Evidence for centennial‐scale Lateglacial and early Holocene climatic complexity from Quoyloo Meadow, Orkney, Scotland

The influence of the North Atlantic on the margins of Europe means the region is particularly sensitive to changes in the ocean–atmospheric system. During the Last Glacial–Interglacial Transition (16–8 cal ka bp) this system was repeatedly disrupted, leading to a series of abrupt and short‐lived shifts in climate. Despite much research, the number and magnitude of these ‘centennial‐scale’ events is not well understood. To address this, we expand upon investigations at Quoyloo Meadow, Orkney, Scotland, one of the best chronologically constrained palaeoclimate records in northern Britain. By coupling stable isotope and chironomid fossil analyses with existing data, this study identifies multiple phases of centennial‐scale disturbance at: c. 14.0, 11.1, 10.8, 10.5, 10.45 and 10.3 cal ka bp, with the events at 14.0 and 10.3 exhibiting a particularly pronounced cold‐climate signature. During the Holocene, the strongest response to climate forcing was at c. 10.3–10.0 cal ka bp, expressed as a two‐stage drop in mean July temperatures, a shift in pollen spectra indicative of ‘less‐stable’ climatic regimes, and a depletion in δ18O values. We interpret this as the first reliably dated incidence of the ‘10.3‐ka event’ in the British Isles and consider the wider impact of this climatic reversal in other Holocene records.


Introduction
Abrupt climatic and environmental changes during the Last Glacial-Interglacial Transition (LGIT; 16-8 cal ka BP) are thought to have been driven by reconfigurations of the ocean-atmospheric system (Broecker and Denton, 1990;Clark et al., 2002;Steffensen et al., 2008). This resulted in millennial-scale oscillations that characterized much of the late Pleistocene (Ruddiman et al., 1989;Dansgaard et al., 1993), but also led to a series of shorter-lived centennial-scale events that preferentially impacted the North Atlantic region (Bond et al., 1997). The best resolved records of this period are the Greenland ice cores and these have become the de facto regional stratotype for comparison (Björck et al., 1998;Lowe et al., 2008;Rasmussen et al., 2014). However, several climatic oscillations have been detected in European and North American archives which have no comparable analogue within the ice-core records (e.g. Marshall et al., 2002;Whittington et al., 2015), suggesting that some climatic changes are more clearly represented in terrestrial records. Despite a wealth of research into the LGIT, shorter-lived 'centennial-scale' events are not routinely identified and may often be overlooked due to a lack of stratigraphic, proxy sampling and temporal resolution within terrestrial archives. The temporal precision afforded by radiocarbon dating during this period, although excellent, has also probably led to the misallocation and conflation of events in some circumstances (e.g. Baillie, 1991;Lowe and Walker, 2000;Blaauw et al., 2007;Blaauw, 2012). Consequently high-resolution, temporally well-constrained records of climate are needed from across the mid-latitudes to understand the frequency and magnitude of these less 'well-defined' events, and to assess the complexities of climate change during, and immediately following, periods of rapid climatic adjustment.
One such archive that fits this remit is Quoyloo Meadow on Orkney Mainland (Fig. 1). Previous work at the site has established the presence of a sedimentary sequence spanning the LGIT (Bunting, 1994), and recent high-resolution investigations by Abrook et al. (2020a) and Timms et al. (2017) have detected a number of abrupt changes in the palynological record that are bracketed by cryptotephra of known age. This site can therefore provide detailed information with regard to the number and timing of centennial-scale climatic events during the LGIT. Here we present climatic proxy data in the form of new stable isotope and chironomid-inferred temperature reconstructions for the LGIT with improved chronological and proxy control, particularly for the early Holocene period. The aim of this study is to better constrain and understand the impact of centennial-scale climatic events on the landscape of northern Britain.

Site Context
The Orkney Isles lie 10-90 km north of mainland Scotland and possess a cool maritime climate with a mean annual temperature of 10.7°C (Met Office, 2020). Their location means they are critically placed to record the glacial and palaeoenvironmental history of northern Britain and the adjacent North Sea basin ( Fig. 1). The Orkney Isles exhibit a complex glacial history (e.g. Rae, 1976, Hall et al., 2016, with cosmogenic dating of glacial erratics and ice-moulded bedrock suggesting final deglaciation from the Devensian ice sheet c. 15-17 ka BP (Phillips et al., 2008;Clark et al., 2012). Previous palaeoenvironmental studies have established that the islands experienced a reduced floral diversity compared to the Scottish mainland during the LGIT (Moar, 1969;Bunting, 1994;Whittington et al., 2015;Abrook et al., 2020a), but were later characterized by a rich woodland environment that provided an important resource for Mesolithic and Neolithic communities (Bunting, 1994;Farrell et al., 2014).
Quoyloo Meadow is situated on Orkney Mainland, the largest of the islands in the archipelago ( Fig. 1; 59.066417/ −3.309333). The basin lies at c. 30 m asl within a topographic depression underlain by Devonian Upper and Lower Stromness flagstones (Mykura et al., 1976). At present a valley mire occupies the site, but previous studies have identified a shallow lacustrine sequence with a sedimentary, biostratigraphic and tephrostratigraphic succession typical of the LGIT (Bunting, 1994;Timms et al., 2017;Abrook et al., 2020a).

Core extraction and sedimentology
The Quoyloo Meadow basin was cored in February 2014 (Timms, 2016). Six 50 × 5-cm overlapping Russian cores were aligned on matching sedimentological units to form the Quoyloo Meadow 1 (QM1) composite sediment stratigraphy with a total depth of 242 cm (Fig. 2). This composite forms the basis of the record reported here, as well as the tephra investigation reported in Timms et al. (2017), and the palynological study of Abrook et al. (2020a). The QM1 sequence was described using the Troels- Smith (1955) classification scheme with bulk sedimentological analyses, including loss on ignition (LOI) for organic content, and an estimation of calcium carbonate and magnetic susceptibility performed following methods outlined by Timms et al. (2017).

Chronology
Ten cryptotephras and one visible ash layer have been reported from the Quoyloo Meadow sequence (Timms et al., 2017. Details of those tephra, their stratigraphic placing and current best age estimates are given in Table 1. This tephrochronological information has been used to generate an age-depth model in OxCal V4.4.2 utilizing the IntCal20 calibration curve (Bronk Ramsey, 2017;Reimer et al., 2020). Model outputs were computed using a 'P_Sequence' depositional model (Bronk Ramsey, 2008) and utilized a variable 'k' parameter which enabled the computation to determine an optimized age-depth relationship (Bronk Ramsey and Lee, 2013).

Chironomids
In total, 35 samples were analysed for chironomid head capsule content. The sampling occurred in two stages with an outline low-resolution scan, and an increased sampling resolution at intervals where temperature fluctuations were initially detected. Extraction of head capsules followed standard procedures described in Brooks et al. (2007). Sediment subsamples of 0.25-3.94 g were gently heated in 10% sodium hydroxide (KOH) at 75°C for 10-15 min. Residual material was passed through 120-and 90-µm sieves under running water, with both size fractions examined for head capsule content at 100× to 400× magnifications. Extracted head capsules were mounted onto glass slides using Euparal, and species identification followed reference guides: Wiederholm (1983), Moller Pillot and Klink (2003) and Brooks et al. (2007).
Two calibration datasets, the Norwegian and the combined Swiss-Norwegian, were tested for the chironomid-inferred mean July air temperature (C-IT; T Jul ) reconstruction at Quoyloo Meadow (Brooks and Birks, 2001;Heiri et al., 2011). The combined Swiss-Norwegian calibration dataset was Figure 2. The QM1 composite sediment stratigraphy from Quoyloo Meadow. Loss on ignition (LOI) and calcium carbonate (CaCO 3 ) content are also shown. The grey horizontal bars mark the primary position of nine cryptotephra and one visible tephra reported in Timms et al. (2017Timms et al. ( , 2019. The abbreviations of the tephra are as follows: TT, Tanera Tephra; BT, Borrobol Tephra; PT, Penifiler Tephra; VA, Vedde Ash; HT, Hässeldalen Tephra; AskT, Askja-S Tephra; AshT, Ashik Tephra; SA, Saksunarvatn Ash; FT, Fosen Tephra; and AnDT, An Druim Tephra. A detailed lithological description is given in Supporting Information Table S2. [Color figure can be viewed at wileyonlinelibrary.com]. selected over the Norwegian for several reasons. First, the incorporation of carbonate lakes from the Swiss calibration dataset means the combined dataset contains records with similar lake chemistry to Quoyloo Meadow during the Holocene. Second, the combined calibration dataset contains a greater number of chironomid taxa/morphotypes and those that are 'rare' in the Norwegian calibration dataset are present in greater abundance. Third, the combined calibration dataset has superior validation statistics compared to the Norwegian dataset (see Supporting Information Table S1).
The selected Swiss-Norwegian combined calibration dataset consists of 274 lakes spanning a temperature gradient of 3.5-18.4°C (Heiri et al., 2011). Nineteen outlier lakes were deleted from the calibration as defined by Heiri et al. (2011). A two-component, weighted averaging-partial least squares (WA-PLS) regression model was used on square-roottransformed fossil data, which gave a root mean squared error of prediction (RMSEP) of 1.4°C, a coefficient of determination (r 2 boot ) of 0.9 and a maximum bias boot of 0.8°C. Samplespecific errors were calculated using 999 bootstrapping cycles.
Five criteria were used to assess the reliability of the C-IT reconstruction: (i) A goodness-of-fit to temperature was assessed by passively plotting fossil samples within a canonical correlation analysis (CCA) of the modern calibration dataset. Fossil samples which had a squared residual distance value exceeding the 90 th and 95 th percentiles of samples in the modern calibration dataset had a 'poor' or 'very poor' fit to temperature, respectively (Telford, 2014a). CCA was conducted on untransformed data in R using the vegan package (Oksanen et al., 2017). (ii) The modern analogue technique (MAT) was employed to assess if fossil samples had good analogues in the modern calibration dataset, with chi-squared used as a measure of dissimilarity. Samples with distances to the closest analogue larger than the 5 th and 10 th percentile of all modern distances were treated as having 'no close' and 'no good' analogue, respectively (Telford, 2014b). MAT analysis was performed on untransformed data using the analogue package in R (Simpson, 2020). (iii) No more than 5% of fossil assemblages should be from taxa that are rare in the modern calibration datasetdefined as having a Hill's N2 value of less than 5 (Heiri et al., 2003).
(iv) In total, 95% of taxa in fossil samples should be present in the modern calibration dataset (Birks, 1998). (v) Samples with a minimum of 45-50 head-capsules are considered reliable (Heiri and Lotter, 2001).
Where these criteria are not met, however, samples are retained but are considered with further caution.

Stable isotopes (δ 18 O and δ 13 C)
Sediment subsamples of c. 0.5 cm 3 were sampled contiguously at 0.5-cm resolution from the marl substrate between 195 and 150 cm. Subsamples were disaggregated with 0.5% sodium hexametaphosphate (NaPO 3 ) 6 , before sieving at 63 µm to remove detrital material and biogenic carbonate contamination, e.g. molluscs and ostracods. Sieved samples were air dried and treated with 10% hydrogen peroxide (H 2 O 2 ) to remove organic material. Samples were homogenized and weighed to a mass of 500-1100 µg, but samples exhibiting <10% CaCO 3 were weighed to c. 1500 µg. δ 18 O and δ 13 C values were determined by analysing the CO 2 evolved from a reaction of the sample with phosphoric acid (H 3 PO 4 ) at 90°C using a VG PRISM series 2 mass spectrometer. Internal (RHBNC) and external (NBS19, LSVEC) standards were run to check for machine precision and drift. All stable isotopic values are quoted with reference to VPDB.

Palynology, charcoal and principal curve analysis
The methodologies for pollen, charcoal and principal curve (PrC) analyses are outlined in Abrook et al. (2020a), where further detail on these data series can be accessed. PrC analysis is a non-parametric statistical method that is particularly sensitive in the identification of gradients in ecological data. PrC analysis, which starts with principal components analysis (PCA) or CA axis 1 scores as the starting curve, is defined as a smooth one-dimensional curve passing through an m-dimensional dataspace (Hastie and Stuetzle, 1989;De'ath, 1999;Simpson and Birks, 2012). To generate the PrC, the starting curve is iterated, using smoothing splines resulting in increased fit to the data. This method was selected as PrCs capture more variance within datasets than standard ordination techniques (Simpson and Birks, 2012).  Timms et al., 2017Timms et al., , 2019. These tephra form the basis of the Bayesian age-model presented in Fig. 3. Tephra age estimates have been recalculated using the cross-referenced Bayesian age-model presented in Kearney et al. (2018) and updated with the IntCal20 calibration curve (Reimer et al., 2020). The age estimate for the An Druim Tephra is derived from a Bayesian age-model of the Lochan An Druim type-site (see Ranner et al., 2005;Timms, 2016) and updated with the IntCal20 calibration curve (Reimer et al., 2020). The input age of the Fosen Tephra is derived from Lind et al. (2013) and given an arbitrary 1σ error range of 100 years due to the absence of a reported age uncertainty (see Timms et al., 2017). The age estimates for the Borrobol and Penifiler tephras were inputted into the model as 'priors' due to their non-normal distribution; hence they are presented here as ranges only. The age estimate of the Tanera Tephra was not used in the construction of the age-model but is instead the result of age-model extrapolation to the base of the QM1 sequence. Ages marked with an asterisk denote the most reliable age estimates for these tephra at present

Sedimentology
The composite sediment stratigraphy, bulk sedimentological indicators and tephrostratigraphy from Timms et al. (2017Timms et al. ( , 2019 is presented in Fig. 2 with a summary outlined below. More detailed sedimentological information is provided in Supporting Information Table S2. A dark grey silty-clay unit, with low organic content (c. 2-7%) is present at the base of the QM1 sequence (Unit 1a/1b). Organic content increases in Unit 1c with the occurrence of highly humified organic detritus (c. 17%). These organic sediments return to lower values (c. 4%) as the sequence transitions into the overlying minerogenic unit. Units 2a to 2c are characterized by a light grey silty-clay (organic content c. 2-5%), with Unit 2b possessing an orange/brown coloration, a feature probably caused by the leaching and dissolution of basaltic glass shards (Timms et al., 2017). Units 3a-3d are characterized by the precipitation of a buff-coloured marl substrate, signalling a significant shift in environmental conditions. Organic values initially peak in Units 3a-3b (c. 11-13%), but begin to fall in Units 3c-3d as carbonate sedimentation becomes more dominant in the basin (CaCO 3 values c. 40-90%). At 160 cm a black horizontal band approximately 0.5 cm thick of fine-grained volcanic ash intersects the marl substrate (CaCO 3 values c. 70%). Above this, marl sedimentation resumes (CaCO 3 values 80-90%) and continues to the top of the studied sequence (Units 5a-5d).

Chronology
The age model derived from the tephrostratigraphic findings of Timms et al. (2017Timms et al. ( , 2019, and updated to include the tephra's current best age estimates (Table 1), exhibits good agreement indices and suggests that the onset of sedimentation in the Quoyloo Meadow basin began at c. 15.30 ± 1.80 cal ka BP (Fig. 3). Age estimates are reported at 2σ uncertainty unless otherwise stated.

Stable isotopes (δ 18 O and δ 13 C)
The δ 18 O and δ 13 C values obtained from the QM1 sequence exhibit a notable degree of variability between samples (Fig. 5). To highlight trends in the data series, a three-point moving average has been applied. Discussion will be based on these moving averages with the acknowledgement that this may limit some interpretations.
The oxygen isotope record is characterized by a general increase in values from −4. 16‰ (194.25 cm) to −3.29‰ (150.25 cm). At several intervals, however, this rising trend is punctuated by short-term decreases in values, labelled as QMO-e1 to e4 ( The carbon isotope record exhibits an opposing trend: enriched values typify the base of the carbonate sequence, falling from 3.07‰ at 194.25 cm to 1.20‰ at 150.25.  Abrook et al. (2020a) to include the current best age-estimates for the tephra (Table 1). Age uncertainties are plotted as 95.4% confidence intervals. [Color figure can be viewed at wileyonlinelibrary.com].

Pollen and PrC
The pollen profile at Quoyloo Meadow is divided into 11 local pollen assemblage zones based on variations in principal taxa (  QMP-1 (223-217 cm) and QMP-2 (217-201 cm) are dominated by herbaceous pollen with dominant taxa including Poaceae and Rumex. Within these two zones Betula percentages increase throughout. At 212.5 and 211.5 cm a small inflection in the PrC is noted, which appears to be driven by reductions in Betula and increases in Rumex. Throughout these zones percentages for the non-pollen palynomorph Pediastrum are high (37%).
QMP-3 (210-202 cm) exhibits high percentages of Empetrum nigrum which occurs alongside Pediastrum and later increases in Artemisia.  is predominantly composed of Artemisia and Pediastrum albeit with generally lower percentages than previous zones. Additional taxa throughout the upper levels of QMP-3 and QMP-4 include Poaceae, Cyperaceae, Saxifragaceae and Selaginella selaginoides. Within QMP-3 and QMP-4 the largest shift in the PrC can be observed, signalling a considerable change to vegetation type and cover (Fig. 6).  and QMP-6 (189-184 cm) exhibit successive peaks of Pediastrum, Empetrum nigrum and Betula. Between QMP-7 and QMP-10 (164-136 cm) there is a general trend towards lower Betula percentages but an increase in herbaceous pollen types. Corylus increases from QMP-7 and is dominant by QMP-11 (136-131 cm) alongside Betula. At two additional locations between QMP-5 and QMP-11 a plateau in the general shift towards lower values or short inflections in the PrC were noted by Abrook et al. (2020a). These occur from 188.5 to 185.5 cm and 158.5 to 154.5 cm and appear to be driven by variability in Empetrum nigrum, Betula and Poaceae.

Charcoal
Multiple phases of increased charcoal abundance are identified throughout the profile with charcoal numbers ranging from 1 to 62 fragments cm −3 ( Fig. 6; Abrook et al., 2020a). The most prominent of these phases are identified at 222.5-221.5, 203.5, 197.5, 180-189.5, 154 and 147-144.5 cm. Greater charcoal accumulations appear within the marl sediments in association with zones QMP-6 to QMP-9. Further increased concentrations appear to coincide with changes in vegetation as indicated by the PrC (Abrook et al., 2020a).

Palaeoenvironmental significance of the proxy record (chironomids)
Of the 36 chironomid samples analysed, 20 have a good fit-totemperature while four have a poor fit and 12 have a very poor fit. The majority of samples with a poor or very poor fit-totemperature occur in the early Holocene (QMC-3 and QMC-5; Fig. 7) and are possibly caused by the high abundance of Tanytarsus glabrescens-type (up to 47%) which is present only in 16 lakes of the calibration dataset. The exceptions to this are sections of QMC-4 and -5 between 160-150 and 145-140 cm where a good fit to temperature is seen. The majority of samples in QMC-1 and QMC-2 (LLS and WI; Fig. 7) have good analogues, in contrast to samples in QMC-3, -4 and -5 which have poor or very poor analogues in the modern calibration data. This suggests the Holocene lake established at Quoyloo Meadow was relatively unlike the lakes contained within the calibration dataset. However, WA-PLS inference models have been shown to perform surprisingly well in non-analogous situations Brooks et al., 2012).
All samples have at least the recommended minimum of 45-50 head-capsules except for 212.5, 199.5 and 135.5 cm which have 43, 43 and 41 head-capsules respectively. Despite the low number of head capsules in these particular samples, the taxa are very similar to neighbouring intervals. This suggests that species composition and thus ecological and temperature inferences are not excessively limited or biased by the low recovery rate. A single head-capsule was found at 241.5 cm (Micropsectra radialis-type) and has been excluded from all numerical analysis. The only taxon not present in the modern dataset is Parakiefferiella Type-Swiss (Brooks and Heiri, 2013) which is present in only four samples and forms less than 4.2% of individual sample assemblages. Taxa which are rare in the modern dataset and present in the record include Tanytarsus nemorosus-type, Tanytarsus lactescenstype, Parakiefferiella nigra-type, Nilotanypus and Diamesa aberrata-type. These are only present in six samples and comprise no more than 4% of fossil samples. Therefore, these should not significantly affect the temperature reconstruction.
Present in the C-IT reconstruction are a series of short-lived oscillations labelled QMC-e1-e3 (Fig. 7). T Jul values throughout QMC-1 are c. 12°C, with a decline noted at 214.5 cm (QMC-e1). In QMC-2, T Jul is c. 7°C, notably lower than the previous zone. A large rise in T Jul is present at the end of QMC-2 with values continuing to rise throughout QMC-3, albeit with a slight inflection at 174.5 cm (QMC-e2). A decline of c. 3.8°C is observed in QMC-4 (QMC-e3), with T Jul reaching a minimum of 14.2°C. T Jul recovers in QMC-5 and reaches highs of 16.8°C. Sample-specific errors range from 1.4 to 1.7°C.

Palaeoenvironmental significance of the proxy record (stable isotopes)
The δ 18 O value of lake carbonates is controlled by two factors: (i) the temperature at which carbonate mineralization occurs and (ii) the δ 18 O value of the lake water (Leng and Marshall, 2004;von Grafenstein et al., 2013). In mid-latitude temperate climate regions, such as NW Europe, the modification of the isotopic value of lake waters by processes such as evaporation is minimal (for discussion see Candy et al., 2016). Consequently, in such areas, the δ 18 O value of lake water is strongly linked to the δ 18 O value of precipitation which is, in turn, heavily dependent upon air temperature (Rozanski et al., 1992). In such a system both of the main influences on lake carbonate δ 18 O values are therefore temperature driven, although they work to drive the carbonate isotope values in different directions (Candy et al., 2011). The relationship between oxygen isotope fraction during carbonate mineralization and temperature is well known and is around +0.24‰/−1°C (Leng and Marshall, 2004) effectively meaning that the δ 18 O of a carbonate becomes higher under lower temperatures. The relationship between the δ 18 O of precipitation and air temperature is, for the mid-latitudes, around +0.58‰/+1°C, resulting in rainfall and surface waters having higher δ 18 O values under higher temperatures.
As the magnitude of the air temperature/δ 18 O of precipitation relationship is greater than the magnitude of the mineral fractionation/temperature relationship, the δ 18 O value of lake carbonates in Europe frequently shows a linear relationship with temperature, i.e. the isotope values decrease as temperatures decline and increase as temperatures rise. This can be seen empirically in LGIT studies on NW European lakes which show δ 18 O values decreasing during well-established cold events such as the Loch Lomond Stadial (Marshall et al., 2002), or with cold events observable in independently reconstructed temperature records (Blockley et al., 2018;Abrook et al., 2020b). Such a relationship is also observable in the early Holocene record of Quoyloo Meadow, which shows a long-term trend of increasing δ 18 O values associated with increasing T Jul estimates and short-lived declines in δ 18 O Figure 6. Selected pollen, pteridophyte and non-pollen palynomorphs (NPP) from Quoyloo Meadow. Key taxa are shown with the addition of local pollen assemblage zones (LPAZ) and a principal curve (PrC) with key phases of variability shown in grey (see text for explanation). Macro-charcoal counts are included with areas higher than the average (red) deemed important. All data are presented as percentages of total land pollen (TLP) unless stated (see Abrook et al., 2020a). For lithological information refer to values associated with abrupt decreases in T Jul estimates. It is therefore probable that air temperature changes are the primary control on the δ 18 O value of Quoyloo Meadow lacustrine carbonates. Consequently, the δ 18 O record of the early Holocene at Quoyloo Meadow is interpreted as recording a long-term trend of climate warming, punctuated by a series of short-lived cooling events.
Such an interpretation assumes that the δ 18 O values of lake waters have undergone no major isotopic modification, i.e. through evaporation, during residence within the basin whilst it also assumes that minimal contamination of the isotopic signature by the presence of detrital carbonates (Candy et al., 2015. Both of these factors are considered unlikely to have been significant in the Quoyloo Meadow record. Evaporation of lake waters has a tendency to produce positive co-variance in both δ 18 O and δ 13 C values (Talbot, 1990), with greater evaporation producing higher values in both isotope systems, whereas in the Quoyloo Meadow record δ 18 O and δ 13 C values trend in opposite directions (δ 18 O increase whilst δ 13 C decrease). Equally, short-term declines in δ 18 O values are not matched by corresponding changes in δ 13 C values.
It is also considered unlikely that detrital contamination has had an influence on the isotopic record seen in Quoyloo Meadow. The incorporation of limestone bedrock, which has much higher δ 18 O and δ 13 C values than lacustrine carbonates, into lake sediments can contaminate the isotopic signal, generating particularly high isotopic values (Candy et al., 2015). Whilst dolomitic siltstones and tills with limestone components occur across Orkney (Mykura et al., 1976;Rae, 1976;Hall et al., 2016) it is considered unlikely that these have had an influence on isotopic values. During periods of high detrital inwash, i.e. the Loch Lomond Stadial, carbonate values are 0%, indicating that the surrounding catchment yielded no detrital carbonate during phases of relative landscape instability (see Candy et al., 2016). Furthermore, analyses on samples with low percentage carbonate values, which would, if impacted by detrital contamination, be the most sensitive to limestone inwash and have the highest δ 18 O values, yield the lowest δ 18 O values of the entire record.
It is worth noting two further features of the Quoyloo Meadow δ 18 O record. First, there is a relatively high degree of scatter within the δ 18 O dataset. Whilst a three-point moving average produces a stable and clear signal, there is significant variability between stratigraphically contiguous samples. It is likely that this is a function of the high resolution of the record (Mangili et al., 2010;Tye et al., 2016). The chronology for the early Holocene indicates a sedimentation rate of approximately 1 cm per 30 years, whilst the sampling resolution, 0.5 cm per sample, implies that each isotope sample reflects c. 15 years of marl accumulation. In such a record, inter-year and inter-decadal variability in weather and climate will still have an influence on the isotopic values of individual samples and, consequently, a high degree of isotopic variability is to be expected between samples (see Tye et al., 2016). This is true for any isotope dataset constructed through an exceptionally high-resolution lake record. Second, the isotopic record from Quoyloo Meadow only extends back in time as far as carbonate precipitation occurred within the lake basin. Based on the site chronology this is 11.48 ± 0.48 cal a BP, which means that the Quoyloo Meadow δ 18 O record may not span the rapid increase in δ 18 O values seen in NGRIP from c. 12 to 11.7 ka and, consequently, may not provide a complete record of the very early Holocene. The Quoyloo Meadow δ 18 O record does provide a high-resolution snapshot of a small part of the postglacial warming, but it is difficult to establish whether the well-defined isotopic trends seen in NGRIP directly after GS-1 are also expressed in the Quoyloo Meadow record.
Less attention has been paid to the δ 13 C values of lacustrine sequences in the British Late Quaternary as they typically reflect changing intra-basin processes rather than wider palaeoclimatic trends Blockley et al., 2018). In general, the δ 13 C of lacustrine carbonate represents the dissolved inorganic carbon (DIC) pool of the lake system (Leng and Marshall, 2004). The increasingly negative trend seen in the QM1 sequence is common to a number of early Holocene lacustrine records and has been used to indicate the increasing amount of plant-respired carbon dioxide being supplied to the basin as the landscape revegetates after the Loch Lomond Stadial (Blockley et al., 2018). This interpretation is consistent with the pollen data from this site and is proposed to explain the trend that is seen at Quoyloo Meadow.

Palaeoenvironmental changes during the Dimlington Stadial and Windermere Interstadial
The onset of sedimentation at Quoyloo Meadow appears to have begun during the latter phases of the Dimlington Stadial, an interval characterized by glacial retreat and broadly comparable to the GS-2.1a cold phase (Rose, 1985;. This (chrono-)stratigraphic correlation is proposed via three lines of evidence: (i) the dominance of sterile minerogenic sediments within the lowermost units at Quoyloo Meadow (Fig. 2); (ii) the occurrence of the 'Tanera Tephra' (242 cm; 15.30 ± 1.80 cal ka BP), a cryptotephra found within Dimlington Stadial deposits elsewhere in the British Isles ; and (iii) a single chironomid head capsule of Microspectra radialis-type, an ultra-cold climate indicator which was identified at 241.5 cm (15.23 ± 1.77 cal ka BP). Together this evidence is suggestive of a cold climate and barren landscape, with sediments accumulating relatively soon after deglaciation.
The first viable chironomid sample occurs at 237.5 cm (15.08 ± 1.58 cal ka BP) and provides an estimated mean July temperature of 11.1°C (QMC-1). This T Jul estimate is indicative of the Windermere Interstadial, a period of comparative climatic warmth and broadly equivalent to the GI-1 Interstadial (Coope and Pennington 1977;. The absence therefore of a robust cold climate indicator for the Dimlington Stadial, or a warming trend out of this phase is puzzling, but means that Quoyloo Meadow is similar to most other chironomid records in the British Isles. At present it is only the C-IT reconstructions from Whitrig Bog, Fiddaun and Lough Nadourcan which exhibit a proxy inferred thermal transition from the cold Dimlington Stadial, to the comparative warmth of the Windermere Interstadial (Brooks and Birks, 2000a;Van Asch et al., 2012;Watson et al., 2010). Why this should be is not immediately apparent, but the perseverance of 'dead-ice' within the basin and the inability to extrude the deepest sediments using Russian coring equipment, as suggested elsewhere (e.g. Brooks et al., 2016), does not seem to be a viable explanation given the stratigraphic occurrence and estimated age of the Tanera Tephra.
At Quoyloo Meadow the first viable pollen counts (n = 100 total land pollen) were achieved at 223 cm with the occurrence of pioneering open-tundra grassland communities (QMP-1; 14.38 ± 0.73 to 14.1 ± 0.15 cal ka BP). This delayed colonization of Orkney, despite the chironomid data suggesting relatively high T Jul from 237.5 cm (15.08 ± 1.58 cal ka BP), may relate to the local conditions. The most probable explanation is that retarded soil development and high windshear created instability in the catchment, which only ceased once pioneering taxa were able to stabilize the landscape. This instability and unfavourable conditions for higher plant colonization helps to explain why there is such a high rate of clastic sedimentation through the lower part of the record (Fig. 2), and also why the typical vegetation succession that occurred across Scotland during the Windermere Interstadial (e.g. Rumex-Empetrum-Juniperus; Walker et al., 1994; is not so readily observed on Orkney (e.g. Poaceae and Empetrum; Bunting, 1994;Abrook et al., 2020a).
At 214.5 cm (14.04 ± 0.17 cal ka BP), the occurrence of cold-climate chironomid species increases (Fig. 4). Taxa such as M. radialis-type, Sergentia coracina-type, H. grimshawi-type and P. sordidellus suggest a short-term decline of 1.8°C in T Jul (12.4-10.6°C; see QMC-e1 in Fig. 7) whilst increases of Rheotanytarsus and Pseudosmittia indicate a possible shallowing of the lake basin, expansion of the littoral zone and/or greater influence from stream input . Local changes are further reflected in the palynological record which records a shift to more open taxa at this time (216.5-210.5 cm; 14.09 ± 0.16-13.66 ± 0.90 cal ka BP; Abrook et al., 2020a; Fig. 6). Following the perturbation, T Jul recovers to around 14.1°C (208.5 cm; 13.40 ± 1.10 cal ka BP) and vegetation communities stabilized (Fig. 6). Sedimentation rates reduced significantly through this interval, possibly caused by more dense and mature vegetation cover inhibiting the input of clastic materials into the basin. As a result, only 7 cm of sediment represents the period between the Penifiler Tephra (14.00 ± 0.16 cal ka BP) and the end of the Windermere Interstadial (13.06 ± 1.21 cal ka BP) at Quoyloo Meadow (Fig. 2).

Palaeoenvironmental changes during the Loch Lomond Stadial
Significant changes to the sediments and proxy series are noted at the transition from Unit 1c to Unit 2a (205.5 cm). The organic content of the silty-clay sediments decreases (208-200 cm; Fig. 2), charcoal frequency increases, and changes to the vegetation assemblage indicate the replacement of open heath vegetation (e.g. Empetrum and Betula cf. nana) by taxa more tolerant of cold and unstable conditions (e.g. Salix, Caryophyllaceae, Saxifragaceae and Selaginella selaginoides; Abrook et al., 2020a; Fig. 6). The chironomid assemblage and C-IT reconstruction corroborates this inference of a cooler environment, cold-adapted species such as M. radialis-type, P. austriacus-type and M. insignilobus-type increase, and T Jul falls from 12.0 to 6.0°C (207.5-205.5 cm; 13.26 ± 1.17 to 13.00 ± 1.22 cal ka BP; Figs. 4 and 7). The timing of these changes, and the stratigraphic occurrence of the Vedde Ash at 198 cm (Timms et al., 2017) confirms this interval as the Loch Lomond Stadial, a high-magnitude cold climate event broadly equivalent to the Younger Dryas and GS-1 interval (Lowe et al., 2008. At 199.5 cm (12.20 ± 0.70 cal ka BP) just prior to the deposition of the Vedde Ash, T Jul increases from 7.6°C (201.5 cm) to 9.9°C before returning to similarly low values shortly thereafter (6.8°C; 197.5 cm; 11.93 ± 0.30 cal ka BP). This temporary increase in T Jul coincides with a reduction of macro-charcoal and a change in vegetation communities. Xerophytic taxa such as Artemisia and instability indicators such as Rumex decrease, whilst Empetrum becomes more abundant. The expansion of heathland-type communities and a reduction in charcoal might suggest a brief interval of amelioration and one potentially typified by slightly wetter conditions.

Palaeoenvironmental changes during the early Holocene
The deposition of marl within the lacustrine environment at Quoyloo Meadow indicates increased catchment stability probably driven by rising temperatures. Calcium and bicarbonate supersaturation and precipitation in lake waters is temperaturedependent, and occurs once values reach approximately 10°C (Kelts and Hsu, 1978;Cole and Weihe, 2015). At Quoyloo Meadow this thermal threshold is reached at the boundary between Units 2c and 3a (11.48 ± 0.48 cal a BP). This inference of amelioration is supported by the T Jul reconstruction which shows a marked increase in temperatures from 6.7 to 12.4°C (11.93 ± 0.30 to 11.16 ± 0.37 cal ka BP), along with the replacement of tundralike plant communities with taxa such as Rumex and Salix. Collectively these indicators signal the onset of Holocene conditions.
Between 194 cm (11.44 ± 0.44 cal ka BP) and 184 cm (10.73 ± 0.23 cal ka BP) two minima in δ 18 O values are observed (Fig. 5): QMO-e1 centred on 191.25 cm (11.13 ± 0.37 cal ka BP) and QMO-e2 centred on 186.25 cm (10.78 ± 0.18 cal ka BP). These low values are relatively subdued (<1‰) when compared to other δ 18 O records of the period (e.g. Whittington et al., 1996Whittington et al., , 2015, but taken in isolation, it is possible that decreases of this scale could still reflect temperature oscillations in the order of 2-3°C (Andrews, 2006;Candy et al., 2011). Importantly, however, no such signal can be gleaned from the T Jul reconstruction at the resolution undertaken, which shows a continued, if not a slightly slowed warming through this interval (QMC-2/3; Fig. 7). Low land-pollen concentrations and high levels of Pediastrum at the transition to the Holocene, and continuing during the QMO-e1 depletion, indicate a sparsely vegetated landscape prone to soil erosion (QMP-5;Abrook et al., 2020a). This phase is succeeded by the establishment of a dwarf-shrub heathland (QMP-6), as well as a significant increase in the frequency of macro-charcoal fragments (Fig. 6). These latter changes coincide with the QMO-e2 isotopic depletion. With no clear fluctuation in the T Jul reconstruction, it is possible that the δ 18 O signal reflects changes to the hydrological regime as opposed to air temperature at this point of the record (Leng and Marshal, 2004). A more arid environment would probably have had a negligible impact on the pioneering vegetation established on the landscape during the QMO-e1 depletion, but following heathland development drier conditions during QMO-e2 may have led to drought stress (Bell and Tallis, 1973). This in turn may have increased the availability of biomass on the landscape, subsequently leading to an increase in fire frequency and charcoal production (Fig. 8). At present, our data cannot reconcile the significance of these isotopic excursions, especially with regard to aridity, but further research is addressing this.
Evidence between 184 cm (10.73 ± 0.23 cal ka BP) and 164 cm (10.28 ± 0.23 cal ka BP) indicates a phase of relative stability. Heathland taxa are gradually replaced by a herbaceous grassland, and T Jul continues to rise, reaching a high of 15.6°C at 168.5 cm (10.38 ± 0.27 cal ka BP) (Fig. 7). However, a number of minor perturbations are noted through this interval. At 174.5 cm (QMC-e2; 10.52 ± 0.32 cal ka BP) a reduction of 0.8°C is inferred from the T Jul reconstruction, and whilst no comparative oscillation can be identified in the isotope record, a slight inflection in the PrC dataset can be observed (Fig. 6). Closer scrutiny of the palynological record, however, illustrates that the PrC signal is probably responding to natural variability in the grassland communities (Abrook et al., 2020a). Similarly, a minor depletion (<1‰) in the δ 18 O record at 170.75 cm (QMO-e3; 10.43 ± 0.31 cal ka BP) has no proportional response in T Jul or local vegetation communities. It is therefore probable that these minor perturbations in temperature and δ 18 O values, if genuine, had a limited impact on the environment surrounding Quoyloo Meadow.
The most significant changes in the Quoyloo Meadow proxy series during the Holocene occur between 162.5 and 151.5 cm (10.25 ± 0.19 to 10.07 ± 0.21 cal ka BP) where T Jul displays a two-stage drop in values (Fig. 7). An initial fall of 1.6°C (162.5-161.5 cm; 10.25 ± 0.19 to 10.22 ± 0.16 cal ka BP) is followed by a temporary recovery, and a further decrease of 2.6°C between 159.5 and 156.5 cm (10.19 ± 0.09 to 10.16 ± 0.11 cal ka BP; QMC-e3). Concurrent with this second reduction in temperature is a depletion in the δ 18 O record of −0.52‰ (QMO-e4). The total drop in T Jul of 3.75°C is the largest seen in any Holocene oscillation identified in the British Isles (e.g. Lang et al., 2010;Blockley et al., 2018), and this factor, combined with the −0.52‰, suggests a significant shift in both mean annual and summer temperature variability that is further reflected in the vegetation assemblage. Retractions of Empetrum heath are apparent, and whilst grassland communities were already dominant on Orkney, elevated percentages of Rumex and increased macro-charcoal remains indicate a period of landscape reorganization, and greater incidences of burning (Abrook et al., 2020a).
Following recovery from the preceding event, the rest of the studied Holocene sequence at Quoyloo Meadow shows no evidence of further short-lived fluctuations in climate. The prevalence of grassland communities is gradually reduced as woody vegetation such as Betula and Corylus becomes more established, and temperatures remain stable.

Short-lived 'centennial' climate oscillations
The precision of the Quoyloo Meadow age-model offers an excellent opportunity to make an independent chronological comparison of the proxy series to the NGRIP regional stratotype (Fig. 8). Centennial-scale climatic events have been identified where a change is recorded across multiple lines of palaeoclimatic evidence, or where a change is recorded in a single proxy or data-point, and reasonable justification can be made as to why this may not be reflected in other data series or surrounding datum (Table 2; Fig. 8). Here we consider the timing and regionality of the short-lived climatic events described from Quoyloo Meadow with a specific emphasis on the early Holocene.

Late Pleistocene oscillations (QMC-e1)
The reduction in mean July temperatures observed at Quoyloo Meadow (c. 214.5 cm; QMC-e1; Fig. 7) during the early Windermere Interstadial (c. GI-1), and the shift to more openhabitat taxa, is an event that is also recorded in other lacustrine records across NW Europe (e.g. Brooks and Langdon, 2014;Heiri et al., 2014;Moreno et al., 2014). In the regional Greenland stratotype, three short-lived centennial-scale oscillations have been recorded in the GI-1 interval: GI-1d, GI-1c and GI-1b (Johnsen et al., 1992;Björck et al., 1998;Rasmussen et al., 2014). Determining their chrono-stratigraphic equivalence in terrestrial records has, however, proved problematic due to uncertainties associated with site-specific age-models and the comparability between different proxy series (Lowe and Walker, 2000;Blaauw et al., 2007;Blaauw, 2012). At Quoyloo Meadow, correlation is assisted by the Borrobol (218 cm) and Penifiler (213 cm) tephras, two isochronous markers which bracket the onset and termination Figure 8. A chronological comparison of the Quoyloo Meadow proxy series with the NGRIP δ 18 O regional stratotype. Both records are plotted on a common chronological scale with the coloured 'blurring' illustrating a one sigma uncertainty. Within this uncertainty a correlation can be made between QMC-e1 and GI-1d. Throughout the early Holocene, the Quoyloo Meadow proxy series exhibits a series of 'events' which have no clear analogue within the ice-core record. A possible exception to this is an event at c.10.8 ka, but the possible corresponding oscillation in the NGRIP record has yet to be formally recognized by the ice-core community. This difference between Quoyloo Meadow and NGRIP during the early Holocene suggests the Orkney Isles were responding to forcing factors that were either different from those operating over Greenland, or were more strongly manifest at mid-latitude sites. NGRIP data are from Rasmussen et al. (2014). [Color figure can be viewed at wileyonlinelibrary.com]. of QMC-e1. The precise age estimate for these tephra (see Table 1) provides a well-constrained age for the climatic oscillation (14.04 ± 0.17 cal ka BP), which, if the error range is taken into account, falls within the ice-core age span of 14.08-13.94 b2k for the GI-1d oscillation . On this basis, and given the relationship of these well-dated tephra with temperature responses identified elsewhere (e.g. Matthews et al., 2011;Brooks et al., 2012;Brooks et al., 2016;Candy et al., 2016), we propose that the oscillation detected at Quoyloo Meadow is the equivalent of the GI-1d cooling event in the Greenland ice-core record (Fig. 8). Unfortunately, the limited stratigraphic resolution afforded by the Quoyloo Meadow record through units 1b and 1c (Fig. 2) means that it is difficult to compare the magnitude of this cooling (c. 1.7°C) with other contemporaneous records. Recourse to other sequences which have the capacity to be resolved to a higher resolution, e.g. Crudale Meadow (Whittington et al., 2015;Timms et al., 2018) and Loch of Sabiston (Cochrane, 2020), may offer a better opportunity for further study of this event on Orkney.
Early Holocene oscillations (QMO-e1, e2 and e4; QMC-e3 and e4) During the early Holocene, the Greenland δ 18 O trend is characterized by gradually increasing isotopic values (Vinther et al., 2006;Rasmussen et al., 2007;Fig. 8). Punctuating this trend, however, are a number of major isotopic excursions, the 11.4-ka event, the 9.3-ka event and the 8.2-ka event (Rasmussen et al., 2007). These events are consistently expressed in DYE-3, GRIP and NGRIP, and hence represent regionally if not hemispherically significant climatic oscillations (Rasmussen et al., 2007;. Some of these events have been identified in European proxy records, but correlation with the Greenland ice-core sequence is uncertain, particularly the 11.4-ka event which has been linked, perhaps incorrectly, with the European Pre-Boreal Oscillation (PBO) (Björck et al., 1996). The PBO, dated to c. 11.2 ka, has been evidenced through greater winter-summer seasonality, a reversion to grassland-type communities and a widespread drying of the climate (van der Plicht et al., 2004;Andresen et al., 2007;Bohncke and Hoek, 2007;Bos et al., 2007;Jessen et al., 2008;Wohlfarth et al., 2017). Chronological uncertainty and variable stratigraphic resolution, however, inhibit establishing the synchroneity of these responses, and whether they can be attributed to the same forcing factors operating over Greenland at this time. It is thus unclear whether the PBO and the 11.4-ka event represent either a single synchronous event, or a temporally and spatially transgressive event, or whether the 11.4-ka event and PBO represent different episodes over Greenland and Europe. It should also be noted that the PBO fits into a broader period of climatic instability described from European sequences dating to between c. 11.25 and 10.5 ka, but which seemingly have no clear analogue within the Greenland ice-core records (Diefendorf et al., 2006;Andresen et al., 2007;Jessen et al., 2008;Fiłoc et al., 2018 and references therein). The Quoyloo Meadow record matches with this continentalwide pattern of early Holocene climatic instability. Isotopic events QMO-e1 (11.13 ± 0.37 cal ka BP) and QMO-e2 (10.78 ± 0.18 cal ka BP) coincide with phases of vegetational change (QMP-5/6; Fig. 6), and increased macro-charcoal is associated with QMO-e2 (Fig. 8). As far as the resolution of the reconstruction allows there is no discernible T Jul response associated with these isotopic events at Quoyloo Meadow. This suggests summer temperature change was not a major component affecting the records, and instead implicates aridity and/or greater winter temperature ranges as principal forcing factors (e.g. Björck et al., 1997;Denton et al., 2005). Indeed, the impact of winter as opposed to summer conditions as drivers of abrupt early Holocene changes has been suggested by a range of data from other sites in Greenland (Andresen et al., 2007), the Faroes (Andresen et al., 2007;Jessen et al., 2008), the Netherlands (van der Plicht et al., 2004;Bohncke and Hoek 2007;Bos et al., 2007), Germany (Mekhaldi et al., 2019) and the British Isles (Blockley et al., 2018;Abrook et al., 2020b). However, summer temperature variability cannot be excluded entirely, as C-IT reconstructions from Star Carr, England, show reductions contemporaneously with isotopic events at 11.4 and 11.1 ka BP (Blockley et al., 2018). Quoyloo Meadow also exhibits summer temperature variability in the early Holocene, but this does not occur until QMC-e2 (10.52 ± 0.32 cal ka BP; Fig. 7). The timing of QMC-e2 and the preceding isotopic events at Quoyloo Meadow, means that, within chronological error, broad correlations can be drawn to the PBO and other phases of environmental instability recorded in European archives at approximately 11.1 (QMO-e1), 10.8 (QMO-e2), 10.5 (QMC-e2) and 10.45 (QMO-e3) cal ka BP (Fig. 8).
Between 10.25 ± 0.19 and 10.07 ± 0.21 cal ka BP the palaeoenvironmental archive at Quoyloo Meadow records its most distinct response to centennial-scale climate forcing. A strong isotopic excursion (QMO-e4), a reduction in T Jul estimates (QMC-e3), and a greater presence of open ground vegetation (QMP-8/9) suggests a major disruption to the local environment (Fig. 8). The precise timing of this event is constrained by the Saksunarvatn Ash and Fosen Tephra, two isochronous markers which act as a distinct tephra couplet within the Holocene (Lind et al., 2013;Timms et al., 2017Timms et al., , 2019. Whilst studies have questioned the chemical uniqueness, and hence validity of the Saksunarvatn Ash as an isochronous marker (Davies et al., 2012;Jennings et al., 2014;Óladóttir et al., 2020), we have found no conclusive evidence of contemporaneous basaltic ash layers (see QM1 175 and 170 in Timms et al., 2017), nor did we identify glass shards bearing low MgO wt% values within our chemical analyses; the latter is a distinctive characteristic of other 'Saksunarvatn-like' tephras deposited within the same time-frame (c. 10.5-9.9 ka BP; Wastegård et al., 2018). We therefore consider the 'Saksunarvatn Ash' to be a reliable stratigraphic and chronological tie-point within the Quoyloo Meadow record. The Saksunarvatn Ash has previously been associated with a climatic perturbation termed the '10.3-ka event' (Björck et al., 2001;Andresen et al., 2007). Evidence for this oscillation has principally come from archives in Europe and North America, which show a significant environmental disturbance at this time (Table 3). On the basis of the tephra evidence and the magnitude of the proxy response at Quoyloo Meadow, we propose a correlation to this event.
In the British Isles very few early Holocene centennial-scale climatic perturbations have been reliably described or dated, and emphasis has historically been placed on providing evidence for the PBO, 9.3-ka and8.2-ka events (e.g. Whittington et al., 1996, 2015;Diefendorf et al., 2006;Marshall et al., 2002Marshall et al., , 2007Lang et al., 2010;Holmes et al., 2015Holmes et al., , 2020Blockley et al., 2018). These events are the largest, most widely detected and often considered the most important in understanding early Holocene climatic complexity. However, these events are only part of the complexity, and a tendency to focus on them has masked the intricacy of local-regional expressions of climate change following largescale reorganization of the ocean-atmospheric system. The evidence presented here would suggest that the event recorded at c. 10.3-10.0 cal ka BP had a greater impact on the climate Table 3. Compilation of palaeoenvironmental site data for which there is evidence of a temperature decline and/or environmental fluctuation at approximately 10.3 ka BP.
The present distribution of evidence would suggest that the 10.3-ka event is infrequently identified in palaeoclimate sequences, and those where it has been found are preferentially located in northern Europe and North America, suggesting an oceanic (North Atlantic) origin, or mechanism of propagation. Data have been primarily extracted from Kaufman et al. (2020)    and landscape of Orkney than any other early Holocene event recorded in the British Isles. The magnitude is approximately double the impact of the PBO recorded at Star Carr (Blockley et al., 2018) and approximately 1.7°C greater than the 9.3-ka event described from Haweswater (Lang et al., 2010). Whilst an oscillation at c. 10.3 ka has been noted at Haweswater (Table 3), it is comparatively subdued when compared to the Quoyloo Meadow record. Why this should be is not immediately clear, and why this event should remain undetected in other climate archives across the British Isles is also uncertain. A possible explanation is that there was a strong latitudinal gradient related to the impact of this event similar to other climatic transitions detected in the British Isles (e.g. Coope et al., 1998;Isarin and Bohncke, 1999;Brooks and Langdon, 2014), and that many climate records that span this interval have not been investigated to a sufficiently high resolution as to detect change.

Climatic drivers in the early Holocene
Freshwater fluxes into the North Atlantic during the early Holocene have long been considered as possible driving mechanisms behind abrupt climate variability (Clark et al., 2001;Fisher et al., 2002;Teller et al., 2002;Nesje et al., 2004). Specifically, multiple outpourings of glacial lake Agassiz-Ojibway into the Atlantic, and periodic drainage of the Baltic are linked to episodes of climate variability dating to between c. 11.7 and 9.0 cal ka BP (Björck et al., 1996;Bodén et al., 1997;Teller et al., 2002;Jessen et al., 2008;Smith et al., 2011). Influxes of fresh water disrupted Atlantic meridional overturning (AMOC), forcing polar waters southwards through increased ocean stratification and inhibiting the northerly advection of heat in the Atlantic Ocean (Björck et al., 1996;Bond et al., 1997;Fisher et al., 2002;Andresen et al., 2007). The 10.3-ka event can therefore be potentially linked to drainage of the lake Agassiz McCauleyville stage (Teller et al., 2002), as well as to a near simultaneous outpouring of the Nedre Glamsjø lake in Norway (c. 10.3 cal ka BP; Longva and Thoresen, 1991;Nesje et al., 2004). The 10.3-ka event was perhaps further influenced by the drainage of the Baltic Ancylus lake shortly thereafter at c. 10.16-9.8 cal ka BP (Björck, 1995;Nesje et al., 2004). The combined impacts of these 'freshening' episodes spanning c. 10.3-9.8 ka BP may have been enough to weaken AMOC and trigger a cooling episode across much of the Northern Hemisphere. In Scandinavia a multi-stage glacial readvance, known locally as the 'Erdalen' events Dahl et al., 2002;Matthews et al., 2008;Shakesby et al., 2020; Table 3), has been attributed to an increase in winter precipitation and by a significant drop in summer temperatures at this time (Dahl et al., 2002). Contemporaneous records from across Europe, North America and the North Atlantic agree with this, and suggest a period of environmental instability, reduced catchment productivity and an increase in ice-rafted debris (IRD event 7) accompanied the shift in temperatures (Table 3; Bond et al., 1997;Björck et al., 2001). An alternative explanation for short-term climate cooling in the early Holocene, other than meltwater, is the role of solar forcing. Evidence of 10 Be and 14 C fluctuations in ice-core and tree-ring records, respectively, suggest that a number of climatic events identified through the early Holocene coincide with solar events (Bond et al., 2001;Muscheler et al., 2004;Adolphi et al., 2014). This mechanism has previously been used to explain the origin of the 10.3-ka event (Björck et al., 2001), and may have been a more prominent contributor to early Holocene climate variability than previously thought (Jessen et al., 2008). This mechanism has also recently been invoked by Mekhaldi et al. (2019) to explain why the PBO appears greater in magnitude and extent than later early Holocene events. A solar minimum coinciding with the PBO is suggested to have 'pre-conditioned' the climate so that it was more susceptible to episodes of meltwater flux. By contrast the 10.3-ka event, whilst also associated with major glacial-lake drainage episodes, had not been 'pre-conditioned' by a solar minimum, but instead occurred at a time of heightened solar activity, effectively muting the signal of the outburst floods. Hence, the 10.3-ka event is less well represented in palaeoclimate records when compared to the PBO phase of instability (Mekhaldi et al., 2019).
At Quoyloo Meadow, however, this is not the case, with the 10.3-ka event notably more pronounced than any preceding short-lived oscillation of the early Holocene, and represented quite clearly by a two-stage structure in mean July temperature (Fig. 7). The later and larger reduction has a coeval response in δ 18 O values, which perhaps suggests the influence of two or more forcing episodes or a divergence between summer and mean annual temperature signals. Whilst speculative, the twostage proxy response detected at Quoyloo Meadow might suggest that an initial meltwater outpouring had only a minimal impact on mean summer temperatures across Orkney, allowing the islands to return temporarily to a stabilized state shortly thereafter. This was followed by a further event(s) which may be represented in the Quoyloo Meadow record by the larger secondary oscillation detected in both the C-IT and δ 18 O records, summer and mean annual indicators respectively. The divergence in proxy response observed may relate to: (i) the magnitude of the outburst floods; (ii) the proximity of the events in relation to the Orkney Isles; (iii) the condition of AMOC at the time; and/or (iv) the susceptibility of the Orkney Isles to change (i.e. how close the system was to climate and environmental thresholds). It is probable that a weakened AMOC from the accumulative effects of several closely timed outpourings may have facilitated a stronger response than expected from a single event, and due to the more 'proximal' location of Orkney and the Faroes to the North Atlantic, this was preferentially recorded here as a larger magnitude event when compared with records from other areas of NW Europe.

Conclusions
The tephrochronological approach adopted and the updated tephra age estimates presented here enable a decadal-level of chronological precision which could not have been achieved at Quoyloo Meadow with radiometric techniques alone. This demonstrates that tephrochronology is a valuable means by which chronologies can be developed in traditionally 'hard-todate' sequences such as carbonate precipitating lakes and those devoid of plant macrofossils for radiocarbon dating. With the number of well-dated (crypto-)tephra that are known to be present in palaeoenvironmental and archaeological records (e.g. Davies et al., 2012;Blockley et al., 2014;Timms et al., 2019), we echo previous calls and recommend this approach be adopted more widely to develop and supplement chronologies from other sequences across the British Isles and NW Europe that span the LGIT.
The Quoyloo Meadow proxy series provides further insight into the climatic variability of northern Britain during the LGIT. In particular, the study presents new biostratigraphic and isotopic data to support evidence of multiple short-lived climatic reversal events impacting the Orkney Isles (see Whittington et al., 2015;Abrook et al., 2020a). These events can be shown to be broadly equivalent to the widely recognized GI-1d and PBO phases of instability. However, we also identify, with varying degrees of confidence, evidence for further climatic instability during the Holocene and dating to c. 10.8, 10.5, 10.45 and 10.3 cal ka BP.
The evidence from Quoyloo Meadow shows that the most pronounced cooling event, other than the Loch Lomond Stadial, occurred between c. 10.3 and 10.0 cal ka BP which is indicated by a drop of 3.75°C in mean July temperatures, a shift in vegetational communities to those indicative of 'less stable' climatic regimes, and a decrease in δ 18 O values. We interpret this as evidence of the '10.3 ka event', a climatic perturbation infrequently identified in early Holocene palaeoenvironmental reconstructions, and is the first reliably dated incidence of this event in the British Isles. This is a significant finding as the magnitude of the event in the Orkney record is considerably larger than any other early Holocene perturbation detected in the British Isles to date. However, we cannot exclude the fact that this magnitude of response may reflect factors either unquantified in the current study or unique to Quoyloo Meadow itself.
The results from Quoyloo Meadow show that not only the LGIT, but also the early Holocene, were both dynamic and complex intervals of climatic and environmental change. Despite research into many LGIT and early Holocene sequences from the British Isles, detecting the exact frequency and magnitude of climatic events remains difficult. Continuing to develop similar records with high chronological precision combined with high stratigraphic resolution will enable better understanding of abrupt climatic events in terrestrial records across NW Europe during these key time intervals. While the Greenland ice-core records continue to be the stratigraphic template for the North Atlantic region, it now seems that they do not exhibit all of the events being discovered in terrestrial palaeoclimate records in Europe. The identification of centennial-scale events such as GI-1d or the PBO, with a wide geographical range, is well known, but the Quoyloo Meadow record has shown that it is also possible to detect more spatially restricted events. Understanding why these events are expressed in this manner, when their driving mechanisms appear to be similar to those that are more strongly represented, is a clear priority for future investigations.

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