Situated between Loch Torridon to the south and Loch Gairloch to the north (Fig. 1), the Redpoint Peninsula comprises a low-relief glaciated landscape of shallow lake basins and ice-moulded bedrock hills largely underlain by arkosic sandstones and conglomerates of the Neoproterozoic Torridonian Group. Today, the Redpoint Peninsula is mantled by extensive blanket-peat cover and thin till-based soils. Climatically, the site's coastal setting means it experiences strongly maritime conditions, with abundant year-round rainfall (∼ 1450 mm), relatively mild temperatures (mean annual max. temp. 12 °C), and limited thermal seasonality (∼ 6 °C) (ECMWF Data). According to regional reconstructions, Late Pleistocene ice-flow across the peninsula would have been in a predominantly westerly to north-westerly direction (Charlesworth, 1956; Hughes et al., 2014; Merritt et al., 2019), with the most recent valley glaciers nourished in the high-relief Torridon mountains south-east of the peninsula (Sissons, 1977; Bickerdike et al., 2016). The most conspicuous and best-studied glacial deposits in the vicinity are those comprising the Redpoint moraine system, which can be traced for ∼ 12 km from the northern slope of Beinn Bhreac in the Shieldaig Forest to the coast at Opinan (Figs. 2 and 3). Together, this system of moraines describes an ice lobe flowing north-west out of the mountainous interior and discharging into Loch Gairloch. First described by Robinson and Ballantyne (1979), the Redpoint moraines occupy a stratigraphic position between the offshore LGM limits of the ice sheet (see Stoker et al., 1994; Bradwell and Stoker, 2015) and those of a small late-glacial icefield that accumulated in the Torridon uplands during the Loch Lomond Readvance (LLR) (Sissons, 1977; Bickerdike et al., 2016). Accordingly, the Redpoint moraines are interpreted as reflecting an episode of stabilisation or readvance of the decaying ice sheet during Termination 1 that is known as the Wester Ross Readvance (Robinson and Ballantyne, 1979; Sissons and Dawson, 1981; Sutherland, 1984).

Correlating the Redpoint moraines with stratigraphically and morphologically similar deposits on the neighbouring Applecross, Gairloch, and Aultbea Peninsulas (Fig. 1), Robinson and Ballantyne (1979) described a resurgent ice margin ∼ 30 km in length and concluded that the Wester Ross Readvance was of at least regional significance. Subsequent studies have sought to establish both the spatial continuation of the complex (Sissons and Dawson, 1981; Sutherland, 1984; Ballantyne et al., 1987; Bradwell et al., 2008) and, crucially, its age. In their original paper, Robinson and Ballantyne (1979) cited a radiocarbon date of 12 810 ± 155 ¹⁴C years (15.3 ± 0.3 cal ka BP (before 1950): Q. 457) on post-glacial sediments at Loch Droma, approximately 40 km inland of the mapped Wester Ross Readvance limits (Kirk and Godwin, 1963), as minimum-limiting age control for the readvance. More recently, Everest et al. (2006) reported six ¹⁰Be surface-exposure ages from boulders on Gairloch sector of the Wester Ross Readvance moraine system; pruned of three outliers, the remaining three ages gave a mean of 16.3 ± 1.6 ka (1σ) with St scaling and assuming zero erosion (Everest et al., 2006). In contrast, Ballantyne et al. (2009) reported 14 ¹⁰Be ages from Wester Ross Readvance moraines spanning the Applecross Peninsula to Achiltibuie (Fig. 1) and reported mean ages of 13.5 ± 1.2 ka (Lm scaling, zero erosion) and 14.0 ± 1.7 ka (De scaling; Desilets et al., 2006; 1 mm kyr⁻¹ erosion). Citing broad agreement between their data and eight ¹⁰Be ages published by Bradwell et al. (2008) from apparently correlative landforms, Ballantyne et al. (2009) attributed the Wester Ross Readvance to the “Older Dryas” event (∼ 13.9–13.7 ka: Björck et al., 1998). Subsequent revisions of the Bradwell et al. (2008) and Ballantyne et al. (2009) data using lower production rates have produced progressively older ages for the readvance: Ballantyne and Stone (2012) reported revised mean ages of 14.3 ± 0.1 ka (1σ) to 15.1 ± 0.1 ka (1σ), reflecting a range of CRONUS-Earth production rates, while Ballantyne and Small (2019) derived a mean of 15.3 ± 0.7 ka (1σ) via an unpublished production rate value attributed to Fabel et al. (2012).

Approximately 2.5 km inside the Redpoint moraines mapped by Robinson and Ballantyne (1979), minimum-limiting age control for deglaciation from the Wester Ross Readvance limits is afforded by the Loch Bad na h-Achlaise sedimentary record (Simms et al., 2022) (Fig. 2). Part of a broader investigation of post-glacial relative sea-level and environmental change (Simms et al., 2022; Taylor et al., 2026), that study reported seven stratigraphically consistent ¹⁴C ages from core LBA18-11 (57.695° N, 5.741° W; Table S1 in the Supplement), which was extracted from the southern margin of the lake (∼ 13 m a.s.l.). The lowermost two ¹⁴C ages record organic sedimentation in the Loch Bad na h-Achlaise basin following the transition from marine to freshwater conditions (Simms et al., 2022): a ¹⁴C date of 13 185 ± 45 ¹⁴C yr BP from 530 cm depth gives a calibrated age of 15.8 ± 0.1 ka BP (sample 208 396) with IntCal20 (Reimer et al., 2020), while a date of 13 240 ± 50 ¹⁴C yr BP from 536 cm depth gives a calibrated age of 15.9 ± 0.1 ka BP (sample 208 397) (Fig. 2; Table S1). Both AMS measurements were made on plant fibres (Simms et al., 2022), the authors noting that they specifically avoided incorporating material that might introduce hardwater effects (e.g., aquatic algae, fine organic detritus, bulk sediment). Five additional ¹⁴C ages located at shallower depths in that same core exhibit strong stratigraphic consistency. Since organic sedimentation could only begin upon deglaciation, the Loch Bad na h-Achlaise ¹⁴C chronology indicates that coastal portions of the Redpoint Peninsula were ice free by at least ∼ 15.9 ka and that, due to the basin's position proximal to the Redpoint moraines, the Wester Ross Readvance culminated prior to that time. Central to our study, the basal ¹⁴C ages from Loch Bad na h-Achlaise also provide minimum-limiting age constraint for the true exposure ages of the boulders sampled for cosmogenic ¹⁰Be assay, all of which are situated above the ∼ 20 m marine limit reported by Simms et al. (2022).

A growing body of chronological evidence suggests that the Wester Ross Readvance occurred during Heinrich Stadial 1 (∼ 17.8–14.6 ka: NGRIP Members, 2004; WAIS Divide Project Members, 2015), though exactly when, for how long, and under what climate conditions remains unclear. At face value, the published ¹⁰Be moraine dates (Ballantyne and Small, 2019) are marginally younger than the minimum-limiting ¹⁴C ages from Loch Bad na h-Achlaise (Simms et al., 2022). Although relatively minor, this discrepancy might reflect real temporal variability along the length of the Wester Ross Readvance moraine system, or it might stem from methodological factors, such as the choice of ¹⁰Be production rate. In this study, we report new ¹⁰Be ages from the Redpoint Peninsula that document local ice sheet behaviour during Termination 1. Comparison of these ¹⁰Be data with the Loch Bad na h-Achlaise ¹⁴C chronology not only affords an upper bound on possible ¹⁰Be production rate values, it also provides a means for assessing the suitability of currently available ¹⁰Be production rate calibration datasets for calculating surface-exposure ages that are compatible with independent age controls. Finally, our dataset constitutes new constraint on the timing and duration of the Wester Ross Readvance, thereby helping place this event within a wider glaciological context.

We mapped the distribution of moraine ridges on the Redpoint Peninsula from satellite imagery and subsequent ground survey in 2024, using the prior mapping work of Robinson and Ballantyne (1979) as a guide, after which we drafted geomorphic maps onto Landmap imagery (2014) in QGIS (3.22.3) (Fig. 2). During fieldwork, we also identified glacial boulders suitable for ¹⁰Be surface-exposure dating, including boulders perched on bedrock surfaces at Creag Bad na h-Achlaise (Fig. 2b) and Maol Ruadh (Fig. 2c), and boulders embedded in the crest of the outermost Redpoint moraine immediately west of Beinn Bhreac (Fig. 4). The latter sampling site coincides broadly with that of Ballantyne et al. (2009) (see Fig. 2 and Sect. 5.3). Because surface-exposure dating relies on uninterrupted exposure of the rock surface, we targeted boulders in apparently stable positions that exhibit no indication of post-depositional movement or burial (e.g., by snow, peat, or vegetation). All samples comprise the upper few centimetres (< 5 cm depth) of material from the top surfaces of arkosic sandstone boulders and were collected using a hammer and carbide-tipped chisel. Sample location and elevation data were recorded with a differential GPS, while the topographic shielding for each boulder was recorded with a clinometer.

All samples were prepared for ¹⁰Be chemistry in the Palaeoenvironmental Research Unit, University of Galway. Following crushing, we boiled the 250–850 µm fraction in 6M HCl to remove metal oxides and NaOH to remove silica cement, after which quartz was isolated via froth flotation. Quartz material was then purified via successive leaching in 2 % hydrofluoric acid (Kohl and Nishiizumi, 1992) and electromagnetic separation; purity of quartz aliquots was verified by inductively coupled plasma optical emission spectrometry (ICP-OES). Beryllium was isolated via a standardised ion-chromatography methodology at the Irish Cosmogenic Nuclide Facility, University of Galway (samples ANT-24-01-03, 06-09, and 13), and the University of Maine Cosmogenic Isotope Lab (samples ANT-24-10-12). Each sample was spiked with a custom-made, low-background phenakite ⁹Be carrier: samples prepared in Galway received the Dartmouth phenakite PK2 carrier (1005 ± 9 ppm) and those prepared in Maine the Phena7 carrier (940 ± 10 ppm). Carrier concentrations are measured repeatedly by ICP-OES to quantify any evaporation effects over time. Finally, BeO was combined with niobium powder and packed into stainless steel targets for accelerator mass spectrometry. Beryllium ratios were measured over four runs between 2024 and 2026 at the Center for Accelerator Mass Spectrometry (CAMS), Lawrence Livermore National Laboratory, relative to the 07KNSTD3110 standard (¹⁰Be / ⁹Be = 2.85 × 10⁻¹²). Procedural blanks have ¹⁰Be concentrations of 8800–16 000 ¹⁰Be atoms per blank (mean 12,759 ± 3778 atoms), representing ∼ 1 % of the ¹⁰Be concentration of a typical sample. All sample and procedural blank beryllium concentration data are reported in Tables 1 and 2, respectively.

Our assessment of current ¹⁰Be production rates is based on the direct and statistical comparison of apparent surface-exposure ages from the Redpoint Peninsula against the independent minimum age for local deglaciation provided by the basal radiocarbon ages from Loch Bad na h-Achlaise (Simms et al., 2022), situated proximal to the ¹⁰Be sample sites (Fig. 2). Recognising that the Loch Bad na h-Achlaise sedimentary record spans the transition from marine to lacustrine conditions, we used the Northern Hemisphere IntCal20 curve (Reimer et al., 2020) to derive calibrated ages due to strong stratigraphic and biological evidence for all dated organic material being of freshwater (rather than marine) origin. Stratigraphically, basal samples 208 396 (530 cm depth) and 208 397 (536 cm depth) are both derived from green-brown organic muds that are compositionally different from the underlying grey, marine-diatom-dominated minerogenic silts (Simms et al., 2022). The higher of the two samples (208 396) is from a layer characterised by 100 % freshwater diatom species (see Fig. S5 of Simms et al., 2022); the absence of salt-tolerant diatoms and the abundance of halophobe species indicate there was minimal, if any, marine input during deposition of this sedimentary unit (Alexander Simms, personal communication, 21 April 2026). The lower sample (208 397), which is indistinguishable in age from 208 396 (Table S1), is from the same sedimentological unit and also strongly dominated by freshwater diatom species (> 90 %), with only trace numbers of salt-tolerant species. Given these diatom assemblages and the strong stratigraphic and chronologic concordance of the two ¹⁴C ages, we conclude it is highly unlikely that the plant fibres dated by Simms et al. (2022) comprised marine carbon, thus justifying our use of the terrestrial calibration curve.

We employed the “R_Combine” function in OxCal 4.4 (version 177), in conjunction with IntCal20 (Reimer et al., 2020), to erive an average calibrated age of 15.9 ± 0.1 cal ka before 1950 (cal ka BP) for samples 208 396 and 208 397 (Table S1; Fig. S1); this corresponds to a 95 % range of 15.7–16.0 cal ka BP. Adjusted for the 1950-sampling age offset (samples were collected in CE 2024), this gives a conservative minimum age of 15 932 ± 72 years (15.9 ± 0.1 cal ka) for the onset of ¹⁰Be accumulation in all our sampled surfaces. Recent age-depth modelling by Taylor et al. (2026) suggests an earlier (∼ 16.7 ka) onset for sedimentation, and thus ice-free conditions, at Loch Bad na h-Achlaise. However, as that is an extrapolated estimate, and given the potential for variable sedimentation rates, we focus our assessment on the absolute age control afforded by the replicable basal ages of Simms et al. (2022).

For this assessment, we compared apparent ¹⁰Be surface-exposure ages calculated via eight production rate calibration data sets developed for boreal mid-latitude settings: the Rannoch Moor (Putnam et al., 2019), Glen Roy (Small and Fabel, 2015), and Isle of Skye & Highlands (Ballantyne and Stone, 2012; Borchers et al., 2016) sites from Scotland; the Chironico site from Switzerland (Claude et al., 2014); the Mount Billingen site from southern Sweden (Stroeven et al., 2015); the western Norway (Goehring et al., 2012) and north-eastern North America (NENA) (Balco et al., 2009) sites; and the primary production rate calibration dataset of Borchers et al. (2016) (hereafter the “Borchers calibration dataset”), which was developed for global applications. Data for each production rate calibration data set were extracted from the ICE-D: Production Rate Calibration Data online database (https://version2.ice-d.org/production%20rate%20calibration%20data/, last access: 15 March 2026); we note that we did not include any samples identified by the original authors as outliers. A ninth production rate calibration, known as the Loch Lomond production rate (LLPR), is currently used in the British Isles. However, at the time of writing the calibration process underlying the LLPR remains unpublished beyond an early reference to an “in prep” study (Fabel et al., 2012) and so we cannot consider it in our assessment. For each production rate calibration data set listed above, we calculated apparent surface-exposure ages with version 3 (v.3) of the University of Washington (UW) online calculator (https://hess.ess.washington.edu, last access: 15 March 2026), which incorporates site-specific air pressure values derived from the ERA-40 Reanalysis (Uppala et al., 2005). Calculated ages are scaled according to the three standard scaling methods employed therein, specifically, the time-independent (“St”) Lal (1991)/Stone (2000) method and the time-dependent Lal/Stone (“Lm”) and “LSDn” (Lifton et al., 2014) methods. We corrected beryllium concentrations for horizon shielding using the UW online topographic shielding calculator (https://stoneage.ice-d.org/math/skyline/skyline_in.html, last access: 15 March 2026). The UW online calculator employs summary statistics to identify likely outliers, specifically by computing the P-value of χ2 value of a population's mean and incrementally removing samples with the highest χ2 values (relative to that mean) until the population's P-value is > 0.01 (https://sites.google.com/a/bgc.org/v3docs/, last access: 15 March 2026).

The rough appearance of rock surfaces in our study area, including the abundance of quartz pebbles protruding 2–3 cm, indicates that post-depositional weathering has occurred throughout our study area. Given the likelihood of variable (and unquantifiable) erosion rates among the eleven target boulders, and given that the resistant mineral is quartz (i.e., the target mineral for ¹⁰Be analysis), we do not correct nuclide age calculations for erosion. We note that the production rate calibration data sets evaluated here also do not include corrections for erosion. Therefore, any common effects of erosion between the boulders sampled here and those sampled for production-rate calibration are inherently incorporated into the reported surface-exposure age calculations. Similarly, we assume that shielding of boulders by snow has been minimal due to (1) the characteristically low winter snow depths experienced on Scotland's west coast and (2) the exposure of this open landscape to strong winds, which typically preclude significant snow accumulation. We do not apply any correction for post-depositional uplift in our exposure-age or production rate calculations. All apparent surface-exposure ages given in Table 3 and Fig. 5 are reported with internal uncertainties that incorporate the analytical error, propagated with machine background, procedural blank, and boron correction uncertainties (where applicable). Site-specific and cumulative age populations depicted in Fig. 5 are reported with both the standard error of the mean (SEM) and external uncertainty (SEM propagated with the 2.7 % Rannoch Moor production rate uncertainty).

The Redpoint moraine system comprises a suite of lateral ridges that, together, describe successive positions of the left margin of a NW-flowing ice lobe. At its seaward end, the moraine complex descends towards the west immediately north of Loch Airigh Uilleim, becoming indistinct at ∼ 30 m elevation above the settlement of Opinan. At its inland extremity, the outer moraine of the complex abuts the north-west ridge of Beinn Bhreac at ∼ 350 m elevation (Fig. 2), less than 100 m west of a separate and stratigraphically younger moraine complex that descends from the same hillside towards the foot of Loch Gaineamhach (Fig. 2). Individual ridges comprising the Redpoint moraine system range from < 1 to 3 m in relief and are typically mantled with sandstone cobbles and boulders (Fig. 6). The few exposures we observed in the moraine ridges revealed a clast-supported matrix dominated by boulders. Both distal to and proximal to the Redpoint moraines, the land surface is extensively peat-covered. Our Maol Ruadh and Creag Bad na h-Achlaise sampling sites both form prominent topographic highpoints underlain by ice-moulded bedrock and mantled with perched glacial boulders (Fig. 3); Maol Ruadh rises ∼ 15 m above the surrounding landscape, while the summit of Creag Bad na h-Achlaise is ∼ 80 m above Loch Bad na h-Achlaise to the north (Fig. 2).

Among their respective site groups, the measured ¹⁰Be concentrations exhibit a high degree of internal consistency. Four samples from Creag Bad na h-Achlaise have concentrations ranging from 6.95 ± 0.20 (× 10⁴) to 7.30 ± 0.17 (× 10⁴) atoms g⁻¹; three samples from the outer Redpoint moraine crest have concentrations ranging from 8.93 ± 0.18 (× 10⁴) to 9.17 ± 0.19 (× 10⁴) atoms g⁻¹ (Table 1); and two samples from Maol Ruadh, outboard of the moraine, have concentrations of 8.47 ± 0.22 (× 10⁴) and 8.75 ± 0.24 (× 10⁴) atoms g⁻¹ (Table 1). Two samples (ANT-24-08 and 09) returned concentrations that are significantly lower than their neighbours (Table 1) and thus were identified by the UW online calculator as statistical outliers; we do not consider those measurements further in this paper. Treated as surface-exposure ages so as to normalise ¹⁰Be measurements for differences in sample elevation, thickness, and horizon shielding, the nine remaining samples (shown in Fig. 5 calculated with the Rannoch Moor production rate) together form a tight normal age distribution with a low reduced chi-squared value (Fig. 5), confirming a degree of variability that, statistically, can be explained by analytical uncertainty alone. Consequently, we can evaluate the eight ¹⁰Be production rates by comparing the composite Redpoint Peninsula population (n = 9) to the minimum-limiting ¹⁴C basal ages from Loch Bad na h-Achlaise. The uncertainties reported for each ¹⁰Be population discussed in Sect. 4.3 include both the standard error of the mean (SEM) and the external uncertainty, the latter being the SEM propagated with the uncertainty of each respective production rate.

Figure 7a depicts the age-depth profile from Loch Bad na h-Achlaise, based on the ¹⁴C dataset of Simms et al. (2022) and described in Sect. 3.3; the combined calibrated age of samples 208 396 and 208 397 (15.9 ± 0.1 ka) is represented by the vertical shaded bars (Fig. 7) and is taken as minimum-limiting ¹⁴C control for deglaciation. Figure 7b illustrates the population of new Redpoint Peninsula exposure ages (n = 9), calculated using the Rannoch Moor production rate calibration data set and the time-independent “St” scaling, relative to the ¹⁴C control. We employ the St method because it exhibited the closest agreement among reference production rates determined from Rannoch Moor and other calibration sites from around the world based on independent chronologies (Putnam et al., 2019); we note, however, that, owing to the proximity of the Redpoint Peninsula to Rannoch Moor, the three default scaling methods used in v.3 of the UW online calculator give results that agree closely with one another. We employ the same approach in Fig. 8, which demonstrates how apparent exposure ages derived from all eight ¹⁰Be production rates (Sect. 3.3) relate to the ¹⁴C control and to one another. All individual ¹⁰Be exposure ages resulting from the various production rates are reported in Table S2. In this assessment, we deliberately included production rates employed in the circum-North Atlantic region that have been calibrated either (i) against independently dated landforms (blue boxes in Fig. 8) or (ii) against surfaces for which the exposure age has been inferred from indirect sources (pink boxes in Fig. 8). Given its broad application as the default production rate in the UW online calculator, we also assessed the output of the Borchers calibration dataset (black box in Fig. 8). With the Loch Bad na h-Achlaise record as an independent geologic benchmark, we consider a production rate to be in stratigraphic accord with the local stratigraphy (i.e., viable) if the derived ¹⁰Be surface-exposure ages are older than or statistically indistinguishable from (within analytical uncertainties) the minimum-limiting ¹⁴C age for deglaciation (15.9 ± 0.1 ka). Where a production rate results in exposure ages that are young relative to the onset of organic sedimentation, this constitutes a poor fit with the local chrono-stratigraphy and identifies the production rate as unrealistically high.

The sedimentary stratigraphy from Loch Bad na h-Achlaise suggests that the two bottommost ¹⁴C ages, although replicable (Fig. S1), probably underestimate the true age of deglaciation; sufficient time had already elapsed for the basin to have transitioned from a marine to freshwater system via glacio-isostatic rebound (Simms et al., 2022; Taylor et al., 2026) and for vegetation to have migrated into the catchment. Therefore, we take the combined basal date of 15.9 ± 0.1 ka as a minimum-limiting benchmark for locally ice-free conditions. Against this benchmark, two of the eight production rate calibrations evaluated (Rannoch Moor and Chironico) produce mean exposure-age values that overlap with the ¹⁴C age control within 1σ (Fig. 8). Of these, only the Rannoch Moor-derived mean age (16.1 ± 0.1 ka) predates the onset of lacustrine sedimentation in Loch Bad na h-Achlaise (15.9 ± 0.1 ka), suggesting that this production rate gives the closest fit with the Redpoint chrono-stratigraphy (Figs. 7 and 8). The remaining six production rates give mean exposure-age values that are younger than the radiocarbon control and thus underpredict the age of deglaciation at this site by varying durations (Fig. 8). Furthermore, although two of the remaining six exposure-age calculations result in some statistical overlap between the ¹⁰Be ages and the ¹⁴C target when considering the outer limits of respective external uncertainties, that overlap primarily reflects the large external uncertainties associated with those particular calibration datasets (Fig. 8). Use of these rates would still produce mean exposure ages that are significantly and systematically younger than the minimum-limiting age of deglaciation determined by independent means. The results depicted in Fig. 8 are corroborated by Welch's unequal-variance t-test statistics (Table S3): the Rannoch Moor, Chironico, and NENA datasets yield mean values that are indistinguishable statistically from the mean ¹⁴C age, as does the Borchers dataset due to its relatively large uncertainties; in contrast, mean ages derived from the Mt. Billingen, Isle of Skye & Highlands, Glen Roy, and Western Norway datasets are statistically different from the ¹⁴C.

The largest ¹⁰Be / ¹⁴C age offsets arise from the Isle of Skye & Highlands primary dataset (Ballantyne and Stone, 2012; Borchers et al., 2016), the Glen Roy dataset (Small and Fabel, 2015), the Mount Billingen dataset (Stroeven et al., 2015), and the Western Norway dataset (Goehring et al., 2012); these give mean exposure-age values 7 %–8 % younger than the basal ¹⁴C control (Fig. 8). One possible cause of this disagreement is the lack of independent, quantitative age control for the calibration surfaces used in those studies. For example, the Isle of Skye & Highlands primary dataset (Ballantyne and Stone, 2012; Borchers et al., 2016) is calibrated against the presumed age of deglaciation of selected cirques following the Loch Lomond Readvance; those authors cited well-dated shifts in distal palaeoecological and Greenland oxygen isotope records as correlatives for cirque deglaciation, yet no direct age control for cirque deglaciation exists (Ballantyne and Stone, 2012). Similarly, the Glen Roy calibration was determined by correlating the 325 m wave-cut shoreline with a varve sequence located ∼ 25 km distant (Small and Fabel, 2015), for which unequivocal age control is absent (see Putnam et al., 2019), while the western Norway calibration dataset relies upon assumed correlation of the timing of moraine-ridge construction and sedimentation changes in a lake that is not associated with the sampled landform. Finally, the Mount Billingen calibration assumes bedrock erosional features correspond to the indirectly dated drainage of the Baltic Ice Lake (Stroeven et al., 2015).

The sub-optimal performance in our study of production rates calibrated against indirectly dated surfaces (Fig. 8) has ramifications for surface-exposure dating generally. Both the UW (v.3) and CREp (https://crep.otelo.univ-lorraine.fr/#/, last access: 15 March 2026) online calculators employ the Borchers primary dataset as the “global” default; yet that compilation incorporates production rates calibrated against both directly and indirectly dated rock surfaces, including one site we have shown to produce exposure ages that are untenably young (Isle of Skye & Highlands). Although the Borchers primary dataset does technically pass the t-test, as noted above, this is on account of the relatively large total scatter of that calibration dataset (6.9 %); our findings suggest that using the Borchers primary dataset will also result in exposure ages that are too young stratigraphically if not statistically (Fig. 8). Confirming that the latter is not a function of geomagnetic distance from our field site (e.g., scaling effects), Fig. S3 and Table S3 show how exposure ages yielded by independently dated calibration datasets from the Southern Hemisphere (Putnam et al., 2010; Kaplan et al., 2011; Kelly et al., 2015) also agree statistically with the minimum-limiting ¹⁴C control. Indeed, the choice of scaling model does not impact the overall relative performance of the eight production rates evaluated with respect to the ¹⁴C benchmark (Table S2), though we do observe that St generally gives the closer fit. In summary, of the eight production rate datasets assessed, the Rannoch Moor dataset gives the best fit between apparent ¹⁰Be ages and ¹⁴C control on the Redpoint Peninsula.

Recognising the potential lag between deglaciation and deposition of organic-bearing lacustrine sediments (see above), we do not attempt to use the Loch Bad na h-Achlaise radiocarbon data as the basis for calibrating a new absolute ¹⁰Be production rate. Instead, the basal ¹⁴C chronology permits us to derive an upper limiting production rate, which is beneficial given that the Rannoch Moor dataset delivers a range of possible values (3.88–3.95 atoms g⁻¹ yr⁻¹) (Putnam et al., 2019), with the “mid-point” rate (3.912 ± 0.106 atoms g⁻¹ yr⁻¹) serving as a convenient reference rate for exposure-age calculation. Employing the nine new ¹⁰Be measurements in Table 1, we used the production rate calibration tool in the UW online calculator (v.3) (https://hess.ess.washington.edu/math/v3/v3_cal_in.html, last access: 15 March 2026) to derive a maximum-limiting production rate for Scotland; all input data for this calculation are provided in Table S4. According to the default scaling models utilised in the online calculator, this approach yields maximum reference sea-level high-latitude (SLHL) production rates of 3.928 ± 0.067 atoms g⁻¹ yr⁻¹ (1.7 %) with St scaling (Lal, 1991; Stone, 2000) and 3.930 ± 0.068 atoms g⁻¹ yr⁻¹ (1.7 %) with Lm scaling (Balco et al., 2008), and a non-dimensional correction factor of 0.774 ± 0.017 (2.2 %) using the LSDn model (Lifton et al., 2014). Use of a production rate higher than this maximum-limiting value will result in surface-exposure ages that are too young with respect to minimum-limiting ¹⁴C dates of deglaciation from Loch Bad na h-Achlaise.

The results presented above confirm that the Rannoch Moor calibration dataset produces surface-exposure ages that are consistent with independently dated Late Pleistocene surfaces in NW Scotland. This production rate is underpinned by ¹⁰Be measurements from 11 boulders located on the crests of recessional moraines of the late-glacial West Highland ice field (WHIF) on Rannoch Moor, central Scottish Highlands (56.63° N, 4.77° W; ∼310–330 m a.s.l.), the age of which is bracketed by ¹⁴C ages (Putnam et al., 2019). Following the pioneering work of Lowe and Walker (1976) and Walker and Lowe (1979), who provided the first radiocarbon evidence for ice-free conditions on Rannoch Moor, Bromley et al. (2014) reported 20 AMS ¹⁴C ages of plant macrofossils (and one beetle) from the basal sections of 13 cores extracted from seven moraine-dammed basins (Fig. 9). Together, the cores cover ∼ 12 km² of the moor's western margin, in which recessional moraines of the decaying WHIF are abundant. Each core comprised a minerogenic basal unit, indicating early postglacial sedimentation, overlain by increasingly organic units and peat (Bromley et al., 2014). Plant remains in the basal clays were interpreted as indicating migration of vegetation into the Rannoch Moor basin. Acknowledging that the process of plant colonisation is neither spatially nor temporally uniform, that study reported ¹⁴C ages ranging from 10 320 ± 270 to 12 480 ± 100 cal yr BP when calibrated with the IntCal13 curve (Reimer et al., 2009) (note: using IntCal20 impacts this calibrated age and the resulting ¹⁰Be production rate calculation by < 1 %). To identify the earliest onset of plant growth, and thus ice-free conditions, Bromley et al. (2014) used three methods: the first takes the single oldest sample in the dataset as representing colonisation; sample OS-99685 (10 550 ± 65 ¹⁴C yr) from core RM-12-3A gave a calibrated age of 12 480 ± 100 cal yr BP and an earliest likely (upper bound of 90 % confidence interval) age of 12 580 cal yr BP (Bromley et al., 2014). The second method takes the oldest replicable ¹⁴C ages as a conservative date for colonisation; samples OS-99977 (10 400 ± 45 ¹⁴C yr), OS-99978 (10 350 ± 40 ¹⁴C yr), OS-89841 (10 300 ± 70 ¹⁴C yr), and OS-89842 (10 500 ± 50 ¹⁴C yr), all from core RM-10-3A, gave an error-weighted mean calibrated age of 12 262 ± 85 cal yr BP and an earliest likely (upper bound of 90 % confidence interval) age of 12 493 cal yr BP (Bromley et al., 2014). Finally, the upper bound of the 90 % confidence interval for all 20 ¹⁴C ages returned a calibrated age of 12 371 cal yr BP for colonisation (Bromley et al., 2014). All of these ¹⁴C dates, regardless of the method of interpretation, yield minimum ages for plant colonisation and thus for deglaciation.

To calculate a ¹⁰Be production rate for Rannoch Moor, Putnam et al. (2019) took 12 480 ± 100 cal yr BP as the minimum-limiting age for the target moraine belt, which is located immediately adjacent to five of the seven core sites of Bromley et al. (2014) (Fig. 9). We note that, regardless of which of the three methods is used to derive a minimum-limiting age, the resultant production rate does not vary by more than 1 %. For a maximum-limiting age, they employed the probable age of 12 700 ± 100 cal yr BP for culmination of the WHIF reported by Bromley et al. (2018) and based on 27 shell ¹⁴C dates from terminal moraines. The derived production rate was reported as a range (Sect. 5.1), with a mid-point value of 3.912 ± 0.106 atoms g⁻¹ yr⁻¹ (Putnam et al., 2019). In a later study, however, Lowe et al. (2019) questioned the original chronologic foundation for the Rannoch Moor calibration. They proposed that the minimum-limiting ¹⁴C ages reported by Bromley et al. (2014) are erroneously old, thereby making the production rate calculated from that site (Putnam et al., 2019) unrealistically low (Lowe et al., 2019). To test that claim, Lowe et al. (2019) revisited Rannoch Moor with a view to evaluating the reproducibility of Bromley et al. (2014)'s dataset.

Of the seven sites investigated by Bromley et al. (2014), Lowe et al. (2019) elected to re-sample the K2 kettle hole that was first investigated by Walker and Lowe (1977) (Fig. 9). Lowe et al. (2019) reported eight new ¹⁴C dates from K2, the lowest of which (OxA-32471) gives a calibrated age ∼ 1 kyr younger than the oldest basal ages of Bromley et al. (2014) from cores RM-10-3A and RM-12-3A, situated 2–3 km south-east of K2. Highlighting the age difference, Lowe et al. (2019) then presented reasons why the latter must be incorrect (see below). We observe, however, that their dating campaign was restricted to a single site and that Lowe et al. (2019) did not attempt to re-sample those core sites from which Bromley et al. (2014) derived their oldest ¹⁴C ages (RM-10-3A and RM-12-3A). We also note that the newer K2 basal ¹⁴C ages of Lowe et al. (2019), which are supported by their pollen and tephra data, are virtually indistinguishable from those reported by Bromley et al. (2014) from the same K2 site, suggesting a high degree of reproducibility between the two datasets.

Highlighting that the five oldest ¹⁴C dates in the Rannoch Moor dataset of Bromley et al. (2014) include Rhacomitrium, Pogonatum, and Sphagnum material, Lowe et al. (2019) questioned the suitability of moss remains for radiocarbon dating. Despite being considered terrestrial species, Lowe et al. (2019) speculated that mosses might still be subject to “inbuilt age errors” due either to some degree of uptake of aquatic carbon or the inclusion of fine-grained carbonate detritus in measured material. Those authors then proposed that their own moss-derived samples from K2 exhibit δ13C values that are “divergent from average values expected from terrestrial land plants” (p. 180; Lowe et al., 2019) and thus are likely contaminated. Citing an apparent age disparity between moss-bearing samples and those comprising angiosperms (Table 1 of Lowe et al., 2019), they concluded that the former are as much as 100 ¹⁴C older than the latter and potentially suspect. We do not consider this a statistically robust experiment, however, being based on only two sets of paired samples. Nor is the apparent age offset significant, being 95 ¹⁴C years in one pairing (OxA-35410 and OxA-35411) but only 55 ¹⁴C years in the other (OxA-35408 and OxA-35409) (Lowe et al., 2019); much greater variability is a common feature of radiocarbon (indeed, of any) chronologies with multiple samples from the same horizon (e.g., MacLeod et al., 2011; Walker et al., 2012).

The δ13C values for samples underpinning the Rannoch Moor record range from -19.8 to -28.0, with those containing moss fragments spanning -20.0 to -26.8 (Bromley et al., 2018). Such values are consistent with other radiocarbon datasets incorporating moss fragments, including several from the British Isles (MacLeod et al., 2011; Walker et al., 2012; Turner et al., 2015). Ultimately, heath mosses are common targets for ¹⁴C dating Quaternary sequences (e.g., Ellis and Tallis, 2000; Ellis, 2008; Blundell et al., 2008; MacLeod et al., 2011; Matthews et al., 2011; Walker et al., 2012; Turner et al., 2015) that, thus far, have not been shown to exhibit species-specific age bias (Humlum et al., 2005; Holmquist et al., 2016). Furthermore, while we agree with Lowe et al. (2019) that detrital carbonate poses a real risk to accurate ¹⁴C measurement, we reiterate that such material is highly unlikely to be a significant constituent of glacial tills underlying Rannoch Moor (p. 180; Lowe et al., 2019); the few calcite-bearing veins reported by Smith and Marsden (1977) south of the moor are highly localised and on a scale dwarfed by the Rannoch Moor granite pluton. There is no documented evidence of such veins anywhere near, or “up-ice” of, the coring sites of Bromley et al. (2014).

The ¹⁴C chronologies underpinning the Rannoch Moor ¹⁰Be production rate imply deglaciation of the WHIF during the early Younger Dryas (12.8–11.6 ka: NGRIP Members, 2004; WAIS Divide Project Members, 2015) (Bromley et al., 2014, 2018), a scenario that is at odds with those studies advocating late-stadial culmination of the LLR. In their examination of the Rannoch Moor ¹⁴C dataset (Bromley et al., 2014), Lowe et al. (2019) cited four works favouring late-stadial deglaciation. First, the ¹⁰Be surface-exposure dataset of Small and Fabel (2016) purportedly records deglaciation of western Rannoch Moor (Fig. 9) no earlier than 11.5 ± 0.6 ka. As discussed by Bromley et al. (2016a) and Putnam et al. (2019), those ¹⁰Be ages were calculated with two problematic production rates: the Glen Roy rate (Small and Fabel, 2015), which lacks independent age control (Sect. 5.1) and which is shown above to produce ages that are incompatible with limiting ¹⁴C ages on the Redpoint Peninsula (Fig. 8), and the unpublished LLPR attributed to Fabel et al. (2012). Second, the Glacial Lake Blane ¹⁴C record from Loch Lomond (Fig. 9) was originally interpreted as indicating maximum extent of the WHIF after ∼ 12 ka (MacLeod et al., 2011). While the geological basis for Glacial Lake Blane is unequivocal, the case for ¹⁴C-dated varves at Croftamie as maximum-limiting age control for the LLR (Rose et al., 1988; MacLeod et al., 2011) is not. Noting (i) the lack of age control for the basal “Wilderness till”, (ii) the dearth of evidence for glacial overriding of the varves, and (iii) the conspicuous absence of the marine “Clyde Beds” (a prominent local indicator of post-LGM marine transgression), Bromley et al. (2018) queried the traditional view of the Croftamie stratigraphy and suggested that the ¹⁴C data of MacLeod et al. (2011) instead constitute minimum-limiting ages WHIF retreat. Third, the Glen Roy-Spean varve database from sites north of Rannoch Moor (Palmer et al., 2010, 2012; Palmer and Lowe, 2017) was invoked by Lowe et al. (2019) as supporting late-stadial deglaciation (Fig. 9). Yet this varve record has also raised questions over absolute age control; Putnam et al. (2019) highlighted its reliance on undated and chemically ambiguous correlations with distal tephras such as the Vedde Ash, which cannot be distinguished compositionally from other late-glacial Katla eruptions (Lane et al., 2012). Fourth, Lowe et al. (2019) cited a marine-geologic record from the Hebridean Sea (Fig. 9), in which the basal unit (MCU6) of core 729 contains cold-water foraminifera and clastic material interpreted as ice-rafted debris (Arosio and Howe, 2018). From this, Arosio and Howe (2018) invoked WHIF tidewater outlets terminating in the Hebridean Sea until ∼ 11.7 ka, a scenario that is hard to reconcile with ice free conditions at the ice field's centre (c.f. Bromley et al., 2014). Acknowledging the value of offshore records for helping elucidate onshore processes, we caution that unit MCU6, which is central to their argument for late-Younger Dryas deglaciation, lacks direct age control. Two AMS ¹⁴C dates on bivalves from the overlying unit MCU7 afford minimum-limiting ages only (Arosio and Howe, 2018); even assuming accurate estimation of the marine reservoir effect and ΔR, the dearth of dates from MCU6 precludes meaningful correlation between core 729 and events at Rannoch Moor, > 90 km to the east. The proposed glacial occupation of the Arisaig coast until ∼ 11.7 ka (Arosio and Howe, 2018) is also inconsistent with replicated ¹⁴C evidence for coastal deglaciation by ∼ 16 ka (Best and Shennan, 2025).

The reappraisal of the Rannoch Moor deglacial record proffered by Lowe et al. (2019) refutes neither the original conclusions of Bromley et al. (2014) nor, by extension, the production rate of Putnam et al. (2019). Further, the results presented here from Loch Bad na h-Achlaise reinforce the validity of the Rannoch Moor production-rate calibration data set. In light of our new results from the Redpoint Peninsula, we stress that choosing any production rate calibration dataset that favours a late-Younger Dryas deglaciation of Rannoch Moor will produce ¹⁰Be ages that are much younger than, and therefore incongruent with, the independent ¹⁴C chronology at Loch Bad na h-Achlaise (Fig. 8).

The Redpoint ¹⁰Be dataset in Fig. 5 includes three samples from the crest of the outer moraine, close to the locations of three samples reported by Ballantyne et al. (2009) (Fig. 2). Recalculated in the same way as our ages, and utilising updated elevations from the Landmap digital terrain model (Landmap, 2014), their samples RM-01, 02, and 03 give revised exposure ages ranging from 15.5 ± 0.6 to 16.4 ± 0.6 ka (mean: 15.8 ± 0.3 ka), in broad agreement with our moraine samples. The published coordinates for a fourth sample, RM-04, are inconsistent with a position on the moraine and so we did not calculate a revised age. While we did not include these recalculated ages in our evaluation of production rates, the agreement between our ¹⁰Be concentrations and those of Ballantyne et al. (2009) (Fig. S2) reinforces the pattern of deglaciation in NW Scotland during Termination 1.

As highlighted in Sect. 4.2, the tight internal consistency among the nine new ¹⁰Be ages confirms that, regardless of production rate choice, the target landforms were all exposed within a very narrow window of time (and within the analytical uncertainty of the data). Since our dataset includes samples from outboard and inboard of the Redpoint moraines – as well as from the moraines themselves – that outcome indicates that the episode of ice-marginal stability represented by the Redpoint moraines was brief relative to the overall span of our chronology (i.e., within the ∼ 200-year standard error of the dataset). Viewed in the broader climatic context of Termination 1, the timing and brevity of the Redpoint moraine-building episode suggests that the Wester Ross Readvance was a prominent yet short-lived interruption to overall ice sheet collapse during Heinrich Stadial 1; there appears to have been insufficient time to accommodate the spatially extensive resurgence of ice envisaged by some prior studies (Robinson and Ballantyne, 1979; Ballantyne et al., 2009; Bradwell et al., 2021). Further dating work on other sections of the complex will establish the validity of this hypothesis.

Our revised chronology for the Wester Ross Readvance supports the conclusions of Simms et al. (2022) that this pause in deglaciation occurred at least ∼ 500 years earlier than previous estimates (Ballantyne and Small, 2019; Bradwell et al., 2021). While it is beyond the scope of this paper to synthesise well-dated moraine chronologies spanning Termination 1, we note that brief yet prominent moraine-building events around the time of the Wester Ross Readvance are salient features of deglacial records from elsewhere in Britain and Ireland (McCabe et al., 2007; Small et al., 2012; Hall et al., 2016; Foreman et al., 2025, 2026), as well as from sites farther afield (Putnam et al., 2013; Bromley et al., 2016b; Hall et al., 2017).