Holocene coastal change at Luce Bay, SouthWest Scotland

Coastal change during the Mid‐ to Late Holocene at Luce Bay, South West Scotland, is examined using morphological, stratigraphic and biostratigraphical techniques supported by radiocarbon dating. Deglaciation left extensive sediments, providing a source for depositional coastal landforms. Glacio‐isostatic uplift resulted in the registration of evidence for former relative sea levels (RSLs), which support the pattern of Holocene RSL change for the northern Irish Sea as determined by shoreline‐based Gaussian trend surface models. The rate of RSL rise was rapid from before ca. 8600 to ca. 7800 cal a BP, but then slowed, changing by <3m over the next 3000 years, a pattern reflected in the convergence of shorelines predicted in the models. By ca. 4400 cal a BP RSL was falling towards present levels. As these changes were taking place, coastal barriers developed and dunes formed across them. In the West of the Bay, a lagoon forming to landward of the barriers and dunes acted as a sediment sink for dune sand. Changes in the coastal landscape influenced the occupation of the area by early human societies. This study illustrates the value of combining an understanding of process geomorphology, RSL and archaeology in studies of coastal change. © 2020 The Authors. Journal of Quaternary Science Published by John Wiley & Sons, Ltd.


Introduction
Luce Bay is a 20-km-wide macro-tidal marine embayment on the northern coast of the outer Solway Firth in SW Scotland ( Fig. 1) with clear but complex inter-relations between Holocene sediment bodies of contrasting depositional environments (Single and Hansom, 1994). Its detailed evolution is bound up with new archaeological insights into the role coastal landscapes played in Neolithic-Bronze Age society (Cowie, 1996;Thomas, 2015;Bradley et al., 2016) yet few local empirical data exist to underpin these insights. The objectives of the work were thus to (i) identify the nature, causes and rates of change of Holocene coastal processes using field and laboratory analyses, (ii) compare relative sea levels (RSLs) at Luce Bay with evidence from neighbouring areas and (iii) examine the intricate effect of coastal changes at Luce Bay on the early human occupation of the area.

Methodology and techniques
Morphological mapping at a scale of 1: 10 000 (cf. Smith et al., 2003aSmith et al., , 2003b was integrated with >300 levelled altitudinal points to Ordnance Datum Newlyn (OD) on closed traverses with accuracies < ±0.05 m. Sediment stratigraphies were recorded from 2.5-cm-diameter Eijelkamp gouge boreholes surveyed to OD (see Supplementary Information) and from the British Geological Survey (NERC) National Geoscience Data Centre (2017) boreholes: the altitudes of some commercial boreholes had to be obtained from LIDAR data with accuracies ±0.05 m. Samples for microfossil analysis and radiocarbon dating were obtained using a 'Russian' closed-chambered corer (Jowsey, 1966) at boreholes previously investigated using an Eijklekamp gouge (thus reducing the possibility of vertical errors in the recorded stratigraphy). Comparison between surveyed altitudes at Luce Bay and the tidal record is facilitated by the availability of tidal values for Luce Bay Platform (Table 1).
Radiocarbon ( 14 C) dating was undertaken on samples of peat from transgressive or regressive overlaps, with the exception of one date (from No. 2 Holdings) given below. All dates were obtained by the accelerator mass spectrometry (AMS) method and were calibrated using Calib 7.1 (Stuiver and Reimer, 1993;Reimer et al., 2009). In the text, dates are given as sidereal (calibrated) age ranges BP (before 1950) to two standard deviations, with their radiocarbon date to one standard deviation in parentheses, thus 8602-8973 cal (7910 ± 30 14 C) a BP: approximate dates are indicated by 'ca.'. The Early Holocene is defined as having occurred 11 700-8200 cal a BP, the Middle Holocene 8200-4200 BP cal a BP and the Late Holocene 4200 cal a BP to present (following Walker et al., 2019).
The coastal marine sediments studied at Luce Bay are considered evidence of former RSLs, and the methodology used in the present investigation follows that used in previous work undertaken by the authors at similar locations in South West Scotland (e.g. Smith et al., 2003aSmith et al., , 2003bSmith et al., , 2006, to enable comparisons to be drawn. Recently, Hijma et al. (2015) have recommended standard protocols for studies of RSL change. In Table 2, the approach used in the present study in measuring RSL change is compared with that of Hijma et al. (2015). Values for the former levels according to the different approaches are given in Table 3, and graphs of the levels determined in the two approaches are shown in Fig. 11. Note that the values for RSL change above Mean High Water Spring Tides (MHWS) in the method used here are closely comparable to the values for RSL change above Reference Water Level in the Hijma et al. (2015) method.
Sea level index points (SLIP) and limiting points (SLLP) in the present work are from former saltmarsh (locally termed merse) and mudflat environments. The error range is based on the measurement of contemporary features and follows the approach of Smith et al (2003a). Full details of each borehole from which samples were taken are given in Supporting Information (tables prefixed by S in this text). In determining regional patterns of glacio-isostatic uplift involving estuarine areas, tidal effects can result in an increase in elevation of comparable index points towards the head of an estuary. In the Solway Firth, altitudes of both mudflat inner margin and saltmarsh inner margin clearly exhibit a rise with the increase in MHWS and Mean High Water Neap Tides (MHWN) towards the head of the estuary. In this area, had OD alone been considered, the tidal factors could have masked or modified evidence of glacio-isostatic or other crustal effects. Figure 2 illustrates the change in tidal parameters from four tidal stations compared with mudflat and saltmarsh heights along the northern shore of the Solway Firth, from Luce Bay in the SW to Redkirk Point in the NE, based on 188 mudflat and 475 saltmarsh inner margin altitudes, taken at 50-m intervals using data for Luce Bay and adjacent areas (Smith et al., 2003a(Smith et al., , 2003b and new data obtained for this study). These observations support the choice of MHWS for comparative studies of RSL change in Scotland by the authors (e.g. Smith et al., 2000). In the same way, reference to the Indicative Meaning as outlined by Hijma et al. (2015) also provides the opportunity to take account of spatial variations in tidal and other effects.

Morphology and stratigraphy
The area studied contains widespread evidence of deglaciation. During the Last Glacial Maximum (LGM), the area was occupied by an ice sheet flowing broadly southward or southsouth-westward into the Irish Sea basin (e.g. McMillan et al., 2011). Decay of this ice sheet left a landscape of extensive glacial deposits. Recent work has indicated that a Late Devensian readvance occurred in the Luce Bay area (McCabe et al., 1998(McCabe et al., , 2005McCabe and Dunlop, 2006), correlated with the Killard Point Stadial, dated at between 17 000 and 18 200 cal a BP. McCabe et al. (1998McCabe et al. ( , 2005 correlated this readvance with Heinrich event I. More recently, BRITICE (Clark et al., 2017) mapping has provided information on patterns of ice flow across the area.
In Fig. 1 the morphological features discussed and the location of the stratigraphical evidence examined are shown. West of Dunragit and landward of Torrs Warren a broad lowland (K) surrounds the Piltanton Burn. To the north, dead-ice hollows (A) and outwash surfaces (B) surround proglacial meltwater channels (C, D, E). Dissected fragments of a surface of sand and gravel at F and G reach consistent altitudes of 15-20 m OD and may indicate a former RSL into which the meltwaters deployed, although exposures in the surface at F and in the extensive gravel pits at I are largely of outwash gravels, the latter exhibiting southward-dipping bedding, with rapid changes in the dominant particle size. The lower reaches of the channels lead into a broad lowland surrounding the Piltanton Burn. The lowland (K) is occupied by silty fine sand, up to 8 m thick in places according to boreholes undertaken by the authors and commercial boreholes, sometimes with silty clay. Along the margins of the lowland, the sand surface is covered in a thin veneer of peat except in the south, where greater thicknesses of peat, in places up to 2 m, occur (Newell in Cowie, 1996 and commercial boreholes). Survey of the sand surface (Fig. 3) at over 200 levelled points discloses altitudes which rise from ca. 6 m OD (2.6 m MHWS) below Piltanton Bridge, to over 10 m OD (6.6 m MHWS) where the Soulseat and Piltanton Burns enter the lowland. The lowland is lowest in the centre and downstream beyond Piltanton Bridge, but rises along the lower reaches of the burns draining into it. The lowland acted as a sediment sink for blown sand as Torrs Warren dunes formed.
East of Dunragit the landscape is one of glaciofluvial deposits (H) locally mantled by blown sand and overlooking a basin area in the west, the Whitecrook Basin (M). Along the northern side of the basin at L, a terrace in silty clay occurs at 8.5-9.5 m OD. To the south, the basin opens out to a terrace in sands and silts along the Piltanton Burn at 6-6.5 m OD, continuing the lowest level in the Piltanton Burn lowland beyond Piltanton Bridge. The Droughduil Motte, a late  Table 2. Comparison between the approaches of the authors and that recommended by Hijma et al. (2015), with the error margins upon which the graphs in Fig. 11  Half the thickness of the 'sample' (assumed here to be half the thickness of the layer). Sample described as 'organic and mud-rich'. In Fig. 11B, the values for sediment consolidation used in the present paper are used. Not estimated.

Indicative range error
High marsh and underlying mudflat surface range in the present paper is used (see above).
Half the elevational range of an indicator. The indicative range at Luce Bay is the range between HAT and MHW, and thus 1.4 m. The indicative meaninghalf the indicative rangeis ± 0.70 m Neolithic artificial mound (Thomas, 2015), underlain by blown sand, overlooks the Whitecrook Basin at the southern end of a ridge (J) extending south from Dunragit. The basin is drained by a narrow incised channel to the Piltanton Burn. More than 120 commercial boreholes, supplemented by our boreholes, record the stratigraphy beneath the Whitecrook Basin. Till widely occurs at the base, overlain by glaciofluvial sands and gravels. Above the sands and gravels, peat, up to 0.6 m thick, is sometimes present, overlain at the margins of the area by dune sand, or by colluvium. Away from the margins, the peat is overlain by up to 0.75 m of clay or silt and then by up to 1.5 m of sands and gravels, which are in turn overlain by one or more horizons of clay or silt, sometimes laminated, which reach a maximum recorded altitude of 10.37 m OD. Dune sand may overlie the highest clay or silt layer, although in some locations the clay or silt is present at the surface (as in terrace L).
A borehole transect ( Fig. 4) in the Whitecrook Basin discloses details of the stratigraphy. Commercial borehole NX15NE1031/ 01, in the west, reached to over 20.42 m depth (reaching − 11.69 m OD), disclosing at the base fine to coarse gravel and sand overlain at −9.17 m OD by laminated clay with horizons of fine sand then at 5.01 m OD by structureless sands and silts to the surface. Eastwards, boreholes undertaken in the present study (prefix W, Fig. 4) reveal horizons of peat and in borehole W8 peat horizons occur at 4.62 m OD and 7.5 m OD, both sampled for radiocarbon dating (see Table 3 and Table S6). Commercial borehole transects elsewhere in this area are interpreted as disclosing coastal barriers with marine/estuarine deposits in the intervening swales. The alignment of each barrier is not clear from available borehole locations. The exceptional depth of silty clay in the Whitecrook Basin may be because this area was in a low-energy environment, protected from high-energy coastal processes by barrier development to the south-east, thus receiving fine sediment during Early-Middle Holocene RSL rise.

Holocene relative sea levels
Evidence for RSLs at Luce Bay is largely provided by stratigraphy. The sites studied are indicated in Fig. 1: Mahaar and Barsolus, in the Soulseat Burn valley at the head of the Piltanton Burn Lowland; Kirminnoch and Mineral Plantation, in a proglacial meltwater channel at the centre of the Piltanton Burn Lowland; and Whitecrook Basin (Including No. 2 Holdings), east of Dunragit.
At Mahaar (Figs 1, 4 and 5), in the peat-filled valley of the Soulseat Burn, sand with some gravel at the base is overlain by peat above which there is a horizon of silts and sands. Within the silts and sands a noticeable horizon of peat is found, while the silts and sands are in turn overlain by surface peat. The detailed stratigraphy at the sampled borehole, 17a, is shown in Table S1. The sediments at Mahaar yielded mixed brackish-marine to brackish-freshwater diatom assemblages (Fig. 5). The assemblages were sparse (there were no diatoms present between 8.50 and 8.82 m OD and counts elsewhere were to 100 total valves) but show a clear brackish-marine component dominated by the benthic Diploneis interrupta and Caloneis westii and a constant brackish-freshwater component of Fragilaria sp. Mahaar represents the most extreme landward extent of estuarine sediments in the Piltanton Burn lowland. The sediment and diatom records and three AMS radiocarbon dates (MA1-MA3, Table 3) indicate that RSL began to rise as peat was replaced by silt at 7609-7729 cal (6839 ± 30 14 C) a BP. RSL then fell as peat developed upon the silt surface. This peat was then replaced by silt at 6797-6987 cal (6052 ± 30 14 C) a BP, although the lack of any diatom assemblage at this point cannot confirm that this is an RSL index point. Nevertheless the presence of brackish-marine to brackish-freshwater diatoms higher in this same silt layer supports it as being of estuarine origin. RSL subsequently fell as peat began to accumulate on this silt at 5296-5461 cal (4618 ± 30 14 C) a BP.
At Barsolus (Figs 1, 4, 6 and 7), 2 km downstream from Mahaar, sediment stratigraphies disclose peat overlain by laminated silt and sandy silt in turn overlain by peat to the surface (Fig. 4). Diatom analyses (Figs 6 and 7) were undertaken at boreholes 1 and 5 (Tables S2 and S3). Radiocarbon dates (BA1-BA3) are listed in Table 3. From borehole 1 at 4.85 m depth where the laminated silt overlies peat, the peat yielded a mixed brackish to freshwater assemblage of Nitzschia navicularis, D. interrupta, Fragilarifroma virescens and Pinnularia intermedia, while overlying laminated silt yielded a marine to brackish assemblage with Paralia sulcata, Podosira montagnei, D. smithii, N. navicularis and D. interrupta. A gradual change in environmental conditions from freshwater to marine/estuarine is indicated, and a date of 8602-8973 cal (7910 ± 30 14 C) a BP was obtained from the top of the peat. Higher in the stratigraphy at 1.5 m depth in borehole 5, a date of 7980-8160 cal (7240 ± 35 14 C) a BP was obtained from the top of peat beneath laminated silt (no diatoms were present). The laminated silt and sandy silt is gradually overlain by herb-rich peat, the minerogenic sediments containing mostly freshwater and brackish-fresh diatoms but with small amounts of marine and marine-brackish diatoms, changing to predominantly fresh and brackish-fresh diatoms in the peat, dominated by Cyclostephanos dubius and      At Kirminnoch (Figs 1, 4 and 8) the valley of a former proglacial meltwater channel is filled with a tapering wedge of silty clay and sand up-valley within peat. Basal fine sand overlain by humified peat occurs with horizons of silty clay, sometimes organic, above which peat occurs to the surface. At its furthest up-valley, the silty clay splits into two horizons. Borehole 2 is representative of the sediment stratigraphy (Table S4): stratigraphic boundaries are all gradual. At borehole 2, diatom analyses (Fig. 8) show freshwater species dominated by Aulacoseria perlabra florinae in the basal peat, but this becomes brackish-marine as indicated by D. interrupta sp., some 80% of the assemblage, a shallow water brackish benthic community as the peat becomes silty, and then to a marine-brackish fauna in the silty clay itself, with D. didyma and D. stroemii comprising 40% of the assemblage, and an increased marine influence is apparent from shallow/ benthic species. The overlying peat becomes freshwater with a well-mixed freshwater assemblage. A radiocarbon date (KH1 ,  Table 3) of 7820-7975 cal (7110 ± 30 14 C) a BP was obtained from peat at the basal silty clay contact and a second date (KH2) of 6305-6410 cal (5640 ± 30 14 C) a BP from peat at the upper silty clay contact.
At Mineral Plantation (Figs 1, 4 and 9), down-valley from Kirminnoch and at the mouth of the channel, organic sediment overlies the silty sand of a terrace which leads into the lowest sandy surface of the Piltanton Burn Lowland. Table S5 lists the detailed stratigraphy at the section examined. A radiocarbon date of 2313-2358 cal (2320 ± 25 14 C) a BP (MP, Table 3) was obtained from the base of the peat at 0.885-0.89 m. Diatom analyses (Fig. 9) show a transition from marine and brackish conditions to that of a freshwater system, possibly a ponded area. A well-developed marine to brackish diatom assemblage typifies the basal sediments with Paralia sulcata and D. interrupta dominating. The presence of the tychopegalic P. sulcata suggests a well-established connection to open water for this marine species to occur in such abundance, as well as having continued freshwater input via the burn, evidenced from the abundance of the freshwater-brackish species F. pinnata. The marine-brackish diatom assemblage continues throughout the banded peat/minerogenic sediment until the peat occurs, when planktonic Cyclotella sp. show the freshwater system becoming dominant, with the increased presence of the many benthic and tychopelagic Fragilaria species showing that this environment was either somewhat ephemeral as a freshwater body or a series of freshwater pools existed with numerous shallow environments.
In the Whitecrook Basin (Figs 1, 4 and 10) close to the Droughduil Burn, a borehole transect between commercial boreholes NXNE1031/1, NX1031/1a and NX1031/8 at the mouth of the embayment below the Droughduil motte is typified by the stratigraphy at borehole 8 (Table S6). Above basal minerogenic deposits, the uppermost 3 m contains a sequence of peat or organic-rich sand, overlain by well-sorted fine-medium sand which becomes structureless, in turn overlain by peat with overlying structureless sands and silts. The top 1 cm of basal organic material was dated at 8184-8332 cal (7430 ± 30 14 C) a BP (WB1 , Table 3) and contained a diatom assemblage, largely brackish-marine in origin, with Caloneis westii, C. sublittoralis, Diploneis didyma, D. smithii, Navicula marina, N. digitoradiata and Paralia sulcata, but also Fragilaria constuens var. venter, F. pinnata and F. ulna, indicating the proximity of the site to drainage from land. The basal 1 cm of peat overlying the structureless sands was dated at 4249-4499 cal (3930 ± 30 14 C) a BP (WB2 ,  Table 3) and the diatom assemblage (Fig. 10) is similarly mixed, although predominantly brackish/marine, with brackish/marine species Caloneis westii, Cocconeis scutellum, C. sublittoralis, Diploneis interrupta, Diploneis smithii and Paralia sulcata. Freshwater species included Pinnularia viridis, P. major, Fragilaria pinnata and F. ulna. The diatom assemblages at both dated horizons indicate transitional marine/estuarine environments.
At No. 2 Holdings (Figs 1 and 4), at the head of Whitecrook Basin, a sample of organic-rich silt overlain by a thin horizon of sand at the surface was dated at 3649-3858 cal (3497 ± 33 14 C) a BP (P, Table 3). The date relates to the commencement of deposition of dune sand at the locality. No diatoms were recovered from the silt, and the limiting date obtained registers the absence of marine/estuarine sedimentation on the valley side above the Whitecrook Basin site.
RSL graphs based on the data described above and according to the methodology upon which the present study was based with the equivalent values from Hijma et al. (2015) and listed in Table 3 are shown in Fig. 11. At the sites examined, the regressive contact dated is probably close to the former mudflat surface, while the transgressive contact is probably close to the former saltmarsh (merse) surface.
RSL at Luce Bay was rising by the Early-Middle Holocene. In the Piltanton Burn lowland, RSL rose from 4.12 to 5.89 m OD The highest recorded Holocene RSL in the area is recorded at Kirminnoch, where RSL had reached 9.23-11.16 m OD (5.89-7.76 m MHWS) by 6305-6410 cal (5640 ± 30 14 C) a BP. At Barsolus, the rise reached between 6.52-8.40 m OD (3.12-5.00 m MHWS) (the early date) and 7.73-9.53 m OD (4.33-6.13 m MHWS) (the last date). At Mahaar, where the dates are only broadly supported by diatom evidence, RSL was rising at 7610-7731 cal (6839 ± 30 14 C) a BP at 7.55-9.30 m OD (4.98-5.90 m MHWS) but a possible regressive contact, not dated, may occur before the rise was resumed by the limiting date of 6798-6990 cal (6052 ± 30 14 C) a BP at 7.91-9.64 m OD (4.51-6.19 m MHWS), before falling again by 5298-5642 cal (4618 ± 30 14 C) a BP at 8.27-10.93 m OD (4.87-6.02 m MHWS). To the east, at Whitecrook Basin, the rise had ended by 4300-4560 cal (3930 ± 30 14 C) a BP when RSL fell from 6.66 to 8.36 m OD (3.26-4.96 m MHWS), but unlike the sites at Kirminnoch, Barsolus and Mahaar, the borehole at Whitecrook Basin does not lie on a continuously tapering wedge of sediment, and the local limit of the fine silt and sand is unclear. It is possible that the stratigraphy at No. 2 Holdings, up-slope from Whitecrook Basin, marks the limit of the sediments dated at Whitecrook Basin and the value of 8.80-9.72 m OD (4.60-6.32 m MHWS), dated at 3649-3858 cal (3497 ± 33 14 C) a BP at No. 2 Holdings marks a limiting value for the altitude of RSL at Whitecrook Basin.
The evidence from Luce Bay indicates that the rise of RSL between the transgressive contacts at Barsolus and Whitecrook Basin and those at Kirminnoch and Mahaar amounted to over 5 m between ca. 8600 and ca. 7800 BP cal a BP at a mean rate of 13.8 mm a −1 or 14.2 mm a −1 if Mahaar (where the diatom record is sparse) is excluded. While the rapidity of the rise at this time may have been due in part to the rapid rises recorded by Lawrence et al. (2016) between ca. 8800 and 8200 cal a BP in the Cree valley to the east, this cannot be confirmed in the absence of stratigraphic evidence at Luce Bay for such rises. The RSL rise at Luce Bay was greatly reduced to ca. 1 m between ca. 7800 and ca. 6300 cal a BP, at a mean rate of 0.66 mm a −1 . Subsequently, RSL fell by 2-3 m before 4400 cal a BP. Hence RSL had changed by only ca. 2-3 m for over 3000 years. RSL later fell to ultimately reach present levels, although the fall may have been interrupted around 2500 cal a BP.

Comparison with adjacent areas
RSL changes at Luce Bay can be compared with those from estuarine sites around the northern Irish Sea ( Fig. 1; Table 4).
At the nearby sites of the Cree valley, Girvan, Woodgrange, Newbie Cottages and possibly the Nith valley, two or more episodes of rise then fall in RSL are identified, and it is maintained that a fluctuating rise in the Middle Holocene is very likely, even ignoring the possible fluctuation indicated by the stratigraphy but not dated at Mahaar (see above).
A period of relatively little change in RSL during the Middle Holocene was argued for the Arisaig area by Shennan et al. (2005, p. 104), who explained it as having been caused by changes in global sea level reflecting a gradual, rather than abrupt, cessation of the melting of the Antarctic and Laurentide ice sheets. An alternative interpretation may be derived if Middle Holocene RSL change varied around Scottish coasts, producing displaced shorelines, as has been argued for many years, and the evidence for such shorelines has been widely articulated (e.g. Smith et al., 2012Smith et al., , 2019. In Fig. 12, isobases derived from Gaussian trend-surface modelling, based on over 450 shoreline altitude measurements, each measurement surveyed to an accuracy of < ±0.05 m for four Middle Holocene displaced shorelines, named the Storegga, Main Postglacial, Blairdrummond and Wigtown shorelines (Smith et al., 2012(Smith et al., , 2019 are shown. Individually produced, these models are similar in outline and in the location of their centres, thus showing little change in the spatial pattern of uplift since the shorelines were reached (for an example of two of the models produced individually, see Smith et al., 2012: fig. 6). However, mindful of the fact that the Main Postglacial, Blairdrummond and Wigtown shorelines are diachronous, these models were combined by Smith et al. (2012) with the virtually instantaneously reached (and therefore not diachronous) Storegga shoreline (based upon survey of the demonstrable limit of the Holocene Storegga Slide tsunami) around a common centre and axis, showing little change. In Fig. 13, the shoreline-based isobase models for the Main Postglacial, Blairdrummond and Wigtown shorelines shown in Fig. 12  Coastal change and prehistoric archaeology at Luce Bay

Holocene coastal processes
Holocene coastal change at the head of Luce Bay was driven by the availability of sediment from the decay of the last ice sheet (McMillan et al., 2011) and by the wind climate acting with changes in RSL to redistribute that sediment. As RSL at the head of Luce Bay rose in the Early-Middle Holocene, gravel barriers developed. Sediment for the barriers was derived from coastal areas both east and west of the Bay. At present, flood tide currents at the head of Luce Bay are clockwise and on the ebb tide anticlockwise. If the same situation applied during the Middle-Late Holocene, on the flood tide glacial outwash sediments at the western side of the bay would have been carried eastwards, while on the ebb tide fluvial and glaciofluvial sediments were carried westwards from the Water of Luce delta. Along the coastline east of Dunragit, these opposing currents together enabled barriers to form. At the same time, the wide, shallow Luce Bay, with its large tidal range, provided sediment for winds to blow onshore and create dune fields across the barriers at Torrs Warren and across barriers and glacial sediments east of Dunragit. Today, winds at West Freugh Weather Station (for location, see Fig. 1) are predominantly from the south or north-west. If these wind directions pertained during the Holocene, the southerly wind would have brought sand onshore from Luce Bay to build up the dune systems, while the north-westerly winds would have been responsible for the development of blowthroughs in exposed parts of the systems. However, the availability of sediment and the wind climate alone could not have produced the coastal landscape of today. Changes in RSL were crucial to coastal change in the area.

Dunes and archaeology
South-west of the Piltanton Burn estuary, the Torrs Warren dune field stretches 8 km from Sandhead in the south-west to Ringdoo Point in the north-east (see Fig. 1 above). The Torrs Warren dunes rise in altitude landward towards their inland margin where they reach over 28 m OD: a LIDAR map of part of the Torrs Warren dune field illustrates the main features of these dunes (Fig. 14), in which the main dune field is marked to seaward by distinctive linear features. Single and Hansom (1994) identified beneath the dunes a series of gravel ridges which can be followed intermittently towards the north-east, where they reach 12.5 m OD (9.1 m MHWS) in commercial boreholes NX15SW4025/29 and NX15SW4025/32. While it is not possible to demonstrate that these ridges are continuous, their presence over the area between Clayshant and Mid Torrs (Fig. 1) supports their probable continuity as far as the present estuary of the Piltanton Burn. The archaeology of these dunes provides some indication of their age. While the locations provided in early studies are imprecise, and many finds are portable, it is evident that the older archaeological sites are found towards the landward margin of the Torrs Warren dune field. The earliest radiocarbon-or pottery-dated archaeological site in them, on a 'reasonably stable soil surface' in the high  (Cowie, 1996;Coles et al., 2011) and Mesolithic micrioliths have been reported (Coles, 1964;Coles et al., 2011). These sites lie at unknown depths on earlier deposited sand, as do those containing early Neolithic Carinated Bowl assemblages on Flint Howe (Cowie, 1996). Sand blow across the area is recorded well to the west of High Torrs, in the site at Mye Plantation (Mann, 1903(Mann, , 1905a(Mann, , 1905b. Continued dune development is evident in the formation of linear dunes near the coast, where peat in a dune slack at NX132.541 at ca. 6 m OD provided radiocarbon dates of 1181-1614 cal (1480 ± 110 14 C) a BP at 0.993-1.10 m depth and 2517-3329 14 C (2780 ± 130) a BP at its base at 2.215-2.3 m depth (Newell in Cowie, 1996).
North of the Piltanton Burn estuary, blown sand is widespread. Excavations along the route of the A75 east of Dunragit ( Fig. 1, above) have disclosed blown sand at ca. 10-12-m OD beneath marine sediments and optically stimulated luminscence (OSL) dated at 5100 ± 650 BP (Creswell et al., 2019). At Whitecrook, birch charcoal at 0.17 m depth below a surface at ca. 10 m OD from a palisaded enclosure in blown sand was radiocarbon dated to 5340-5000 cal (4455 ± 40) a BP (Gordon, 2009). At Droughduil Mote, the base of which is ca. 10 m OD, and which probably rests on blown sand, an OSL date from the base yielded 4520 ± 250 BP (Thomas, 2015).
The archaeological evidence at Luce Bay indicates that sand blow and dune development began at least as early as the Neolithic, probably in the Mesolithic, and thus before ca. 6000 BP. The linear dunes at Torrs Warren were forming by ca. 3000 BP, perhaps representing a change in the pattern of sediment accumulation. The altitudes of available dates for archaeological sites in the Torrs Warren dune field are plotted in Fig. 11. They record the occupation of the area as RSL fell.

Coastal change
Evidence from sites studied for RSL in this paper show that the rise in RSL in the Early Holocene at Luce Bay reached its maximum at ca. 6300 cal a BP. The similarity in height of the gravel barriers south-west of Genoch Mains with the limit of silty clay at Mahaar, Barsolus and Kirminnoch (Fig. 1) implies that the lagoon along the Piltanton Burn (K, Fig. 1) formed at that time. The lagoon became increasingly isolated from the Bay as the barriers in the west reached higher outwash terraces at Genoch Mains and the only outlet became the present estuary of the Piltanton Burn at Piltanton Bridge. Figure 15 shows the inferred maximum Holocene RSL in the area, and is derived from the altitude of the base of surface peat which had accumulated without apparent lacunae over demonstrably estuarine sediment at Kirminnoch, the surface of the sediment being at 9.23-11.16 m OD (5.89-7.76 m MHWS) at 6305-6410 cal a BP. At this time, the lagoon along the lower reaches of the Piltanton Burn (see Fig. 3) extended southward to the west of Torrs Warren. Here, the extensive peat moss is underlain by a surface of silty clay at 7.90-9.90 m OD according to boreholes from the British Geological Survey database, the altitudes of those boreholes determined in the present work by Lidar. Newell (in Cowie, 1996) recorded a surface of clay at 8 m OD beneath 2.5 m of peat in the same area. At Mye, towards the south-west of this area (Fig. 15), Mann (1903, 1905a, 1905b) described the excavation of sunken anthropogenic structures in a surface of sand and shallow peat of around 15 m OD, probably a few metres above the edge of the main peat area. A report of the excavation, which was some 2-2.5 m in depth, disclosed peat at the surface, with sand (possibly blown sand) and blue clay beneath. Inserted into the clay were concentric rings of inverted faceted stakes of Alnus, Betula and Corylus, one Alnus sample subsequently dated to 4185-4505 cal (3193 ± 39 14 C) a BP (Sheridan, 2002). The purpose of the structures is unknown but to function at all, sea level controlling the groundwater level had to be below 13-14 m OD. Along the seaward edge of the area of peat around Mye, boreholes in the British Geological Survey database support the observation of Mann (1903Mann ( , 1905aMann ( , 1905b) that blown sand partially overlies the silty clay along its valley side margins, indicating that ultimately the sea withdrew. At Kirminnoch, Mahaar and Whitecrook Basin RSL fell by ca. 2-3 m from before ca. 7000 to ca. 4400 cal a BP. However, while the fall in RSL resulted in the lagoon along the Piltanton Burn and its extension to the south around Mye shrinking in extent, elsewhere the shoreline position changed little. The slender peninsula upon which the Droughduil Motte now stands may well have been briefly inundated by exceptionally high tide levels, but remained essentially a fixed feature on the coast. Bradley et al. (2016) have argued that such fixed points on the coast, protected by gravel bars, served as trading places, 'maritime havens', for sea-going communities in the Neolithic period. Luce Bay is one of their examples, one of the most productive areas for archaeological finds in Scotland (Cowie, 1996). Bradley et al. (2016)  More spatially detailed reconstruction of RSL data is critical in understanding the precise relationship between specific monuments and coastal change in Luce Bay. The base of the Droughduil motte is at 10 m OD (6.6 m MHWS). It was positioned south of and apart from the major later Neolithic monumental complex at Dunragit at the west end of the Whitecrook Basin. It is dated by OSL to the later Neolithic (4470 ± 270 cal a BP: Thomas, 2015, Figure 1). Our reconstruction of the maximum position of RSL in the Holocene shows that during the late Neolithic the motte was joined to the mainland by a narrow (in places <100 m wide) neck of land to the north (Fig. 15) which may have lain perhaps no more than 2 m above the level of the highest tides, and thus 'magically' appearing as a separate island (Tipping et al., 2015). The spatial pattern of later prehistoric settlements surrounding the Piltanton Burn lowland (Fig. 3) might also be related to the maintenance of high groundwater tables long after 6000 BP. Later prehistoric archaeological sites, including the only excavated site at Fox Plantation, dated to 3990-4410 cal a BP (MacGregor et al., unpublished) are not only above the valley floor but are confined to high ground to the north and west of Piltanton Burn, ringing the lowland (Fig. 3)  crop-marked archaeological site, a sub-rectangular, probably Iron Age but unexcavated enclosure, is found on the valley floor (Historic Environment Scotland. Canmore Database 79044).
In the east of the Piltanton Burn lowland, dating of organic material overlying the silty sand at Mineral Plantation at 2313-2358 cal (2320 ± 25 14 C) a BP may mark a reduction in the rate of RSL fall. The lowland below ca. 6 m OD bears the hallmarks of relatively recent flooding, with abandoned meanders of the Piltanton Burn visible through much of the area. The river was straightened and embanked before the first Ordnance Survey map of the area (1843), and such changes would have steepened its course: before these changes, the river would have had a gentler gradient and thus been subject to flooding both from the estuary and from high river flow. Since installation in 1988, the Barsolus River Gauge recorded a maximum level of 8.97 m OD (Scottish Environment Protection Agency), higher than the surface of most of the lowland, and combined with the highest astronomical tide at the Luce Bay Platform of 4.1 m OD, it is likely that the lower part of the lowland would have been inundated occasionally, especially before the changes made to the channel of the Piltanton Burn.

Conclusions
Middle to Late Holocene coastal changes at Luce Bay were marked by the emplacement of barrier systems across which dune fields developed as RSLs rose and fell. Morphologically, the area is marked by the Torrs Warren dunes and the remarkable sediment sink in the Piltanton Burn lowland to landward, explained by the abundance of sediment and the large tidal range in the Bay, exposing extensive sandflats. Stratigraphically, the RSL record is notable for the 3000-year period of relatively little change in the Middle Holocene, reflected here in the convergence of two Holocene shorelines. These coastal changes have had important impacts on later prehistoric occupation of the area. A relatively stable coastline during the Middle Holocene, with a prominent peninsula at Dunragit, influenced the development of a remarkable ceremonial complex and associated dwellings. In turn, coastal archaeology can inform and constrain contemporary environmental conditions induced by RSL change. Luce Bay is an area where coastal change and archaeology are closely linked and where knowledge of each contributes to the other in an understanding of the changing coastal environment.

Supporting information
Additional supporting information may be found in the online version of this article at the publisher's web-site.   Table 3