Application of Non-Destructive Techniques on a Varve Sediment Record from Vouliagmeni Coastal Lake, Eastern Gulf of Corinth, Greece

: During the last few decades, X-ray attenuation systems have been established as standard techniques in geosciences and as excellent scientiﬁc tools for the analysis of sedimentary facies and structures. In the present study, we use two non-destructive and high-resolution systems (computed tomography, X-ray ﬂuorescence) to address all sedimentological facies and structural characteristics on a 6 m long, partly laminated sediment record, from Vouliagmeni lake, located at the eastern part of the Corinth Gulf, Greece. Vouliagmeni lake is the deepest coastal lake in Greece, and its location is of great importance since it is located in one of the most tectonically active regions in the world. The chronological framework of the retrieved sediment sequence spans the last 12,000 years, with alternations of laminated and non-laminated sections. The annual accumulation of the laminated sequences was determined through the radionuclide concentration of 14 samples. Scanning tomography was performed with a medical CT scanner and a µ CT scanner, aiming to compare the potentials and variations of both methods. Lamination boundaries, event layers (turbidites) and sediment deformations were distinguished through processing the extracted 3D rendered volumes, after applying ranges depending on Hounsﬁeld (HU) values. Microscopic analysis revealed three main layer types in the varve sequences that were examined, attributing to summer / spring, autumn and winter deposits. Statistical analysis of the elemental composition, magnetic susceptibility (MS) measurements and HU values revealed three clusters / facies, reﬂecting climatic and in-lake hydrological changes. Cluster A emulates sedimentation during arid conditions, while Cluster B reﬂects humid conditions with increased precipitation and erosion. Cluster C represents sequences with homogenous Ca-rich sediment. Our proposed workﬂow highlights the possible correlation between the non-destructive variables that were measured, but also the variations and applications of each method and software used during this study.


Introduction
Holocene archives of the Eastern Mediterranean region are shaped by complex topography and climate heterogeneity, combined with increasing human impact on the landscape since the

Core Sampling
During the field campaign in April 2018, one 6 m long sediment core and a 4 m long parallel one were retrieved from the deepest part of the lake (~46 m) (38 • 1.740 N 22 • 52.920 E) using a Usinger piston corer system [79] and a wire-operated second platform at the lake bottom. Coring equipment was provided by the Institute of Ecosystem Research, Kiel University, Germany. After the extraction, the cores were cut into 1 m sections, split, and macroscopically examined for the first overview concerning sediment structure, color and texture, and then stored in cool rooms (4 • C) in the Department of Geology, University of Patras, Greece.

Sedimentology/Mineralogy
Standard sedimentological analyses were conducted on 60 samples from the sediment core (~10 cm resolution). Grain size distribution was established through a Malvern Mastersizer, Hydro 2000, with the classification established according to [80]. CaCO 3 content was determined using a FOG II/Digital Hand-Held Calcimeter (BD Inventions) using a modified method from [81] and [82]. Magnetic susceptibility measurements were conducted through a Bartington MS2E system, with a step size of 1 cm throughout the sediment sequence. Sediment color was determined by extracting the RGB spectrum through ImageJ software.
X-ray diffraction (Brucker D8 Advance) with Cu-K-alpha radiation (λ = 1.5418 A) and Nickel filter was used for the characterization of the mineralogical composition of the sediment samples. Crystallographica Search-Match v 2.0.3.1 (©Oxford Cryosystems Ltd, UK.) software was used for the qualitative analysis of the samples, with the identified minerals further verified with TOPAS software v. 3. The semi-quantitative analysis of the minerals was performed via the same software, excluding the clay minerals due to the error of the method. For the quantification of clay minerals, the Area method was used [83]. The final percentages of the minerals in the sediment samples (including the clay fraction) were obtained via normalization. Due to the detection limit of the XRD system (2-3%), minerals that were <3% in the sediment samples were not detected, and for this reason, only semi-quantitative analysis is possible.

CT Scanning/Thin Sections
Medical computed tomography was conducted for each core segment (1 m) separately, using a Toshiba Aquilion Prime CT scanner, at the University of Patras, Greece. Acquisition parameters were set as follows: 0.5 mm slice thickness, 0.3 mm slice interval, helical rotation with pitch factor 0.637, 120 kV and 350 mA. For enhancement of the output data, during the rendering process, the soft tissue and bone algorithms were used. Each algorithm provides a different setting; thus, both approaches were examined and compared for better visualization of the sediment structures. Each core segment scanned (~1 m) produced 3500-4100 DICOM files when the scanner was set at the highest possible resolution.
Micro-CT scanning was performed at the Hellenic Centre for Marine Research (HCMR) using a Skyscan 1172 micro-tomograph (Bruker, Kontich, Belgium). The scanner uses a tungsten X-ray source which is equipped with an 11 PM CCD camera (4000 × 2672 pixel). The sample was scanned at a voltage of 100 kV and a current of 100 µA with a combination of aluminum and copper filter and a pixel size of 13.79 µm for a half rotation of 180 • . Projection images were reconstructed into cross-section images using the SkyScan's NRecon software (NRecon, Bruker, Kontich, Belgium), which implements a modified Feldkamp's back-projection algorithm. Subsequently, the reconstructed images were loaded into the software CT Analyser v. 1.18.4.0 + (CTAn, Bruker, Kontich, Belgium) to calculate the mean grayscale value, which represents the relative density of the sample. Furthermore, 3D analysis of the scanned sample was performed by using the custom processing plugin of CTAn software to calculate the 3D thickness of the sample. For optimization of the exported data, all samples during both CT scanning methods were scanned "wet" [45].
Appl. Sci. 2020, 10, 8273 6 of 25 Three representative thin sections (4.3 × 2.3 cm) were prepared using standard techniques, including embedding in epoxy resin, and grinding, aiming to underline potential variations and advantages/disadvantages between CT scanning and microscopic analysis. Microscopic examination was performed through a petrographic microscope at 100-400× magnification.

Elemental Composition
Downcore elemental variations were measured by an Avaatech X-Ray Fluorescence core scanner, on the split core surfaces, at the Institute of Geosciences, Kiel University. The core segments were covered with a high-purity polypropylene film and then scanned with 5 mm resolution and a rhodium X-ray source. Two separate runs were conducted, the first at 10 kV and 10 sec exposure time for the elements Al, Si, S, Cl, K, Ca, Ti, Mn, Fe and a second scan at 30 kV and 15 sec exposure time for the elements Zn, Rb, Sr, Zr. Measured elemental intensities were all plotted as ratios rather than absolute concentrations, to avoid closed-sum effects [84,85].

CT Scan Workflow
Core segments were first scanned with a Nikon line scan camera, and high-resolution digital photos were acquired. These photos were used for a first evaluation of the laminated sequence, with reference (top/bottom depth) of each laminated section. Since the Lake Vouliagmeni core presents distinct organic (black) and calcite (white) laminations, RGB colors and a grayscale profile were also extracted through ImageJ software, providing an even clearer distinction of laminated sections' boundaries.
Each point scanned through CT is characterized by a specific "signal", which is expressed in Hounsfield units (HU). Medical CT scanners, similar to the one used in this study, are set with a HU value of −1000 for air and 0 for water. Apart from sediment internal structural characteristics, the correlation between HU and the sediment density/atomic composition can provide a fast overview and distinction between sedimentological facies, thus sections where different sedimentological processes prevail in the lake. For recognition of each laminae boundary, the HU ranges of the calcite/white and organic/black laminations were detected through INOBITEC medical software and SedCT software [86]. During the examination in all core sections, the PVC tube from the core, as well as the first 1 mm of the core surface, were excluded from the 2D rendered volume, to prevent miscounting of HU values due to oxidization and sediment deformation from core splitting. By presenting each time the HU range of each lamination type while excluding the non-laminated deposits, in the 3D model, the exact lamination number was measured by the automated software Cybis Coordinate Recorder (CooRecorder; http:/www.cybis.se/cbeewing/index.htm) and BMPix and Peak tools [87]. Additionally, laminations were counted manually by extracting 2D image sections, setting a scale and then measuring each lamination separately. A total of~3230 laminations were calculated in the core by using both proposed methods. No significant variations were observed between the different software used, except some minor miscounts of Cybis Coordinate recorder, in sections where lamination boundaries in the core surface were deformed during splitting.
Event sedimentation into the system was examined through 4 main criteria on the non-laminated deposits, including (1) HU values, (2) magnetic susceptibility, (3) Mn content and (4) potential inverse grading. Then, 3D models were constructed by selecting regions of interest (ROI) in the Avizo Fire (Thermo Fisher) software and by projecting them with the frequency distribution and grayscale colourmap. Both HU and MS present high values at the base of the turbidite deposits [88][89][90].

Core Chronology
The chronological framework of the core was established through 10 accelerator mass spectrometry (AMS) radiocarbon samples, analyzed in Poznan Radiocarbon Laboratory (Poland) and Beta Analytic, Dublin (Ireland) ( Table 1). Three different types of sample material were chosen due to the lack of datable organic material in some sections. Whenever it was possible, charcoal samples or sediment with high organic compounds were preferred for dating. In sections where this was not possible, bivalve shells (Cerastoderma glaucum) were selected with the criteria of being intact and showing no sign of transport. Shell samples were also thoroughly washed with distilled water and placed in an ultrasonic bath for around 30 min, removing any residue sediment from the valves. Due to the lack of data concerning the exact reservoir correction on the Gulf of Corinth, all shell samples were treated with a ∆R correction of 89 ± 58 years [91]. Table 1. List of radiocarbon samples from Lake Vouliagmeni core. Calibrated ages are presented in median values and were determined using IntCal13 calibration dataset [92]. * = Marine reservoir correction as suggested by [91]. For the investigation of the annual mechanism of deposition in the laminated sections, 14 dried sediment samples (1 cm thickness) were tested for radionuclides concentration in teleDOS Labs (ISO 11929:2010). The artificial radionuclides 137 Cs and 241 Am, as well as the natural 238 U decay series, were measured using Broad Energy Germanium detectors (BEGe-5030 with carbon window) of Canberra-Eurisys.

Core Description
The Vouliagmeni core is characterized by a complex stratigraphy, with alternating laminated and non-laminated deposits ( Figure 2). Sediment color (RGB) presented in % of all spectrums, has the highest values in the white (CaCO 3 ) laminations and lowest values in the non-laminated deposits ( Figure 2). The 8-bit image used for the grayscale analysis also displays the maximum values (~200) in the same regions. Laminated sediments occur throughout the sediment core but present a higher frequency from 220 to 430 cm. The mean thickness of the laminated sections in the core is around 9 cm.
Appl. Sci. 2020, 10, x FOR PEER REVIEW 8 of 26 in the same regions. Laminated sediments occur throughout the sediment core but present a higher frequency from 220 to 430 cm. The mean thickness of the laminated sections in the core is around 9 cm. Grain size does not present significant variations in the sediment sequence, with silt and clay fractions ranging from 55 to 70% and 35 to 50%, respectively ( Figure 2). Sand fraction was only recorded in the upper part of the sediment core (50-150 cm) ranging from 1 to 10% ( Figure 2). The mineralogical composition of the non-laminated sections shows high variability, with quartz, calcite, halite, aragonite, albite and clay minerals displaying the highest compound. Quartz and calcite comprise the main mineralogical facies through the sequence and generally follow the same trend ( Figure 2). Lower values of both minerals occur in sections 120-140, 320-330 and 450-470 cm. Halite values range from ~0 to 10%, except for a single sample at 130 cm with a value of 60%. Aragonite and albite range from ~0 to 8% and ~0 to 7%, respectively, with no significant changes in the sequence. Clay minerals range from ~1 to 20% with a mean value of 12%.
The CaCO3 content measured in the non-laminated parts ranges from ~0 to 45%, with the highest values recorded at 320 and 380 cm ( Figure 2). Carbonate precipitation seems to be constant in the system, as indicated by most measurements conducted on the core, but continuously altering Grain size does not present significant variations in the sediment sequence, with silt and clay fractions ranging from 55 to 70% and 35 to 50%, respectively ( Figure 2). Sand fraction was only recorded in the upper part of the sediment core (50-150 cm) ranging from 1 to 10% ( Figure 2). The mineralogical composition of the non-laminated sections shows high variability, with quartz, calcite, halite, aragonite, albite and clay minerals displaying the highest compound. Quartz and calcite comprise the main mineralogical facies through the sequence and generally follow the same trend ( Figure 2). Lower values of both minerals occur in sections 120-140, 320-330 and 450-470 cm. Halite values range from~0 to 10%, except for a single sample at 130 cm with a value of 60%. Aragonite and albite range from~0 to 8% and~0 to 7%, respectively, with no significant changes in the sequence. Clay minerals range from 1 to 20% with a mean value of 12%.
The CaCO 3 content measured in the non-laminated parts ranges from~0 to 45%, with the highest values recorded at 320 and 380 cm ( Figure 2). Carbonate precipitation seems to be constant in the system, as indicated by most measurements conducted on the core, but continuously altering regarding crystallization process. Samples examined for micro and macrofauna remains were all barren, except for five samples at 85, 290, 470, 570 and 590 cm, where minor assemblages of intact juvenile Cerastoderma glaucum shells were recognized and used for radiocarbon dating.

Laminated Sections
X-ray radiographs were composed for each core segment individually, and the mean HU value for each slice was extracted through SedCT software [86]. Colorized and grayscale images (Figure 3a) generated through the software reflect the interaction between the beam of gamma rays with the sediment and are highly dependable on the alternating density of non-laminated and laminated deposits. Sections with blue-based colors indicate lower HU values and thus lower density, whereas areas with more yellow/red colors indicate sections of more dense material (Figure 3a). HU values range from 400 to 700 for the laminated sequences and from 900 to 1400 for the non-laminated deposits.
Appl. Sci. 2020, 10, x FOR PEER REVIEW 9 of 26 regarding crystallization process. Samples examined for micro and macrofauna remains were all barren, except for five samples at 85, 290, 470, 570 and 590 cm, where minor assemblages of intact juvenile Cerastoderma glaucum shells were recognized and used for radiocarbon dating.

Laminated Sections
X-ray radiographs were composed for each core segment individually, and the mean HU value for each slice was extracted through SedCT software [86]. Colorized and grayscale images (Figure 3a) generated through the software reflect the interaction between the beam of gamma rays with the sediment and are highly dependable on the alternating density of non-laminated and laminated deposits. Sections with blue-based colors indicate lower HU values and thus lower density, whereas areas with more yellow/red colors indicate sections of more dense material (Figure 3a). HU values range from 400 to 700 for the laminated sequences and from 900 to 1400 for the non-laminated deposits. Since the mineralogical composition does not remain constant in all laminae, minor variations can be expected depending on the HU values; nevertheless, the distinction between the laminated and the non-laminated part can be easily obtained through the 3D model (Figure 3b). A two-scale approach was used for better visualization and distinct laminae counting. In larger-scale images, the mean HU value for all CaCO3-rich deposits was obtained by using INOBITEC medical software; thus, these sections ( Figure 3b) were separated from the rest of the material. Visualization of the annual layers (varves) was enhanced by just focusing on regions of interest (ROI) (Figure 3c). Since the mineralogical composition does not remain constant in all laminae, minor variations can be expected depending on the HU values; nevertheless, the distinction between the laminated and the non-laminated part can be easily obtained through the 3D model (Figure 3b). A two-scale approach was used for better visualization and distinct laminae counting. In larger-scale images, the mean HU value for all CaCO 3 -rich deposits was obtained by using INOBITEC medical software; thus, these sections ( Figure 3b) were separated from the rest of the material. Visualization of the annual layers (varves) was enhanced by just focusing on regions of interest (ROI) (Figure 3c). Thickness/counting of each lamination can be then measured through the same software or by extracting 2D images from each section.
The µCT data, when compared to the medical CT models, indicate an estimated error in lamination boundaries of around 0.5% since the resolution of both methods exceeds the minimum requirements for lamination distinction. The structural or "sphere fitting" thickness model (Figure 4) constructed through CTan plug-in software was compared to the rendered volume, for a more coherent distinction between laminations. extracting 2D images from each section.
The μCT data, when compared to the medical CT models, indicate an estimated error in lamination boundaries of around 0.5% since the resolution of both methods exceeds the minimum requirements for lamination distinction. The structural or "sphere fitting" thickness model (Figure 4) constructed through CTan plug-in software was compared to the rendered volume, for a more coherent distinction between laminations. The varve cycle in Vouliagmeni lake primarily consists of three distinct layers with variance in mineralogical composition, as indicated by the microscopic analysis. White carbonate laminae consist of aragonite and are deposited in dry summer conditions when evaporation in the lake water is high. Precipitation of coarser crystals, in the gray sublayer (Figure 4), is linked to the beginning of runoff events during autumn, reaching the maximum intensity in the dark, organic-rich laminae deposited during winter months. Minerogenic grains like quartz and calcite, mixed with organic residues, are The varve cycle in Vouliagmeni lake primarily consists of three distinct layers with variance in mineralogical composition, as indicated by the microscopic analysis. White carbonate laminae consist of aragonite and are deposited in dry summer conditions when evaporation in the lake water is high. Precipitation of coarser crystals, in the gray sublayer (Figure 4), is linked to the beginning of runoff events during autumn, reaching the maximum intensity in the dark, organic-rich laminae deposited during winter months. Minerogenic grains like quartz and calcite, mixed with organic residues, are the main components of the dark laminae. Laminae mineralogical composition and absence/presence of laminae are strongly connected to the precipitation/evaporation regime in the lake; thus, the varves of the Lake Vouliagmeni are characterized as endogenic.

Non-Laminated Sections
The nature of the non-laminated section deposition was examined separately through the HU/density model, MS measurements and Mn content (Figure 5a). Six main event deposits/turbidity flows were distinguished (128-136, 139-145, 148-158, 343-352, 492-496, 505-520 cm), interrupting the laminated sections and presenting the highest values of HU, MS and Mn (Figure 5a). High-density sediment sections, like the one occurring at 100-120 cm, were not classified as event deposits, since they did not present any high values of MS (Figure 5a). Turbidite layers that were recognized are not characterized by inverse grading but by a more homogenous distribution of the coarser fraction (Figure 5b).

Non-Laminated Sections
The nature of the non-laminated section deposition was examined separately through the HU/density model, MS measurements and Mn content (Figure 5a). Six main event deposits/turbidity flows were distinguished (128-136, 139-145, 148-158, 343-352, 492-496, 505-520 cm), interrupting the laminated sections and presenting the highest values of HU, MS and Mn (Figure 5a). High-density sediment sections, like the one occurring at 100-120 cm, were not classified as event deposits, since they did not present any high values of MS (Figure 5a). Turbidite layers that were recognized are not characterized by inverse grading but by a more homogenous distribution of the coarser fraction (Figure 5b). Post-depositional soft-sediment deformation structures (SSDS) were detected in four laminated sections (176-178, 380-381, 436-437, 543-544 cm). Folding (Figure 5c), as well as convolute bedding, were the two main groups of SSDS recognized. Laminae structural characteristics in the deformed sections could still be observed. Folding laminae were observed to be followed by higher density deposits with thickness ranging from 0.4 to 1.3 cm.

Hierarchical Clustering
From the 14 elements measured through XRF scanning, calcium is the dominant one, representing around 45% of the total counts. Calcium in the lake can derive from limestone weathering in the catchment or through autochthonous precipitation inside the lake. Elements measured with the second (19.6%) and third (10%) higher abundance were Fe and Si, respectively. Signals obtained from those elements can be associated with an allochthonous origin from the Flysch formations and the coastal deposits in the study area.
The correlation between elemental concentrations, HU values and MS was expressed through a hierarchical clustering (Pearson correlation distance) heatmap ( Figure 6). Three main clusters were exported from the analysis, mirroring different prevailing climatic and in-lake physiochemical conditions. Measured elements were distinguished in Cluster A and Cluster C, with detrital originated elements and carbonate groups composing them, respectively. HU values and MS measurements are expressed in Cluster B and are primarily associated with highly dense material as well as event layers. Post-depositional soft-sediment deformation structures (SSDS) were detected in four laminated sections (176-178, 380-381, 436-437, 543-544 cm). Folding (Figure 5c), as well as convolute bedding, were the two main groups of SSDS recognized. Laminae structural characteristics in the deformed sections could still be observed. Folding laminae were observed to be followed by higher density deposits with thickness ranging from 0.4 to 1.3 cm.

Hierarchical Clustering
From the 14 elements measured through XRF scanning, calcium is the dominant one, representing around 45% of the total counts. Calcium in the lake can derive from limestone weathering in the catchment or through autochthonous precipitation inside the lake. Elements measured with the second (19.6%) and third (10%) higher abundance were Fe and Si, respectively. Signals obtained from those elements can be associated with an allochthonous origin from the Flysch formations and the coastal deposits in the study area.
The correlation between elemental concentrations, HU values and MS was expressed through a hierarchical clustering (Pearson correlation distance) heatmap ( Figure 6). Three main clusters were exported from the analysis, mirroring different prevailing climatic and in-lake physiochemical conditions. Measured elements were distinguished in Cluster A and Cluster C, with detrital originated elements and carbonate groups composing them, respectively. HU values and MS measurements are expressed in Cluster B and are primarily associated with highly dense material as well as event layers. During XRF scanning, elements intensities are acquired through repeated and continuous counts on the sediment core. The random error that can occur during measurements, combined with the constant radiation used on all core segments that were scanned, does conform to a Poisson distribution [93]. However, in the heatmap (Figure 6), three different analytical methods with variations in radiation and exposure parameters were used; thus, the dataset does not demonstrate a normal/Gaussian distribution. For this reason, log (base 10) values rather than z-scores were used for each variable.

Bayesian Age-Depth Model
The chronological framework of the core covered approximately the last 12,000 years and was established through Bayesian age-depth modeling, using the R package Rbacon (v.2.3; [94]) and the terrestrial calibration curve IntCal13 [92] (Figure 7). The construction of the age-depth model was based on the 10 radiocarbon dates that were obtained from shells, organic-rich sediment and charcoal ( Table 1). The core top age was set to 2018 CE, which was the year of coring. Turbidite layers that were recognized in the sediment sequence were assigned as "slumps" (event deposits), in the Rbacon script. Laminated sequences' accumulation rates were not taken into account in the final age-depth model since it would have been impossible to add anchor dates for all of them.
Appl. Sci. 2020, 10, x FOR PEER REVIEW 14 of 26 Figure 6. Hierarchical clustering heatmap of Pearson correlation coefficients between non-destructive techniques results for each depth (cm) measured. The 3 main clusters that were exported are presented at the top of the heatmap.
During XRF scanning, elements intensities are acquired through repeated and continuous counts on the sediment core. The random error that can occur during measurements, combined with the constant radiation used on all core segments that were scanned, does conform to a Poisson distribution [93]. However, in the heatmap (Figure 6), three different analytical methods with variations in radiation and exposure parameters were used; thus, the dataset does not demonstrate a normal/Gaussian distribution. For this reason, log (base 10) values rather than z-scores were used for each variable.

Bayesian Age-Depth Model
The chronological framework of the core covered approximately the last 12,000 years and was established through Bayesian age-depth modeling, using the R package Rbacon (v.2.3; [94]) and the terrestrial calibration curve IntCal13 [92] (Figure 7). The construction of the age-depth model was based on the 10 radiocarbon dates that were obtained from shells, organic-rich sediment and charcoal ( Table 1). The core top age was set to 2018 CE, which was the year of coring. Turbidite layers that were recognized in the sediment sequence were assigned as "slumps" (event deposits), in the Rbacon script. Laminated sequences' accumulation rates were not taken into account in the final age-depth model since it would have been impossible to add anchor dates for all of them.  Radionuclide measurements were conducted on 14 samples, covering the uppermost 30 cm of the sediment core. Unsupported 210 Pb inventory was extrapolated according to [95] and by assuming a constant influx of 210 Pb into the system since no event-linked sedimentation was recognized in this section [96,97]. The concentration limit was assigned at 17.5 cm with an age of~150 yrs. Deposition of 137 Cs was detected at its maxima at the depths of 1.5, 4.5 and 7.5 cm. The fallout of the Chernobyl accident in 1986 CE was recorded at 1.5 cm depth, followed by the nuclear weapon tests in the 1960s. The co-presence of 241 Am and 137 Cs supports the connection between 137 Cs peaks and the nuclear tests [98,99]. From 4.5 to 7.5 cm, 8 ± 2 seasonal laminations were measured, responding to 4 years, which is in agreement with the 137 Cs chronology.

Non-Destructive Proxies
The comparison of medical CT with µCT scanning reveals that both techniques are an effective and fast way for sediment core internal structure characterization. Laminae distinction, as well as event layers, could easily be determined through the proposed workflow. The main advantage compared to thin-section micromorphology examination is that, during thin section preparation, lamination disturbance and distention could lead to inaccurate measurements of lamination boundaries/thickness [45]. Processing of the acquired data is also much faster compared to thin section preparation and resin impregnation on the sediment samples, which would require much time to cover the whole sediment sequence. However, in the Lake Vouliagmeni sediment core, laminae sublayers could only be detected through microscopic analysis, which leads to the conclusion that the use of both methods is essential for thorough microstructural examination.
Density is a crucial parameter in X-ray interaction with matter [100][101][102][103]. Since CT scanners "express" this interaction with HU values, optimizing the output 3D rendered volumes with specific boundaries can provide information on the desired parameter. On the studied core, we used a range of HU values that would represent the white/aragonite-rich laminations and exclude the brown/black/organic-rich deposits. The principle behind this proposed method is similar to the one used by [104] to distinguish gastropod shells and calculate sediment versus shell percentage on a lagoonal sediment core. The main drawbacks during CT analysis are that, depending on the lamination thickness, older CT systems may not provide the required resolution to address lamination boundaries. Furthermore, due to the massive data amount produced from even 1 m of the sediment core from the CT scanner (~4000 DICOM files), standard personal computers may not be adequate to create the 3D rendered volumes.
Chemical weathering in the lake system is interpreted by variations in Rb/Sr ratio [105][106][107][108][109] (Figure 8). During matrix weathering and transport, Rb is a common substitution of potassium (K) in K-feldspars that comprise the crystal lattice of clay minerals. Alkaline elements like Sr and Ca enter the lake catchment either from an allochthonous source (carbonate weathering) or from autochthonous sedimentation inside the lake in the form of SrCO 3 [110,111]. Both geochemical proxies present a near symmetrical distribution in the sediment sequence ( Figure 8), with higher values indicating wet conditions in the study area, triggering extensive erosion and sediment flux in the lake.
In limestone-rich environments, calcium precipitates in the form of Ca 2+ and CO 3 2− ions, especially under anoxic/hypoxic conditions [112][113][114][115][116]. The solubility of CaCO 3 decreases during dry periods, thus forming carbonaceous laminated deposits [117] if lake bottom conditions are favorable. Positive balance between precipitation and evaporation is characterized by higher values of Ca/Sr ratio [110] and lamination frequency (number of laminae/cm) in the sediment sequence ( Figure 8), with aragonite accumulation occurring during dry seasons. Mn and Fe are positively affected by redox potential in the catchment, with Fe ions precipitating earlier than Mn ions due to decreased stability. Higher concentrations occur in oxidizing environments [108,[118][119][120][121] or events in which the lake bottom is re-oxygenated.
High values of Zr/Rb and Si/Ti are associated with coarser sediment material and biogenic silica (diatoms) in the sediment sequence [40,[122][123][124][125][126], while the clay fraction is mostly represented by allochthonous Rb accumulation. A correlation between elements with high atomic numbers like Zr and HU values has been recorded by [88], at the coarse base of turbidite deposits. The Lake Vouliagmeni sediment sequence is characterized by the dominance of the finer sediment fraction (silt/clay) including quartz and calcite crystals, indicating the dominance of chemical weathering in the catchment.

Paleoenvironmental Interpretation
The Lake Vouliagmeni sediment record displays continuous sedimentary cycles of endogenic varve sedimentation, interrupted by non-laminated deposits (Figure 8). The lake is located in a tectonic, fault-controlled, karstic coastal landscape, and due to the interaction between the lake and the open sea, the distinction between the marine and lacustrine character of the varves is difficult to determine [19]. The occurrence of typically identical varve sections throughout the sediment sequence and the presence of Lithophaga shells at the west coast of the lake [73] verify that the lake always interacted with the Gulf of Corinth. Additionally, the preservation/deposition of varves in the sediment sequence depends on sporadic water column density, stratification and anoxic conditions in the lake bottom that would decrease bioturbation [18].
During the formation of Vouliagmeni lake at the onset of the Holocene period [72], lake water conditions were not favorable for varve preservation until~10800 cal BP (Figure 8). Cluster C dominance in this period reflects high concentrations of Ca, Sr and S in the homogenous sediment. High seismicity rates during the Pleistocene-Holocene transition and the constant uplift of the Perachora peninsula during this period [72] must have been responsible for the laminae folding at~10,000 cal BP. Perennial humid conditions, inferred from a Rb/Sr increase, commence from~10,800 until~8000 cal BP and reach local maxima at~9000 and~9500 cal BP. An increase in water level due to the rising sea level [72] and possible high discharge of the karstic aquifer, suggested by high Ca/Sr values [127] during this time, is in agreement with other Eastern Mediterranean studies [128][129][130]. Two flood turbidites recorded during this time account for the re-oxygenation of the lake bottom water, as indicated by the increase in the Mn/Fe ratio. Both turbidite events are mainly associated with Cluster B, which reflects increased MS and HU values and increased precipitation in the area and sediment flux into the lake.
Between 8000 and 7200 cal BP, the absence of varves highlights the positive balance between precipitation and evaporation ratio. Humid climatic conditions that could lead to a decrease in lake water salinity and alternation of water pH (rise to~9) combined with possible wind-driven mixing, as shown by a small increase in the Mn/Fe ratio [131], preclude stable stratification of the water column. Extensive freshwater inflow into the system could also lead to dislocation of the hypoxic/anoxic layer at the lake bottom, causing oxidation of the bottom surface that could potentially lead to bioturbation. Similar patterns of anoxic zone movement into upper layers have been recorded in the Amvrakikos Gulf [132,133] and Aetoliko lagoon [134] in Western Greece. A transition to an arid phase around 7200-6800 cal BP, as suggested by a low Rb/Sr ratio, has also been recorded in other sites of the Eastern Mediterranean region [135,136]. Cluster A, dominant during this phase, expresses arid conditions and is composed of detrital originated elements. Increased lamination frequency during this period is in comparison with the anoxic status indicated by low Mn/Fe (Figure 8). Coarser sediment indicated by high Zr/Rb and Si/Ti (Figure 8) may have accumulated at the lake bottom through a seismic event that also produced the laminae deformation at this point. However, no distinct event deposits were recognized in the sequence. Multi-proxy diagram including geochemical profiles of Rb/Sr Ca/Sr, Mn/Fe, Zr/Rb and Si/Ti, lamination frequency as measured through CT analysis, clustering stratigraphy exported through the constructed heatmap, Eastern Mediterranean precipitation and temperature model for mid-to late-Holocene as suggested by [137] and Turbidite deposits, as well as SSDS, recognized in the core. Turbidite deposits and SSDS were assigned with additional 2σ error as suggested by the age-depth model. From 6500 to 4000 cal BP, the high frequency of laminated deposits is attributed to increased weathering of the limestone-rich catchment and saturation of the lake water with CaCO 3 . Eastern Mediterranean precipitation and temperature model (EMP and EMT) [137] present their maxima at around 4000 cal BP, reflecting dry and hot phases. In warm summer months, when the lake water temperature increases, CaCO 3 becomes more soluble [18] and precipitates in the form of aragonite [117]. In the course of this period, destruction of an early Helladic settlement in the area at around 4300 yrs BP [72] corresponds to the flood turbidite layer recognized around this time. Heavy precipitation combined with the steep geomorphology of the area seem to have been the triggering mechanisms for the hyperpycnal flow [138].
The transition to the late Holocene period (3000 cal BP to present) is accompanied by an increase in sedimentation rate into the system, probably caused by extensive urban development [139,140] that led to landscape modifications such as deforestation, agriculture, drainage of water bodies, etc. A dry period recorded from 3600 to 3000 cal BP during the late Bronze age civilization expansion [15,130,[141][142][143] is reflected in Lake Vouliagmeni by a distinct decrease in the Rb/Sr ratio. The period after 2000 cal BP can be characterized as relatively stable compared to the early-and mid-Holocene periods. Lamination frequency presents the highest increase during this time, possibly due to increased eutrophication in the lake system, caused by extensive anthropogenic activity. The rupture of the fault system around 1950 cal BP is the cause of laminae deformation at this time and the triggering mechanism of the turbidite deposit.

Conclusions
The Vouliagmeni lake sediment record is formed through a complex interplay between prevailing environmental conditions and in-lake sedimentary mechanisms. Using a multi-proxy approach that included core-logging non-destructive techniques and standard sedimentological analysis, varve formations and sedimentological facies were determined. Varved sections were distinguished from the non-laminated deposits, after applying HU boundaries during 3D reconstruction. Sediment semi-quantitative density, as recorded through CT scanning, was used for event layer recognition as well as soft sediment deformations. Sedimentological processes in the lake can be categorized into three main clusters/facies mirroring variations in climatic conditions and sediment flux. Clusters A and B were characterized by non-continuous laminated sediments, during the arid and humid phases, respectively. Cluster C, occurring mostly at the early stage of the lake, reflects homogenous, non-laminated, Ca-rich deposits. The alternation between well-varved and non-varved deposits, combined with the high susceptibility of the area in tectonic calamities, defines Vouliagmeni lake as one of the most promising archives of Greece for the study of paleoenvironmental/paleoclimatic transitions along with the active tectonism of the eastern Gulf of Corinth during the Holocene period.
Author Contributions: Conceptualization P.A. and A.E.; XRF scanning A.E.; CT scanning A.E.; µCT scanning K.K. A.E. and XRD analysis E.G. and A.E.; writing-original draft preparation A.E; review and editing, A.E., P.A., I.U., J.S., E.G. and K.K.; sampling and field campaign were supported by A.E., P.A., I.U. and J.S. P.A designed the project. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.