer Hebrides Ice Cap extended an unknown distance westward across the northern Hebrides Shelf, and a southern zone, dominated by SW flow of the Hebrides Ice Stream towards the shelf edge.

7.5.1. The Outer Hebrides and adjacent Hebrides Shelf

The timing and pattern of retreat of the Outer Hebrides Ice Cap on the Hebrides Shelf is poorly constrained. Peacock et al. (Reference Peacock, Austin, Selby, Graham, Harland and Wilkinson1992) reported radiocarbon ages of 22.5±0.3 ¹⁴C ka (26.7–26.0 cal ¹⁴C ka) for a shell from glacimarine deposits distal to moraine banks in the St Kilda basin ∼60 km W of North Uist, and 15.3±0.2 ¹⁴C ka (18.3–17.9 cal ¹⁴C ka) for another from proximal glacimarine deposits on the same moraine banks. These dates indicate that the moraine banks were deposited by the last ice sheet, and that the ice margin retreated from this location before ∼18 ka. For the southern part of the Outer Hebrides Ice Cap, Small et al. (Reference Small, Benetti, Dove, Ballantyne, Fabel, Clark, Gheorghiu, Newall and Xu2017a) reported mean TCN ages of 18.9±1.1 ka for the deglaciation of Mingulay (but acknowledged that a single age of 17.3±0.9 ka may indicate later deglaciation) and 17.1±1.0 ka for the deglaciation of Barra, 27 km further N. A single (recalibrated) TCN age of 16.3±0.9 ka obtained for a bedrock sample from a col in South Uist (Stone & Ballantyne Reference Stone and Ballantyne2006) suggests prior deglaciation of much of the southern Outer Hebrides (Fig. 19).

Figure 19 Uncertainty-weighted means of terrestrial cosmogenic nuclide exposure ages indicating the timing of deglaciation, calibrated using the protocol outlined in section 2. Statistical outliers are excluded. Each age is followed by the full (±1σ) uncertainty, and the number of individual ages in the sample is shown in brackets. Single ages on the Outer Hebrides are shown for completeness, but should be treated with caution.

Three consistent TCN ages suggest that low ground in southern Harris was deglaciated by ∼17.9 ka (unpublished data), implying that by that time the retreating ice cap had split into a southern component occupying the Uists and a northern component centred on the mountains of N Harris. Stone & Ballantyne (Reference Stone and Ballantyne2006) obtained TCN ages for bedrock samples on low ground and cols in N Harris, but disparities amongst these suggest that most are compromised by nuclide inheritance. The two youngest (recalibrated) exposure ages indicate deglaciation by 17.3±0.9 ka on low ground and by 16.6±0.8 ka on a col at 425 m OD, and a single age of 15.6±0.8 ka for a sample from the summit of Oreval (662 m OD) suggests that ice may have persisted over high ground as late as 16–15 ka (Fig. 19). No data have been published for the timing of ice retreat on Lewis but, as noted in section 6.2.1, the early retreat of the Minch Ice stream and the limited extension of the Outer Hebrides Ice Cap west of Lewis (Bradwell & Stoker Reference Bradwell and Stoker2015a) suggest the early deglaciation of the coastal fringes of Lewis, followed by a gradual ice retreat to the mountains of Harris.

8.1. The NW Highlands and western seaboard

As noted in section 7.1, the configuration of ice-marginal moraines mapped by Bradwell & Stoker (Reference Bradwell and Stoker2015a) suggests a very early deglaciation of Cape Wrath and the early development of a marine embayment in the North Minch, with the ice margin subsequently retreating eastward into the fjords of the NW mainland (Fig. 12). The nature of the retreat in the Loch Ewe to Summer Isles area has been beautifully captured in bathymetric surveys (Stoker et al. Reference Stoker, Bradwell, Wilson, Harper, Smith and Brett2006, Reference Stoker, Bradwell, Howe, Wilkinson and McIntyre2009), which reveal numerous large crescentic or crenulate cross-fjord moraines 10–20 m high and up to 3 km long (Fig. 20) and smaller ridges interpreted as De Geer moraines, a submarine landsystem indicative of pulsed retreat of a grounded ice margin. Shells in cores recovered between the Summer Isles and the mouth of Loch Broom provide minimum deglaciation ages of 14.0–13.6 cal ¹⁴C ka and 13.8–13.6 cal ¹⁴C ka. From this and other evidence, Stoker et al. (Reference Stoker, Bradwell, Howe, Wilkinson and McIntyre2009) inferred deglaciation of this area at ∼14.5–13.0 ka. Recalibrated TCN ages for the Wester Ross Readvance, however, imply much earlier deglaciation, as outlined below.

Figure 20 Multibeam bathymetry of the seafloor in the vicinity of the Summer Isles, outer Loch Broom, showing multiple recessional moraines formed by readvances that interrupted ice margin retreat. From Bradwell et al. (Reference Bradwell, Stoker, Golledge, Wilson, Merritt, Long, Everest, Hestvik, Stevenson, Hubbard, Finlayson and Mathers2008a) Journal of Quaternary Science 23, 401–407. © 2008 NERC. Reproduced with permission from John Wiley & Sons Ltd.

The oldest TCN exposure ages in Wester Ross relate to erratic boulders deposited on the summit blockfields of Maol Chean-dearg, Beinn Liath Mhór and Slioch at altitudes of 916–967 m OD (Fabel et al. Reference Fabel, Ballantyne and Xu2012; Figs 14, 19, 22). Excluding outliers, these yielded (recalibrated) uncertainty-weighted mean ages of 16.0±0.8 k, 16.1±0.8 ka and 16.1±0.8 ka respectively, suggesting that mountain summits emerged from the thinning ice sheet as nunataks at ∼16 ka.

Most dating evidence for the retreat of the ice margin in NW Scotland relates to the moraines that mark the limit of the Wester Ross Readvance (WRR), which is intermittently defined by terrestrial moraines between the Applecross Peninsula and Achiltibuie (Robinson & Ballantyne Reference Robinson and Ballantyne1979; Sissons & Dawson Reference Sissons1981; Sutherland Reference Sutherland1984; Figs 21, 22), and designated individually as the Applecross, Redpoint, Gairloch, Aultbea and Achiltibuie moraines. Everest et al. (Reference Everest, Bradwell, Fogwill and Kubik2006) obtained six TCN ages for erratics on the Gairloch moraine, but even after the rejection of two dates as outliers, these yielded (recalibrated) ages of 20.4±3.6 ka to 14.1±2.1 ka, which permit no valid conclusions. Conversely, eight consistent TCN ages with an apparent average age of ∼13.5 ka were reported by Bradwell et al. (Reference Bradwell, Stoker, Golledge, Wilson, Merritt, Long, Everest, Hestvik, Stevenson, Hubbard, Finlayson and Mathers2008a) for samples from boulders on the Achiltibuie moraine, a lateral moraine above Little Loch Broom, and another in the Loanan Valley in Assynt. They concluded from these results and the record of submarine moraines between Loch Ewe and Loch Broom that “substantial dynamic ice caps existed in NW Scotland between 13 and 14 ka” (Bradwell et al. Reference Bradwell, Stoker, Golledge, Wilson, Merritt, Long, Everest, Hestvik, Stevenson, Hubbard, Finlayson and Mathers2008a, p. 401), a conclusion further developed by Stoker et al. (Reference Stoker, Bradwell, Howe, Wilkinson and McIntyre2009).

Figure 21 Wester Ross Readvance moraines: (a) the Applecross moraine crosses the photograph from left to right between the two lochans; (b) the central section of the Gairloch moraine, which is composed of Torridon Sandstone boulders and occasional Lewisian Gneiss erratics. The former glacier was located to the right (east) of the ridge.

Figure 22 Moraines marking the limits of the Wester Ross Readvance, showing the sites sampled by Bradwell et al. (Reference Bradwell, Stoker, Golledge, Wilson, Merritt, Long, Everest, Hestvik, Stevenson, Hubbard, Finlayson and Mathers2008a; sites 1, 2 and 4) and Ballantyne et al. (Reference Ballantyne, Schnabel and Xu2009a; sites 3, 5, 6 and 7) for cosmogenic ¹⁰Be exposure dating of boulders on moraines. From Ballantyne & Stone (Reference Ballantyne and Stone2012), Journal of Quaternary Science 27, 297–306. © 2011 John Wiley & Sons Ltd.

This proposition was initially supported by 14 ¹⁰Be exposure ages obtained by Ballantyne et al. (Reference Ballantyne, Schnabel and Xu2009a) for boulders on the Applecross, Gairloch, Redpoint and Achiltibuie moraines. These yielded apparent mean ages of 14.0–13.5 ka, almost identical to the eight TCN ages reported by Bradwell et al. (Reference Bradwell, Stoker, Golledge, Wilson, Merritt, Long, Everest, Hestvik, Stevenson, Hubbard, Finlayson and Mathers2008a), and indicate that all the dated moraines represent an isochronous, or nearly isochronous, readvance of the ice margin. Subsequent recalibration of all 22 TCN ages using a locally-derived ¹⁰Be production rate, however, showed that the initially reported ages are too young (Ballantyne & Stone Reference Ballantyne and Stone2012). Recalibration of these ages using the protocol adopted here yields an uncertainty-weighted mean age of 15.3±0.7 ka for the WRR (Fig. 23). This revised age implies a much earlier retreat of the ice margin across low ground in Wester Ross, consistent with Lateglacial pollen-stratigraphic evidence from Loch Droma on the watershed inland from Wester Ross. Though the radiocarbon age of 15.6–15.1 cal ¹⁴C ka obtained by Kirk & Godwin (Reference Kirk and Godwin1963) for this site (Fig. 24) may be compromised by ‘old' carbon residues, the associated pollen assemblages appear to span the entire Lateglacial Interstade (Pennington et al. Reference Pennington, Haworth, Bonny and Lishman1972). If so, most or all of the low ground in Wester Ross must have been deglaciated between the WRR at 15.3±0.7 ka and the onset of the Lateglacial Interstade at ∼14.7 ka. Deglaciation of low ground further N was probably earlier. An age of 15.9–15.2 cal ¹⁴C ka for organic material near the base of a core from Cam Loch in Assynt is minimal for deglaciation (Pennington Reference Pennington1975). Similarly, Boomer et al. (Reference Boomer, von Grafenstein and Moss2012) obtained an age of 14.4–14.1 cal ¹⁴C ka from near the base of a core from Loch Assynt, and inferred from their age–depth model that deglaciation occurred before ∼17 ka.

Figure 23 Summed normal kernel density estimates of recalibrated cosmogenic ¹⁰Be exposure ages and their uncertainties for boulder samples from Wester Ross Readvance moraines. Upper line: all 22 samples; weighted mean age 15.28±0.69 ka. Middle line: 14 ages originally reported by Ballantyne et al. (Reference Ballantyne, Schnabel and Xu2009a); weighted mean age 15.31±0.70 ka. Lower (dashed) line: eight ages originally reported by Bradwell et al. (Reference Bradwell, Stoker, Golledge, Wilson, Merritt, Long, Everest, Hestvik, Stevenson, Hubbard, Finlayson and Mathers2008a); weighted mean age 15.23±0.71 ka.

Figure 24 Key terrestrial and nearshore radiocarbon ages that constrain the timing of ice-sheet deglaciation, showing the extent of glacier ice during the Younger Dryas Stade (∼12.9–11.7 ka). Ages are expressed as calibrated ±1σ age ranges (cal ¹⁴C ka). Most nearshore and coastal ages were obtained from molluscs or foraminifera, and inland ages represent the oldest date obtained from organic material in cores recovered from lakes or infilled depressions. For a few sites two or more ages have been aggregated and the combined ±1σ age range is shown. All ages represent minima for deglaciation. Ages shown with a question mark may be too old.

Retreat of the ice margin along the NE coast of Skye, first southwards along the Sound of Raasay and then eastwards towards the mainland, is constrained by TCN ages. Two consistent TCN ages for ice-abraded bedrock at 450 m OD in southern Trotternish (Stone et al. Reference Stone, Ballantyne and Fifield1998; Fig. 19) give a recalibrated mean age of 16.6±1.2 ka, and five samples from boulders on a medial moraine near Broadford yield a recalibrated mean age of 16.0±0.8 ka (Small et al. Reference Small, Rinterknecht, Austin, Fabel, Miguens-Rodriguez and Xu2012). The latter age implies that the ice margin backstepped on to the adjacent mainland after ∼16 ka, consistent with the timing of the WRR (15.3±0.7 ka) on the adjacent Applecross peninsula.

Elsewhere on Skye, there is stratigraphic and geomorphological evidence that the retreat of the mainland ice sheet was succeeded by an expansion of a residual ice cap centred over high ground in the south-central part of the island (Benn Reference Benn1997). In lower Glen Brittle, the up-valley termination of a high Lateglacial shoreline coincides with end moraines marking the limit of ice advance down the glen (Walker et al. Reference Walker, Ballantyne and Lowe1988). Four cosmogenic ³⁶Cl exposure ages obtained from boulders on these moraines give a weighted mean age of 17.4±1.2 ka (Small et al. Reference Small, Rinterknecht, Austin, Bates, Benn, Scourse, Bourlès and Hibbert2016), implying prior disengagement of mainland ice from the Skye ice cap, but the range of individual ages is wide (19.4±1.7 ka to 15.5±1.7 ka). In southern Skye, bathymetric mapping of Loch Scavaig has revealed a pronounced arcuate submarine end moraine that impinges on the adjacent island of Soay, and samples from erratic basalt boulders on the Soay moraine have produced four ³⁶Cl ages of 16.4±1.5 ka to 14.6±1.5 ka, with a weighted mean age of 15.1±1.0 ka. The closeness of this age to that of the WRR moraines (15.3±0.7 ka) suggests contemporaneity, although Small et al. (Reference Small, Rinterknecht, Austin, Bates, Benn, Scourse, Bourlès and Hibbert2016) cautioned against definitive correlation of the two because of uncertainties in the derivation of the mean exposure age for the Soay moraine.

Farther south, the TCN age of 17.5±0.9 ka obtained for the Ross of Mull (Small et al. Reference Small, Benetti, Dove, Ballantyne, Fabel, Clark, Gheorghiu, Newall and Xu2017a; Fig. 19) is consistent with a basal age of 16.0–15.7 cal ¹⁴C ka for the onset of organic sedimentation in nearby Loch an t-Suidhe (Walker & Lowe Reference Walker and Lowe1982), the latter being minimal for deglaciation. North of Mull, a radiocarbon age of 16.8–16.2 cal ¹⁴C ka for a shell in a core recovered from outer Loch Sunart (Baltzer et al. Reference Baltzer, Bates, Mokeddem, Clet-Pellerin, Walter-Simonnet, Bonnot-Courtois and Austin2010) indicates prior deglaciation of the Ardnamurchan Peninsula and western Moidart (Fig. 24). On Jura, a medial moraine formed by a landslide onto the thinning ice sheet yielded a TCN age of 16.5±0.8 ka (Small et al. Reference Small, Benetti, Dove, Ballantyne, Fabel, Clark, Gheorghiu, Newall and Xu2017a), consistent with TCN ages obtained for nearby postglacial rockslide deposits (∼15.4–13.7 ka), which are minimal for deglaciation (Ballantyne et al. Reference Ballantyne, Wilson, Gheorghiu and Rodés2014). Numerous recessional moraines recorded on the seafloor of the Firth of Lorne and Sound of Jura indicate the subsequent oscillatory retreat of a grounded ice margin (Howe et al. Reference Howe, Dove, Bradwell and Gafeira2012; Dove et al. Reference Dove, Arosio, Finlayson, Bradwell and Howe2015, Reference Dove, Finlayson, Bradwell, Howe and Arioso2016).

The available geochronological evidence for the western seaboard therefore indicates deglaciation of the extreme NW, and of parts of Skye and Mull, before ∼17 ka. By ∼16 ka, the margin of mainland ice had backstepped from most of the Inner Hebrides and the westernmost mainland peninsulas, and mountain summits were emerging from the ice sheet. By 15 ka, following the WRR, the mainland ice margin had probably retreated east of the coastline, though residual ice masses may have occupied the mountains of Skye and possibly Mull. The subsequent deglacial history of this sector is difficult to assess, as many of the fjords and valleys of western Scotland were reoccupied by glaciers during the Younger Dryas Stade.

8.2. Caithness, NE Scotland and eastern Scotland

8.2.2. The Moray Firth and NE Scotland

Lithostratigraphic and flowset evidence (Merritt et al. Reference Merritt, Auton, Connell, Hall and Peacock2003; Hughes et al. Reference Hughes, Clark and Jordan2014) indicates that following the LLGM there was major reorganisation of ice flow in NE Scotland. Merritt et al. (Reference Merritt, Connell and Hall2017) envisaged that within the period ∼22–19 ka, predominantly cold-based ice flowing E from the eastern Grampians and Cairngorms was surrounded along coastal areas by a pincer movement of ice flowing E then SE from the Moray Firth and NE-flowing ice from Strathmore (Fig. 25a). They concluded that initial deglaciation on land occurred near the confluence of these three ice masses, inland from Peterhead, but was succeeded by a localised readvance, the Logie–Buchan Readvance. They also suggested that the subsequent disengagement of Grampian and Strathmore ice led to the deglaciation of the east coast from Peterhead to the Tay estuary by ∼20.9–20.1 ka (Fig. 25b), as indicated by radiocarbon ages obtained from foraminifera in raised marine muds at Lunan Bay (McCabe et al. Reference McCabe, Clark, Smith and Dunlop2007; see below).

Figure 25 (a) Pattern of ice movement in eastern Scotland following opening of a marine embayment to the Witch Ground Basin (red hexagon) depicted by Merritt et al. (Reference Merritt, Connell and Hall2017). (b) Subsequent retreat of Strathmore ice and deglaciation of eastern Scotland depicted by the same authors. Dashed lines represent flow directions, and thick lines with open diamonds denote ice divides. From Merritt et al. (Reference Merritt, Connell and Hall2017) Journal of Quaternary Science 32, 276–294. © 2016 John Wiley & Sons, Ltd.

The extent of Moray Firth ice during the period 22–19 ka is uncertain. At St Fergus, on the Buchan coast near Peterhead, a moraine formed by a localised readvance of Moray Firth ice from the E or NE contains glacitectonised marine sediments, and a radiocarbon age of 15,320±200 ¹⁴C a BP (18.0–17.5 cal ¹⁴C ka) obtained for a shell in raised marine silts indicates prior ice-free conditions (Hall & Jarvis 1989) and provides a minimum age for the readvance. Organic material within a core recovered from marine silts in the Moray Firth itself, ∼17 km NE of Banff, yielded radiocarbon ages ranging from 20.9–19.3 cal ¹⁴C ka to 17.7–16.4 cal ¹⁴C ka (Harkness & Wilson Reference Harkness and Wilson1979), the youngest age suggesting that the ice margin had retreated to this position by ∼17 ka. Subsequent ponding of lakes along the N coast of Buchan implies that the margin of the Moray Firth lobe lay along this coast after deglaciation of inland areas, and there is evidence for a readvance near Elgin, which Merritt et al. (Reference Merritt, Connell and Hall2017) assigned to ∼15 ka, although earlier deglaciation seems likely. A further oscillation of the retreating ice margin occurred at Ardesier near Inverness, where a moraine containing glacitectonised glacimarine silts indicates that the retreat of a calving tidewater margin in response to rising sea level was interrupted by a readvance, tentatively assigned to ∼13 ¹⁴C ka (∼15.6 cal ¹⁴C ka) by Merritt et al. (Reference Merritt, Auton and Firth1995). The ice margin then retreated into Loch Ness, at that time a fjord open to the sea. Recessional moraines in the north of the loch indicate the punctuated retreat of a grounded ice margin, followed by deposition of a large moraine (the Foyers rise) midway down the loch, then a rapid retreat towards Fort Augustus, where evidence of subsequent events is lost under Younger Dryas deposits (Turner et al. Reference Turner, Woodward, Dunning, Shine, Stokes and Ó Cofaigh2012).

8.3. Deglaciation of southern and central Scotland