The geomorphology of the study area was mapped to identify processes influencing sedimentation in Lake Vatnasetvatnet and to determine potential sediment sources. Avalanche-prone terrain was identified by delineating potential snow accumulation and trigger zones, avalanche tracks, and snow avalanche depositional landforms. Field mapping was conducted during two field seasons, supplemented by LiDAR-derived elevation models (0.25 m resolution; The Norwegian Mapping Authority, https://www.hoydedata.no) to generate high-resolution hillshade and slope maps. Changes in vegetation and avalanche track activity were assessed using aerial imagery from different years, dating back to 1971 (https://www.norgeibilder.no, last access: 14 January 2025).

The bathymetry and sediment distribution of the lake were surveyed along four west-east and three north-south transects (Fig. 2) using ground-penetrating radar (GPR) equipped with a 25 MHz antenna and GPS. The GPR data were processed using the online version of Mala Vision, employing three filters: DC Offset, background removal, and extensive amplification with a gain of 300. A velocity of 33 m ns⁻¹ was applied to calculate water depth and sediment thickness. The GPR data facilitated the selection of coring sites in undisturbed areas with a flat bottom. Three piston cores (VAP109: 294 cm, VAP209: 255 cm, VAP309: 217 cm) were retrieved from the lake ice in 2009 using a modified piston corer (Nesje, 1992) with a 6 m long, 110 mm wide PVC tube. In 2022, an additional sediment core (VAG122: 117 cm) was obtained using a UWITEC gravity corer with a 2 m long, 90 mm wide plastic tube to secure an undisturbed sediment-water interface. Cores were sectioned, sealed, and stored at 4 °C.

The sediment cores were split, photographed, and visually described to document sediment colour, grain size variations, and lithological units. Physical properties were measured on 1 cm³ bulk samples collected at 0.5 cm intervals from VAP109 and VAP209. Each sample was weighed before and after overnight drying at 105 °C to determine dry bulk density (DBD) and water content. Loss-on-ignition (LOI) was conducted at 550 °C and reweighed to quantify organic matter content. Based on the DBD and LOI results, 59 and 70 samples from potential event layers in VAP109 and VAP209 were selected for grain size analysis. The samples were treated with hydrogen peroxide (H₂O₂) to remove organic material and analysed using a Micrometrics SediGraph 5120. The results were processed using GRADISTAT (Blott and Pye, 2001) in Microsoft Excel.

Magnetic susceptibility (MS) was measured at 2 mm intervals using a Bartington MS2E sensor on a GEOTEK Multi-Sensor Core Logger (MSCL) (Gunn and Best, 1998) to detect minerogenic input variations. X-ray fluorescence (XRF) scanning was performed to analyse the geochemical composition of the sediments. Cores VAP109, VAP209, and VAG122 were scanned at 30 kV and 50 mA with 1 mm resolution using an ITRAX core scanner (COX Analytics) (Croudace et al., 2006). To account for down-core variations in water content and organic material, elemental count rates were normalised against the total incoherent and coherent scattering (Int + Coh) (Kylander et al., 2011; Davies et al., 2015). Subsequently, the data was standardised to Z-scores (Z=(x-μ)/σ) to ensure consistent comparison across proxies and sediment records. The one mm-resolution XRF data were smoothed using a 5-point moving average to reduce noise. Computed tomography (CT) scanning was conducted to generate 3D density models of sediment structures in VAP109 and VAG122 using a ProCon XRAY CT-ALPHA. A helical scan was performed with 2400 images per rotation at 126.5 kV and 800 µA, using a 267 ms exposure time and a 0.5 mm Cu-filter to minimise beam hardening (Brooks and Di Chiro, 1976). The CT projections were reconstructed using 2-binning to enhance contrast and reduce file size, yielding a final voxel resolution of ∼ 98 µm. A CT greyscale dataset was extracted along a spline through the sediment cores at 0.1 mm resolution, enabling direct comparison with the XRF dataset for event detection.

For radiocarbon dating, 10 sediment samples were extracted from the sediment cores (Table 1). After sieving, terrestrial macrofossils were identified and dried overnight in sterilised vials before being sent to the Poznań Radiocarbon Laboratory for Accelerator Mass Spectrometry (AMS) analysis. The final age-depth model was constructed using the rBacon package (Blaauw and Christen, 2011) in RStudio v. 4.3.1, employing the IntCal20 northern hemisphere calibration curve (Reimer et al., 2020). Before modelling, distinct minerogenic layers, representing potential event deposits, were identified following the methodology outlined in Sect. 3.4. Layers thicker than 3 mm were classified as instantaneous deposits in rBacon. Calibrated ages were estimated at 0.1 cm intervals to match the resolution of the XRF and CT data.

We applied two independent approaches to identify the potential event layers in the sediment record. Snow avalanches introduce abrupt pulses of minerogenic material into lakes (e.g., Nesje et al., 2007), resulting in distinct sedimentary signatures that sharply contrast the background sedimentation. To detect these rapid shifts in the sediment composition, we calculated the rate of change (RoC) for XRF elements by dividing the change in the parameter (δy) by its corresponding time/depth interval (δt) (Støren et al., 2010; Birks, 2012). To further validate these results, we applied interactive thresholding to CT greyscale data, which reflects sediment density, using Thermo Scientific Avizo 2020.2 software (Røthe et al., 2019b; Hardeng et al., 2024). The proportion of sediments exceeding predefined density thresholds was then quantified (vol %) to identify depth intervals corresponding to avalanche layers. Visually distinct minerogenic layers were used to calibrate the RoC and CT greyscale thresholds for event detection.

The Vatnasetvatnet catchment (Fig. 2) is a glacially sculpted landscape with thick (> 1–1.5 m) till deposits and mires in low-lying areas. The till cover thins with increasing elevation toward Vasslifjellet (1056 m a.s.l.), where large areas of exposed bedrock are present. A well-defined, north-oriented cirque is located 300 m east of Lake Vatnasetvatnet along the catchment's eastern boundary. The boundary between the cirque and the lake catchment is marked by a prominent ridge measuring 8–11 m in height and ∼ 250 m in length. The ridge consists of sub-angular boulders embedded in a matrix of finer-grained sediments. Downslope from the ridge, a channel-like depression approximately 5 m deep extends toward the lake basin.

Slope processes are concentrated along the southern margin of Lake Vatnasetvatnet, where two well-defined avalanche tracks (Tracks 1 and 2) descend from the northern slopes of Vasslifjellet (Fig. 2). Both tracks originate in north-facing accumulation zones with slope gradients of 30–50° (Fig. 3). From these trigger zones, the tracks extend downslope through treeless corridors where bedrock protrusions and a patchy, thin debris cover indicate ongoing surface erosion (Fig. 4A). The tracks terminate in runout zones composed of coarse colluvial fans that extend into the lake basin. Trees are absent from the central parts of the colluvial fans, and tree size/age increases progressively towards the outer edges (Fig. 4D). The colluvial fans comprise angular to sub-angular gravel, pebbles, and boulders, with finer sediments perched atop the coarser material (Fig. 4C). We observed snapped trunks, uprooted trees, and patches of sand and gravel atop the vegetation cover along Avalanche Track 1 (Fig. 4B). Both tracks contain small, intermittent streams that drain limited sections of the Vasslifjellet mountain face (0.35 and 0.12 km²). Given their limited drainage area, the contribution of these streams to the influx of allochthonous minerogenic sediments into the lake is minimal.

Lake Vatnasetvatnet consists of two distinct basins (maximum depths of 33 and 31 m) separated by a submerged ridge at ∼ 21 m water depth (Fig. 2A). The GPR profiles (Fig. 2B) reveal relatively thick (> 6 m) sediment infill in both basins, with the highest accumulation in the western basin. The radar stratigraphy displays a continuous sediment sequence over a strong basal reflector, likely representing bedrock or compacted glacial deposits (till). Along the southern lake margin, where the avalanche tracks terminate, chaotic and hyperbolic reflections indicate coarser, unsorted deposits, likely associated with the colluvial deposits below the avalanche tracks.

Four sediment cores were retrieved from Lake Vatnasetvatnet: VAP109 (294 cm) and VAG122 (117 cm) from the western basin, and VAP209 (255 cm) and VAP309 (217 cm) from the eastern basin (Fig. 2). The sediment record was divided into two units visually (Fig. 5): Unit A, comprising organic-rich gyttja interspersed with distinct minerogenic layers (potential event layers 1–25 in the visual sediment log, Fig. 5), and Unit B, which is a finely laminated inorganic silt deposit.

VAP309, retrieved near Avalanche Track 2 in the eastern basin (Fig. 2), contains thick (> 10 cm) layers of coarse minerogenic sediments (coarse sand and gravel) with plant fragments. Sharp erosional boundaries indicate sediment remobilisation, likely due to its proximity to the colluvial fan. In contrast, VAP209 cored at a more distal position in the eastern basin, exhibits stratigraphy comparable to VAP109 from the western basin, with distinct minerogenic layers 1-19 correlatable between the two records (Fig. 5). While both VAP109 and VAP209 provide high-resolution and undisturbed sediment sequences, VAP109 is longer and better situated to capture a more balanced record of avalanche deposits, preserving both smaller and larger events. Additionally, geomorphological evidence (Sect. 4.1) confirms recent avalanche activity in Avalanche Track 1, supporting the focus on the western basin for avalanche reconstruction.

To recover an undisturbed sediment-water interface, gravity core VAG122 was retrieved near the coring location of VAP109 (Fig. 2). The two cores were correlated based on broadly matching sedimentological features, including a sequence of distinct minerogenic layers followed by a more homogeneous sediment section without distinct layers. The visual correlation of overlapping sections was not entirely conclusive due to slight discrepancies between the records, which may be attributed to (1) the location of VAG122 closer to Avalanche Track 1, potentially capturing events absent in VAP109, or (2) differential compaction resulting from varying coring techniques. That said, the correlation is supported by extrapolation of the age-depth model from VAP109, which dates the top of the core to 432 cal yr BP, indicating that approximately 21 cm of sediment is missing. The uppermost section of VAG122 was used to fill the missing interval to account for this; the resulting composite core, VAPG (315 cm), serves as the primary sedimentary record for this study.

A Principal Component Analysis (PCA) was conducted to support sediment fingerprinting and aid the classification of samples with similar characteristics into stratigraphic units and facies. Two separate PCAs were conducted, the first for the entire core (Fig. 6A) and the second for just Unit A (Fig. 6B). The analyses included a selection of XRF elemental data (K, Ca, Sr, Ti), the Inc / Coh-ratio, MS, greyscale data, and bulk measurements (LOI, water content, and DBD). Before analysis, the high-resolution scanning data were resampled to match the resolution of the bulk measurements by averaging values within corresponding 0.5 cm intervals. The first PCA identified two primary clusters (Fig. 6A), which align with the visually defined Units A and B. Samples from Unit B are confined to the lower part of the core (315–260 cm), while Unit A samples correspond to the upper section (above 260 cm, Fig. 5). The first principal component (PC1) accounts for the vast majority of the variance in the dataset (82 %). It reflects a gradient in minerogenic content, with strong positive correlations to Sr, Ca, K, Ti, DBD, MS, and greyscale values. Conversely, LOI, water content, and the Inc / Coh ratio negatively correlate with PC1, as they are associated with more organic-rich sediments. Samples from Unit B are separated from the minerogenic layers in Unit A by having a generally higher minerogenic content (reflected in higher PC1 scores) and along PC2, which primarily reflects higher MS values.

The second PCA (Fig. 6B) aimed to separate the visually identified minerogenic layers from the gyttja in Unit A. The results show a similar trend as the first PCA, where PC1 account for most of the variability (78 %), while PC2 account for 9 %. Samples from these minerogenic layers plot markedly higher along the PC1 axis than gyttja samples, and reflect short-lived pulses of minerogenic input that we interpret as event deposits. Setting an exact threshold between gyttja and event layer samples is challenging, likely due to the 0.5 cm sampling resolution, which can result in mixed samples. Nevertheless, gyttja is characterised by organic-rich sediment, while the event layers exhibit markedly higher minerogenic input. Ti shows the strongest correlation with PC1 (r=0.98), thus serving as the most effective proxy for identifying event layers.

Unit B (315–260 cm, Fig. 5) is characterised by finely laminated silt with a high mean dry bulk density (DBD) of 1.56 g cm⁻³. Loss-on-ignition (LOI) and water content are consistently low, averaging 0.31 % and 13.6 %, respectively, indicating minimal organic content and a high degree of compaction. MS values are elevated (∼ 94.2 SI ×10-5) relative to Unit A. XRF elemental count rates (Z-scores) are generally high (> 0) but display substantial variability, reflecting the fine laminations. The XRF count rates have a prominent trend, where the Z-score baseline of Ti decreases from ∼ 3 in the bottom to ∼ 0 near the transition into Unit A. The incoherence-to-coherence (Inc / Coh) ratio is inversely correlated with the XRF elemental count rates, exhibiting low values (mean: -1.63) with fluctuations corresponding to the laminations. CT greyscale values remain consistently high, indicative of the high density of the sediments. The upper boundary at 260 cm is sharp, marking an abrupt transition into Unit A.

Unit A (0–260 cm, Fig. 5) consists of dark brown gyttja with relatively high organic content (> 30 %) and water content (> 70 %), resulting in low DBD (< 0.30 g cm⁻³) and MS (mean 21.3 SI ×10-5). The LOI is notably higher (> 40 %) in the lower part of Unit A (260–180 cm), and a weak decreasing trend in water content down-core reflects gradual sediment compaction. XRF elemental count rates generally show lower Z-scores relative to Unit B (< 0), while the Inc / Coh ratio remains relatively high (mean: 0.35).

The minerogenic layers (1–9 cm thick) interspersed within the gyttja in Unit A exhibit distinct physical and geochemical characteristics compared to the surrounding gyttja and represent instantaneous deposits, i.e., event layers (Fig. 7). Their high minerogenic content is reflected by elevated DBD values (0.35 to 0.88 g cm⁻³, mean 0.48 g cm⁻³), with correspondingly lower LOI (2.9 % to 27.2 %, mean 14.2 %) and water content (31.3 % to 71.7 %, mean 62.8 %). Several layers contain macrofossils, which is reflected by the high variability in LOI values. MS varies widely between 8 and 121 SI × 10⁻⁵, with higher values generally associated with the coarser samples. XRF elemental counts show elevated values. Grain size distributions indicate a dominance of coarse silt (25 %–38 %), medium silt (18 %–30 %), and a variable proportion of very coarse silt and sand (> 10 %), with overall poor sorting (sorting index ∼ 3.2). The layers consist of a coarse-grained basal layer (sand), in some cases with irregular lower boundaries, indicating potential erosion (Fig. 7). The basal layer grades into a silty deposit with dispersed organic fragments, which is capped by a fining upwards sequence of fine silt and clay.

Ten macrofossil samples containing leaves, bark, birch fruits and fern leaves were extracted from VAP109 for AMS radiocarbon dating. The resulting radiocarbon ages (Table 1) are in stratigraphic order, with progressively older ages down-core. The oldest sample, obtained at 235.5 cm depth (256.5 cm in the composite VAPG core), has a median calibrated age of 10 109 cal yr BP.

Before age-depth modelling, event layers were identified using the two approaches outlined in Sect. 3.4: the rate of change (RoC) of XRF elemental count rates and interactive thresholding of CT greyscale data (Fig. 8). As Ti shows the strongest correlation with PC1 (r=0.98), it was selected as the most effective proxy for event-layer detection. RoC analysis was therefore applied to the Ti record, with the threshold for event detection set at the 95th percentile (rationale in section 5.3). For CT thresholding, we classified depths with more than 40 % minerogenic sediment as event layers. Layers > 3 mm thick were marked as instantaneous deposits in the rBacon age-depth model to improve chronological accuracy and better reflect true depositional ages (Fig. 8). Following the establishment of the age-depth model, the data were resampled to their corresponding age intervals, and the RoC of Ti was recalculated using the change in age as δt. The results of the CT greyscale thresholding were also plotted on the age scale to reconstruct the fraction (vol %) of snow avalanche deposits over time.

The results from the two approaches were summed into 100-, 200-, and 500-year intervals to assess long-term variability in event frequency and influx of snow avalanche detritus (Fig. 9). The two reconstructions show a strong correlation (r=0.76) for the 100-year summed data, indicating consistency between the methods.

A total of 187 event layers were identified using the RoC approach, and the reconstructed avalanche frequency (Fig. 9) reveals significant variations throughout the Holocene. The early to early-mid Holocene (> 6500 cal yr BP) was characterised by a low frequency, with 0–1 events per 500 years. A marked increase occurred between 6500 and 5100 cal yr BP, reaching 11 events per 500 years, followed by a decline between 5000 and 4200 cal yr BP (1–2 events per 500 years). From 4200 to 2300 cal yr BP, avalanche frequency gradually increased from 4 to 21 events per 500 years. The highest activity occurred in the late Holocene (< 2300 cal yr BP), with 23–37 events per 500 years. This period exhibits strong fluctuations, with event frequencies ranging from 1 to 16 events per 100 years. The sedimentation rate (Fig. 8C) varied between 7 and 29 cm per 500 years, mirroring the reconstructed event frequency and the influx of minerogenic sediments. Periods of low sedimentation rates coincide with low event frequencies, while peak sedimentation rates correspond to peak event frequency.

Lake Vatnasetvatnet is well-suited for recording episodic slope processes due to its isolated setting and lack of significant fluvial inflow. Steep terrain is confined to the south of the lake, making it relatively straightforward to distinguish potential sediment sources. Postglacial landscape evolution has been limited, with only small, intermittent streams and no evidence of substantial fluvial reworking.

The small cirque located approximately 300 m to the east (Fig. 2) likely hosted a local cirque glacier during the final stages of deglaciation in the early Holocene. Its north-facing orientation is favourable for preserving glacial ice by promoting snow accumulation from the prevailing southwesterly winds and reducing solar radiation, similar to the sites described in Larsen et al. (1984). The morphology, composition, and position relative to the cirque suggest that the observed ridge represents a lateral moraine. The shallow depression extending downslope from the ridge likely functioned as a meltwater outlet feeding into the lake basin.

Like the cirque, the two well-defined avalanche tracks (Figs. 2–4), with associated accumulation and trigger zones, have a north-facing aspect – sheltered from the predominant southwesterly winter storms, creating optimal conditions for slab formation through wind redistribution of snow. Slope gradients range from 30 to 50° (Fig. 3), which is gentle enough to permit snow accumulation yet steep enough for avalanches to trigger (Martinelli, 1974; Luckman, 1977). The tracks are treeless corridors with visible signs of active surface erosion, indicative of frequent avalanche activity. The absence of trees in the central parts of the runout zones, coupled with progressively older trees toward the margins, suggests regular, channelised avalanches punctuated by less frequent high-magnitude events with a broader impact. Additional evidence of recent activity includes snapped trees and gravel deposits perched atop the vegetation. The colluvial fans exhibit characteristic avalanche-derived sedimentology with poorly sorted material, including sub-angular and angular gravel, pebbles, and boulders. Gravel and pebbles are deposited in unstable positions in many places due to the melting of sediment-laden avalanche snow (Luckman, 1978; Blikra and Nemec, 1998). No morphological indicators of debris flow or fluvial activity, such as levees, lobate deposits, or channelised erosional features (e.g., Sharp, 1942; Blikra and Nemec, 1998), were observed along the avalanche tracks or within the fan systems.

These combined geomorphological and sedimentological evidence points to snow avalanches as the dominant high-energy slope process within the lake catchment and the primary source of allochthonous minerogenic sediment to Lake Vatnasetvatnet since deglaciation. Although fluvial processes likely contribute fine-grained material during extreme precipitation events, neither the catchment morphology nor the colluvial sedimentology supports fluvial or debris-flow events capable of producing the observed distinct minerogenic event layers.

The finely laminated silt in the lower part of the core (Unit B, Fig. 5) is interpreted as glaciolacustrine sediments deposited during the deglaciation of the Vatnasetvatnet catchment. The consistently high DBD values (> 1.5 g cm⁻³) and low organic content, as indicated by LOI (∼ 0.3 %), are characteristic of glaciolacustrine sedimentation (e.g., Karlén, 1976; Nesje et al., 1991; Bakke et al., 2005b). Additionally, Unit B exhibits high XRF elemental count rates, elevated MS values, and peak CT greyscale values, all of which are indicative of the dense, minerogenic sediments associated with glacial environments (Fig. 6). Fluctuations in the XRF parameters and MS are the primary drivers for the samples' position along the PC2 axis (Fig. 6). This variability is prominent in the upper half of Unit B and likely reflect pulses of sediment supply, potentially related to seasonal meltwater influx and varve formation (e.g., Regnéll et al., 2019). The trend in decreasing XRF count rates towards the transition into Unit A reflects a decrease in the mean grain size, possibly indicating a gradually more stagnant glaciolacustrine setting as the glacier retreated. The abrupt boundary between Units A and B represents the end of glacial influence, marking the transition from a glaciolacustrine environment into a lacustrine environment. The last influence of glacial processes was likely the small glacier inhabiting the cirque directly east of the catchment. The minimum age of this final deglaciation is constrained to ∼ 10 100 cal yr BP, which is supported by previous studies from the Folgefonna glacier (Bakke et al., 2005a; Røthe et al., 2019a).

Unit A is characterised by organic-rich gyttja with clay, which we interpret as the background sedimentation. The high LOI values indicate that the in situ production of organic material dominates the background sedimentation. Higher LOI values in the lower part of Unit A suggest that the organic production rate was higher during the early Holocene, likely related to the elevated summer temperatures during the Holocene Thermal Maximum (HTM) (e.g., Bjune et al., 2005; Seppä et al., 2009).

In contrast, the event layers have distinct physical and geochemical signatures that reflect their high-energy depositional origin. The PCA (Fig. 6) separates these layers from the background sedimentation along PC1, which correlates strongly with variables indicative of minerogenic input, most notably Ti (r=0.98). The event layers are poorly sorted, with grain sizes ranging from sand to clay, and commonly exhibit a structured sequence: a basal, sand-rich unit with an irregular lower boundary. This basal unit grades into silt-rich material with dispersed macrofossils (e.g., twigs, grass, leaves), which is capped by a fining-upward sequence of silt and clay (Fig. 7). The presence of terrestrial organic material and elevated LOI values (up to 27.2 %) indicates that these events entrained vegetation during downslope transport. The observed stratigraphy reflects rapid settling from a turbulent sediment cloud, with coarse clasts and organic fragments deposited first, followed by finer suspended particles.

We interpret the predominant depositional process of the event layers as wet snow avalanches, most likely during late winter or spring melt episodes. Wet snow avalanches are dense, water-saturated flows and highly effective erosional agents, capable of mobilizing material ranging from fine particles to boulders and terrestrial organic matter (Rapp, 1960). Although the event layers share sedimentological similarities with deposits produced by other high-energy processes such as floods or debris flows (e.g., Gilli et al., 2013), such similarities alone are insufficient to uniquely identify the responsible depositional process. The characteristics of event deposits is highly site-specific and depends on local sediment sources, slope morphology, and basin geometry (Rubensdotter and Rosqvist, 2003). Consequently, interpretation must be grounded in a combined assessment of sedimentological characteristics and the geomorphological and hydrological constraints of the catchment.

Fluvial processes are strongly constrained by the catchment setting and are therefore unlikely to generate deposits comparable to the observed event layers. Lake Vatnasetvatnet drains a small, steep alpine basin with a very limited drainage area and no capacity for sustained channelised inflow. Snowmelt in such settings drain gradually and diffusely across slopes rather than concentrating into peak discharge events. Any flood capable of mobilising sediment would therefore need to be rainfall-induced, but even during intense rainstorms, sediment mobilisation is restricted to the narrow gullies that contain limited sediment and lack soil and vegetation cover. The surrounding slopes are mantled by compact glacial till stabilised by vegetation, offering limited erodible material for running water. Although water saturation of glacial till can trigger slides or debris flows, we observed no geomorphological or sedimentological evidence for such processes on the slopes or within the colluvial fans (Sect. 5.1). While minor sediment transport along intermittent streams may occur during extreme precipitation events, these sources are insufficient to explain the volume, grain-size range, and organic content of the event layers observed at the coring site. In contrast, snow avalanches affect broader sections of the slope and efficiently entrain both sediment and terrestrial organic material, consistent with the presence of macrofossils in the event layers. As outlined in Sect. 5.1, the key argument for an avalanche origin in this study lies in the catchment characteristics rather than in the sedimentary structures alone. The morphology and sedimentology of the colluvial fan entering Lake Vatnasetvatnet are consistent with avalanche-dominated systems described in previous studies (e.g., Rapp, 1960; Luckman, 1978; Blikra and Nemec, 1998).

As an independent line of evidence, the PCA identifies only two sedimentary populations within Unit A: background sedimentation and minerogenic event layers, with no indication of multiple event-layer facies. If several depositional processes had contributed substantially through time, a higher degree of compositional variability among the event layers would be expected (Vasskog et al., 2011; Hardeng et al., 2022). Instead, the uniformity of the event-layer population supports repeated deposition by a single dominant process.

Dry snow slab and powder avalanches, while almost certainly occurring throughout the Holocene, are less likely to produce comparable deposits because they are generally less erosive and tend to glide along planes within the snowpack rather than eroding surface sediments and vegetation (Rapp, 1960; Nesje et al., 2007). The age–depth model shows no evidence of significant hiatuses, indicating that the sharp basal contacts reflect compaction or settling dynamics rather than erosion and reworking. Their irregular morphology further argues against deposition as dropstones from melting lake ice (Luckman, 1975; Nesje et al., 2007), instead supporting direct sediment delivery by water-saturated snowflows, potentially breaching lake ice upon impact (Sabatier et al., 2022).

Together, the geomorphological setting, sedimentological characteristics, macrofossil content, and the consistent properties of the event layers through time support repeated deposition by wet snow avalanches as the dominant process shaping the sedimentary record of Lake Vatnasetvatnet.

Quantifying snow avalanche activity from lacustrine sediment records requires a robust methodological approach to distinguish episodic event layers from background sedimentation. In this study, we applied two independent detection methods: (1) rate of change (RoC) analysis of Ti as a proxy for abrupt minerogenic influx and (2) interactive thresholding of CT greyscale values to identify and quantify density-based variations in sediment composition (Fig. 8). Snow avalanches introduce sudden pulses of minerogenic material into the lake basin, significantly altering the depositional conditions (Sabatier et al., 2022). Our detection methods are based on the premise that event layers are abruptly deposited, exhibit a distinct density contrast relative to the background sedimentation, and show strong co-variability with peak values in both XRF elemental counts and CT greyscale data (Figs. 5 and 6).

Rate of change (RoC) analysis has proven to be an effective tool for detecting the abrupt sedimentary shifts associated with episodic events (e.g., Støren et al., 2010; Røthe et al., 2019b; Johansson et al., 2020; Hardeng et al., 2022). Applying consistent criteria for defining avalanche layers eliminates the need for subjective assessment of individual layers throughout the sediment record. Selecting the RoC threshold requires careful calibration to avoid under- or overestimating event frequency. For this study, we computed the RoC of Ti, a parameter well suited for distinguishing minerogenic layers from gyttja due to its low signal-to-noise ratio and the sharp contrast between baseline (background) and peak values (Figs. 5 and 6). Multiple RoC thresholds were tested against the sediment log and CT images. The 95th percentile was selected as the optimal threshold, capturing all visually identified minerogenic layers while minimising noise effects.

CT scanning is a well-suited method to identify the stark density difference between the inorganic, dense avalanche layers and the low-density organic gyttja. Although initially developed for medical applications, CT scanning has become a valuable tool for identifying and quantifying deposits and structures in sediment cores, including tephra (van der Bilt et al., 2021), ice-rafted debris (Cederstrøm et al., 2021), varves (Ballo et al., 2023), floods (Støren et al., 2010; Hardeng et al., 2022; 2024), and snow avalanches (Fouinat et al., 2017; Røthe et al., 2019b). Similar to setting the RoC threshold for distinguishing background sedimentation from event layers, defining the CT greyscale threshold for minerogenic sediment involves a degree of subjectivity. Additionally, CT greyscale values are not directly comparable between core sections, as each scan generates its unique greyscale distribution (van der Bilt et al., 2021). As such, a separate threshold calibration for each core section is required. To aid the process, we used the visual core descriptions and bulk samples to identify depths containing minerogenic sediments, ensuring consistent and reliable threshold calibrations.

Despite the inherent subjectivity in defining threshold values, both methods allow for objective and reproducible event detection once the threshold is established. The RoC approach quantifies the frequency of discrete snow avalanche layers, while CT thresholding provides the relative proportion of minerogenic sediment input over time. The strong correlation (r=0.76) between these independent reconstructions supports the robustness of our methodology and reinforces the validity of a multi-proxy approach for quantifying snow avalanche frequency.

The snow avalanche reconstruction from Lake Vatnasetvatnet (Fig. 9) reveals significant centennial to millennial-scale variability throughout the Holocene. This variability exhibits strong co-variability with regional glacier dynamics, flood frequencies in river systems in southern Norway, and broader North Atlantic climate trends (Fig. 10), highlighting the interconnected nature of these systems. However, caution is warranted when interpreting avalanche frequency solely as a proxy for climatic conditions. The lake's sensitivity to external forcing is modulated by local factors such as vegetation cover, topography, and sediment availability (Rubensdotter and Rosqvist, 2009). In particular, the position of the local tree line may influence avalanche release, as vegetation can stabilize the snowpack and reduce avalanche frequency. Shifts in vegetation may result from natural climate variability but can also reflect anthropogenic land-use changes, such as grazing or deforestation, which may alter slope stability and snow dynamics (Rapuc et al., 2024).