Suspended sediment export from large Alpine catchments (
Erosion processes and sediment dynamics in Alpine catchments are determined by geological, climatic, and anthropogenic factors. Geological forcing is one of the main drivers of sediment production and landscape development, through crustal thickening, deformation and isostatic uplift, and glacier inheritance (e.g. England and Molnar, 1990; Schlunegger and Hinderer, 2001; Vernon et al., 2008). Glacier inheritance influences sediment production and transport as demonstrated by a strong spatial association between sediment yield and past and current glacial cover (Hinderer et al., 2013; Delunel et al., 2014). Almost continuous temperature-driven glacier recession in the European Alps since the late 19th century (Paul et al., 2004, 2007; Haeberli et al., 2007) has maintained large parts of the landscape in early stages of the paraglacial phase, where unstable or metastable sediment sources (Ballantyne, 2002; Hornung et al., 2010) can maintain high sediment supply rates. Anthropogenic impacts on sediment yields are more recent, and on a global scale largely related to land-cover change through intensified agriculture and the trapping of sediment in reservoirs (e.g. Syvitski et al., 2005). Land-use changes mainly impact fine-sediment production (e.g. Foster et al., 2003; Wick et al., 2003), while river channelization, flow regulation, water abstraction, and sediment extraction have caused a general reduction in sediment yield and consequently led to sediment-starved rivers worldwide (Kondolf et al., 2014). In Alpine catchments, in addition to trapping in reservoirs, sediment transfer is also disturbed by flow abstraction at hydropower intakes. The reduction of sediment transport capacity downstream of intakes and the periodic flushing of locally trapped sediment has severe impacts on the sediment budget (e.g. Anselmetti et al., 2007) and downstream river ecology (e.g. Gabbud and Lane, 2016).
Here we focus on the dominant role of climate in sediment production and transfer in Alpine environments (e.g. Huggel et al., 2012; Zerathe et al., 2014; Micheletti et al., 2015; Palazón and Navas, 2016; Wood et al., 2016). The premise behind this work is that to explain impacts of changes in climate on Alpine catchment suspended sediment yield, it is necessary to consider both transport capacity and sediment supply. Sediment supply depends on many factors, most importantly the spatial location of sediment sources (e.g. lithology, distance to outlet, connectivity) and the specific processes of sediment production (e.g. hillslope erosion, glacial erosion, release of subglacially stored sediment, channel bed and bank erosion, mass wasting events) and transport (e.g. hysteresis).
In this study we look at specific sediment sources and the hydroclimatic conditioning of their activation (e.g. precipitation, runoff, and air temperature) with a process-based perspective with the aim to infer the possible effects of changes in hydroclimate, such as increases in temperature and/or precipitation intensity on suspended sediment dynamics. We identify four main sediment sources typical of Alpine environments: glacial erosion, hillslope erosion, channel bed or bank erosion, and mass wasting events (e.g. rockfalls, debris flows). Climatic conditions, specifically precipitation and air temperature, contribute to the activation of these four sediment sources through different processes and at different rates. Erosive processes of abrasion, bed-rock fracturing, and plucking at the base of glaciers provide proglacial areas with large amounts of sediment (Boulton, 1974). Due to glacial erosion, discharge from subglacial channels has high suspended sediment concentrations (e.g. Aas and Bogen, 1988). Temperature-driven snow and ice melt in spring and summer, as well as intense rainfall on snow-free surfaces, may lead to entrainment from proglacial areas provided they are connected to the river network (Lane et al., 2016). Hillslope erosion driven by overland flow and rainfall erosivity may be exacerbated in Alpine catchments by permanently or partially frozen ground (Quinton and Carey, 2008). In summer and autumn, when Alpine catchments are largely free of snow, intense rainfall may erode large amounts of sediment and transport it in rills and gullies to the river network. Intense rainfall is also responsible for triggering mass wasting events, such as debris flows and landslides, where a large mass of sediment is delivered to the channel network instantaneously (e.g. Bennett et al., 2012). Flow conditions (e.g. shear stress, stream power) then determine the sediment transport capacity and in-stream sediment mobilization along rivers, and hence its transfer to downstream locations.
Map of the upper Rhône Basin with topography (DEM resolution
is 250
The close link between precipitation, air temperature, runoff, and the activation–deactivation of sediment sources in Alpine catchments becomes critical in the context of climate change. Alpine regions represent a sensitive environment in relation to current rapid warming. In Switzerland, together with glacier recession, a reduction in snow-cover duration and mean snow depth has been observed during the last 30 years (e.g. Beniston, 1997; Laternser and Schneebeli, 2003; Scherrer et al., 2004; Marty, 2008; Scherrer and Appenzeller, 2006). Although current effects of climate change are less clear for precipitation (Brönnimann et al., 2014) than for temperature, a sharp reduction in the number of snowfall days has been observed at many meteorological stations in Switzerland (Serquet et al., 2011).
The upper Rhône River Basin draining into Lake Geneva in Switzerland is at the centre of our investigation. The basin has experienced a rise in air temperature that coincided with a rise in suspended sediment concentrations in the mid-1980s. Our main objective is to explore the presence of the signal of a warmer climate in the suspended sediment dynamics of this regulated and human-impacted Alpine catchment. In this work, we refer to fine sediment as the sediment transported in suspension. To investigate the potential causes of the observed increase in suspended sediment concentration, we conceptualize the upper Rhône Basin as a series of spatially distributed sediment sources that are activated or deactivated by hydroclimatic forcing. In addition to discharge (transport capacity), we consider four main hydroclimatic variables: (a) ice melt runoff (IM), which evacuates accumulated fine sediment, the product of glacial erosion, through subglacial channels (e.g. Swift et al., 2005); (b) snow-cover fraction (SCF), which influences ice melting onset, impacts ice-melt efficiency through albedo, and may result in more rapid erosion and sediment production through an increased glacier basal velocity (e.g. Herman et al., 2015); (c) snowmelt runoff (SM) from snow-covered areas, which may generate downstream hillslope erosion and channel erosion (e.g. Lenzi et al., 2003); and (d) effective rainfall (ER), defined as liquid precipitation over snow-free areas, which leads to hillslope erosion, mass wasting, and also, due to enhanced discharge, channel erosion (e.g. Bennet et al., 2012; Meusburger and Alewell, 2014). Our aims are the following: (a) to estimate daily basin-wide ice melt, snow-cover fraction, snowmelt, and effective rainfall over the Rhône Basin for the last 40 years; and (b) to analyse these variables with the goal to provide statistical evidence for possible reasons for the rise in suspended sediment concentrations in the mid-1980s.
List of the variables analysed: observed SSC and hydroclimatic
variables originating from measurements (
The upper Rhône Basin is located in the southwestern part of Switzerland,
in the central Swiss Alps (Fig. 1). It has a total surface area of
5338 km
The catchment has been strongly affected by anthropogenic impacts during the last century. The main course of the Rhône River has been extensively channelized for the purposes of flood protection: levees were constructed and the channel was narrowed and deepened in the periods 1863–1894 and 1930–1960 (first and second Rhône corrections). Due to the residual flood risk that affects the main valley, a third project was started in 2009 with the main objectives to increase channel conveyance capacity and river ecological rehabilitation (Oliver et al., 2009). In addition, significant gravel mining operations are carried out along the main channel and many tributaries. Since the 1960s, several large hydropower dams have been built in the main tributaries of the Rhône River. The total storage capacity of these reservoirs corresponds to about 20 % of the mean annual streamflow (Loizeau and Dominik, 2000). Flow impoundment, water abstraction, and diversion through complex networks of intakes, tunnels, and pumping stations have significantly impacted the flow and sediment regime of the catchment. Flow regulation due to hydropower production has resulted in a considerable decrease in discharge in summer and increase in winter (Loizeau and Dominik, 2000). That said, the construction of dams and the start of hydropower operation has coincided with a drop in the suspended sediment load of the main Rhône River measured at Porte du Scex in the 1960s (Loizeau et al., 1997; Loizeau and Dominik, 2000).
Two sub-catchments of the upper Rhône Basin are used for the
calibration and validation of the ice-melt model: the Massa and the Lonza
sub-catchments (Fig. 1). The Massa is a medium-sized basin (195 km
Our objective is to explore the potential effect of climate on suspended
sediment dynamics in the upper Rhône Basin during the period 1975–2015. To
this end, we analyse observed and simulated hydroclimatic and sediment-transport
variables as listed in Table 1: mean daily temperature
We use a snowmelt model to predict SM and SCF over the entire basin, because
snow station measurements are sparsely and irregularly distributed and a
physical consistency between precipitation and air temperature as climatic
driving forces and snowmelt and snow cover as response variables is needed.
The spatially distributed temperature index method (degree-day model) was
used due to its simplicity, low data requirements, and demonstrated success
on daily temporal scales over large basins (e.g. Hock, 2003; Boscarello et
al., 2014). The degree-day approach also matches the coarse spatial
(250 m
The snowmelt model includes snow accumulation and melt. On the grid scale,
precipitation
Similar to snowmelt, ice melt is also simulated with a temperature index
(degree-day) model on grid cells that are identified as glacier covered. The
daily ice melt
We perform the calibration and validation of the snow and ice-melt model
parameters in sequence, since the snow-covered surface is required for
ice-melt estimation on glaciers. The snowmelt factor
The objective function for calibration is based on a combination of mean
absolute error and true skill statistic. The mean absolute error MAE is
estimated as
We calibrate the ice-melt factor
The optimal value of
Here, nd is the number of observation days from June to October,
It should be noted that in this study we neither consider glacier evolution,
i.e. changes in ice thickness due to accumulation and melt, nor glacier ice
flow. Neglecting glacier retreat raises the possibility that we
overestimate the ice-melt contribution over the study period. To quantify the
potential effect of glacier retreat, we compare our simulations with time
series produced from the Global Glacier Evolution Model (GloGEM), a model
accounting for both the mass balance and glacier evolution (Huss and Hock,
2015). For comparison, we use total monthly runoff generated from glacierized
surfaces of the upper Rhône Basin, simulated with GloGEM for the period
1980–2010. GloGEM computes the mass balance for every 10 m elevation band
of each glacier, by estimating snow accumulation, snow and ice melt, and
refreezing of rain and melt water. The response of glaciers to changes in
mass balance is modelled on the basis of an empirical equation between ice
thickness changes and normalized elevation range parameterized as proposed by
Huss et al. (2010). Normalized surface elevation changes
We use the non-parametric Pettitt test (Pettitt, 1979) for the detection of
the time of change (year-of-change) in the air temperature data. We then test
the other variables (SSC,
In our catchment, SSC is sampled intermittently (twice per week). This might
have an effect on the change detection analysis of the hydroclimatic
variables. We estimate this potential effect by considering the hydroclimatic
variables SM, IM, ER, and
For precipitation and air temperature we use spatially distributed datasets
provided by the Swiss Federal Office of Meteorology and Climatology
(MeteoSwiss). Total daily precipitation and mean, minimum, and maximum daily
air temperature are available on a
We use daily discharge data measured by the Swiss Federal Office for the Environment (FOEN) at three gauging stations: Porte du Scex (available since 1905), Blatten Bei Naters (available since 1931) and Blatten (available since 1956) (Fig. 1). For suspended sediment concentration, two in-stream samples per week collected by FOEN at Porte du Scex are available since October 1964 (Grasso et al., 2012).
In this work, we focus on sediment transported in suspension. Previous
analysis on the grain size distribution of suspended sediment at the outlet
of the upper Rhône River reports a bimodal distribution, with mode
diameters equal to 13.7
We use snow-cover maps derived from satellite imagery for the upper Rhône
Basin over the period 2000–2008 processed in previous research (Fatichi et
al., 2015). We use the 8-day snow-cover product MOD10A2 retrieved from the
Moderate Resolution Imaging Spectroradiometer (MODIS) (Dedieu et al., 2010)
for the calibration and validation of the snowmelt model. MOD10A2 is provided
at a 500 m
The surface covered by glaciers is assigned based on the GLIMS (Global Land
Ice Measurements from Space) Glacier Database (Fig. 1). Ice-covered cells
identified based on the GLIMS data of 1991 show that more than 10 % of the
upper Rhône Basin as covered by ice with a total glacier surface of
almost 620 km
We use a digital elevation model (DEM) with 250 m
The snowmelt factor, calibrated following the procedure described in
Sect. 3.3, is
Comparison between observed (red circles) and simulated (light blue
lines) snow-cover fraction (SCF) of the upper Rhône Basin for five
different elevation bands. Simulations are computed with calibrated snowmelt
factor
The ice-melt factor, calibrated following the procedure described in
Sect. 3.3, is
Although in our hydrological model we do not include glacier evolution, the
annual runoff volumes (SM
Map of average snow permanence during the period 2000–2008,
expressed as the fraction of time in which pixels are snow-covered
(snow cover duration fraction, SCDF[0–1]):
Comparison of mean monthly observed (dark blue) and simulated
(light blue) discharge for the period 1975–2015:
Runoff (snowmelt
Observations for the period 1965–2015 of
Monthly differences between the period after and before the
year-of-change (1987–2015 and 1965–1986) of:
Simulations for the period 1975–2015 of mean annual variables:
Mean monthly differences (between 1987–2015 and 1975–1986) in variables
Relative contribution of snowmelt (SM), rainfall (
Empirical cumulative distribution functions of total daily
basin-averaged SM
Mean annual air temperature shows a clear and statistically significant
increase in 1987 (
The change in air temperature around 1987 coincides with statistically
significant changes in mean annual suspended sediment concentration
(Fig. 6c). After the abrupt warming, mean annual suspended sediment
concentrations are roughly 40 % larger than before: average values have
risen from
While the upper Rhône Basin underwent an abrupt warming around 1987, mean annual precipitation (Fig. 6b) and mean monthly precipitation (Fig. 7b) did not change significantly in time. Likewise, mean annual discharge does not show any statistically significant change in 1987 (Fig. 6d). Mean monthly discharge (Fig. 7d) is characterized by a small statistically significant increase in winter (November–February) runoff, most likely due to increased snowmelt and possibly changes in hydropower generation.
Mean annual simulated snowmelt (SM) shows a decreasing tendency during the
last 30 years (Fig. 8a). The reduction in snowmelt after 1987 occurs
mostly in summer and early autumn (Fig. 9a) mainly due to poor snow cover
(Fig. 9b). However, except July and September, the changes in all months are
within the 95 % confidence interval. The increase in snowmelt in March
and April is due to warmer temperatures in spring. Results are coherent with
the temporal evolution of simulated snow-cover fraction, which is also
gradually decreasing (Fig. 8b), especially in spring and summer (Fig. 9b).
Statistical analysis reveals a step-like reduction of more than 10 % for
mean annual values of snow-cover fraction in 1987 (
Although mean annual and monthly precipitation were shown not to change
significantly in the mid-1980s, effective rainfall (ER) on snow-free areas
has increased, especially in early summer (Figs. 8d, 9d). Effective rainfall
increases in conjunction with decreases in snow-cover fraction, and a
statistically significant jump is identified in 1987 (
Our results show that the temporal evolution of ice melt is consistent with suspended sediment concentration rise. Although the change is rather gradual on the annual scale (Fig. 8c), the step-like increase in ice melt is evident in the ice-melting season (May–September) and reaches highest magnitudes in July and August (Fig. 9c) in conjunction with rises in suspended sediment concentration in those months (Fig. 7c).
The simultaneous increase in ice melt and decrease in snowmelt suggests that
the abrupt warming has led to important alterations of the hydrological
regime. To quantify this alteration, we compute the relative contribution of
rainfall, snow, and ice melt on the sum of these three components in July and
August. The average relative contribution of ice melt has almost doubled after
1987 (from
The empirical cumulative distribution functions of total daily basin-averaged
SM, IM, ER, and
The value of the snowmelt factor
Despite the large regional and temporal variability that characterizes
ice-melt factors, comparison with previous studies confirms that the
calibrated value (7.1 mm day
Considering that the aim of this study is to evaluate long-term changes in hydro-climatology and sediment dynamics of the upper Rhône Basin and not the short-term variability in ice melt on the daily scale, we consider the snowmelt and ice-melt model performances as satisfactory. In addition, we show that although our model does not account for glacier retreat, it does not overestimate the ice-melt contribution during the period 1975–2015. However, considering climate projections further into the future, and glaciers that continue to retreat, the issue of future ice-melt contribution will need to be revised. Under climate change, even the largest glacier in the basin, the Aletsch Glacier, is expected to shrink at a rate where its ice-melt contribution would start decreasing before 2050 (Farinotti et al., 2012; FOEN, 2012; Brönnimann et al., 2014).
Abrupt temperature jumps, such as the one we observed in the upper Rhône basin, rather than gradual changes in air temperature have been observed globally (e.g. Jones and Moberg, 2003; Rebetez and Reinhard, 2008). Observations indicate that Switzerland has experienced two main rapid warming periods in the past, with the 1940s and 1980s being the warmest decades of the last century (Beniston et al., 1994; Beniston and Rebetez, 1996). The simultaneous increase in temperature and suspended sediment concentration indicates that changes in climatic conditions may effectively impact sediment dynamics, especially in Alpine environments where temperature-driven processes, like snow and ice-melt, have a strong influence on the basin hydrology. The statistically significant change in the SSC variance supports the finding that processes related to fine sediment regime of the upper Rhône Basin have been altered by changing climatic conditions, resulting in greater concentrations and higher variability of suspended sediment reaching the outlet of the basin.
Conversely, differences in precipitation before and after 1987 are within the 95 % confidence interval and are not statistically significant. Differences in discharge are also not statistically significant except in winter, when the suspended sediment concentration does not show changes. Therefore, it is very unlikely that the abrupt increase in suspended sediment concentration around mid-1980s in July and August is caused by changes in mean precipitation and/or discharge.
Our simulations of snow cover and melt are in agreement with snow observations across Switzerland. The decreasing tendency in snow cover after the mid- or late 1980s has been demonstrated for the Swiss Alps (Beniston, 1997; Laternser and Schneebeli, 2003; Scherrer et al., 2004; Scherrer and Appenzeller, 2006; Marty, 2008). Snow depth, number of snowfall days, and snow cover show similar patterns during the last century: a gradual increase until the early 1980s, interrupted in late 1950s and early 1970s, and a statistically significant decrease afterwards (Beniston, 1997; Laternser and Schneebeli, 2003). Previous analyses also state that the reduction in snow cover after mid-1980s is characterized more by an abrupt shift than by a gradual decrease (Marty, 2008), in agreement with our simulations. The reduction in snow-cover duration, which is observed to be stronger at lower and mid-altitudes than at higher elevations, is mainly the result of earlier snow melting in spring due to warmer temperatures (Beniston, 1997; Laternser and Schneebeli, 2003; Marty, 2008). Moreover, by analysing 76 meteorological stations in Switzerland, Serquet et al. (2011) demonstrated a sharp decline in snowfall days relative to precipitation days, both for winter and early spring, showing the impact of higher temperature on reduced snowfall, independently of variability in precipitation frequency and intensity. Therefore, despite the high complexity that characterizes snow dynamics in the Alps (Scherrer et al., 2006, 2013), the dominant effect of temperature rise on snow-cover decline after the late 1980s has been clearly shown (Beniston, 1997; Marty, 2008; Serquet et al., 2011; Scherrer et al., 2004; Scherrer and Appenzeller, 2006).
The increase in potentially erosive rainfall is partially confirmed by
recent observations. Rainfall erosivity, expressed by the
Enhanced ice melt is coherent with the observed acceleration of Alpine glacier
retreat after the mid-1980s. Ground-based and satellite observations, combined
with mass balance analysis, reveal that current rates of glacier retreat are
consistently greater than long-term averages (Paul et al., 2004, 2007;
Haeberli et al., 2007). Estimations of glacier area reduction rates indicate
a loss rate for the period 1985–1999, which is 7 times greater than the
decadal loss rate for the period 1850–1973 (Paul et al., 2004).
Investigations with satellite data and in situ observations suggest that the
volume loss of Alpine glaciers during the last 30 years is more
attributable to a remarkable down-wasting rather than to a dynamic response
to changed climatic conditions (Paul et al., 2004, 2007). Haeberli et
al. (2007) estimated that glaciers in the European Alps lost about half of
their total volume (roughly 0.5 % year
Most importantly, runoff coming from glaciers is notoriously rich in sediments. Very fine silt-sized sediment resulting from glacier erosion is transported in suspension most often as wash load (Aas and Bogen, 1988). Proglacial areas generally represent rich sources of sediment due to active glacier erosive processes of abrasion, bed-rock fracturing, and plucking (Boulton, 1974; Hallet et al., 1996). Glacier retreat discloses a large amount of sediments available to be transported by proglacial streams. Moreover, change in climatic conditions and specifically temperature-driven glacier recession and permafrost degradation may initiate specific erosional processes that consequently enhance sediment supply in proglacial environments (Micheletti et al., 2015; Micheletti and Lane, 2016; Lane et al., 2016).
As shown in Sect. 5.3, the ice-melt increase is highest in July and August (Fig. 9c) in agreement with the jump in suspended sediment concentration (Fig. 7c), while ER rise occurred mainly in June and July (Fig. 9d). We then conclude that the significant increase in ice melt detected in the mid-1980s (Figs. 8c, 9c, 10) is likely to be the main cause of the sharp rise in suspended sediment concentration entering Lake Geneva; this occurs through a combination of (1) increased discharge originated in proglacial environments, which implies higher suspended sediment concentration; (2) a larger relative contribution of sediment-rich ice melt compared to snowmelt and precipitation fluxes; and (3) intensified sediment production and augmented sediment supply in proglacial areas due to rapid ice recession.
The interpretation of increases in suspended sediment concentration may be
complicated by anthropogenic drivers and changes in the mid-1980s. Three main
anthropogenic activities may have potentially influenced the suspended
sediment regime of the upper Rhône Basin: river channelization,
construction of reservoirs and hydropower operations, and gravel extraction
along the main stream and tributaries. However, the second and last large
channelization project was completed in 1960 (Oliver et al., 2009), much
earlier than the observed increase in SSC. Likewise, the largest reservoirs
in the catchment have been in operation since 1975 (Loizeau and Dominik,
2000). Therefore, it is unlikely that these two anthropogenic factors have
contributed to the SSC rise detected in the mid-1980s. The same holds for gravel
mining activities. Annual volumes of gravel extracted from the Rhône,
provided by the Valais Cantonal Authorities as differences from the average over the
period 1989–2014, do not show any significant correlation with mean annual
suspended sediment concentration (
Our results show that even in highly human-impacted and regulated catchments such as the Rhône Basin, a strong climatic signal in hydrological and sediment dynamics can persist. This also suggests that the decrease in fine-sediment load at the outlet of the upper Rhône Basin observed in the 1960s on the basis of sediment cores recovered in the Rhône delta region and reported by Loizeau et al. (1997), could be the result of a combined effect of hydropower system development, as it has been hypothesized (Loizeau et al., 1997; Loizeau and Dominik, 2000), but also reduced ice-melt loads due to colder temperatures at the time. The cooling period, which occurred between the 1950s and late the 1970s (e.g. Beniston et al., 1994) was characterized by colder and snowy winters (e.g. Laternser and Schneebeli, 2003) and has been accompanied by reduced ice-melt rates, glacier advance, and positive glacier mass balances (Zemp et al., 2008; VAW-ETH, 2015).
The climate signal in sediment dynamics takes on particular importance in the context of climate change projections into the future. Despite the large uncertainty, future projections under different climate change scenarios show a common tendency for Switzerland, characterized by a shift from snow-dominated to rain-dominated hydrological regime, reduced summer discharge, increased winter discharge, reduced snow cover, and enhanced glacier retreat (Bavay et al., 2009; Jouvet et al., 2011; Brönnimann et al., 2014; Fatichi et al., 2015; Huss and Fischer, 2016). In contrast to these hydrological predictions, changes in sediment fluxes are highly uncertain due to the complexity and feedbacks of the processes involved, inherent stochasticity in sediment mobilization and transport, and large regional variability in sediment connectivity across the Alpine landscape (Cavalli et al., 2013; Heckmann and Schwanghart, 2013; Bracken et al., 2015; Lane et al., 2017).
The aim of this research was to analyse changes in the hydroclimatic and suspended sediment regimes of the upper Rhône Basin during the period 1975–2015. We show an abrupt increase in basin-wide mean air temperature in the mid-1980s. The simultaneous step-like increase in suspended sediment concentration at the outlet of the catchment, detected in July and August, suggests a causal link between fine sediment dynamics and climatic conditions. Two main factors link warmer climate and enhanced SSC: increased transport capacity and increased sediment supply resulting from spatial and/or temporal activation–deactivation of sediment sources. Our results show that transport capacity, through discharge, is not likely to explain the increases in SSC because no statistically significant changes in the mid-1980s are present in Rhône Basin discharge, neither on the annual nor monthly timescales. The suggestion is that the impact of warmer climatic conditions acts on fine sediment dynamics through the activation and deactivation of different sediment sources and different sediment production and transport processes.
To understand sediment supply conditions, we analyse the temporal evolution of three main sediment fluxes: (1) sediments sourced and transported by snowmelt along hillslopes and channels; (2) sediments entrained and transported by erosive rainfall events over snow-free surfaces, including hillslope, channel bank erosion, and mass wasting events; and (3) fine sediment fluxes generated by glacier ice-melt. The fluxes of snow and ice melt together with snow-cover fraction and rainfall are analysed to detect changes in time and their coherence with changes in SSC.
Our results show that while mean annual precipitation does not show any evident change between the periods before and after the SSC jump in the mid-1980s, potentially erosive rainfall clearly increases over time especially in June and July, but not in August. On the other hand, ice melt has significantly increased due to temperature-driven enhanced ablation. Statistically significant shifts in ice melt were identified for summer, with highest increases in July and August, in accordance with the rise in SSC. Concurrently to the temperature and SSC rise, the relative contribution of ice melt to total annual runoff (sum of rainfall, snow, and ice-melt) presents a significant increase in the mid-1980s, substantially altering the hydrological regime of the Rhône Basin. Based on these results, we propose that climate has an effect on fine sediment dynamics by altering the three main fluxes of suspended sediment in the Rhône Basin, and that ice melt plays a dominant role in the suspended sediment concentration rise in the mid-1980s through (1) increased flow derived from sediment-rich subglacial and proglacial areas, (2) a larger relative contribution of sediment-rich ice melt compared to snowmelt and precipitation, and (3) increased sediment supply in hydrologically connected proglacial areas due to glacier recession. While snowmelt has decreased, the reduced extent and duration of snow cover may also have contributed to the suspended sediment concentration rise through enhanced erosion by heavy rainfall events over snow-free surfaces.
Because changes in SSC are not consistent with changes in discharge and transport capacity, our work emphasizes how the inclusion of sediment sources and their activation through different processes of production and transport is necessary for attributing change. This analysis also demonstrates that climate-driven changes in suspended sediment dynamics may be significantly strong even in highly regulated and human-impacted catchments such as the upper Rhône Basin, where sediment fluxes are affected by flow regulation due to hydropower production and by grain-size dependent trapping in reservoirs. This has consequences for climate change impact assessments and projections for Alpine catchments with hydropower systems, where climate change signals are sometimes thought to be secondary to human regulation. Although at this stage we cannot reliably conclude in which direction sediment fluxes will change in the future, our paper clearly shows that a more process-based understanding of the connections between hydrological change and the activation of sediment sources will provide us with a better framework for analysing and attributing changes in sediment yields in Alpine catchments in the future.
Data analysed in the current study are available from the Swiss Federal Office for the Environment (FOEN, streamflow and suspended sediment concentration), the Swiss Federal Office for Topography (Swisstopo, DEM), and the Swiss Federal Office of Meteorology and Climatology (MeteoSwiss, gridded datasets of temperature and precipitation). Daily catchment-averaged data (observed and simulated) used in the analysis are available from the author.
The supplement related to this article is available online at:
AC and PM designed the methodology. AC developed the code and carried out simulations and computations. AC prepared the manuscript with contributions from all co-authors.
The authors declare that they have no conflict of interest.
We thank Christoph Frei (Federal Office of Meteorology and Climatology MeteoSwiss) for providing us with experimental temperature and precipitation datasets and for suggestions on the right use of MeteoSwiss gridded data and the application of statistical tests. We also thank Daniel Farinotti (Swiss Federal Institute for Forest, Snow and Landscape Research WSL, Department of Civil, Environmental and Geomatic Engineering ETH Zurich) for providing us with GloGEM simulations, for the fruitful discussion on glacier retreat and glacier dynamics and for kindly revising the manuscript. The Federal Office of the Environment (FOEN) provided discharge and suspended sediment concentration data. We thank Alessandro Grasso (FOEN) for the explanation on the SSC data collection procedures. Finally, we would like to thank the Valais Cantonal Authorities for supplying information on gravel mining extraction. This research was supported by the Swiss National Science Foundation Sinergia grant 147689 (SEDFATE). Edited by: Laurent Pfister Reviewed by: two anonymous referees