The influence of supraglacial debris on the rate and spatial distribution of glacier surface melt is well established, but its potential impact on the structure and evolution of the drainage system of extensively debris-covered glaciers has not been previously investigated. Forty-eight dye injections were conducted on Miage Glacier, Italian Alps, throughout the 2010 and 2011 ablation seasons. An efficient conduit system emanates from moulins in the mid-part of the glacier, which are downstream of a high melt area of dirty ice and patchy debris. High melt rates and runoff concentration by intermoraine troughs encourages the early-season development of a channelized system downstream of this area. Conversely, the drainage system beneath the continuously debris-covered lower ablation area is generally inefficient, with multi-peaked traces suggesting a distributed network, which likely feeds into the conduit system fed by the upglacier moulins. Drainage efficiency from the debris-covered area increased over the season but trace flow velocity remained lower than from the upper glacier moulins. Low and less-peaked melt inputs combined with the hummocky topography of the debris-covered area inhibits the formation of an efficient drainage network. These findings are relevant to regions with extensive glacial debris cover and where debris cover is expanding.
Debris-covered glaciers are prevalent in mountainous regions such as the Pamirs and Himalaya (Scherler et al., 2011; Bolch et al., 2012), Caucasus Mountains, Russia (Stokes et al., 2007), and the Western Alps (Deline et al., 2012) and the extent and thickness of debris-cover on glaciers is increasing in many regions (Bolch et al., 2008; Bhambri et al., 2011; Lambrecht et al., 2011; Kirkbride and Deline, 2013). Glacier-runoff is important for downstream water resources, especially during dry seasons (Xu et al., 2009; Maurya et al., 2011). The ablation of ice has a non-linear relationship to the thickness of the overlying debris, with the exact relationship determined by the debris thermal and radiative properties. The relationship between ablation and debris thickness has been derived for several different glaciers and surface covers (e.g. Østrem, 1959; Mattson et al., 1993; Kirkbride and Dugmore, 2003). The dominant effect is a reduction in the melt rate compared with that of bare ice where debris is continuous and more than a few centimetres thick (Brock et al., 2010), with recent hourly energy balance modelling suggesting the debris causes attenuation of the diurnal melt signal (Fyffe et al., 2014).
On debris-free temperate glaciers, dye-tracing studies have demonstrated that the seasonal evolution of the hydrological system,
characterised by increasing efficiency over time, is closely linked to the increase in volume and daily amplitude of surface
meltwater inputs associated with the upglacier retreat of the seasonal snowline (Nienow et al., 1998; Willis et al., 2002;
Campbell et al., 2006). Understanding the nature and evolution of the glacial drainage system is important because it controls how
meltwater inputs impact glacial dynamics (Mair et al., 2002), with the glacial dynamic response affecting erosion rates (Hallet
et al., 1996). However, only Hasnain et al. (2001) have carried out dye tracing on a debris-covered glacier, focussing on the
autumn close-down rather than the spring evolution of the hydrological system, and not dealing explicitly with the influence of
debris cover. Direct investigation of englacial conduit systems within debris-covered glaciers (e.g. Gulley and Benn, 2007) have
not yet revealed the morphology of inaccessible regions, or gauged the efficiency of the entire system. Considering the strong
influence debris has on surface ablation rates (Nicholson and Benn, 2006; Lejeune et al., 2013; Fyffe et al., 2014) extensive
debris cover can be expected to influence the morphology and evolution of a glacier's hydrological system, but the nature and
extent of this impact is not currently known. Based on field investigations at an alpine debris-covered glacier, this study
therefore has two aims:
to understand the influence of debris cover on the daily amplitude and magnitude of surface meltwater input to the glacial
drainage system; to determine the morphology and seasonal evolution of the englacial and subglacial hydrological system and its relationship
to the spatial distribution of supraglacial debris cover.
Miage Glacier is situated in the Western Italian Alps (Fig. 1). It originates from four main tributaries, the Mont Blanc, Dome,
Bionassay and Tête Carrée Glaciers, which form steep icefalls prior to joining the main tongue. As the main tongue enters
Val Veny it bends eastwards before splitting into the large northern and southern lobes and smaller central lobe. The glacier area
is 10.5
Field data were collected at Miage Glacier over two ablation seasons, from 5 June 2010 to 13 September 2010, and from 4 June 2011 to 16 September 2011.
The main outflow stream from the glacier exits the northern lobe, while very little drainage exits the southern lobe. Discharge
was monitored at a gauging station directly downstream of the northern portal (Fig. 1). Stage was measured using a pressure
transducer mounted in a well attached to a large, stable boulder (see Table 1 for details). The Onset HOBO pressure
data were compensated using air pressure data from Mont de la Saxe, 7.6
Prior to conducting a dye trace, the discharge and velocity of the chosen supraglacial stream (
The dilution gauging velocity is the distance between injection and detection points divided by the time between injection and peak of the concentration curve. This gives the average water velocity, a preferable measure of velocity than the float method. Therefore, discharges measured using the velocity-area method were adjusted using the ratio of dilution to float velocity found from simultaneous measurements.
Supraglacial streams and their catchments were defined by applying Arnold (2010)'s lake and catchment identification algorithm
(LCIA) to a digital elevation model (DEM). The algorithm calculates surface slope and direction of steepest descent (flow
direction) for each cell. Sinks (potential lakes) are defined as cells with no lower neighbours, with the algorithm using the flow
direction matrix to find the upstream cells that feed to that sink. The catchment outlet is determined as the lowest cell on the
catchment boundary, with each cell lower than this within the catchment flooded with water to identify lakes. The algorithm also
determines the flow pathways between each catchment allowing the entire supraglacial stream and lake network to be defined. This
supraglacial algorithm is favoured over most others because it does not rely on the artificial filling of sinks before calculating
the flow routing. Arnold (2010) provides detailed model methods. The DEM was derived from airborne LiDAR surveys in 2008 (provided
by Regione Autonoma Valle d'Aosta, VDA DEM hereafter) and has a spatial resolution of 2
Ice thickness data was required to calculate the conduit closure rates (see Appendix). The ice thickness is calculated as the
difference between the surface and bed elevation. The VDA DEM was used to give the surface topography. A map of the bed topography
in Deline (2002) (based on Carabelli, 1961; Casati, 1998; Lesca, 1974), was scanned, georeferenced, digitised and
interpolated into a raster with a 25
Three meteorological stations were located on the glacier. The lower and upper weather stations (LWS and UWS hereafter) were full energy-balance stations situated on continuous debris cover, with the ice weather station (IWS) measuring only air temperature on an area of dirty ice (Fig. 1). Details of the instruments installed on LWS, UWS and IWS are given in Brock et al. (2010) and Fyffe et al. (2014).
In total 48 dye injections were conducted into 16 surface streams. All dye traces were carried out using 21 % rhodamine WT
liquid dye. Between 40 and 280
Although it was intended to use injection points which led directly into a moulin, this often was not possible, especially where debris cover was thick. Streams often flow beneath the debris, making it difficult to inject dye. In some cases, difficulty in accessing moulins due to ice cliffs meant an injection point was used further upstream. During 2011 the execution of repeat injections at individual points was emphasised. Five injection points were chosen, two on the lower glacier debris zone (S5 and S7), and three on the upper glacier debris zone (S12, S14 and S15) (see Fig. 1). The three upper points were intended to be spread equally along the glacier, but following an extensive search the only moulins found were all in a relatively small area. The parameters calculated for each dye-breakthrough curve are given in Table 2.
The injection point into S5 was into a stream 446
An overview of the air temperature, discharge and precipitation in both years is given in Fig. 2. On average, air temperatures in
the June to August period were similar in 2010 and 2011 (11.1 and 10.5
The mean
Dye trace parameters for all 2010 and 2011 injections are reported in Tables 4 and 5, with dye return curves shown in Figs. 6 and 7. For ease of reference, injections into S9 and above will be termed upper glacier traces (zone of patchy debris and bare ice), while those into S8 and below will be termed lower glacier traces (continuously debris-covered ice).
Generally, the water entering the glacier via the main moulins around the upper limit of continuous debris cover travelled quickly
to the proglacial stream, with mean
A striking result is that average
Lower glacier traces in early June were generally slow (e.g. traces into S1, S3, S5 and S7 had
Between June and July 2011
In September the S3_090910
Most upper glacier traces in June (into S10, S12, S13, S14) had
Comparing June and July traces, the S15_280711
The September traces into S12, S14 and S15 showed faster
In the region of the glacier between approximately 2300 and 2500
Surface relief decreases downglacier due to the gravitational redistribution of debris down moraine flanks into the troughs. This
inverts relief development by reversing the ablation gradient down the moraine flanks, reducing the systematic spatial variation
in debris thickness, and eventually resulting in the hummocky topography of the lower tongue (Fig. 4). Consequently, there is less
potential for the formation of an integrated channel network on the continuously debris-covered zone, resulting in a chaotic,
local stream network with hollows which may lead to pond development. Consequently, catchments tend to be smaller than upstream
(Fig. 5), demonstrating that continuous debris cover can constrain catchment size. Melt beneath a continuous debris cover is less
than that of clean or dirty ice, in 2010 averaging 0.019 m we day
Fast, peaked and low dispersion dye return traces from the upper glacier indicate that a channelized system connects surface streams originating on clean and dirty ice, above the continuously debris-covered zone, to the proglacial stream. This was the case even in early June 2010 when the glacier was snow-covered well below the elevation of the upper moulins.
It is widely accepted that the seasonal evolution of a temperate glacier's hydrological system is caused by an increase in the magnitude and amplitude of inputs into the system, initiated by the switch from snow to ice melt, which causes pressure fluctuations large enough to destabilise the hydraulically inefficient distributed system into a more efficient discrete channel system (e.g. Nienow et al., 1998; Willis et al., 2002; Campbell et al., 2006). The question of how a channelized network draining the upper glacier moulins could be established prior to the depletion of the winter snow cover could be explained by two factors: (a) the channels did not completely close over the winter; or (b) early season snowmelt inputs were sufficiently large. Both of these possibilities will be evaluated in turn.
Conduit closure calculations estimate that the main conduit system is likely to have closed over the winter (Appendix). Although there is some uncertainty in the ice thickness values, the modelling suggested it would take only 6–9 days for the conduits emanating from S12 and S14 to close, depending upon the ice density and whether they fed into separate or one combined conduit. Furthermore, if the subglacial conduit was broad and low rather than semi-circular (as suggested by the form of the proglacial stream outlet), closure rates would be faster than those estimated (Hooke et al., 1990).
Runoff generated by the large catchment areas supplying the S12 and S14 moulins combined with topographic flow concentration
(Sect.
The
Normally, it would be expected that increased melt inputs between the early and mid-ablation season would result in
a progressively more efficient channel network. The slower and more dispersed July traces could be due to increased conduit
roughness, caused by a smaller discharge allowing boulders and cobbles on the conduit floor to decrease flow velocity (Gulley
et al., 2012). However June and July proglacial discharges were similar and the degree of dispersion seen was less in June. Rapid
changes in flow velocity can also result from inflow modulation and/or changes in the channel geometry (Nienow et al., 1996;
Schuler and Fischer, 2009). However similar patterns were observed at three different moulins traced at similar times on different
days (Table 5), so it is unlikely that inflow modulation over short time periods was the cause of the differences between the
July and August traces. More plausible is that cold weather between 17 and 27 July 2011 (Fig. 2b, maximum daily
temperatures were generally below 10
The September 2011 traces into the upper glacier moulins (Fig. 7d–f), suggested the drainage system remained more
efficient than in late July but slightly less efficient than in early
August. Air temperatures remained high throughout August 2011 (mean LWS air
temperature in July was 9.4
The drainage system beneath the continuously debris-covered zone was far
less efficient than the upper debris-free area. Traces into S1, S3, S5 and
S7 had slower
Traces into S3, S5 and S7 showed evidence of drainage system evolution
(Fig. 7a–c). Certain peaks of the dye breakthrough
curves became more prominent or coalesced over the season, suggesting
certain flow paths began to dominate within a more integrated network.
Therefore the hydrological network did increase in efficiency, but not to
the extent that water was transferred as quickly as from the upper glacier
moulins. Later in the season there was evidence that the efficiency of the
hydrological network decreased, e.g. a decrease in
The role of debris in reducing meltwater inputs below the critical discharge
at which channels develop (Hewitt and Fowler, 2008) appears crucial in
inhibiting channelisation. Low ablation rates (around 0.02 m we day
These results imply the coexistence of an inefficient drainage system beneath the continuously debris-covered zone with an efficient channelized system which emanates from the upper glacier. Distributed and channelized systems are known to coexist, for instance on Haut Glacier d'Arolla away from the preferential axis of drainage (Nienow et al., 1996), on the western side of Midtalsbreen, southern Norway (Willis et al., 1990), and within the smaller drainage catchment of the South Cascade Glacier, USA (Fountain, 1993), but unusually on Miage Glacier the distributed system occurs downglacier of the channelized network and is the main system of transferring melt on the lower glacier, even though a conduit system exits within the same drainage catchment. However, the proportional distance water has travelled in the efficient and less efficient systems is not known and the systems may not merge until close to the snout.
On the lower glacier it is envisaged that the link between the supraglacial
stream and the main subglacial channel is the inefficient part of the
system. It is this part which causes the lower
Sediment layers are commonly found beneath debris-covered glaciers, due to high rates of sediment supply (Maisch et al., 1999; Hewitt, 2014). It is likely that a layer of sediment underlies the lower glacier (Pavan et al., 1999, cited in Deline, 2002), and if this is thick and highly porous it will likely further inhibit conduit formation, since a sediment wedge downglacier of a hard bed can stall channelisation (Flowers, 2008).
This is the first extensive investigation of the structure and seasonal evolution of the hydrological system of a debris-covered glacier using dye tracing techniques.
Forty-eight dye injections were conducted into 16 surface streams distributed across both debris-free and debris-covered areas of Miage Glacier over the 2010 and 2011 summers. The return curves were analysed in conjunction with supraglacial stream discharge measurements, meteorological data, proglacial stream discharges and topographical analysis of a DEM. The main findings are that:
The upper ablation zone, exhibiting patchy debris cover and high surface melt rates, is connected to the main proglacial stream via an efficient channelized system, which is established early in the season when snow-cover is still extensive, and maintained throughout the ablation season. The majority of meltwater from the lower continuously debris-covered area is drained via an inefficient network which may feed gradually into the main channelized network, although on occasion streams make a direct connection with the main conduit system. Significant and rapid changes in capacity and efficiency of the main channelized network may occur mid-season in response to meltwater supply fluctuations. Although the drainage network beneath the continuously debris-covered zone increased in efficiency between the early and mid-season, it did not become as efficient as the upglacier system. The spatial distribution of debris influences hydrological system development in important and contrasting ways, through its influence on both melt rates and surface topography. First, the establishment and maintenance of an efficient channelized network emanating from moulins draining the upper ablation zone is promoted both by very high ablation rates on patchy debris and dirty ice areas and the topographic concentration of flow into large channels within the moraine troughs. This topographic enhancement is a direct consequence of the large difference in melt rates between medial moraines, which are insulated by thick debris, and the high melt rates of the dirty ice in the intermoraine valleys. Second, the small discharges and low amplitude hydrographs of streams draining the continuously debris-covered area result from both low and attenuated melt peaks beneath thick debris and the hummocky topography which restricts catchment and stream size. This produces dispersed low magnitude melt inputs, preventing water pressure fluctuations becoming great enough to destabilize the distributed system beneath.
These interpretations contrast with conclusions from similar dye tracing
studies conducted on debris-free glaciers. In particular, on Miage Glacier: (i) the formation of the channelized network is not related to the position
of the snowline and (ii)
These findings have implications for those glaciers which are becoming increasingly debris covered (Bolch et al., 2008; Bhambri et al., 2011; Lambrecht et al., 2011) since the debris is likely to influence melt water travel times and therefore the proglacial runoff signal. Debris thickness and spatial extent at Miage Glacier is similar to debris-covered glaciers in mountain ranges such as the Himalayas (Rounce and McKinney, 2014; Schauwecker et al., 2015) and Alaska (Kienholz et al., 2015) hence these findings have relevance to regions where debris-covered glaciers are extensive and common.
Conduit closure rates were calculated by integrating Eq. (7) in Hooke (1984, cited in Nienow et al., 1998). The time,
To understand the sensitivity of the calculations to
In all simulations the largest distance from the gauging station at which the conduits would take 4 months to close was between
1820 and 1844
This work was performed while C. Fyffe was in receipt of a studentship from the School of the Environment, University of Dundee. The authors thank the University of Worcester for kindly funding the publication of this paper. The authors would like to thank J. Holden for the loan of a Seapoint Rhodamine fluorometer and P. Nienow for advice on performing dye tracing studies. F. Brunier from Regione Autonoma Valle d'Aosta kindly provided air pressure data from Mont de la Saxe. Students from the University of Dundee, Northumbria University, Aberdeen University and Cambridge University as well as L. Gilbert provided invaluable help in the field. We would also like to thank M. Vagliasindi and J. P. Fosson of Fondazione Montagna Sicura for excellent logistical support at the field site. The VDA DEM was kindly provided by Regione Autonoma Valle d'Aosta (Modello Altimetrico Digitale della Regione Autonoma Valle d'Aosta aut. n. 1156 del 28 August 2007).
Details of supraglacial and proglacial stream instruments.
Parameters calculated for each dye breakthrough curve.
Mean supraglacial discharges (
Dye trace parameters for all injection points in 2010, for definitions see Table 2.
Mean
Dye trace parameters for all 2011 dye injections. The
Map of Miage Glacier showing location of monitoring stations, lakes and dye tracing points. Inset shows location of Miage Glacier in the Alps. “IWS” is the ice weather station, “UWS” the upper weather station, “LWS” the lower weather station and “GS” the gauging station.
Meteorological conditions and discharge during the
Topographic influence on supraglacial hydrology. The top panel gives an overview of catchment topography. The left inset shows the clear along-glacier ridge and valley topography associated with the central, eastern and western moraines on the upper tongue. The right inset shows hummocky topography on the lower glacier. Both insets show contours at 10 m intervals. Source: Regione Autonoma Valle d'Aosta DEM.
A map of the outlines (shown as white lines) of the modelled supraglacial catchments.
Dye return curves from streams that were only traced once. Note that vertical and horizontal scales differ between subplots.
Repeat dye return curves from single injection points, where
Relationship between the distance to gauging station and