Over the past decades, X-ray computed tomography (CT) has been increasingly applied in the geosciences community. CT scanning is a rapid, non-destructive method allowing the assessment of relative density of clasts in natural archives samples. This study focuses on the use of this method to explore instantaneous deposits as major contributors to sedimentation of high-elevation lakes in the Alps, such as the Lake Lauvitel system (western French Alps). This lake is located within a very steep valley prone to episodic flooding and features gullies ending in the lake. This variety of erosion processes leads to deposition of sedimentary layers with distinct clastic properties. We identified 18 turbidites and 15 layers of poorly sorted fine sediment associated with the presence of gravels since AD 1880. These deposits are respectively interpreted as being induced by flood and wet avalanche. This constitutes a valuable record from a region where few historical records exist. This CT scan approach is suitable for instantaneous deposit identification to reconstruct past evolution and may be applicable to a wider variety of sedimentary archives alongside existing approaches.
At their introduction to the field around 50 years ago, X-ray radiographs were initially used to explore the
internal structure of sediment cores (Bouma, 1964; Baker
and Friedman, 1969) in order to optimize the opening process or even explore
bioturbation structures in the sediment (Howard, 1968). One of the
technical problems was the loss of information with respect to depth, as the
radiographs are a planar representation of a 3-D structure. A recent review of
computed tomography (CT) scans in the geosciences (Cnudde and Boone, 2013)
demonstrates the growing application possibilities of X-ray technology as
well as the limits of the technique. Improvements in CT scanning allowed
exploration of complex sedimentary structures through 3-D reconstructions,
leading to improvement compared to classic 2-D imaging
(Pirlet et al., 2010; Bendle et al., 2015). The
method is based on the relative density of each voxel (i.e., volumetric
pixel) constituting the chosen sample. The position of each voxel is set on
a
High-elevation lakes situated in mountain areas are often characterized by
elevated and highly variable sedimentation rates (Arnaud et
al., 2016). The variety of erosion processes caused by chemical and
mechanical weathering as well as rock breaking by frost action creates
heterogeneous grain size elements. Extreme climatological events can trigger
several high-energy transport mechanisms which could induce deposition of
these elements in lake sediments. Depending on the processes, extreme events
may induce different sedimentary structures containing coarse grains
(Arnaud et al., 2002; Sletten et al.,
2003; Nielsen et al., 2016). Fluvial events such as floods are able to
transport very large quantities of sediment in a short period of time
(Sturm and Matter, 1978; Jenny et al.,
2014). In recent years, they have also been largely identified as a major
sedimentary source in high-elevation lakes
(Giguet-Covex
et al., 2012; Glur et al., 2013; Wilhelm et al., 2013, 2015; Wirth et al., 2013). As floods are formed by heavy precipitation events, the
torrential stream transporting sediment in suspension will enter into the
lake and create a density current resulting in a characteristic deposit
called turbidite (Gilli et al., 2013). The density difference
between subaerial flow and lake water can create different underwater flow,
but each type will result in a coarse-grain base with a fining-upward trend
(Sturm and Matter, 1978). In certain cases, lake surroundings
may include gullies orienting subaerial flow into the water. Two
mass-wasting types of transport related to these gullies were identified in
high-elevation lakes' sediments. The first type is debris flow triggered by
water transport but with a lower water content compared to floods
(Postma, 1986; Dasgupta, 2003). Its transport capacity is thus
increased and they will form specific deposits in underwater environments
due to their higher density and sediment cohesion (Mulder and
Alexander, 2001). Typical deposits are composed of a load cast layer
containing a basal erosive surface, which is overlain by a fining-upward
layer comprising the finer sediment fractions
(Sletten et al., 2003; Irmler et
al., 2006). The second main type of input can be attributed to wet-snow
avalanches that occur mostly over springtime. Wet-snow avalanches are
typically observed in steep alpine valleys where the slope exceeds
28
All of these high-energy processes led to the presence of coarse-grained deposits, and methods used to identify and count the coarser elements have been based on wet sieving of successive layers of sedimentary cores, which is a both time-consuming and destructive method (Seierstad et al., 2002; Sletten et al., 2003; Nesje et al., 2007; Vasskog et al., 2011). In this study, we propose a complementary method to grain size analysis to better characterize these coarse grains in a simpler, faster and non-destructive way based on the use of CT scanning. This provides an ideal context in which to test our novel X-ray CT based technique and application on sediment cores from Lake Lauvitel located in the Oisans Valley (western French Alps).
Lake Lauvitel (44
The core LAU11P2 (76 cm) was retrieved using a short UWITEC gravity corer to
obtain a well-preserved interface, and LAU1104A (104.5 cm) was retrieved
using a piston corer with a 90 mm sampling tube at the same location. The
cores were split lengthwise and photographed at high resolution (20 pixels mm
CT scanning was performed at Hopitaux Universitaires de Genève (HUG)
using a multidetector CT scanner (Discovery 750 HD, GE Healthcare,
Milwaukee, WI, USA). The acquisition parameters were set as follows: 0.6 s
gantry rotation time, 100 kVp, 0.984:1 beam pitch, 40 mm table feed per
gantry rotation, and a
Grain size measurements were carried out on the core using a Malvern Mastersizer 800 particle sizer at a lithology dependent sampling interval. Ultrasonics were used to dissociate particles and to avoid flocculation. Several layers of gravel-sized mineralogic particles were identified (Fig. 2a) in the LAU1104 sediment core. To obtain a quantitative estimate of these particles, we passed samples through a 1 mm mesh and wet-sieved the sediment at variable intervals from 1 to 3 cm depending on the gravel concentration. The number of particles > 2 mm and macro-remains present in the sieve was counted for each interval in the core LAU1104A.
The chronology of the Lake Lauvitel sediment sequence is based on
short-lived radionuclide measurements. The short-lived radionuclides in the
upper 75 cm of core LAU11P2 were measured using high-efficiency, very
low background, well-type Ge detectors at the Modane Underground Laboratory
(LSM; Reyss et al., 1995). The sampling
intervals followed facies boundaries, resulting in a non-regular sampling of
approximately 1 cm. Twelve thick beds (at depths of 10.4–12.7, 17.3–19,
22.9–24.8, 29.7–30.9, 38–39, 40.6–42.4, 43.1–44.2, 45.7–50, 54.5–56.9,
60.4–62.5, 64.1–66 and 67.2–68.3 cm) were not analyzed because they were
considered to be instantaneous deposits or part of an instantaneous deposit
(see Results).
The core lithology is composed of three facies (Fig. 2a). Facies 1 (F1) is
silty-clay, dark-brown, finely laminated layer. It is interbedded by two
other facies that are almost always associated with each other: facies 2
(F2) is a normally graded bed from coarse sand to silt, sometimes with an
erosive base; this facies is always associated with a thin white clay-rich
layer, facies 3 (F3), on the top. Figure 2b presents typical normally graded
beds with grain size distribution (in red) characterized by a median grain
size (Q50) of 44.1 and a mode of 81
The CT scan analysis is based on relative density expressed on the histogram
(Fig. 3a) representing the frequency of each of 1–255 levels of grey (0 is
not shown on the graph due to overrepresentation corresponding to the
background signal). Three modes representing the most frequent values are
apparent in the histogram and are associated with certain types of
sediment. The first mode is centered on the 106 value. After selecting this
mode, we isolated the numerical values in order to map them by using the
plugin. The corresponding elements in the sediment core were small OM
macroremains such as a
To compare objects counted numerically and objects counted manually, we need
to know the size limit in units of volume (voxels), which is equivalent to
2 mm diameter holes in a sieve. In 2-D, a particle is retained in the sieve
only if at least two sides are 2 mm in length, meaning at least two sides
are 4 pixels long. Therefore, a particle of 16 (4
In the LAU1104A sediment core, a total of 456 gravel clasts equal to or
larger than 13 voxels were identified for a total of 112 683 mm
We then compared the 3-D Object Counter results and the coarse grains
recovered from the sediment cores in slices of variable thickness ranging
from 1 to 3 cm. The depth 97–98 cm had no gravel > 2 mm in either
the manual or numerical counting (Fig. 3b, d). When considering a large
amount of gravel, the manual and numerical counting methods showed
differences. For depths 15–18, 42–44, 44–46, 51–52, and 72–73 cm, the number
of gravel clasts was always underestimated by the numerical counting. As the
3-D Object Counter plugin is identifying objects from one pixel and its eight
neighbors in 2-D and its 26 neighbors in 3-D (Bolte and
Cordelieres, 2006), the identification of objects could vary especially
because of the noise treatment and when the object size is close to the
image resolution. The numerical counting result is slightly underestimated
compared to the manual counting result (30 % on average). However, depths 5–7 and 46–48 cm showed an overestimation by the numerical
counting (77 % on average). Considering the resolution, it is possible
that a certain number of aggregated sand grains could have been considered
gravel by the numerical counting method, leading to an overestimation. This
could be explained by the presence of flood deposits in these two depths
(Fig. 3b). Aggregated sand-sized elements would be considered by numerical
counting as larger elements. In addition, the sand-sized elements are
rounder and would go through the sieve, as opposed to an angular particle of
similar volume which would be retained in the sieve. Overall, from this
comparison between the numerical and the manual counting and accounting for
the previously mentioned CT scan bias, we obtained a relatively
well constrained positive correlation (
The OM counting identified 7413 objects, spread throughout almost every
part of the sediment core. The largest OM element found in the core was
6949 voxels in size, corresponding to 1732 mm
The
A number of distinct layers, including normally graded beds, are identified in the Lake Lauvitel sedimentary record. Analysis of median grain size (Q50) and the coarser 10th percentile (Q90) parameters from within graded beds leads us to consider these to be turbidites caused by heavy rainfall in the watershed (Støren et al., 2010; Giguet-Covex et al., 2012; Wilhelm et al., 2012, 2013, 2015; Gilli et al., 2013). Gravels were found in the upper part of the flood deposit that are associated with receding torrential activity (Gilli et al., 2013). These gravels are unlikely to have originated from the torrential activity due to the distance from the delta. The presence of gravel in the turbidites could possibly be attributed to debris flow activity resulting in an dense cohesive underflow transforming in a turbidite layer (Weirich, 1988). However, we do not observe a load clast at the base of the deposit as is typical of what a debris flow would exhibit, but instead our results show sparse gravel presence in the upper part of the deposits (Fig. 2). Gravels within flood deposits could be linked to temporary tributaries only active during heavy precipitation, for example flows transmitted through avalanche corridors over summer. A similar pattern of gravel distribution is also observed in the homogeneous fine annual sedimentation (Fig. 2). In these layers, the sorting is similar to that of the annual sedimentation without gravel. This could be explained by the gravel elements falling either directly into the lake or onto the frozen lake surface and subsequently producing drop stones as it melts away.
Sum of gravels > 13 voxels at 5 mm intervals identified in the LAU1104A sediment core since AD 1880 without the normally graded beds. The dashed line represents the threshold number from which avalanche periods are identified (highlighted in blue). Exceptional winters found in the bibliography are represented in red (Allix, 1923; Jail, 1970; Ecrins National Park internal report, 1978). EPA (Enquête Permanente sur les Avalanches) number of avalanches per path since AD 1950 is in green, and interannual mean value is in black (Eckert et al., 2013). The avalanche record for the past century is from tree rings in the nearby Romanche river valley (Corona et al., 2010).
Fifteen layers are identified exhibiting a high proportion of gravel elements accompanied by poorly sorted fine grains of multi-modal grain size distribution (Fig. 2). Similar features have also been observed in Norway, where they are attributed to avalanche induced depositions (Blikra and Nemec, 1998; Seierstad et al., 2002; Nesje et al., 2007; Vasskog et al., 2011). Wet avalanches occur in spring, when warmer temperatures lead to a loss of cohesion and instability in the snowpack. Given that the lake ice does not thaw until late spring, avalanches could either be deposited on ice or enter directly in the water as observed during the 1 May 2015 avalanche (Fig. 1). During this event, the snow flow originated from the C1 corridor in the northern part of the lake containing the upper basin and was thus unlikely to have any sedimentary connection to the coring site in the deeper basin. Snow avalanche detrital material can be integrated into lacustrine sediments in two ways. In the case of a frozen lake, surface avalanche deposits are spread across the ice and subsequently drop to the lake sediment from drifting ice. If an avalanche occurs while the lake is ice-free, the avalanches directly enter into the water, where particles are concentrated in a more restricted area closer to the avalanche corridor. The presence of fine sediment in between gravels could thus be originating from previously deposited particles and/or from the avalanche. Consequently, we consider them as annual sedimentation in our age model. Given that avalanche deposit can be a very local phenomenon, the coring point must be directly beneath the avalanche corridor and thus capture both drop stones and direct avalanche deposits in order to record the maximum number of events. In our record, we identify an avalanche deposit as multiple gravel elements at the same sediment depth, as opposed to a single element that could be related to a single rock falling from steep slopes. In order to better understand this deposition processes, multiple cores spatially dispersed in the deeper lake basin would give a better overall estimation.
After establishing the age model to the LAU1104A sediment core, we are able
to express the gravel abundance per 5 mm slice, for the interval from AD 1880 to present (Fig. 5). The gravel abundance goes from zero to almost
20 gravel elements per 5 mm deposited in the lake floor. A total of 456 gravel
elements were identified in the sediment core, 217 of which were identified
outside flood layers. Despite this, they represent a total of 106 922 mm
Our X-ray CT-based counting method is well suited for this type of
lacustrine sediment because density difference between fine silty and coarse
gravel elements is quite significant. The resolution of the CT scan allowed
identification of the centimeter-sized gravels found in sediment cores.
However, in this study the resolution of the CT scan was limited to a pixel
resolution of only 500
CT scanning is a well-established technique in medical diagnosis and has been
used for several geoscience-related studies in recent times. The principle
of the analysis is based on differences in the relative densities of an
object. This study explores the possibility of using a novel X-ray CT-based
approach to analyze distinct deposits in lake sediments. The analysis
highlighted the presence of denser > 2 mm mineralogical particles
in the silty sedimentary matrix, as well as the abundant organic matter
which could be a useful tool for sampling macroremains for
The grain size data, gravel counts and age model from this study
are
available at
The authors declare that they have no conflict of interest.
Laurent Fouinat's PhD fellowship was supported by a grant from Ecrins National Park, Communauté des Communes de l'Oisans, Deux Alpes Loisirs and the Association Nationale de la Recherche et de la Technologie (ANRT). The authors wish to thank Ecrins National Park for their authorization for sampling and assistance during the field work. The authors thank the Laboratoire Souterrain de Modane (LSM) facilities for the gamma spectrometry measurements, Hopitaux Universitaires de Genève (HUG) for the CT scan analysis, and Timothy Pollard for correcting the English of this paper. Edited by: V. Galy Reviewed by: two anonymous referees