Surface Pond Energy Absorption Across Four Himalayan Glaciers Accounts for 1/8 of Total Catchment Ice Loss.

Glaciers in High Mountain Asia, many of which exhibit surface debris, contain the largest volume of ice outside of the polar regions. Many contain supraglacial pond networks that enhance melt rates locally, but no large-scale assessment of their impact on melt rates exists. Here we use surface energy balance modeling forced using locally measured meteorological data and monthly satellite-derived pond distributions to estimate the total melt enhancement for the four main glaciers within the 400-km2 Langtang catchment, Nepal, for a 6-month period in 2014. Ponds account for 0.20 ± 0.03 m/year of surface melt, representing a local melt enhancement of a factor of 14 ± 3 compared with the debris-covered area, and equivalent to 12.5 ± 2.0% of total catchment ice loss. Given the prevalence of supraglacial ponds across the region, our results suggest that effective incorporation of melt enhancement by ponds is essential for accurate predictions of future mass balance change in the region.

Introduction

The glaciers in High Mountain Asia contain ∼15,000 km3 of ice, the largest reservoir outside of the polar regions (Grinsted, 2013) and play a crucial role in generating and regulating river discharges necessary for hydropower, irrigation, and sanitation for millions of people (Bolch et al., 2012; Immerzeel & Bierkens, 2012). The region's glaciers vary in surface characteristics, climatic setting, and sensitivity to climate change (Sakai & Fujita, 2017), but the majority of large valley glaciers exhibit a thick debris mantle over their ablation areas. Thick debris reduces surface melt (e.g., Östrem, 1959; Nicholson & Benn, 2012) and alters the glaciers' dynamic response to climate warming (e.g., Anderson & Anderson, 2016; Scherler et al., 2011). Despite the general ablation‐reducing effect of thick surface debris, large‐scale studies have identified comparable rates of thinning between debris‐covered and debris‐free glaciers in High Mountain Asia (Brun et al., 2017; Gardelle et al., 2013; Kääb et al., 2012).

Translating thinning rates into rates of melt for debris‐mantled and debris‐free areas is difficult due to differences in (1) glacier hypsometry and consequent meteorological setting (e.g., Ragettli et al., 2016; Vincent et al., 2016) and (2) glacier geometry and consequent flux divergence (Banerjee, 2017; Brun et al., 2018; Nuimura et al., 2017). Recent research has highlighted the role of supraglacial ponds and ice cliffs as areas of enhanced surface melt (e.g., Immerzeel, Kraaijenbrink, et al., 2014; Pellicciotti et al., 2015; Salerno et al., 2017), partially compensating for the reduced ablation over the debris‐mantled areas. The spatial variability of these features leads to a heterogeneous pattern of surface ablation and drives a feedback of glacier surface evolution facilitating the development of new ponds and cliffs (Benn et al., 2012).

Melt rates at ice cliffs have been measured using photogrammetric and geodetic methods (Immerzeel, Kraaijenbrink, et al., 2014; Brun et al., 2016; Thompson et al., 2016; Watson, Quincey, Carrivick, et al., 2017) and have been modeled with increasing complexity (Buri, Miles, et al., 2016; Buri, Pellicciotti, et al., 2016; Han et al., 2010; Reid & Brock, 2014; Sakai et al., 1998, 2002). The effects of supraglacial ponds on ablation are more difficult to measure directly or model, but like ice cliffs, they also lead to local enhanced surface lowering for debris‐covered glaciers (Benn et al., 2001; Miles et al., 2016; Röhl, 2006; 2008; Sakai et al., 2000; Salerno et al., 2017; Watson, Quincey, Smith, et al., 2017).

Ponds represent localized surface drainage inefficiency in the linked supraglacial/englacial drainage network of debris‐covered glaciers, and fill and drain seasonally, moderating glacier discharge (Benn et al., 2012, 2017; Irvine‐Fynn et al., 2017; Liu et al., 2015; Miles, Steiner, et al., 2017; Miles, Willis, et al., 2017; Narama et al., 2017). Due to their strongly positive surface energy balance, ponds warm seasonally and contribute to local surface ablation (Benn et al., 2001). Additionally, some atmospheric energy absorbed by ponds is delivered englacially via pond drainage to drive internal ablation along conduits (Benn et al., 2012; Sakai et al., 2000). The remaining melt energy is often dissipated through the postmonsoon and winter, although some ponds do maintain a winter heat reservoir (Watson et al., 2017; Xin et al., 2011). Thus, ponds can lead to considerable surface and internal ablation; although this has been evaluated for individual ponds on a few glaciers, it remains to be quantified at the glacier and catchment scales.

In this study we aim to provide the first glacier‐ and catchment‐scale estimates of supraglacial pond surface energy balance and potential ablation for Himalayan debris‐covered glaciers. This is achieved by leveraging unique data sets of supraglacial pond seasonality and near‐surface meteorology in the Langtang catchment of Nepal to drive a supraglacial pond energy‐balance model.

Methods

We estimate the net surface energy balance of all supraglacial ponds for a typical ablation season within a large glacierized Himalayan catchment, the 400‐km2 Langtang catchment (Figure 1), which contains four valley glaciers varying in size, elevation, and debris cover. We use extensive local meteorological measurements and Landsat‐derived monthly distributions of supraglacial ponds to drive a pond surface energy balance model (Miles et al., 2016) and use a Monte Carlo approach to account for uncertainty in key model parameters.

Figure 1

Location of the Langtang Valley in Nepal (inset), and principal glaciers in the valley, displaying median results for the cumulative April‐October surface energy balance of supraglacial ponds in each 50‐m elevation band, expressed as equivalent meters of surface melt in that band (left). Cumulative surface energy balance results for all 5,000 parameter sets are shown for each glacier as equivalent surface melt and as a portion of the mean annual 2006–2015 net ablation (right). All debris area refers to the equivalent thinning if spread over the glaciers' debris area, All DCG area instead uses the total glacier area for all glaciers that exhibit debris‐covered tongues, and All glacier area corresponds to all glaciers in the catchment regardless of debris cover. Glacier outlines are updated from the RGI 6.0 (Pfeffer et al., 2014) as in Ragettli et al. (2016). RGI = Randolph Glacier Inventory.

The study period is 15 April to 15 October 2014, the period when pond surfaces are most commonly thawed and for which local meteorological data are available. Meteorological observations were made at an automatic weather station located in nearby Kyanjing village for January–October 2014 (Figure 1), and we leverage an in‐depth understanding of near‐surface meteorology in the catchment based on a dense network of additional stations (Heynen et al., 2016; Immerzeel et al., 2014; Shea et al., 2015; Steiner & Pellicciotti, 2016). Based on these observations we transfer the meteorology observed at the automatic weather station to pond locations (see supporting information).

The study 0‐m elevation bins I think the information in the reference is sufficient to point readers to the CGIAR‐CSI without spelling the whole thing out. Shuttle Radar Topography Mission digital elevation model (Jarvis et al., 2008), and each glacier's monthly ponded area in each bin was prescribed based on Landsat TM/ETM+ pond measurements for 1999–2013 (Miles, Willis, et al., 2017). For each bin of each glacier, the mean monthly pond area from the Landsat archive was assumed to represent conditions in the middle of each month for 2014, and these were then linearly interpolated to hourly values for the 15 April to 15 October study period, producing a time series of pond surface area distributed along each glacier's elevation range.

The pond surface energy balance model of Miles et al. (2016) is applied to each 50‐m elevation band across each glacier's debris‐covered area. A full model description is available in supporting information. The model adjusts observations of incident radiation based on the ponds' topographic setting (Buri, Pellicciotti, et al., 2016; Ridley et al., 2010) and uses the bulk aerodynamic method with a stability correction to model turbulent fluxes. We employ a Monte Carlo approach to account for uncertainty in parameters that are difficult to constrain: albedo, emissivity, roughness length, and temperature of the water surface, as well as topographic view factors for each pond location. We construct 5,000 independent sets for these parameters, varying them randomly within their likely ranges (see supporting information; Brewster, 1992; Chikita & Joshi, 2000; Divine et al., 2015; Hicks, 1972; Jin et al., 2011; Katsaros et al., 1985; Sakai et al., 2009; Yamada, 1998; Yu et al., 2010) and calculate the energy balance for the entire catchment using each parameter set. We additionally test the sensitivity of the model to a ±10% change in temperature lapse rates, as well as an empirical correction for on‐glacier wind speeds.

For each glacier, and for the basin as a whole, we consider the cumulative pond surface energy balance as the ponds' ablation potential (Figure 1). To convert this ablation potential into portions of mass loss, we calculate the mean emergence velocity for each glacier's debris‐covered area using a flux gate approach (supporting information; Cuffey & Paterson, 2010; Huss et al., 2007; Tadono et al., 2014; Vincent et al., 2016). These calculations are based on Glabtop2 ice thickness estimates (Frey et al., 2014) and high‐confidence velocity fields derived from 2015 and 2016 RapidEye orthoimages using ImGraft (Dehecq et al., 2015; Messerli & Grinsted, 2015; Planet Team, 2017). We use the mean emergence velocity values to convert thinning rates observed by Ragettli et al. (2016) for the 2006–2015 period into mean annual ablation rates (Table 1). These calculations assume that pond coverage in 2014 was similar to the average coverage throughout 1999–2013, that the 2015–2016 annual velocity data set is similar to 1999–2013, and that ablation in 2014 was similar to the mean in the period 2006–2015. Consequently, these values represent long‐term averages of the portion of thinning that could be attributable to pond‐associated ablation rather than corresponding to a specific year. In addition, the calculations assume that all energy absorbed by supraglacial ponds leads to glacier melt; in this sense it is an upper bound on the role of supraglacial ponds. In reality, energy left in persistent ponds that do not drain during the monsoon will instead dissipate over winter, while other ponds may drain rapidly and advect unused energy out of the glacier.

Table 1

Characteristics of the Debris‐Covered Glaciers in the Langtang Valley and Cumulative Surface Energy Balance for Supraglacial Ponds in 2014

	Area (km2)		Elevation (m a.s.l.)		Pond density	Thinning	Emergence	Pond April–October surface energy balance			E F
Glacier	Glacier	Debris	Min	Mean	% Debris area	(m/year)	(m/year)	m3×105 Ice	DCA m lowering	% DCA ablation	—
Shalbachum	11.7	2.8	4,218	4,607	0.7%	1.30 ± 0.20	0.07 ± 0.02	2.6 ± 0.4	0.09 ± 0.01	6.6% ± 1.4%	9 ± 2
Lirung	6.1	1.2	4,025	4,287	0.6%	1.67 ± 0.59	0.16 ± 0.1	1.4 ± 0.2	0.11 ± 0.02	6.0% ± 0.2%	10 ± 4
Langtang	52.8	17.8	4,468	4,944	1.3%	0.91 ± 0.05	0.39 ± 0.04	44.0 ± 7.0	0.25 ± 0.04	19.2% ± 3.2%	15 ± 2
Langshisa	21.7	4.4	4,526	4,884	0.5%	1.16 ± 0.23	0.48 ± 0.09	6.0 ± 1.0	0.14 ± 0.02	8.5% ± 1.9%	17 ± 4
All debris		26.2	4,025	4,906	1.0%	1.02 ± 0.18	0.43 ± 0.06	54.5 ± 8.6	0.20 ± 0.03	13.8% ± 2.9%	14 ± 3

Note. Minimum and mean elevation are for the DCA of each glacier. Pond density is the April–October mean pond coverage as a percent of the DCA (Miles, Willis, et al., 2017). Thinning is the mean observed lowering rate for each glacier's debris‐covered area for 2006–2015 (Ragettli et al., 2016). Emergence is the mean emergence velocity over the entire DCA derived from flux gate calculations (supporting information). The ponds' cumulative surface energy balance is expressed as the volume equivalent if all energy was used to melt ice, as a mean ice melt rate for the glacier's DCA and as a fraction of the ablation (combination of thinning and emergence, all here in meters per year) in this area. E F is the melt enhancement factor for ponds relative to all other surface types (equation (1)). For all calculations ρ =900 kg/m3, and we show our results as μ ± 2σ for the 5,000 model runs. DCA = debris‐covered area.

Based on the modeled ablation potential, we calculate a pond‐associated ablation enhancement factor (E F) for each glacier and for the entire study area as the ratio of the rate of pond energy absorption to average ablation rates for the entire debris‐covered area, where M is the atmospheric energy absorbed by ponds annually (m3/a ice melt), is the mean ponded area (m2), is the mean annual surface lowering (m/year; Ragettli et al., 2016), and is the mean emergence velocity (m/year; see supporting information), both for the debris‐covered area.

Results

Overall Pond Energy Absorption

We find that ponds are a major avenue of atmospheric melt energy for debris‐covered glaciers. Across the catchment, pond‐absorbed energy equates to 0.20 ± 0.02 m (expressed here and after as μ ± 2σ) melt over the entire debris‐covered area, accounting for 13.8 ± 2.9% of the annual ablation for the entire debris‐covered area. Ponds thus could account for 12.5 ± 2.0% of net annual mass loss for the total glacierized area in the valley including mass loss for entirely debris‐free glaciers.

Supraglacial pond energy absorption varies between glaciers by several orders of magnitude depending on each glacier's size, hypsometry, and monthly pond densities (Table 1). The whole‐valley energy absorption is dominated by Langtang Glacier's ponds, which absorbed enough energy to melt 4.4 × 106 m3 of ice over the 6‐month study period, equivalent to 0.25 ± 0.04 m of surface melt if distributed across the glacier's debris‐covered area. Compared to the 2006–2015 observed thinning, corrected for mean emergence velocity, Langtang Glacier's ponds are responsible for 19.2 ± 3.2% of the net ablation across its debris‐covered area (Table 1). Ponds account for 6–9% of the net ablation for the three other major debris‐covered glaciers.

Pond melt enhancement factors also vary between glaciers, ranging from 9 ± 2 for Shalbachum Glacier to 17 ± 3 for Langshisa Glacier (Table 1). Across the whole valley, ponds absorb energy at a rate of 14 ± 3 times that of the debris‐covered area. Emergence velocity is a primary source of uncertainty in this calculation due to the regional scarcity of direct measurements of emergence or ice thickness.

Altitudinal and Seasonal Changes in Pond Energy Balance

Our results show a marked altitudinal decrease in mean April–October net pond surface energy balance (Figure 2). The primary driver of this altitudinal decrease is the latent turbulent heat flux, which decreases 100 W/m2 between 4,150 and 5,400 m a.s.l., changing from an atmospheric source to a sink of energy at ∼4,550 m. The sensible turbulent heat flux also shows a moderate decrease with altitude but remains low in magnitude. These reductions with altitude are due primarily to the air temperature lapse rates, which control water surface temperatures. Shortwave and longwave radiative fluxes do not vary significantly with altitude as their topographic controls are modeled stochastically. Despite the latent energy sink at higher elevations, the net surface energy balance is strongly positive for the entire modeled elevation range. Consequently, high‐altitude ponds can remain thawed and absorb energy even when mean daily air temperatures are below freezing (above 5,150 m for our study period in the Langtang Valley), suggesting high‐altitude ponds still act as seasonal recipients of atmospheric energy.

Figure 2

Pond surface energy balance results show a slight decay in Q for the altitudinal range of each glacier (top). Ponded area shows considerable variability between glaciers and with elevation. Ponds could account for up to 0.5 m of ablation for individual elevation bands (bottom).

The pond surface energy balance is strongly affected by seasonal variations in meteorological conditions associated with the South Indian Monsoon. We find that seasonal differences in pond surface energy fluxes are of a greater magnitude than the altitudinal variations, and the daily mean net surface energy balance remains positive at all elevations for the entire study period (Figure 3). Monsoon‐driven decreases in net shortwave radiation due to cloud cover are overcome by increases in net longwave radiation and latent heat fluxes due to warming air temperatures. Diurnal fluctuations are common for all surface energy fluxes, but they vary distinctly with elevation and between the premonsoon and monsoon (Figure 3). Diurnal variations in net shortwave receipts are often reduced in the monsoon compared to the premonsoon due to cloud cover, whereas net longwave receipts increase in magnitude and decrease in diurnal variability due to higher temperatures and increased cloud cover. Little seasonal change is evident in the sensible fluxes, but between the premonsoon and the monsoon the latent fluxes greatly reduce in diurnal variability and, for lower elevations, switch from an energy sink to a source. The resulting net surface energy balance has negative nighttime values at all elevations in the premonsoon, indicating regular refreezing, but is always positive in the monsoon.

Figure 3

Daily mean surface energy balance components at 4,500, 5,000, and 5,400 m at Langtang Glacier for the full study period from our median model run (left). Shaded boxes correspond to the periods from May (middle) and August (right) showing hourly results, highlighting the difference in diurnal energy fluxes between the premonsoon and monsoon. Energy fluxes shown are shortwave balance I (top row, with lower altitude results plotted beneath the 5,400‐m data set), longwave balance L , sensible (H), and latent (E) turbulent fluxes, and the net surface energy balance Q . Note differing y‐axis scales for clarity.

Discussion

Spatial Variations in Pond‐Absorbed Energy

The extent of supraglacial ponding is the primary control on the magnitude of pond energy absorption, while altitudinal differences are of secondary importance (Figure 2). Consequently, glaciers with high pond density at lower altitudes are the strongest recipients of energy via ponds. This is the case across individual glaciers as well: Areas with greater pond concentration will be the most affected, as demonstrated by several elevation bands on Langtang Glacier that absorb melt‐equivalent energy >0.2 m (maximum 0.47 m, Figure 1). Such energy absorbed by ponds will lead to local surface ablation while the ponds are present, and drive internal ablation down‐glacier when ponds drain, leading to a variable, punctuated expression of surface lowering across the debris‐covered areas of the glaciers reported here and Himalayan glaciers more generally (Immerzeel, Kraaijenbrink, et al., 2014; Thompson et al., 2016).

As pond extent is a primary control on overall glacier mass loss (Salerno et al., 2017), we assess the prevalence of ponding in this study relative to other debris‐covered glaciers in order to consider the regional implications of our findings. The average concentration of supraglacial ponds in the debris‐covered areas of the Langtang catchment ranges from 0.5% to 1.7%, with a mean of 1.0% (Table 1). This is within the range observed for other glaciers, such as in the Everest region, where pond densities vary from 0.2% to 6.3% (Watson et al., 2016). Considering the entire Hindu Kush Himalaya mountain range, Gardelle et al. (2011) found pond extent was highly variable, covering 0.01% to 0.4% of total glacier area (we find 0.3% for our domain). Thus, while the error for our Landsat‐derived pond coverages is not insignificant (Miles, Willis, et al., 2017), our results are appropriate for the Central Himalaya, and there may be considerable variability in the overall role of ponds throughout the region.

Comparison to Other Studies and Limitations

Our results indicate that ponds absorb sufficient energy to account for an impressive 12.5 ± 2.0% of annual glacier mass loss in the Langtang catchment of Nepal despite covering only 1% of the catchment's debris‐covered area, or 0.3% of the total glacier area. Our energy‐balance approach is the most sophisticated yet to be applied to Himalayan supraglacial ponds and the only one to have been applied at the catchment scale, but three previous estimates of pond‐associated ablation are useful for comparison. Sakai et al. (2000) attributed 3.4% of Lirung Glacier's total ablation to ponds although they covered only 0.43% of the debris‐covered area, but they used a less sophisticated energy balance model. Focusing on the terminus area of the Tasman Glacier, Röhl (2008) attributed 10% of all ice loss to supraglacial ponds, although that study estimated only the effects of ponds at the glacier's surface (it did not estimate internal ablation), and Tasman Glacier has a very different meteorological setting compared to Himalayan glaciers. Thompson et al. (2016) estimated an upper bound of Ngozumpa Glacier's internal ablation of 9% based on the surface subsidence associated with submarginal conduits; this would account only for energy advected into the glacier by pond drainage.

Our energy‐balance approach makes key advances over these previous studies in terms of formulation and scale of analysis. Nonetheless, many dynamic local processes affecting pond temperature could not be represented, such as natural or wind‐driven pond overturning and losses of energy to proximal ice cliffs (Miles et al., 2016), as the model is built to suit application over glacier and catchment scales rather than individual ponds. We have attempted to account for input data limitations through a Monte Carlo approach with a wide range of parameter sets, and our results are very consistent (μ ± 2σ = 12.5 ± 2.0% of the whole catchment ice loss). We have also tested the model's response to reduced or enhanced air temperature lapse rates; both lead to less than 10% change in the mean values from our 5,000 parameter sets (supporting information). The on‐glacier wind correction has a more substantial effect (15% change in mean value) underlining the importance of locally representative meteorological forcing. Opportunities exist for future improvement of the model and its application; the model could be adapted to directly model local subaqueous melt, or to represent pond surface freeze‐thaw cycles, which occur seasonally and diurnally at higher elevations. Additional measurements are needed to better prescribe some input data, especially for supraglacial pond albedo and surface temperature.

The lack of direct emergence velocity or ice thickness measurements for debris‐covered glaciers makes it difficult to convert thinning rates to ablation over the debris‐covered area. Emergence velocities are usually <0.5 m/year over the lower portion of the debris‐covered area but can be considerable near the clean ice transition (Immerzeel, Kraaijenbrink, et al., 2014; Nuimura et al., 2017; Rounce et al., 2018; Vincent et al., 2016). Our flux gate method is useful for estimating the magnitude of emergence velocity but is dependent on ice thickness modeled using Glabtop2 (Frey et al., 2014). Glabtop2 performed reasonably well in the ice thickness model intercomparison experiment (Farinotti et al., 2017), with a mean bias of −34% for glaciers, but all models' performance varied greatly between glaciers. Consequently, direct measurements of ice thickness and emergence velocity are a high priority for the region.

Outlook for Modeling Pond Ablation at Large Scales

Our results show the potential for ponds to cause significant surface and internal ablation, which has rarely been represented in models of glacier mass balance and evolution. The 14% of net ablation for debris‐covered glacier areas attributable to supraglacial ponds is in addition to the heightened ablation at ice cliffs, which account for an additional 23–24% of ablation for this area (e.g., Brun et al., 2018). Efforts to date have represented ponds and cliffs as areas of reduced debris thickness (Ragettli et al., 2015) or as areas with a constant melt enhancement factor (Kraaijenbrink et al., 2017, who used a factor of 10). These efforts are limited by the static representation of ponded area; although the use of an enhancement factor is a practical approach to include the first‐order effects of pond‐associated mass loss within large‐scale models, such studies should consider a range of E F values to account for differences between glaciers and ensure they use seasonally representative pond distributions.

A more sophisticated approach, as in this study, could be applied across the entire region using downscaled regional climatology, but there is a need for representative seasonal pond coverage data sets for other catchments in the region. Supraglacial pond cover varies seasonally, but most assessments of supraglacial ponding have focused on the postmonsoon, when atmospheric stability leads to cloud‐free imagery but when pond cover can be at its lowest due to seasonal drainage (Liu et al., 2015; Miles, Willis, et al., 2017; Müller, 1968; Narama et al., 2017). Moreover, surface ponds can be small and highly transient features, appearing and disappearing over periods of a few days to a few years, so such data sets must be of high resolution in both space and time (Benn et al., 2017; Miles et al., 2017; Watson et al., 2016).

Conclusions

We use surface energy balance modeling, forced by local meteorological data and monthly satellite‐derived supraglacial pond distributions to estimate the total pond‐associated ablation for four glaciers within the Langtang catchment of Nepal, for a 6‐month period in 2014. Ponds enhance melt rates by a factor of 14 ± 3 compared with other surface types, and account for 0.20 ± 0.03 m/year of surface lowering, equivalent to 12.5 ± 2.5% of total catchment ice loss. This is the first assessment of the total impact of pond‐generated ablation at the catchment scale.

Pond energy absorption will lead to ablation at the surface of the glaciers locally and inside the glaciers along conduits and has a strong spatial and seasonal variability forced by the glacier‐specific spatiotemporal distribution of ponds. Pond energy absorption is also controlled by the seasonal variability in meteorological conditions and the altitudinal decrease of air temperature, but the mean daily pond surface energy balance remains strongly positive for the entire debris‐covered area and for the entire study duration. Our results confirm supraglacial ponds as major contributors to mass loss for debris‐covered glaciers in the Central Himalaya, not only responding to but also contributing to the decline in longitudinal gradient for heavily debris‐covered glaciers. In addition to the need to better constrain the ablation due to supraglacial ice cliffs and the thickness of these glaciers, our study highlights the need for improved pond occurrence data sets measured at high spatial and temporal resolution in order to accurately assess the evolution of debris‐covered glaciers in the 21st century.

Supporting Information S1

Click here for additional data file.