A 4,000 year debris-flow record based on amphibious investigations of fan delta activity in Plansee (Austria, Eastern Alps)

The frequency of debris flows is hypothesized to increase in recent decades with enhanced rainstorm activity. Geological evidence to test this tendency for prehistoric times is scarce due to incomplete sediment records, complex stratigraphy, and insufficient age control especially in Alpine environments. In lacustrine archives, the link between onshore debris-flow 15 processes and the depositional record in lake depocentres is poorly investigated. We present an amphibious characterization of alluvial fan deltas and a continuous 4,000 year debris-flow record from Plansee (Tyrol, Austria) combining Light detection and ranging (LiDAR) data, swath bathymetry, and sediment core analyses. The geomorphic investigation of two fan deltas in different developmental stages revealed a sediment delivery ratio of 7.9 % for the juvenile fan and no sediment transport into the lake on the mature fan within a 3-month summer period (May 2019–August 2019). Event deposits were dated and 20 categorized according to their causal mechanism in a transect of four sediment cores. Debris flow-induced turbidites feature a more gradual fining-upward grain-size trend and higher TOC and δ13C values compared to earthquake-induced turbidites. Over the last 4,000 years, the record containing 138 debris flow-induced turbidites reveals four different debris-flow activity phases. Phase 1 (2050–1960 before the common era; BCE) depicts the second highest observed event frequencies. Phase 2 (1960 BCE–1550 common era; CE) shows large recurrence intervals. Phase 3 (1550–1905 CE) displays a gradual increase of event 25 frequency. Phase 4 (1905–2018 CE) exhibits a debris-flow frequency increase between 1908 and 1928 CE, followed by the overall highest debris-flow frequency between 1928 and 1978 CE, and lower debris-flow frequencies since 1978 CE, which still exceed those of phase 1 to 3. Most remarkably, we find a ~7-fold increase of debris-flow frequency compared to the reference period 1700-1900 CE. The triggering of debris flows is more controlled by short intense rainstorms than for any other mass movement process and we demonstrate that lacustrine debris-flow records provide a unique inventory of hazard30 relevant rainstorm frequencies over decades, centuries, and millennia. In a calibration period of 7 decades, we can show that the debris flow-induced turbidite record matches with the previously published debris-flow volume increase derived from aerial photography coincident to a pronounced rainstorm frequency increase. Here we show a millennium-scale debris-flow record that documents a ~7-fold increase in debris-flow frequencies in the 20th and 21st century coincident to 2-fold enhanced rainstorm activity in the Northern European Alps and provide a novel basis for systematic non-stationary estimation of future 35 debris-flow frequencies in a changing climate.

al. (2020) showed that the past 3 decades were among the most flood-rich periods in Europe in the past 500 years. Many of the largest debris flows in the Alps in the past 20 years were triggered by intense rainfall in summer or fall when the snowline was elevated (Rickenmann and Zimmermann, 1993). Schlögel et al. (2020) investigated in situ and satellite-based climate data 55 in South Tyrol and observed an increase in the average annual duration of rainfall events (+1.1 hours per year) and debris-flow occurrence (+1.2 events per year from 1998 to 2018). Other studies expect an increasing magnitude of debris flows due to an increased availability of loose sediment, longer return periods and presumably fewer, but more intense rainfalls in summer (Stoffel et al., 2014;Stoffel, 2010).
However, the frequency of debris flows over longer time scales is difficult to estimate as terrestrial inventories rarely provide 60 stratigraphically distinct and continuous evidence of subsequent debris flows. There is a lack of continuous long-term datasets to evaluate if Central Europe is in a period of high debris-flow activity compared to the preceding millennia and if long term variations in debris-flow activity follow climatic trends (Stoffel et al., 2005;Irmler et al., 2006). Obtaining such regional perspective is especially challenging due to the local imprint of debris flows and a potentially local trigger process, such as convective storms. The investigation of debris-flow frequencies and magnitudes in lacustrine environments may provide 65 reliable data on prehistorical changes because the typical continuous sedimentation regime can lead to a complete highresolution archive in which evidence for individual debris flows is preserved (Irmler et al., 2006). Despite being widely used for reconstructing past river flood activity (e.g. Gilli et al., 2013;Schillereff et al., 2014;Wilhelm et al., 2019), lacustrine sediments remain an underexplored archive for debris-flow studies. This results from three challenges related to these lacustrine inventories: (i) Debris flows are in many studies not distinguished from other sources of coarser grained sediments 70 such as floods and landslides outside and inside the lake, (ii) debris-flow volumes could not be quantitatively assessed as only lake bottom sequences in sediment cores were analysed, and (iii) there is a lack of long-term instrumental debris-flow data that covers the period with less human interference.
To address these challenges, we here combine on-and offshore investigations to reconstruct debris-flow dynamics at Plansee, a mid-elevation lake in the Alps, that acts as a natural, continuous sedimentary archive (Oswald et al., 2021). Dietrich and 75 Krautblatter (2017) provided evidence for an increase in debris-flow activity since 1980 CE near the lake site. Their results from air photogrammetry studies suggest that on the slopes surrounding the lake, the mean debris-flow volume per year after 1980 CE exceeds the yearly volume from the 1940s to the 1970s by a factor of 3. These recently increased debris-flow rates exceed the Late Glacial rate by a factor of 2 to 3 (Dietrich and Krautblatter, 2017), which may link to the nearly doubled frequency of heavy rainfall events (≥35 mm d -1 ) from 1920 to 2010 CE in the Plansee area (Hydrographischer Dienst Tirol, 80 2020).
In this study, two types of alluvial fan deltas are investigated at Plansee, representing the end members of premature and fully developed geomorphological evolution, hereafter referred to as "juvenile" and "mature" fan deltas. The juvenile fan delta is a high-sloping (>20°) semi-conical shaped deposit adjacent to a trough-like channel cross-section. In this early stage of fan development, the deposition of material reduces the capacity of the channel, so shifts in the channel course resulting in 85 interfingering deposits are likely to occur over time, therefore gradually building up the semi-conical shape (e.g. Sass and Krautblatter, 2007). The mature fan delta displays a late stage of fan development, where the morphology has flattened out following a long period of sediment delivery and the terrestrial profile extends into the lake. It is connected to a large catchment and shows fluvial influence. The large delta depicts a lobate-shaped fan and lower mean slope angles compared to the juvenile fan delta. 90 Instantaneous deposits are a major contributor to lacustrine sedimentation in Plansee (Oswald et al., 2021). Given the setting, it is expected that detrital sediment is supplied to the main basin of Plansee nearly exclusively by episodically occurring debris flows, because no permanent river inflow as potential source of flood deposits is present. When entering a lake, debris flows incorporate more water and turn into a turbulent high-density current (Lowe, 1982). The resulting sediment deposits at the lake bottom are referred to as "df turbidites" hereafter. 95 We investigate subaerial and subaquatic sediment dynamics of a juvenile and mature debris-flow system and tackle the following research questions: (i) How can df turbidites be distinguished from other sediment sources in inner-alpine lakes? (ii) What is the ratio of terrestrial and subaqueous deposition of recent debris flows on juvenile and mature fan deltas? (iii) How are the geomorphic expressions of debris flows related in terrestrial digital elevation model (DEM) and bathymetry data? (iv) Can we systematize the subaqueous deposition pattern of debris flows? (v) Can we infer seasonality of debris-flow activity 100 and can we decipher the frequency in the last millennia to reveal the recent peak activity of debris flows?

Study site
The inner-alpine lake Plansee (surface area: 2.78 km 2 ; maximum depth: 78 m) is located in the Northern Calcareous Alps in Austria, North Tyrol (47°28'10'' N, 10°48'00'' E; 976 meter above sea level, m a.s.l.) and was formed in a glaciated alpine trough valley. Plansee is predominantly surrounded by numerous alluvial fans, talus slopes and fan deltas, which subdivide it 105 into two main basins (Fig. 1). In the SW, Plansee and Heiterwangersee, originally separated by a large fan delta, were connected by a 300 m long canal in 1908 (Hibler, 1921).The lake has two permanent and several episodic inflows draining the catchment and one outflow in the NW. Since 1902, the lake is used as a reservoir for hydropower generation causing artificial seasonal lake level fluctuations of up to 5 m with the lowest level at the end of March (Pighini et al., 2018;Bundesministerium für Soziales, Gesundheit, Pflege und Konsumentenschutz, 2020). As this study only refers to episodic debris flows and their 110 deposits, the human interference on their volume only affects a few percent of the contributing catchments along the shoreline of Plansee and is already in the lower depositional domain of the debris-flow channels. While the human interference may influence the continuous background sedimentation in the lake quite a bit, the episodic debris flows eroding materials from 10,000 m² large catchments way above the lake will not be relevantly influenced and this study only refers to these episodic layers. 115 The mountains surrounding Plansee consist of intensely jointed (cm to dm scale) Upper Triassic lagoonal dolomites (Hauptdolomit), mechanical erosion of which provides a vast amount of loose sediment to the upper catchment areas, which can be remobilized during extreme precipitation events, e.g. in the form of debris flows along incised ditches and canyons.
The slopes in the direct vicinity of the lake are dominated by numerous fan deltas of different developmental stages and subordinated talus slopes (Fig. 1). The fan deltas overlie a local glacial till and reach up to 25 m thickness near the lake shore 120 (Dietrich and Krautblatter, 2017), which highlights the vast amount of remobilized sediment derived from only small and local catchments with catchment areas ranging from ~0.05 to 1.5 km². The steep forested slopes are prone to episodic gravitational mass-transport processes propagating into the up to 78 m deep lake.
Previous work onshore Plansee investigated a debris-flow fan with electrical resistivity tomography combined with orthophoto analysis of the last century (Fig. 1d) and deduced 10 times increased debris-flow volumes in the 1990s compared to 1947-125 1952 CE and linked this to enhanced rainstorm activity (Dietrich and Krautblatter, 2017). This provoked the research question whether such an increase in debris-flow activity can be validated in a continuous millennial-scale lacustrine record. Previous limnogeological work on Plansee investigated subaquatic mass-wasting events recorded in the stratigraphy based on subbottom profiles and a 7 m long sediment core in the main basin, and inferred five severe Holocene earthquakes (local magnitude ML ≥5.3; Oswald et al., 2021). Furthermore, the sedimentary sequence in the main basin of Plansee can also archive extreme 130 sediment transport events in high resolution and high continuity, which sets the stage for the herein presented study on debrisflow activity.
The annual mean precipitation reaches 1,700 mm in the study area with a pronounced summer rainstorm precipitation maximum obtained from two nearby meteorological stations recording since 1923 CE (Höfen, Berwang; Hydrographischer Dienst Tirol, 2020). The relative frequency of heavy rainfall events (≥35 mm d -1 ) at the nearest meteorological station 135 "Berwang" has increased on average by 10 % per decade from 1920 to 2010 CE (Hydrographischer Dienst Tirol, 2020) raising the hypothesis of also increased debris-flow activity since then. Two extraordinary large cyclonic rainstorms with overall damage of hundreds Mio. US$ in May 1999and August 2005CE (Barredo, 2007 also hit the study area with a peak daily sum precipitation of 180 mm and 130 mm, respectively, measured at the weather station in Berwang ~7 km SE of Plansee (Hydrographischer Dienst Tirol, 2020). 140  (Dietrich & Krautblatter, 2017) used for characterization of geomorphic features. The juvenile (b, e) and mature (c, f) fans were repeatedly investigated by terrestrial laser scan measurements in 2019 (see Fig. 4, 5). Onshore DEM derived from Land Tirol (data.tirol.gv.at). The coupled study of debris-flow systems on land and underwater delivers new insights into geomorphic expressions from catchment to depocentre. We investigate two representative types of fan deltas on the southern shore of the lake -a juvenile, steep debris cone and a mature, low angle fan delta (Fig. 1b, c) -by repeated terrestrial laser scanning (TLS) measurements and a multibeam bathymetric survey in 2019. TLS offers a precise method for quantifying short-term volume changes on alluvial fan deltas, while bathymetric surveys reveal the subaquatic deposition patterns. Furthermore, a transect of four ~1.5 m 155 sediment cores from the juvenile fan delta towards the 78 m deep depocentre of the main basin provides the means for deriving the relative deposition of debris flows from the lake shore to the basin. Radiocarbon dating of the sedimentary succession is used to establish an age-depth model. In this study, (i) debris-flow volumes are calculated, (ii) three types of sediment deposits are differentiated, iii) the spatial extent of subaquatic high-density currents is determined, and (iv) long-term sedimentation patterns are analysed. 160

Terrestrial LiDAR (light detection and range) data acquisition and processing
Topographic surveying of debris-flow volumes was conducted on two alluvial fan deltas bordering the lake. Two digital terrain models derived from consecutive TLS were compared for each fan delta. Computing DEM difference rasters is a straightforward and commonly applied method to detect topographic surface changes (Bremer and Sass, 2012;Abellán et al., 2009). The fan deltas were scanned from five scan positions on May 10 th and August 22 nd 2019 using a Riegl VZ-400 laser 165 scanner (long range mode, near-infrared wavelength, measurement range 1.5-600 m, accuracy 5 mm, precision 3 mm, measurement rate 42,000 pts s -1 , beam divergence 0.3 mrad; RIEGL Laser Measurement Systems GmbH, 2017). Data processing of the point clouds was executed with RiSCAN Pro (v. 2.9). The point spacing ranges from 0.035 cm to 10.5 cm, depending on the scanner position. After a coarse error removal, all point clouds of a survey date were coarsely registered with four corresponding points in two consecutive scans and were then fine registered by a multi station adjustment, which uses 170 planar patches of the point clouds, resulting in a 3.1 cm mean deviation of the 3D distances. Vegetation was automatically eliminated by filtering the point cloud with a multidimensional terrain filter, followed by manual removal of remaining shrubs.
A 2.5 D digital terrain model was derived by triangulation. Distances between the May and August models were measured perpendicular to the XY plane using the 'surface comparison' tool. The resulting point cloud was rasterized with the software Cloud Compare (v. 2.11.0; Grid step 0.12; average cell height and scalar field values). The differential volumes (erosion and 175 deposition) over the 3-month period were calculated by creating volumetric meshes with RiSCAN Pro. The Sediment Delivery Ratio (SDR) was calculated to determine the proportion of sediment entering the lake, following Eq. (1) (Lu et al., 2006).
3.3 Limnogeological data 3.3.1 Swath bathymetry 180 High-resolution bathymetry data was acquired in October 2019 by a Kongsberg EM2040 multibeam echo sounder (University of Bern) operating at 300 kHz in a 1 by 1 degree beam-width configuration. For positioning, a Leica GX1230+ GNSS receiver was used in combination with real-time kinematic corrections (RTK) provided by EPOSA (real-time positioning service Austria, EPOSA, 2021). Sound correction was based on continuously monitored surface sound-velocity and nine vertical velocity profiles recorded at least twice per day using a Valeport MiniSVP probe. Speed of sound in the water column ranged 185 from ~1,452 m s -1 at the surface to 1,424 m s -1 in the deepest basin, and 1,446 m s -1 to 1,428 m s -1 in Heiterwangersee. The recorded raw data have been processed in Caris HIPS/SIPS 9.1 software. During processing, all auxiliary sensor data (motion sensor, heading sensor, GNSS sensor) are merged, reviewed and manually corrected if necessary. Daily lake level changes (4-6 cm in 24 hours with respect to a reference level of 976.0 m a.s.l.) were corrected using data from a local gauging station (E-Werke Plansee). The resulting point cloud (~200 million points) was reviewed and different algorithms for rasterizations were 190 tested, resulting in a bathymetric map with 1 m horizontal and a few decimetres vertical resolution. Besides depth information, the amplitude of the backscattered acoustic signal was calculated based on a median temperature of 4.7° C, a median sound velocity of 1,425 m s -1 and an assumed salinity of 0 ppt from a vertical sound velocity profile taken in the deepest part of the lake. Backscatter data provides rough estimates on the sediment grain size by coarser grain sizes yielding higher amplitude values (Beyer et al., 2007;Hilbe et al., 2011). Topographic openness maps were calculated with SAGA GIS and combined 195 with analytical hillshades for enhanced visualisation of the geomorphic features ( Fig. 2) or combined with a color-ramped shading representing the water depth ( Fig. 1) using QGIS Desktop 3.10.5 (see also Daxer et al. (2020) and references therein for a detailed description). Interpretation of subaquatic geomorphic landforms was carried out following Strasser et al. (2020) and references therein.

Sediment core analyses 200
Four ~1.5 m long sediment cores with 63 mm diameter were retrieved in 2018 and 2019 CE using a gravity coring system equipped with a manual percussion system (Table S1). The selected coring locations follow a transect from the depocentre of the main basin towards the juvenile fan at intervals of 50-75 m ( Fig. 1; Fig. S1). In the lab, sediment cores were split lengthwise, macroscopically described and imaged using a Smartcube® Camera Image Scanner. A core-to-core stratigraphic correlation was conducted based on sedimentary facies and distinct marker layers. Event deposits of ≥1 mm thickness were 205 primarily characterized and mapped in all four cores by macroscopic identification directly on the core surface in combination with color-and contrast-enhanced core images (Automatic Histogram Equalization).
We macroscopically identified, mapped, and correlated event deposits of ≥1 mm thickness in all four sediment cores following the sedimentological criteria outlined in Sect. 3.3.3 and based on Irmler et al. (2006), and measured the thickness of each event deposit. Event deposit thicknesses of the uppermost sediments were corrected based on water content for comparable 210 measurements with deeper deposits with lower water content.
Laser-diffraction grain size analyses were performed on event deposits using a Malvern Mastersizer 3,000 in combination with a Hydro Sight module for visualization and quality control of the dispersion. Samples were measured without any chemical pretreatment, as the amount of organic material is negligible in the clastic-dominated sediments. Samples were taken using a toothpick with a resolution of up to 2 mm dependent on deposit thickness. The measurement was started at an obscuration of 215 ~10 %, followed by sonication (60 s, 70 %). The grain-size distribution was calculated following international standards (ISO 13320: 2020). The particle size statistics were calculated with the software GRADISTAT (Blott and Pye, 2001). The fraction of median grain size (Q50) and the coarser 90 th percentile (Q90) parameters within graded beds was used to further characterize event deposits (Wilhelm et al., 2013). Heatmaps of closely spaced grainsize data were calculated using the kriging method in Surfer 11. 220 For organic geochemistry, the sediment cores were described and afterwards sampled in 1 cm resolution. After lyophilisation, 1-2 cm³ of each sediment sample were homogenized with an agate mortar. 3.0±0.3 mg of each sample was weighed into silver capsules for decalcification on a hot plate (70° C) using 5 % HCl until no effervescence was observed. After complete drying, capsules were folded and total organic matter content (TOC, wt%) and carbon isotope composition of bulk organic matter ( 13 CTOC) were determined using an elemental analyser (NC 2,500, Carlo Erba, Italy) coupled to an isotope-ratio mass 225 spectrometer (DeltaPlus, Thermo-Finnigan, USA). Elemental standards Atropine and Cyclohexanone 2,4dinitrophenylhydrazone were used for calibration of carbon content, and IAEA-CH-7 and USGS41 for isotope calibration.
Additionally, a lab standard (peptone) was used for linearity correction and isotope calibration. All isotope values are reported in the common -notation.
Bayesian age-depth modelling of radiocarbon ages using the R-software Bacon v 2.4.3 (Blaauw and Christen, 2011) combined with ages derived from the peak fallouts in 1986 CE and 1963 CE of the radionuclide 137 Cs and by constant flux-constant sedimentation rate (CFCS) modelling of excess 210 Pb activities using the R-package SERAC (Bruel and Sabatier, 2020).
Radiocarbon ages were calibrated with IntCal20 (Reimer et al., 2020) and reported in years before the common era (BCE) or common era (CE). 235 For the age-depth modelling, event deposits >5 mm were removed to obtain an event-free sediment depth. Radiocarbon samples are derived from organic macro remains in finely laminated intervals of background sediment (Table S2).

Sedimentary event identification
Debris flows form concentrated density flows in a lake which deposit turbidites with distinct sedimentological characteristics at the lake bottom, such as color, texture, grainsize evolution, and organic content (Sletten et al., 2003;Irmler et al., 2006). Df 240 turbidites are sharp-bound units with a fining-upward grainsize trend, often bearing terrestrial macro-remains (Sletten et al., 2003). In contrast, strong earthquake shaking triggers multiple subaquatic slumps that evolve into turbidity currents and generate a turbidite in the depocentre (e.g Schnellmann et al., 2002), hereafter referred to as "eq turbidite", with distinct sedimentological characteristics, such as a homogeneous grainsize trend and an in-lake geochemical fingerprint. In Plansee, previously interpreted eq turbidites are 5 to 35 cm thick amalgamated turbidites indicating deposition of multiple turbidity 245 currents within a short time. All other event deposits were interpreted as flood-or debris flow-related (Oswald et al., 2021).
This initial interpretation will be tested in this study by more in-depth analyses based on above-mentioned conceptual approach.

Spatial and temporal analyses of df turbidites
Thickness measurements of individual events in several cores provide means to visualize the spatial extent and geometry of the respective event deposit (cf. Moernaut et al., 2014), which can provide insights for sediment dynamics or source of the 250 turbidity current. Therefore, we calculated the thickness distribution (in %) for each individual event by its thickness in a core relative to the accumulated thickness of the deposit in all four cores. Moreover, deposit thickness of flood-induced hyperpycnal flow deposits have been calibrated in several alpine environments to represent flood intensity (e.g. Czymzik et al., 2013;Wilhelm et al., 2016) which we herein also aim to test its applicability for df turbidites. Additionally, we calculated annual occurrence rate of df turbidites using a central running sum with a band width of 21 years to obtain changes in debris-flow 255 frequency. The cumulative thickness over time involves both the thickness and frequency of df turbidites and its slope provides information on df turbidite accumulation rate per year.

On land and underwater characterization of the alluvial fans and fan deltas
The investigated alluvial fan deltas form conical to lobate sediment accumulations protruding into the lake in front of funnel-260 to cirque-shaped catchments (Fig. 1). Onshore, the fan deltas can be subdivided into an active part and a partially active part dependent on geomorphic characteristics and vegetation type (Fig. 1e-f and 2). The active part is characterized by braided lobes and channels, cutting into older fan deposits, and is mostly vegetation free (Fig. 2a, d). In contrast, partially active parts have clearly smoother topography due to gravitational hillslope processes and are abundantly occupied by shrub and tree vegetation including pines. There, debris-flow activity is indicated by small lobes and channels close to the active channel 265 levee representing sediment spill-overs during large debris-flow events exceeding the channel capability. The onshore mean slope angle of the alluvial fan deltas varies from 10° for the lobate, mature fan delta to 24° for the cone-shaped, juvenile fan delta. Independent of these parameters and maturity of the fan itself, the mean slope angle shifts to ~30° once a fan submerges into standing water. The subaquatic geomorphology of alluvial fan deltas has a greater variety in geomorphic features than its onshore part (Fig. 2a,d). In general, three subaquatic landforms are subdivided: the coastal zone, an active fan delta, and an inactive fan delta.
The coastal zone is characterized by a 5-10 m broad area of a high backscatter signal indicating abundant coarse (sand-gravel) sediments (Fig. 2a, d). A coarse-grained coastline occurs almost everywhere independent whether the part of the fan is active 280 or inactive. The subaquatic active fan delta is generally characterized by a bulge in respect to the general arcuate trend of the fan (Fig. 2a, b, d). In the detailed view of Fig. 2b, d, the bulge consists of numerous stacked and braided debris-flow lobes.
Some of the lobes have intermediate to high backscatter signals indicating coarse-grained and thus recent debrites, whereas older debrites are expected to be covered in fine-grained lacustrine mud and thus exhibit low backscatter values. This observation is in accordance with detailed mapping of onshore debris-flow deposits which have recently occurred between 285 1947 and 2014 CE (Dietrich and Krautblatter, 2017), as the subaquatic continuations of these mapped debris-flow deposits exhibit high backscatter values (Fig. 2a). In contrast, the inactive fan delta is characterized by a more regular subaquatic morphology dominated by parallel downslope-oriented gullies and crests (Fig. 2a, c, d). The backscatter signal is low at the gentle slopes representing inactivity of coarse detrital sedimentation and already sufficient coverage with fine-grained lacustrine mud. Intermediate back scatter signals occur along gullies and at the basin near the slope-break (Fig. 2c). At the 290 latter location, coarse sediments are deposited by debris-flow activity or remobilisation and transportation of coarse coastal material mixing with fine-grained lacustrine basin floor sediments. The cause of abundant coarse material in gullies is interpreted to be coastal erosion and related density flows which either deposit coarse coastal material or erode a possible finegrained lacustrine sediment cover and can thus expose old, coarse debrites. Independent of its underlying process, the more or less regular pattern of gullies and crests seems to develop at the inactive fan delta due to subaquatic gravitational slope 295 processes, which cannibalize the previously deposited debris-flow lobes and thus fundamentally alter the subaquatic geomorphology created by debris-flow processes. Funnel-shaped landslide scars also occur at inactive fan delta slopes and locally the corresponding subaquatic landslide deposit can still be observed in bathymetric data (red and blue dashed lines in  The western juvenile fan delta is small (0.11 km²), steep (24° on average) and has a steep, conical shaped apron with a dominant, deeply incised active channel eroding into previous fan deposits (Fig. 3a). The apron contains an inactive to partly active channel and a backfilled channel (dashed and dotted line in Fig. 3a, respectively). The fan catchment (0.18 km 2 ) is funnel-shaped (Fig. 1a). In the subaquatic realm, the fan has a slightly arcuate shape due to its prominent active fan delta in the middle of two inactive areas. Numerous stacked debris-flow lobes built up the active part indicating active fan progradation. 315 A small delta-failure event occurred in the active part and is indicated by a subaquatic landslide deposit and its corresponding scar (Fig. 3a). In the basin near the basin-slope transition an irregular, hummocky morphology shows relicts of older and larger fan delta failures buried by lacustrine mud.
The eastern mature fan delta is much larger (0.36 km²) and depicts a lobate-shaped fan with a low mean slope angle (10°). The large catchment (1.19 km²) is bowl-shaped and extends over several side valleys (Fig. 1a). It has one distinct active channel in 320 the centre and several smaller diffuse currently inactive channels (Fig. 3b). On a topographic profile along the main flow axis, the onshore part of the fan has a concave shape with a trend to lower slope angles towards the lake. The subaquatic morphology is characterized by an overall arcuate apron shape at the slope-basin transition (Fig. 3b). A few smaller bulges against the general trend represent the current or rather recent active fan part prograding in the lake. The inactive fan part is dominated by parallel crests and gullies and shows a few recent debris-flow lobes, which contrasts to the lack of these lobes for inactive parts 325 of the steeper fans e.g. the juvenile fan (Fig. 3a). These lobes often match with the subaerial diffuse channels and thus might represent deposits of the over-spilling events (orange arrows in Fig. 3b). Additionally, the inactive fan part contains a few subaquatic landslide scars (red dashed lines in Fig. 3b). The currently active fan part has only a few stacked debris-flow lobes in the upper part of the subaquatic slope, while most of the lobes occur in the lower slope in front of an incised gully (blue arrows in Fig. 3b). This implies that gully formation partially overprinted the debris-flow lobes and potentially hints that, in 330 recent times, most sediment gets accumulated onshore on the shallow fan and only during extraordinary large events a significant debris-flow mass reaches the subaquatic slope.
In comparison, the juvenile fan is characterized by a smaller fan area, a larger average slope angle and a smaller fan-catchment size ratio than the mature fan. The mature fan is connected to a large catchment and talus is removed by a small perennial stream in addition to episodical sediment transport. The terrestrial profile of the mature fan extends into the lake, creating a 335 low inclination depositional area, which contrasts the convex shape of the juvenile fan. The latter shows active progradation in the subaquatic realm, whereas the mature fan displays less signs of recent debris-flow activity.

TLS-measured net topography change of alluvial fans
The net topography changes of the onshore fan delta surfaces range between -2.6 m and +1.8 m from May to August 2019.
The total mobilized volume during the 3-month period is 1.9 times higher on the steep, juvenile fan delta compared to the flat, 340 mature fan delta. This difference is potentially linked to the maturity, catchment topography, and connectivity (i.e. sediment throughput) of the individual fan delta.
The steep, juvenile fan delta displays a maximum of 1.6 m increase in height and a maximum erosion depth of 2.6 m (Fig. 4).
Erosion and deposition are balanced for the confined fan delta and the channel geometry was drastically altered during the investigated period. The fan topography on May 10 th displayed a shallow, continuous channel which has eroded previous 345 deposits on the fan apron and was partially backfilled in the lower third of its extent. Since most subaquatic depositional lobes are found in elongation of this channel (see Sect. 4.1), it is assumed to be the dominant pathway for channelized flows. The second survey on August 22 nd reveals backfilling and overtopping of the former channel (Fig. 4). There are two zones of deposition in the lower fan area: a funnel-shaped debris accumulation in the terminal part bordering the hiking trail and a widespread, more proximal zone of deposition, which represents progressively short and wide debris flows. The erosional zone 350 extends from a small ravine to a deeply incised U-shaped channel cutting the talus above the newly developed zone of deposition.

360
Stratified deposits of alternating coarser and finer layers representing interfingering channels were exposed on the scarps and the new channel displays abundant coarse clasts. A second zone of erosion formed below the trail, indicating sediment transport into the lake. Over the whole debris-flow track, the eroded volume (628 m 3 ) exceeds the deposited volume (578 m 3 ), causing a sediment flux into the lake of 50 m 3 and a sediment delivery ratio of 7.9 % for the investigated period.
During the 3-month period, onshore sediment deposition dominated the shallow, mature fan delta (Fig. 5)

Event type differentiation
Lacustrine sedimentation in the main basin of Plansee is characterized by dark-grey to ochre, finely laminated clayey silts with abundant detrital carbonates and subordinate contents of diatoms and organic matter (background sediment; Fig. 6a). The background sediment contains 1.5-2.3 wt% total organic carbon (TOC) with C/N ratios between 13 and 19 and δ 13 C values 385 between -28.0 and -30.4‰ ( Fig. 6b; Table S3). This indicates a mixture between lacustrine organic matter of algal origin (typically with C/N ratios <10) and terrestrial organic matter with TOC/TNmolar ratios typically exceeding a value of 20 (Meyers and Teranes, 2001). Two different types of event deposits are intercalated in the background sediment, which can be macroscopically and analytically distinguished (Fig. 6a-c).
One event type consists of a grey, homogenous coarse silt turbidite with a thin fining upward base as e.g. present at core depth 390 116 cm in Fig. 6a. This event type, here referred to as "eq turbidite", was related to events of multiple subaqueous masswasting caused by strong seismic shaking (Oswald et al., 2021) and occurs only at three stratigraphic levels in the short cores ( Fig. 7a), corresponding to earthquakes at ~2150 BCE, ~1050 BCE, and 1930 CE (Namlos M 5.3 earthquake). Eq turbidites yield around 0.7-2.3 wt% TOC and δ 13 C values between -26.9 and -29.2 ‰ ( Fig. 6b; Table S3). Relatively constant lightness L* and density values support the homogeneous character of eq turbidites except the deposits of eq 3 in Plan19-03 and Plan19-395 04 where a subaqueous landslide deposit characterized by contorted strata overlain by a graded turbidite was cored, and which corresponds to a strong earthquake at ~2150 BCE (Oswald et al., 2021;Fig. 7a  finely laminated background sediment intercalated by debris flow-induced turbidites (df turbidites) and earthquake induced turbidites (eq turbidites). Df turbidites have a gradual fining-upward grainsize trend, whereas eq turbidites are homogeneous deposits on top of a thin coarse base. Eq turbidites show broader grain-size distributions (poorer sorting). b) TOC/δ 13 C plot of samples from all cores shows general differences for df-and eq turbidites and similarities between df turbidites and the (predominantly clastic) background sediment. c) D50/D90 grainsize plot of df-and eq turbidite samples from all cores highlights the 405 different evolution of df-and eq turbidites. Colored ellipses represent 95 % confidence ellipse.
The second event type (df turbidite) is characterized by brown to ochre colored, up to 6.5 cm thick detrital deposits, which abundantly occur throughout the sediment cores (Fig. 6a). Df turbidites generally have a sharp or irregular coarse-grained base (coarse silt to fine sand) overlain by a progressive fining upward sequence and a fine-grained (fine silt) top. Some of the df turbidites have a coarsening upward trend at the base with the maximum grainsize in the lower-middle part followed by a 410 normal grading (Fig. 6a). Such grainsize trends in detrital deposits indicate waxing and subsequent waning of flow-energy transporting the terrestrial sediments into the lake during a single high discharge event (e.g. Gilli et al., 2013). Terrestrial organic macro-remains often occur bedding-parallel aligned at the base or in the middle part of the deposits. TOC values ranging from 1.0 to 2.8 wt% are higher compared to eq turbidites ( Fig. 6b; Table S3). The lower TOC contents in eq turbidites are potentially caused by the decomposition of organic matter on subaquatic slopes prior to the earthquake-induced 415 remobilization of the slope deposits, whereas higher TOC contents in df turbidites show no sign of decomposed organic matter (Vandekerkhove et al., 2020). On average, df turbidites also show slightly lower δ 13 C values from -25.5 to -29.9 ‰. In addition, the grain size evolution patterns of df-and eq turbidites derived from a D50 versus D90 diagram are different (Fig. 6c). Eq turbidites have a steeper trend, suggesting a poorly sorted turbidite caused by subaqueous mass movements (Wilhelm, 2012;Wilhelm et al., 2016). In contrast, df turbidites mainly follow a much less steep trend (Fig. 6c), which is commonly related to 420 well-sorted density flow deposits induced by flood events (Wilhelm et al., 2016). All above mentioned observations and characteristics of these event deposits suggest the interpretation that df turbidites evolved from terrestrial debris flows. The potential of misinterpreting df turbidites as river flood induced turbidites, which could have similar characteristics (Gilli et al., main inflowing rivers are trapped either in Heiterwangersee or in the easternmost subbasin in Plansee and do not reach the 425 studied main basin (Fig. 1a).

Spatio-temporal distribution of df turbidites
Sedimentation processes and temporal occurrence of df turbidites offshore the juvenile fan delta were investigated in four short cores with 50-75 m spacing, forming a proximal-to-distal transect from the juvenile fan delta towards the main basin depocentre (Fig. S1). The two most distal cores (Plan18-10 and Plan19-02) are located in the flat depocentre at 77 m water 430 depth with low slope angles (<1°), whereas the more proximal cores Plan19-03 (76 m water depth, 2° slope angle) and Plan19-04 (75 m water depth, 3° slope angle) are located on the gentle slope towards the juvenile fan. The most proximal core Plan19-04 is located close to the slope break. The short cores share the overall lithostratigraphic succession, which allows for crosscorrelation of the 138 identified event deposits (Fig. 7a, Table S4). A 2D spatial distribution of df turbidites is obtained by the deposit thickness distribution in all four cores (Fig. 7b), which provides potential information on sedimentation processes and 435 source areas. In addition, a temporal distribution of df turbidites of the last 4,000 years is derived from combined radiocarbon and short-lived radionuclides dating of the most distal core Plan18-10 ( Fig. 8, Fig. S2).
The stratigraphic succession present in the short cores can be subdivided into four main phases based on lithostratigraphic units and occurrence pattern of df turbidites (Fig. 7).
Phase 1 depicts a short phase (2050( -1960  Most df turbidites are relatively thicker in the two distal cores compared to the two more proximal cores (Fig. 7b), indicating a ponding geometry of the turbidite body resulting from sediment-laden density flows, i.e. underflows (Gilli et al., 2013).
Several df turbidites highlight this ponding-deposition character by more than 70 % thickness distribution in the two distal cores (grey arrows Fig. 7b). Extreme examples for pronounced distal sediment deposition are present in sequences in phase 4 (df 11-16) and at the end of phase 1 (df 128-129) where the events exclusively occur in the distal cores. In contrast, several 450 df turbidites are thicker in the proximal cores and occur throughout the sequence (black arrows Fig. 7b). An exclusive deposition in proximal cores only occurs during phase 1 (df 130-138; Fig. 7b), for which no df turbidites are present in the distal core Plan18-10. 455 Fig. 7: Core-to-core correlation of distal (left) to proximal (right) short cores in respect to the juvenile fan delta. Four lithostratigraphic phases (ages in CE/BCE) can be distinguished based on the initial core images, bulk density (violet) and lightness (orange) and dated by combined radiocarbon (red stars) and short-lived radionuclide ( 210 Pb and 137 Cs; dark-red stars) dating (Fig.  S2). Each of the 138 identified debris flow-induced (df) and 3 earthquake-induced (eq) turbidites are cross correlated in the four cores and measured for their thickness. b) Thickness distributions of each df turbidite in the four short cores (individual thickness 460 relative to accumulated thickness in the four cores) are color-coded based on the corresponding core and show different deposition patterns for different events and phases.
For further analyses on past debris-flow frequency and intensity, the core Plan18-10 is considered as the most representative due to its location in the depocentre where the potential erosive power of underflows is lowest and by containing the most complete event stratigraphy deduced from core correlation. This location is also considered to potentially best represent event 465 intensity derived from thickness measurements due to the overall ponding geometry of df turbidites. The few missing events (df 130-138) are projected to the master core Plan18-10 to guarantee record completeness for further analyses. In general, the temporal distribution and thickness of df turbidites (Fig. 8a) in combination with derived statistics (recurrence interval, mean thickness, and deposition rate (Table 1) strongly vary in phases 1-4. The oldest phase 1 is characterized by 480 abundant df turbidite occurrence represented by relatively higher event frequencies with recurrence interval of 9.1 years ( Fig.   8a; Table 1) and by containing the thickest event deposits of the whole record. High event frequency and thick event deposition is also reflected by the steep cumulative thickness trend with a debris-flow deposition rate of 1.0 mm a -1 . Phase 2 is characterized by sporadic df turbidites, centuries of event quiescence, the largest recurrence interval of 44.3 years, and a low debris-flow deposition rate of 0.08 mm a -1 . The few events in this phase have average thickness reflected by vertical steps in 485 the cumulative thickness plot (Fig. 8a), indicating that low event frequency does not imply smaller event thicknesses. Phase 3 is represented by a relatively gradual increase of event frequency reflected by e.g. the decreased mean recurrence interval of 25.4 years and an increased debris-flow deposition rate of 0.2 mm a -1 compared to phase 2. In addition, the event thickness is increasing in phase 3 represented by the quasi-exponential trend in the cumulative thickness (Fig. 8a).
The youngest phase 4 exhibits the highest abundance of df turbidites with a mean interval of 3.3 years (Table 1), which is also 490 reflected by the steepest slope in cumulative event thickness with debris-flow deposition rates of 1.8 mm a -1 and 1.1 mm a -1 at the beginning and end of phase 4, respectively. In detail, phase 4 can be subdivided into three sub-phases based on the occurrence rate and cumulative thickness (grey inlet in Fig. 8a). Phase 4.1 is represented by a strong and fast frequency increase between 1908 and 1928 CE, followed by a period of highly frequent debris-flow events between 1928 and 1978 CE (phase 4.2) and the current phase 4.3 since 1978 CE, characterized by lower frequencies relative to phase 4.2 but still by far higher 495 than in phases 1-3. Relatively, debris-flow frequency in 4.2 increased by a factor of 9 compared to the reference period 1700-1900 CE and by a factor of 17 compared to phase 2. In phase 4.3, debris-flow frequency increased by a factor of 5 compared to the reference period 1700-1900 CE and by a factor of 10 compared to phase 2.  The weather station in Berwang, 7 km SE of the coring site has recorded daily precipitation sums since 1900, which comprises the whole phase 4 (Fig. 8b). The two most pronounced rainstorm events in 1999 and 2005, corresponding to long lasting precipitation and river floods in several countries in central Europe (e.g. Barredo, 2007), temporally coincide within age uncertainties with two df turbidites (df 5 and df 6). However, event thickness does not correlate with daily precipitation sum 505 indicated by an only intermediate thickness of df 5 and df 6. In addition, not all df turbidites of the last century can be unambiguously linked to a measured rainstorm event, potentially due to age uncertainties, limits in macroscopic df turbidite detection, and in transport efficiency of debris flows (see discussion Sect. 5.1 and 5.3).

Data quality 510
Methodological limitations in our study principally relate to i) data quality and ii) first-order interpretation and calculations.
Error sources in processing terrestrial laser scanning could derive from shrub vegetation coverage that were cut out from the point cloud and from the roughness of the terrain. Due to the poor vegetation cover in the debris-flow channel and multiple scanning positions only minor deviations are expected for the volume calculation, the roughness was addressed by using and the lowest end of the newly formed channel could not be entirely assessed, the erosion volume could be slightly underestimated by a few percent, but TLS certainly covered the main incisions. Erosion and deposition most likely account for few single events, but consecutive minor redistribution could not be excluded.
In terms of bathymetric mapping, the uncertainty is derived from a combination of accuracy calculations for individual sensors (navigation, orientation, motion compensation), latency and sensor errors, and the application of sound velocity profiles (point 520 observations) over the entire basin. Generally, the uncertainty is larger in areas with i) steep subaqueous terrain with irregular morphology resulting in low point density for these parts, ii) deep flat lake floor morphology with little data overlap from independent survey stripes, and iii) where large beam angles are needed (central beams are more accurate then outer beams).
Regarding the df turbidites, the sediment cores only represent a complete archive of deposits from debris-flow events that are capable to enter the lake. Low magnitude events which only create terrestrial deposits are not detectable as independent df 525 turbidites in sediment cores. Instead, subsequent debris flows of larger magnitude can incorporate previous terrestrial debris deposits by sediment entrainment, so over time most of the total debris-flow volume ends up in the sediment record of the lake. On the prevalent juvenile fan type, most debris flows exceed the sediment volume or discharge threshold required to reach the lake, transport sediments up to the coring site, and deposit a sufficiently thick event layer. Df turbidites remain macroscopically undetected if their layer thickness is below the typical lamination thickness of the background sedimentation 530 and in this case would require detailed microfacies investigations. Stacked event layers could potentially correspond to multiple debris-flow surges of a long-lasting high-discharge event or to coeval debris-flow activity at different fans during the same event. Thus, identification of the clay cap, representing the post-event deposition of a suspension cloud, is crucial to disentangle stacked event layers. Erosion of underlying sediment is generally negligible, as most df turbidites have a coarse silt to fine sand maximum grainsize and the coring sites are located far from the slope break where the erosion potential is highest (Fig.  535   S1). The investigated transect provides first-order insights to spatially investigate underflow deposition of the basin and debrisflow activity of the proximal juvenile fan, but several more cores in transects to other fans would be required for a holistic view on the deposition pattern of each df turbidite.
The presented age-depth model results in 95 % uncertainty range of a few years and decades at the near-surface sediments and increases to a few centuries deeper down (Table S2). Therefore, we refrain from detailed comparisons of the Plansee 540 record with other flood deposit archives for discussing potential climatic drivers on debris-flow activity (see Sect. 5.4). This would require microscopic event identification, further 14 C dating and detailed investigations (and counting) of potential annual mixed, organic-clastic varves.

Terrestrial and subaqueous depositional patterns
This amphibious investigation provides evidence of debris-flow activity on the juvenile and the mature fan delta during the 545 summer months of 2019. The difference in net transport volume and sediment delivery into the lake relates to their stages in fan development. Juvenile and mature fan deltas essentially differ in the following structural components: fan and catchment geometry and their connectivity, slope, vegetation cover, and the ruggedness of the terrain.
The 7.9 % sediment delivery ratio of the juvenile fan indicates high connectivity between the small catchment and the fan delta, which enables a greater intake of rock debris (Kaitna and Huebl, 2012) and facilitates high throughput towards the lake. 550 The juvenile fan delta displays high topographic heterogeneity of the terrain and a lack of intermediate storage. Therefore, deposits currently located in the depositional area are rapidly subjected to consecutive transport to the lake. The bathymetric survey uncovered stacked lobes providing evidence for recent debris-flow activity and active fan progradation. Since the juvenile fan delta type is prevalent at Plansee, most debris flows enter the lake and form underwater deposits.
Between May and August 2019, the debris-flow channel was backfilled and overtopped, followed by avulsion and the 555 formation of a new channel (cf. Haas et al., 2016). Initially, onshore deposition reduced the channel gradient. The channel was then progressively filled up with sediment, and avulsion was initiated, diverting the next debris flow to a steeper flow path, where it formed a new channel. The erosion depth of up to 2.6 m suggests the concentration of a flow in the newly formed ravine, where it transformed into a debris flow by impinging on a dam of loose debris (fire-hose effect, after Theule et al., 2015). The steep slope of the juvenile fan delta facilitates channel scouring, sediment entrainment and transport into the lake. 560 The channelized sediment transport combined with quasi-cyclic avulsion of the geomorphologically active sector create the semi-conical shape of the juvenile fan delta (Haas et al., 2016).
The mature fan delta exclusively experienced terrestrial deposition during the investigated period. It features a longer debrisflow track and a more stable morphology compared to the juvenile fan delta. Large forested areas between screes and bare rock faces in the catchment buffer surface run-off and sediment transport, and the shallow topography of the channel and the 565 fan delta affect the transport distance. Fluvial erosion likely accounts for bedload transport and the removal of unconsolidated sediment otherwise available for entrainment. Low slope angles, widespread sediment deposition on the unconfined, flat fan delta and intermediate storage in the large, bowl-shaped catchment contribute to the lack of sediment transport into the lake over the summer of 2019. Low-magnitude debris flows terminate in dry, coarse sediments at the flow margins. Abundant coarse clasts spread over the entire cross-section at the foot of the apron are remnants of this zone of high frictional resistance 570 (see Kaitna and Huebl, 2012). However, the massive subaquatic volume of the fan delta shows that over time, the largest portion of debris-flow sediment accumulates in the lake basin, and the onshore fan thickness represents only part of the total transported volume. Subaquatic deposits from over-spilling events provide evidence for widespread lateral sediment distribution. The active part of the mature fan delta shows overprinted debris-flow lobes in the subaquatic realm, which prove that terrestrial deposition dominates recently. From the onshore morphology changes and the bathymetric investigation, we 575 conclude that in recent times, there is a higher magnitude threshold for debris flows protruding into the lake on the mature fan delta.

Deposition patterns of df turbidites
The majority of the investigated df turbidites have ponding sediment bodies indicating overall underflow deposition (e.g. Gilli et al., 2013) induced by debris-flow activity of the adjacent fans. Besides this general deposition trend, several df turbidites 580 show major or even exclusive deposition either in the distal or in the proximal cores, which demand for different interpretations.
From the limited perspective of this transect of four cores, we interpret that df turbidites with major deposition in the distal (basin) cores (grey arrows in Fig. 7b), are potentially related to high-energy flows, which bypassed the proximal sites.
Alternatively, debris-flow activity occurred on another fan delta and our transect cannot resolve its real source. Exclusive deposition in the two distal basin cores only occurs in a sequence from 1965 to 1973 CE ("D" bracket in Fig. 7b), during which 585 there were extensive street constructions at the northern side of the basin. Therefore, we interpret this phase of exclusive distal basin deposition to be related to low-energy, human-induced detrital input from the northern shoreline and thus, df turbidites 11-16 likely reflect anthropogenic impact and not debris flow-induced turbidites. In contrast, df turbidites with a major thickness in the fan-proximal cores (black arrows in Fig. 7b) form wedge-shaped sediment bodies near the slope break and can be explained by low-energy debris-flow activity on the juvenile fan. This only occurred in a short sequence ("P" bracket in 590 Fig. 7b) immediately after the ~2150 BCE earthquake, which is expressed as multiple subaqueous mass-wasting deposits overlain by the eq 3 turbidite ( Fig. 7a; Oswald et al., 2021). This earthquake was interpreted to trigger the large-scale catastrophic rockslides at Eibsee and Fernpass, both within 15 km distance, and thus can also have triggered small-scale mass movements in the catchment of Plansee, causing an enhanced availability of loose sediment. Such additional sediment availability in a lake catchment can lead to enhanced detrital input in the lake in the aftermath of a strong earthquake 595 ('postseismic landscape response' cf. Howarth et al., 2016). Accordingly, we interpret the sequence of exclusive proximal event deposition in the aftermath of the ~2150 BCE earthquake to represent highly sediment-concentrated but low energy flows from the adjacent juvenile fan caused by seismically induced enhanced sediment availability in the catchment. Following our observations, it took about 100 years for the landscape to recover and return to equilibrium state after the ~2150 BCE earthquake. 600

Driving forces of debris-flow activity
The Plansee area offers a number of prerequisites for conducting a study on debris flows, which include (i) large sediment availability, (ii) juvenile highly connective fan morphology, (iii) small changes in vegetation cover, and (iv) little human influence in the debris-flow release zones compared to other catchment areas in the European Alps. Thus, the temporal fluctuations are presumably mostly controlled by (v) a few earthquakes and (vi) climate forcing. 605 i) The intensely jointed dolomites (Hauptdolomit) surrounding Plansee are prone to hydro-mechanical erosion forming a quasi-constant and practically infinite supply of loose debris. In this transport and not weathering-limited setting, the slopes are highly sensitive towards short, intense precipitation triggering the release of debris flows. During phases of postseismic landscape response, the precipitation threshold to initiate sediment transport could be lowered (see

Sect. 5.3). 610 ii)
Debris-flow deposition is dependent on fan morphology i.e. maturity of a fan. The prevalent fan type at Plansee displays a juvenile highly connective fan morphology, which facilitates high throughput towards the lake. Debris flows on juvenile fans bypass the rather steep fan delta and deposit offshore, which contrast the debris-flow deposits on mature fans, where only extraordinary large events deposit offshore. Since most fans are still in the juvenile stage of development, subaquatic deposition of most events is ensured and changes in fan morphology can likely be ruled 615 out as a major controlling factor for changing debris-flow activity of the last 4,000 years. In this closed sediment system, the largest part of all mobilized sediment ends up in the lake basin. iii) The percentage of vegetation cover on debris-flow fans is mainly regulated by the debris-flow activity (Dietrich and Krautblatter, 2017). The debris covered slopes in the study area are mostly free of vegetation as a consequence of high debris-flow activity. The majority of the catchment is vegetation-covered, which generally stabilizes a slope 620 (Barker, 1995). There are no significant human-induced vegetation changes documented, as there is no agriculture in the area, only minor use of forest, and no documented major forest clearance or wildfires. We therefore infer that there were no significant changes of vegetation before or during the period of increasing debris-flow frequency in the last century (phase 4).

iv)
Human interference on Plansee is minor compared to other Alpine lake environments. The influence of human activity 625 is limited to the lakeshore level. Artificial lake level changes since 1908, wave action from the operation of a cruise ship since 1927, and construction works on the street nearby the northern shoreline may have had an impact mainly on coastal erosion. While shoreline processes contribute to the background sedimentation, it is possible that they create subaquatic detrital layers which are difficult to distinguish from other event deposits. Debris flows can entrain sediment from the coastal zone, which increases the sediment volume reaching the lake basin. The construction of 630 ripraps and retention basins since the mid-19 th century possibly led to a slightly increased threshold for sediment delivery into the lake by debris flows. Regarding the frequency of episodic debris flows, human interference plays a subordinated role since their zone of release lies above the altitude of human influence.

v)
Strong earthquake shaking causing mass movements can fundamentally change the sediment availability in the catchment and is interpreted as the causal factor for increased debris-flow activity in phase 1 (Sect. 5.3). Such a 635 postseismic increased debris-flow activity could also be interpreted for eq 1 at 1930 CE but in much lower magnitude (~1.8-fold frequency increase) and extent (~10 years) than for eq 3 at ~2150 BCE. However, conclusive interpretation on postseismic landscape response of eq 1 is not possible, due to stacked effects of the contemporary human influence.
In contrast, after eq 2 at ~1050 BCE no significant increase in debris-flow activity can be observed. Eq 2 was inferred to be located more to the south and had likely less local intensity at Plansee (Oswald et al., 2021). Thus, we conclude 640 that only the strongest shaking events well above local intensity of VI generate a distinct postseismic landscape response in a lacustrine record, as the M 5.3 Namlos 1930 CE earthquake has no clear landscape response.

vi)
Changes in precipitation patterns through climate forcing are the main factor controlling debris-flow activity (Jomelli et al., 2019) and intense rainfall is the most important trigger mechanism for their release. Since the Plansee catchments offer large material supply, the system is highly sensitive towards changes in precipitation. Variations in 645 the precipitation pattern are directly reflected in the well preserved and highly resolved sediment archive of the last 4,000 years. Temporal coincidence of two df turbidites with the two heaviest rainstorms of the century in 1999 and 2005 CE (Fig. 8b) let us infer that regional (advective) rainstorms lasting over several days are a trigger mechanism of debris flows also in the Plansee region. This is also supported by coincidence of several outstanding thick df turbidites of phase 2 e.g. in the first and eighth century BCE, a period for which enhanced flood activity has also been 650 documented in the record of Ammersee (Czymzik et al., 2013) or Mondsee (Swierczynski et al., 2013). However, for the last century in phase 4, not all rainstorms triggered a macroscopic df turbidite and not each df turbidite has a corresponding rainstorm event (Fig. 8b). This is quite expectable since local storm cells have diameters of a km or less. Potential mismatches might also be due to the age error of df turbidites. Such a relationship between short extreme convective precipitation and debris-flow activity in the Alps is commonly observed in recent times (e.g. 655 Schneuwly-Bollschweiler and Stoffel, 2012).
The region has experienced a drastic 2-fold increase in convective precipitation frequency between 1920 and 2010 CE with an average increase of 10 % per decade (Hydrographischer Dienst Tirol, 2020). The 40 km distant Hohenpeissenberg Meteorological Observatory recorded a 2-fold increase in days with precipitation ≥30 mm from 1879 to 2000 CE (Fricke and Kronier, 2002). Before comparing these rainfall records with our debris-flow record, 660 human influences in the 20 th century need to be considered. The presented debris-flow record is likely overestimated in phase 4.1 and 4.2 (1908( -1978 due to artificial lake level changes, coastal erosion, and road constructions, but the record is likely underestimated in phase 4.3 (since 1978 CE) due to preventive constructions in the northern lake part (see iv above). Without the possibility of quantifying these human influences, we infer a mean frequency of phase 4.2 and 4.3 to be a best estimate, showing a ~7-fold increase in phase 4.3 (1978-2018 CE) compared to phase 3 665 (1700-1900 CE) coincident with the instrumentally documented increase in rainstorm activity. Moreover, we herewith provide sedimentological evidence for the increased debris-flow activity in the 20 th century observed on differential LIDAR data from several fans at the northern shore of the lake (Dietrich and Krautblatter, 2017). Their proposed 10-fold increase from 1947-1952CE to 1987-2000 CE is also in the order of magnitude of the corrected 7-fold increase inferred from our record. An increasing rainstorm frequency since the 20 th century is also observed in 670 several lacustrine flood records in the Alps (e.g. Glur et al; Swierczynski et al., 2013) and historically documented river floods in central Europe (Blöschl et al., 2020), pointing to regional changes in the atmospheric circulation patterns. However, the absence in the Plansee record of other historic periods with enhanced rainstorm activity documented in these records let us infer that debris-flow activity in small catchments is strongly controlled by local high-intensity convective precipitation events. A temperature rise in the course of ongoing global warming can cause 675 increased convective rainfalls (Trapp et al. 2013) and therefore likely leads to increased rainstorm-triggered debrisflow activity.

Conclusion 680
The sedimentary infill of Plansee holds a well-preserved Alpine archive of Holocene debris-flow activity in a system of high permanent debris production on juvenile, highly connected fan morphologies. Here, we present a 4,000 year continuous record of debris-flow dynamics and an unprecedented amphibious characterization of debris-flow fans. The ratio of terrestrial and subaqueous deposition from debris-flows during the summer of 2019 on morphologically stable and unstable alluvial fan deltas are determined. In order to reconstruct the sedimentation dynamics in the lake basin, we distinguish the causal processes of 685 different turbidites based on sedimentological and geochemical characteristics and their spatial occurrence in a transect of four sediment cores. Previously non-observed geomorphic expressions of debris flows from source to sink revealed by TLS and bathymetric investigations emphasize the importance of amphibious research when studying depositional patterns. This multidisciplinary approach can be applied to study sediment dynamics in other lakes which are influenced by mass movements.
The highlight of this study is sedimentological evidence for a massive increase of debris-flow frequencies in the 20 th and 21 st 690 century. Numerous empirical studies and global climate models attribute an enhanced hydrological cycle and increase in frequency and/or magnitude of heavy precipitation to climate forcing. The temporal coincidence of increasing debris-flow frequency at Plansee with enhanced rainstorm activity in the 20 th century provides further evidence for the direct link between climate change and debris-flow activity.

Data availability
The TLS data, bathymetric data and grain size data from this study are available upon request.

Competing interests
The authors declare that they have no conflict of interest.