At 11:40 on 12 June 2016, three landslide-like features began flowing almost simultaneously down a south-facing slope of Lhotse Glacier, followed by large amounts of discharging water from three apparent englacial conduits and one supraglacial stream (Fig. 1c, 2a). At the same time, approximately 200 m northwest of these landslide-like features, large amounts of sediment-laden water was observed to be discharging into the main channel from multiple englacial conduits and supraglacial channels (Fig. 2b). Around 12:10, an additional supraglacial torrent and two supraglacial streams, located up-glacier and to the east of the initial observations, joined the floodwater discharging from this initial area. The discharging water immediately began ponding and quickly breached the pond, allowing the floodwater to propagate downstream and join the pre-existing main channel in addition to creating a secondary channel down the southern lateral moraine (Figs. 1c, 2b). During this time, channel banks composed of ice and debris were severely undercut as the floodwater melted the surrounding ice as well.

The main channel continued to flow downstream until it re-entered englacial conduits (Fig. 1c), which created an “ice bridge” that allowed investigators to cross the secondary and main channel after the peak flow started subsiding around 12:26. At 16:22, discharge below Chukhung was measured to be 32±14 m3 s-1. Peak discharge was estimated retroactively to be 210±43 m3 s-1. This estimate is considered to be conservative since it uses average velocity measurements taken 4 h after peak discharge. Interestingly, this estimate agrees well with an empirical approach for predicting peak discharge based on glacier-bed area (Fountain and Walder, 1998), which predicts the peak discharge to be 38–1500 m3 s-1 based on a glacier area of 6.825 km2 for Lhotse Glacier (Arendt et al., 2015). A best-estimate hydrograph (Fig. 1b) was reconstructed based on the photos of the water level at the ice bridge showing a peak flow of 210±43 m3 s-1 at 12:26, followed by a gradual falling limb such that the discharge returned to normal conditions within 24 h. The shape and timing of the hydrograph are consistent with the 1985 glacial lake outburst flood from Dig Tsho (Vuichard and Zimmerman, 1987), although the peak flow from Lhotse Glacier was significantly smaller. Based on this hydrograph, the overall flood volume was estimated to be 2.65×106 m3 and 1.88–3.45 × 106 m3 for the estimated low and high bounds, respectively. Minimal damage was caused to the community of Chukhung, which community members credited to the recently constructed gabions (Fig. 2c). The main damage was the loss of a pedestrian bridge, an outbuilding, and small amounts of floodwater in the courtyard of one lodge. The Supplement provides footage of the observed events.

A detailed field assessment of Lhotse Glacier was conducted to reconstruct the glacier outburst flood by identifying potential flood pathways, englacial conduits, sinkholes, and drained ponds (Fig. 1c). Satellite imagery from 14 May 2016 revealed a sizeable supraglacial pond (27∘54′20′′ N, 86∘53′27′′ E) with an area of 4900 m2 located directly beneath a large bare ice face (∼10–20 m) that was considerably smaller during our field assessment (Fig. 2d). This pond also had fine, wet sediment along its slopes in addition to a series of bare ice, sinkholes, and englacial conduits located immediately downstream, which could have facilitated its drainage. This was the pond located the furthest up-glacier that appeared to have recently drained, although a detailed assessment of all the supraglacial ponds and terrain up-glacier was not possible due to time limitations.

This ponded water likely entered a series of englacial conduits and potentially supraglacial pathways before entering another supraglacial pond located ∼ 200 m down-glacier (Fig. 1c). This second supraglacial pond had similar indicators of having recently drained (Fig. 2e), although the satellite image does not show a large supraglacial pond. It is possible that meltwater filled the pond between the time that the satellite image was acquired and the glacier outburst flood. A collapsed englacial conduit was observed between these two ponds (Fig. 1c) in addition to a series of sinkholes along with an entrance to an englacial conduit located immediately downstream of the pond (Fig. 2h). Based on recently uprooted and displaced alpine shrubs, the flood appeared to continue downstream, where it branched into multiple paths (Fig. 1c). The southern branch appears to have entered a third supraglacial pond (Fig. 2f), which had similar indicators and large sinkholes. Downstream of this third pond was a small valley that was littered with areas of clean ice and deep crevasses (Fig. 2i). It appears that this supraglacial pathway and englacial conduits fed into the flood torrent that joined the initial discharge at 12:10 (Fig. 1c). The other branch showed signs of supraglacial and englacial pathways in the form of bio-indicators, sinkholes, and englacial conduits as well, which appear to have contributed to the heavy flow that was observed discharging into the main channel as well (Fig. 2g).

Satellite imagery provides unique opportunities to observe the contribution of supraglacial ponds to these glacier outburst flood events; however, it is important that this imagery is acquired immediately before and after the event as these supraglacial ponds experience large temporal and spatial changes (Fig. 3). In order to estimate the potential flood volume associated with the drainage of supraglacial ponds, an area-to-volume relationship was used (Cook and Quincey, 2015). Based on the change in areal extent between 14 May and 29 October 2016, the drained volume from the furthest supraglacial pond up-glacier (Figs. 1c, 2d) was 0.01×106 m3. This volume is 2 orders of magnitude less than the estimated flood volume of 2.65×106 m3, which suggests that the drainage of a single supraglacial pond contributes very little to the overall flood volume. In fact, if all of the 274 supraglacial ponds (0.21 km2) that were present on Lhotse Glacier on 14 May 2016 drained completely, the potential flood volume would only be 0.52×106 m3. This provides strong evidence that a significant amount of the flood water was stored in the glacier's subsurface.

The glacier outburst flood on 25 May 2015 also originated from Lhotse Glacier and occurred overnight (L. Sherpa, personal communication, 9 June 2015). Satellite imagery from 8 May, 25 May, and 7 June 2015 reveals a large supraglacial pond (0.036 km2) filling during the period 8–25 May and draining completely during 25 May–7 June (Fig. 3c, d, e). The drainage of this supraglacial pond could have contributed up to 0.17×106 m3 to the 2015 glacier outburst flood. Community members reported that the 2016 event was larger than the 2015 event. A similar outburst event was also reported to have occurred in early May 2016 in the vicinity of the “crampon put-on point” (5600 m) of Island Peak (6189 m) that damaged sections of the high and low basecamp regions (P. T. Sherpa, personal communication, 18 June 2016).

The field observations immediately following the 2016 glacier outburst flood suggest that some of the source water was from the drainage of supraglacial ponds; however, the satellite imagery analysis revealed that the drainage of supraglacial ponds alone could not account for the entire flood volume. Therefore, the water that was unleashed during the 2016 glacier outburst flood was likely stored in both the glacier's subsurface and in supraglacial ponds. Once the flood was initiated, the melting of ice both from the channel banks and in the englacial conduits caused these outlet pathways to grow, which likely contributed more water to the total flood volume in addition to opening more efficient pathways for the stored water to drain.

Potential triggering mechanisms for these glacier outburst floods include dam failure, the rapid drainage of stored lake water through hydraulically efficient pathways, and/or catastrophic glacier buoyancy. The sudden discharge observed during the 2016 event (Fig. 1b) suggests that the trigger was most likely dam failure or the rapid drainage of stored lake water, since catastrophic glacier buoyancy typically has a hydrograph with a more gradual rising limb (Fountain and Walder, 1998).

Dam failure would require an englacial conduit to be temporarily blocked, which could occur if meltwater refroze in the conduits over the winter (Gulley et al., 2009a) or if passage closure processes caused an englacial conduit to close (Benn et al., 2012). The former blockage scenario seems more likely since these glacier outburst floods have occurred in back-to-back years and the refreezing of meltwater is an annual process. During the early melt season the subsurface drainage system is distributed and inefficient, which provides opportunities for water to accumulate englacially (Fountain and Walder, 1998). Dam failure may then occur if the hydrostatic pressures in the englacial conduits exceed the cryostatic pressure that was previously constraining the stored water, thereby causing the dam to rupture (Richardson and Reynolds, 2000). Alternatively, as water accumulates in the englacial conduits, the changes in water pressure can cause these conduits to grow in an unstable manner, thereby causing drainage to occur (Fountain and Walder, 1998). This progressive enlargement is similar to piping failures and the failures of ice-dammed lakes (Richardson and Reynolds, 2000).

The rapid drainage of stored lake water through hydraulically efficient pathways is another plausible triggering mechanism that commonly occurs for supraglacial ponds in the Everest region (Benn et al., 2012). Field observations of supraglacial ponds (Fig. 2d, e) revealed that there were englacial conduits located at the end of both of these lakes that likely helped facilitate their drainage. This link between the englacial conduits and supraglacial ponds is not surprising as near-surface water storage on glaciers can result from water accumulating in englacial conduits (Fountain and Walder, 1998). Once these ponds come in contact with an englacial conduit or a highly permeable layer, the warm pond water can cause significant internal ablation that helps facilitate the drainage of additional stored water. The drainage of supraglacial ponds that was observed for the 2015 and 2016 events supports this theory; however, as previously discussed, the drainage of supraglacial ponds alone likely accounts for a small fraction of the total flood volume.

This suggests that the most feasible triggering mechanism is likely some form of dam failure resulting from the material blocking the englacial conduits being overburdened or failure resulting from the progressive enlargement of englacial conduits. The timing of these events, which occurred around the start of the monsoon season, further supports this triggering mechanism as this provides ample time for these englacial conduits to fill with meltwater or precipitation prior to dam failure. It should not come as a surprise that this time of year is also when supraglacial pond cover is at its highest (Miles et al., 2016) as this may be indicative of the amount of water stored englacially as well. In fact, it is possible that the large supraglacial pond that filled immediately before the 2015 glacier outburst flood (Fig. 3c, d) was the surficial expression of the englacial conduits accumulating too much water, which could explain the pond's short lifespan once the englacial conduits drained. This may also explain how the second supraglacial pond (Fig. 1c, 2e) was not apparent in satellite imagery on 24 May 2016 but appeared to have drained recently based on field observations (Fig. 3a, b); i.e., the pond likely filled between 24 May 2016 and the glacier outburst flood. On the other hand, the most up-glacier supraglacial pond (Fig. 1c, 2d) was present in the imagery and had been growing since 2011 (Watson et al., 2016), which indicates that the rapid drainage of supraglacial ponds through hydraulically efficient pathways may also be contributing to these glacier outburst floods as well, albeit contributing a smaller volume than the water stored englacially.