Ongoing climate change and associated glacier retreat is
causing rapid environmental change, including shifts in high-alpine
landscapes. Glacier lakes, which can form in topographical depressions left
behind by glacier retreat, are prominent features within such landscapes.
Whilst model-based estimates for the number and area of future glacier lakes
exist for various mountain regions across the world, the exact morphology
and temporal evolution remain largely unassessed. Here, we leverage a
recently released, measurement-based estimate for the subglacial topography
of all glaciers in the Swiss Alps to provide an estimate about the number,
size, time of emergence, and sediment infill of future glacier lakes.
The topographical information is based on 2450 km of measured ice thickness
profiles, whilst the temporal evolution of glaciers is obtained from a
glacier evolution model forced with an ensemble of climate projections. We
estimate that up to 683 potential lakes with an area
As global temperatures continue to rise, worldwide mountain glaciers are rapidly shrinking (Hock et al., 2019). Glaciers in the Alps are also affected, with further loss of glacier volumes being inevitable (Marzeion et al., 2018; Zekollari et al., 2020). Depending on climate scenario, the Alps may lose up to 94 % of their 2020 ice volume by the end of the 21st century (Zekollari et al., 2019), with important consequences for the appearance of new landscapes (Orlove et al., 2008). Retreating glaciers produce environments dominated by erosion and deposition, consisting of hills, sinks, and overdeepenings amongst other features (e.g. Cook and Swift, 2012). Overdeepenings, i.e. confined topographical depressions caused by erosion, may entirely fill with sediment transported by glaciers and their proglacial streams or with water giving rise to new glacier lakes (Frey et al., 2010; Mölg et al., 2021). Predicting which of the two scenarios will materialize is difficult, since the spatio-temporal dynamics of glacial erosion are governed by a complex interplay of processes that are difficult to quantify and thus to anticipate (e.g. Lane et al., 2017). In their recent review, Carrivick and Tweed (2021) highlighted that, at the worldwide scale, specific sediment yields from glacierized catchments can span as many as 5 orders of magnitude, with sediment yields in the European Alps ranging from dozens to thousands of tonnes per year and square kilometre. Moreover, the sediment yields are strongly controlled by glacierization itself (Hinderer et al., 2013), adding a significant temporal dependence on the corresponding estimates.
Despite the difficulty in predicting their emergence, the interest in potential new glacier lakes has recently been on the rise. Indeed, such lakes have been identified to pose both risks and opportunities (Haeberli et al., 2016; Anacona et al., 2018). On the one hand, glacier lakes represent a potential hazard for downslope populations and infrastructure if they burst (Frey et al., 2010; Emmer et al., 2014; Anacona et al., 2018), with a number of studies aimed at identifying already-existing hazardous lakes (e.g. Bolch et al., 2012; Huggel et al. 2002; Veh et al., 2018; Zhang et al., 2022) and at clarifying whether any change in the frequency of their outbursts can be detected at large scales (e.g. Veh et al., 2019). On the other hand, new glacier lakes are of relevance for high-mountain biodiversity (Čiamporová-Zaťovičová and Čiampor, 2017; Tiberti et al., 2019), have been shown to hold significant hydropower potential (Ehrbar et al., 2018; Farinotti, et al., 2019), might be attractive to tourists (Purdie, 2013; Welling et al., 2015), or might serve as reservoirs for artificial snow production in ski areas (Haeberli et al., 2016) and other water management purposes (Farinotti et al., 2016; Brunner et al., 2019).
In light of the above relevance, accurate knowledge about the size, the
distribution, and the time of formation of glacier lakes is key. At the
worldwide scale, there are multiple examples of inventories of
already-existing glacier lakes (e.g. Komori, 2008; Gardelle et al., 2011;
Zhang et al., 2015; Emmer et al., 2016; Petrov et al., 2017; Drenkhan et
al., 2018; Shugar et al., 2020; Wang et al., 2020), with most studies
concurring that the formation of glacier lakes has been accelerating during
the past decades (see also the review by Carrivick and Tweed, 2013). In the
European Alps, inventories of existing proglacial lakes are available for
Austria (Emmer et al., 2015; Buckel et al., 2018), various parts of Italy
(Galluccio, 1998; Salerno et al., 2014; Viani et al., 2016), and Switzerland
(Mölg et al., 2021). According to the latter, Switzerland hosted 987
proglacial lakes in 2016, covering an area of
Several studies that aim to predict future glacier lakes exist, notably for
(parts of) High Mountain Asia (Linsbauer et al., 2016; Kapitsa et al., 2017;
Zheng et al., 2021), the Andes (Colonia et al., 2017; Drenkhan et al.,
2018), or the European Alps (Linsbauer et al., 2012; Magnin et al., 2020;
Viani et al., 2020; Gharehchahi et al., 2020; Otto et al., 2022). For the
Swiss Alps, Linsbauer et al. (2012) estimated that between 400 and 600 new
glacier lakes could form if glaciers were to vanish entirely, with a total
area in the order of 50 to 60 km
In this study, we rely on such extensive ice thickness surveys to present a new estimate for the potential formation of future glacier lakes in the Swiss Alps. More specifically, we rely on the Swiss-wide subglacial topography recently released by Grab et al. (2021) on the basis of almost 2500 km of ground-penetrating radar (GPR) surveys and use it to detect the location and size of subglacial overdeepenings that could give rise to glacier lakes after their retreat. In contrast to previous studies, we also quantify the timing of the potential lake formation. We do so by combining the ice-free subglacial topography with results from the Global Glacier Evolution Model (GloGEM; Huss and Hock, 2015) forced by state-of-the-art climate model projections. This provides insights into the water volumes that could be retained in future glacier lakes under different climate scenarios. For the first time, we also aim at roughly quantifying future sedimentation rates which affect the overdeepenings after glacier retreat, thus providing indications for the long-term persistence of the emerging lakes. The resulting estimates for the temporal evolution of future glacier lakes provide a first glimpse into how alpine landscapes might change throughout the 21st century.
The geographical extent of this study is given by the Swiss Glacier
Inventory 2016 (SGI2016; Linsbauer et al., 2021), which can be divided into
the four large river catchments of Switzerland, i.e. Rhine, Rhône, Po, and
Inn (Fig. 1). The SGI2016 is an inventory of all Swiss glaciers that has
been produced on the basis of high-resolution aerial images acquired between
2013 and 2018 (centre year: 2016). The 1400 inventoried glaciers cover an
area of 961 km
Overview of the five largest potential lakes. The lake volume expected to be ice free until 2050 and 2100 (neglecting sedimentation) is given for SSP245 and is expressed in percentage of the potential total lake volume. Square brackets provide confidence intervals for the given quantities (see Sect. 3.3). Note that two of the five potential lakes are predicted for the area presently occupied by Grosser Aletschgletscher.
The subglacial topography of all considered glaciers is taken from Grab et al. (2021). This topography is based on both extensive helicopter-borne GPR measurements (Rutishauser et al., 2016; Langhammer et al., 2019a, b) and ground-based GPR data. In total, 2450 km of GPR profiles collected on 251 different glaciers and covering 81 % of the SGI2016 glacier area were available (cf. Table 1 in Grab et al., 2021). Since the density of the GPR measurements varies between glaciers and since direct ice thickness measurements are not available for all glaciers, Grab et al. (2021) used two different ice thickness models to obtain a spatially complete estimate: the Glacier Thickness Estimates model (GlaTE; Langhammer et al., 2019a) and the Ice Thickness and Volume Estimation based on Observations model (ITVEO; Huss and Farinotti, 2012). Both of these models are based on principles of ice flow dynamics and are designed to make optimal use of the information contained within sparse measurements of ice thickness. The performance of both models has been assessed within ITMIX2 (Farinotti et al., 2021), and since the average of several ice thickness models has been shown to yield the most robust results (Farinotti et al., 2017), the final dataset was obtained by averaging the ice thicknesses estimated by GlaTE and ITVEO. Henceforth, we will refer to the subglacial topography obtained by subtracting this average ice thickness from the SwissALTI3D surface topography as the “mean bedrock topography”. The spatial resolution of this topography is 10 m, whilst an estimate for the local vertical accuracy (based on the distance to the next GPR measurement and the difference between the estimates of GlaTE and ITVEO) is provided as separate information to the dataset. For further details on the methodology, refer to Grab et al. (2021).
Our detection of potential lakes within the glacier extent of the SGI2016 is
based on the above-mentioned mean bedrock topography. Following previous
studies, we define potential lakes through the detection of overdeepenings
in the subglacial topography, i.e. we only consider bedrock-dammed lakes and
neglect potential ice-dammed lakes or lakes dammed by moraines not resolved
by the GPR data. The overdeepenings are detected by applying the tool
“Fill” from the ArcGIS “Hydrology Toolset” (Esri, 2022) to the mean
bedrock topography. This operation yields a spatially distributed dataset in
which every overdeepening within the currently glacierized area is filled.
We assume the difference between this dataset and the original mean bedrock
topography to represent the depth of potential future glacier lakes and
detect their extents by drawing polygons around connected areas with depth
Being constrained to the area within the SGI2016 glacier outlines, the above procedure fails to identify already-existing glacier lakes that are in contact with glacier ice and that might further expand in future (one of the most prominent examples is the lake presently in front of Rhonegletscher, at the source of the Rhône river). To include such cases in our analyses, we visually inspected aerial images from Swisstopo (Swisstopo, 2022). We detected 15 lakes that are presently in contact with ice and are dammed by rock or sediment. To include them, we manually extended the lake polygons obtained with the procedure described above. Since the bathymetry of the added lake portions is generally unknown, we pragmatically assumed their depth to be equal to the mean depth of the remaining lake portion (i.e. the lake portion that is presently glacierized and that has a depth estimate based on the available ice thickness information).
To distinguish between individual lakes, a lake identifier (lake ID) is defined. The lake ID is composed of the SGI-ID (i.e. the glacier identifier of the SGI2016), followed by the rank of the lake volume within the particular glacier. For example, the largest lake of the glacier with SGI-ID “B40-07” (that is Fieschergletscher), is named “B40-07-01”, the second largest is named “B40-07-02”, and so on. This nomenclature is used throughout the article and in the data that we provide as digital supplement (see the data availability section at the end of the paper).
For estimating the timing of future lake formation, we rely on glacier
retreat projections scenarios based on GloGEM (Huss and Hock, 2015). The
model describes the main processes determining glacier surface mass balance
(snow accumulation, ice melt, refreezing) and computes annual surface
elevation change – and thus glacier retreat or advance – based on a
mass-conserving parameterization (Huss et al., 2010). The current ice
thickness distribution is taken from Grab et al. (2021), and GloGEM is
applied to all glaciers of the SGI2016 with a glacier-specific calibration
based on observed ice volume changes between 2000 and 2019 (Hugonnet et al.,
2021). For computational reasons, the model is discretized into 10 m elevation bands but results on area and thickness changes in individual
bands are extrapolated to the same
For each of the 56 climate models considered and for each potential glacier
lake, we compute the year when
In order to obtain a first-order estimate for the time required to fill overdeepenings becoming ice free with sediments, we propose a simple approach that is tightly connected to the GloGEM results. Accounting for the spatio-temporal dynamics of sedimentation of new glacier lakes is crucial as (1) many overdeepenings will be sedimented a few years after their formation (e.g. Mölg et al., 2021) – especially in smaller examples – and (2) erosion and sediment transport rates in glacial environments are known to be extreme (for compilations of such rates, see Hinderer et al., 2013, or Carrivick and Tweed, 2021). In this respect, glacial overdeepenings represent important sediment traps (e.g. Geilhausen et al., 2013; Bogen et al., 2015), and the connectivity of the fluvial system emerging after glacier retreat must be considered since lower-lying areas might be deprived of sediment input once new traps come into existence above them (Micheletti and Lane, 2016; Lane et al., 2017). A detailed description of the sediment processes that affect deglacierizing areas is out of reach for regional- to global-scale glacier models, and considerable simplifications are necessary. The approach described below attempts to quantify the most relevant drivers of sediment yields into proglacial basins and to capture the corresponding spatio-temporal dynamics.
We parameterize the sediment volume transported into a glacial lake via (i)
the sediment load per unit volume of water (
(1) Subglacial abrasion scales with glacier flow speed (e.g. Herman et al.,
2015) and is parameterized as a non-dimensional index
(2) The deglacierized area is known to be a major erosion source due to
exposed unconsolidated sediment (e.g. Delaney et al., 2018b). Analogously to
(1), we parameterize this effect with an index
(3) Glacial and periglacial erosion is most powerful in headwalls (MacGregor
et al., 2009; Sanders et al., 2012; Hartmeyer et al., 2020). For each
glacier, we compute the area and the average slope of the headwall, here
defined as the non-glacierized area above the glacier's median elevation
lying within the glacier's hydrological catchment (the latter is defined
based on the DEM extending beyond the presently glacierized surfaces). The
index
Finally, the indices for the three considered factors are averaged to yield
In the case that multiple potential glacier lakes are exposed at the same time, we assume all of them to be directly linked with each other and to transfer water and sediments from the higher lakes to lower ones. Two-dimensional aspects of connectivity are neglected for simplicity. The topmost exposed proglacial overdeepening receives a sediment yield corresponding to Eq. (1), computed using all runoff originating from above the respective elevation. Lower-lying potential lakes, however, are only fed by sediment yields computed based on runoff from the elevation interval up to the next higher potential lake, which typically results in much smaller sediment input. For each potential lake, the volume that is still free from sediments is annually updated, and as soon as it is completely sedimented, it is no longer considered as a sediment sink. This can then cause the sediment yields to rise again in lower-lying lakes.
Our results are affected by various uncertainties, including uncertainties in the lakes' morphology (i.e. location, total area, and total volume) and temporal evolution (i.e. year of formation and rate of sedimentation). The following paragraphs describe how each of these uncertainties is estimated.
The location and areal extent of the individual potential lakes are
determined by the subglacial topography. This means that, in general, the
number of lakes, the lake extents, and the lake locations computed on
the basis of the subglacial topography generated with GlaTE (cf. Sect. 2)
might be different from the ones computed by using the subglacial topography
generated with ITVEO. The same is true when comparing the lakes computed on
the basis of the mean bedrock topography with either of the results from
GlaTE or ITVEO. Because of these differences, it is not possible to
establish a one-to-one relation between lakes generated on the basis of the
individual bedrock topographies, and we thus estimate the uncertainty in lake
extents by aggregating the results at the level of individual glaciers. More
specifically, we compute (1) the total area of lakes obtained for a given
glacier by using the GlaTE bedrock topography and (2) the total area of
lakes obtained for the same glacier by using the ITVEO topography, and we use
the relative difference of these totals (
The uncertainty in lake volume is also controlled by the uncertainty in the
subglacial topography. The dataset by Grab et al. (2021) provides
information about the local uncertainty of the bedrock elevations (
Since the total volume of each lake (
Since we do not expect the results by Grab et al. (2021) to be systematically
biased towards either
We estimate the uncertainty in the year of lake formation as the standard deviation of the 13 different years of lake formation resulting from the 13 GCMs considered for each SSP. The year of lake formation will in fact also be affected by other uncertainties such as the subglacial topography (e.g. Farinotti et al., 2017), as well as structural uncertainties in the glacier model itself (e.g. Huss and Hock, 2015). However, several studies have shown that climate model uncertainty largely dominates over glacier model uncertainty (e.g. Marzeion et al., 2018, 2020). We therefore refrain from propagating uncertainties from subglacial topography and the glacier model through all climate scenarios.
Estimating the uncertainty in sedimentation rates is highly challenging as only few data are available for direct benchmarking. We account for uncertainties in our first-order estimates in the same way as for the year of lake formation, i.e. by determining the standard deviation of the results provided by the 13 GCMs considered for each SSP. We deem it likely that actual uncertainties are larger, especially when considering that (i) the underlying processes are only described by highly simplified parameterizations and that (ii) systematic uncertainties due to the chosen model parameters come into play as well. We argue that our simplified approach is defensible as our present understanding does not allow for providing a more complete uncertainty assessment.
In total, we detected 683 potential lakes with an area larger than 5000 m
Cumulative area
To obtain a regional picture, we aggregated the potential lakes among the four main river catchments of Switzerland (Fig. 1). The region with the highest number of potential glacier lakes (447) is the Rhône basin, where their volume corresponds to 2.2 % of the total glacier volume at present. In the Rhine and Inn catchments, 185 and 26 potential lakes are found, respectively. The total lake volume in these catchments corresponds to 1.5 % (Rhine) and 1.4 % (Inn) of the present-day glacier volume. The region with the lowest number of potential glacier lakes (24) is the Po catchment, where the total lake volume is equivalent to 0.6 % of today's glacier volume.
In general, the size of the potential lakes is related to the size of the glacier, meaning that the largest potential lakes are also found beneath the largest glaciers. Our results indicate that the four glaciers with the highest potential lake volume might give rise to up to 117 new lakes, accounting for 53 % of the anticipated total volume in Switzerland. The five largest potential lakes of the Swiss Alps are characterized in Table 1. In the first instance, the correlation between lake size and glacier extent can be attributed to the fact that larger glaciers have more erosive power, and thus they can produce larger overdeepenings (Iken and Bindschadler, 1986).
Overview of the total potential lake volume anticipated to emerge by the years 2050 and 2100 when accounting for sedimentation (first two columns), the percentage of the total lake volume that might be lost because of sedimentation until 2100 (as % of the potential volume), as well as numbers of potential lakes appearing (because of glacier retreat) and disappearing (because of sedimentation) until the year 2100 (last two columns).
Depending on the climate scenario, the total volume of potential glacier
lakes that will be exposed until the end of this century can strongly vary
(Table 2). Within the first half of this century, the projected glacier
retreat and thus the formation of new glacier lakes is not particularly
sensitive to the chosen climate scenarios (Fig. 3a), in contrast to the
situation after 2050. SSP585, for example, shows a rapid formation of new
lakes after 2050. This formation only decelerates towards the end of this
century, when most glaciers will have vanished entirely. SSP245 instead
leads to an almost linear increase in the volume of potential new lakes,
with the formation of additional
The rate of formation of new glacier lakes differs within the four main
river catchments used to aggregate the results (Figs. 1 and 3b). After 2040,
the rate is approximately constant in all catchments, the highest rates
being predicted for the Rhône and Rhine catchments (annual increase of
potential lake volume of
According to our results, sedimentation of new glacial lakes is an important
process that has the potential to fill smaller bedrock overdeepenings.
Depending on the climate scenario, we find that between
Our simple model for catchment erosion and the provision of sediment to new
glacier lakes (see Sect. 3.2) shows complex spatio-temporal dynamics. At
the scale of the entire Swiss Alps, we expect sedimentation input into newly
formed lakes to increase until about 2050 (with a range of
Temporal dynamics of lake volume growth and modelled
sedimentation for the example of Gornergletscher (SGI-ID B56-07; see Fig. 1 for location), harbouring the third-largest new glacier lake in
Switzerland (see Table 1). Results refer to a selected GCM amid the climate
scenario SSP585.
The complexity of the temporal sedimentation dynamics is best illustrated at
the scale of an individual new glacier lake. Figure 5 shows the volume
evolution of the large lake at the snout of Gornergletscher. In a climate
scenario implying high emissions (SSP585), the lake is expected to first
appear around the year 2025, to experience fast deglaciation especially
after 2040, and to be completely ice-free by about 2060 (Fig. 5A). Whereas
modelled overall basin erosion rates show a increase to about 0.9 mm yr
With close to 700 detected overdeepenings, the number of potential new
glacier lakes in the Swiss Alps is significant. Although in terms of area
and volume a relatively small proportion of lakes determines the totals (the
largest 60 lakes account for 50 % of the total area and 80 % of the
total volume, respectively; Fig. 1), this high number of emerging new lakes
is set to change the appearance of the alpine landscape – together with
glacier retreat, which causes the emergence of these lakes in the first
place. A number of impacts can be anticipated from that, including the
dynamics of the affected periglacial ecosystems, the potential anthropogenic
use of the newly emerging lakes (e.g. for recreational use, water management
purposes, or hydropower production), or the change in natural hazard
potential as significant water masses come to lie in areas surrounded by
steep topography and are thus prone to gravitational mass movements (such as
rockfalls, avalanches, or even landslides). The latter impacts can be further
exacerbated by the fact that high-elevation sites are undergoing rapid
changes, notably implying the degradation of stabilizing agents such as
permafrost (e.g. Haeberli et al., 2017). An in-depth analysis of the
possible implications of newly emerging glacier lakes has been conducted in
the frame of the Swiss National Research Programme 61 (NELAK, 2013), and it
is beyond the scope of our analysis to summarize the findings here (for such
a summary, refer to Haeberli et al., 2016). What our results show, however,
is that part of this evolution is set to take place in any case, i.e.
independently of the climate evolution of the next few decades. Indeed, as
illustrated by Fig. 3a for example, the temporal evolution of the emerging
new lakes is virtually independent of the chosen climate scenario until ca.
2050. This is congruent with the anticipated glacier evolution in the Alps
(e.g. Zekollari et al., 2019; Compagno et al., 2021) and is directly related
to the glacier's response time (Zekollari et al., 2020), i.e. the time
required by glaciers to adapt their geometry to given climate conditions. In
practice, this means that some
This large number of potential new lakes calls for a reflection about their future role and implications. Whilst it might be anticipated that the large majority of potential new lakes will not trigger major debates and will naturally become part of the newly emerging landscapes, the situation could be different for some of the larger lakes. Here, the potential for conflicting interests can be seen, spanning from a strict preservation of their state due to their importance in terms of emerging natural habitats to their exploitation for commercial use. Recent debates around this topic – mostly happening in the context of artificial reservoirs rather than natural lakes (e.g. Kellner, 2019, 2021; Kellner and Brunner, 2021) – have focused on the potentials for multipurpose usage, i.e. the potential of a given lake to satisfy different needs. Such needs range from the generation of hydroelectricity (e.g. Ehrbar et al., 2018), to the alleviation of water scarcity (e.g. Brunner et al., 2019), and to the management in terms of flood prevention (e.g. Volpi et al., 2018), but they are most often centred on anthropogenic interests. It seems desirable to extend these considerations to aspects that are more difficult to quantify in terms of economic value, notably including the ecosystem services provided by lakes (e.g. IPBES, 2019) and their role in the colonization of areas becoming ice free by plants and other living beings. Whether such considerations will gain momentum remains to be seen, but they seem important in light of the scale of the changes to come.
In the presentation of our results, we stressed the wording “potential glacier
lakes”. This is to indicate that even for overdeepenings that are detected
through GPR data with confidence, a considerable uncertainty remains about
whether a lake will actually form. The actual formation can indeed be impeded by the
local topography or geology. On the one hand, narrow outlet channels, which
can neither be resolved by the GPR measurements nor be anticipated with the
used ice thickness estimation approaches, can be sufficient to prevent a
predicted glacier lake forming at all. On the other hand, overdeepenings
eroded into porous or fissured bedrock might never fill with water because
of leakage through the underlying rock. In their recent study, Mölg et al. (2021) determined that only about 40 % of the area contained within
overdeepenings that have been exposed by glacier retreat since the Little
Ice Age have actually filled with water to form a glacier lake. If this is
assumed to be a characteristic value, it would mean that about
A further process that might hamper actual lake formation is refilling with sediment. Although the literature addressing the interplay between sediment dynamics and bedrock overdeepenings is vast (e.g. Hooke, 1991; Alley et al., 2003; Cook and Swift, 2012; Swift et al., 2021), our study is the first that attempts to capture this process in the context of the formation of potential new glacier lakes. It does so by using a simple approach that takes into account relevant processes and their spatio-temporal changes. Although a more process-based understanding is needed to increase the reliability of the results, and although our approach has not been evaluated in combination with the morphometry of the potential new lake basins, our assessment clearly suggests that sedimentation of new proglacial lakes in bedrock overdeepenings can be important – especially for small lakes in the first years after their formation. At the scale of the entire Swiss Alps, our model indicates that – depending on the climate scenario – between 11 % and 32 % of the potential new lake volume might be filled with sediments by 2100, with higher relative sedimentation for low-emission scenarios. Although other factors, such as the actual volume of overdeepenings and whether they will actually fill with water (see above), are likely more important, it is clear that sediment input into new proglacial lakes will systematically reduce their volumes. Neglecting this effect thus results in an overestimate for the volume of potential future glacier lakes. Our approach shows complex spatio-temporal dynamics of the sediment input (Fig. 5) and captures effects such as the erodibility of sediments in the upstream catchment, the transport of fine-grained materials by water runoff, and the dependence on upstream sediment sinks given by other glacier lakes. Our model has to be understood as a first-order approach to characterize these dynamics and recognize the weak validation of our approach. The problem here is that a validation at the scale of the Swiss Alps is almost impossible, given that only a small amount of data – mostly referring to individual and small basins – are available. A rigorous validation would not only require long-term time series of catchment-wide erosion rates for a representative set of glaciers but also a partitioning of the individual sediment sources within the individual basins. Such information is extremely rare as it is very difficult to devise measurements that would enable such a partitioning in first place – let alone maintaining them over a period of time that would allow for establishing a relationship between a change in partitioning and glacier retreat. Together with the realization that sediment fluxes from high-mountain areas have significantly increased over the past decades (Li et al., 2021), our results might be a motivation to enhance our understanding of the related processes, thus spurring further studies in this domain.
Cumulative
Despite the unusually comprehensive set of GPR data available to determine
the location and size of the potential future lakes, our results are
critically dependent on the methods used to infer the subglacial topography
in glacier areas that are not covered by the direct measurements. In this
respect, we assess the robustness of our estimates by comparing the results
obtained when using the two independent methods used by Grab et al. (2021),
i.e. GlaTE and ITVEO. When applying the same thresholds for the detection of
potential lakes as used for the mean bedrock topography (i.e. maximal lake
depth
In general, ITVEO tends to produce more pronounced topographical features,
whilst the subglacial topography generated with GlaTE is smoother (Fig. 7b).
This difference in smoothness results in a higher number of potential lakes
being detected with ITVEO than with GlaTE (Fig. 7a), but since the additional
lakes are generally small, they contribute little to the total volume of
potential lakes (Fig. 6a). With an average depth of 32.2 m, GlaTE also
produces slightly deeper lakes than ITVEO (average depth
The differences between GlaTE and ITVEO are dependent on the available GPR
coverage. In particular, the two approaches show larger differences with
increasing distance to GPR measurements (Fig. 7b). On glaciers with low GPR
coverage, the differences can be large, with some cases showing estimated
potential lake volumes differing by up to a factor of 2 (e.g. Glacier du
Trient, SGI-ID B90-02; Glacier de Corbassière, B83-03; Unterer
Theodulgletscher, B56-28; Bisgletscher, B58-08; Hüfifirn, A51d-10; or
Obers Ischmeer, A54l-31). In general, however, the total volume of potential
lakes detected on a glacier-by-glacier basis by GlaTE or ITVEO is very
similar, with the comparison visualized in Fig. 6c showing a correlation
coefficient as high as
All of the above considerations indicate that although uncertainties still exist, our approach is reliable and robust when identifying potential glacier lakes that might emerge in future.
Our study is not the first one trying to quantify the emergence of future
glacier lakes in the Swiss Alps. Linsbauer et al. (2012) applied the
“Glacier bed Topography (GlabTop)” model to estimate the subglacial
topography of the glaciers contained within the Swiss Glacier Inventory 2000
(Paul, 2007) and assumed future lakes to form in the detected bedrock
overdeepenings. By applying a threshold for the lake area of
More recently, Gharehchahi et al. (2020) used an approach named “volume and
topography automation” to estimate the size and distribution of future
glacier lakes in the Rhône catchment. They assessed the performance of their
approach to be more reliable on glaciers larger than 5 km
We quantified the number, area, and volume of the potential new glacier lakes that may form in Switzerland due to ongoing glacier retreat. Relying on the recently released subglacial topography by Grab et al. (2021), we systematically detected subglacial overdeepenings and characterized them in terms of both location and shape. In contrast to previous estimates (Linsbauer et al., 2012; Gharehchahi et al., 2020), our analysis is based on extensive GPR measurements, which increases the robustness of the results.
In total, we detected 683 potential glacier lakes with a size larger than
5000 m
The results reiterate the rapid changes that have to be expected in alpine regions and highlight that relatively little time is available to adapt to these changes. Whilst glacier lakes can be of high ecological relevance and might be attractive for a number of purposes ranging from water management to recreational use, they also bear some hazard potential – notably including potential outbursts upon failure of their embankments or flood waves triggered by mass movements stemming from the steep slopes that often surround alpine glacier lakes (Haeberli et al., 1989, 2017; Carrivick and Tweed, 2013; Haeberli and Drenkhan, 2022). A quantification of the changes in such processes remains beyond the scope and the capabilities of our study, but we suggest that the related consequences should be at the centre of attention in follow-up impact assessments.
In terms of methodology, our analysis is strongly based on the unique dataset that is available for the subglacial topography of the Swiss Alps (Grab et al., 2021). This notwithstanding, our approach for estimating the future evolution of potential glacier lakes is transferrable, at least in principle, to other glacierized regions on Earth. This is particularly true for our first-order analysis of future lake sedimentation rates – as long as average sediment production rates can be estimated for the region of interest. Indeed, our results are based on a combination of modelling results that are available at the global scale (Huss and Hock, 2015) and morphological characteristics that can be retrieved from surface characteristics. Although such an approach entails some important simplifications – such as neglecting the influence of bedrock lithology in the estimated sedimentation rates – our results may serve as a base for future studies. Since glacier lakes are set to become an important element of our future landscapes, we argue that such studies can be of high relevance.
The results generated within this study are available in digital format at
DF and MH conceived the study. TS and RE performed the lake volume calculations with the help of EH and DF. MH performed the GloGEM simulations and modelled the future sediment inputs. DF designed the figures, which were realized with the help of TS, RE, and MH. TS, DF, and MH drafted the manuscript, to which all authors contributed.
The contact author has declared that none of the authors has any competing interests.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We are thankful to Melchior Grab, Hansruedi Maurer, Andreas Bauder, and the many other collaborators that led to the release of the ice thickness distribution of all Swiss glaciers, which is of central importance for our study. Similarly, we are thankful to Andreas Linsbauer and the members of the Glacier Monitoring Switzerland (GLAMOS) Office for providing the Swiss Glacier Inventory 2016. The work benefited from some early thought exchanges with Thomas Kissling, as well as spontaneous feedback on the article's preprint from Wilfried Haeberli. The thorough and constructive comments by Greta Wells and Jan-Christoph Otto significantly improved the manuscript during review.
This research has been supported by the Swiss National Science Foundation (grant no. 184634).
This paper was edited by Susan Conway and reviewed by Jan-Christoph Otto and Greta Wells.