Thermal and Physical Investigations Into Lake Deepening Processes on Spillway Lake, Ngozumpa Glacier, Nepal

water

Article Thermal and Physical Investigations into Lake Deepening Processes on Spillway Lake, Ngozumpa Glacier, Nepal

Ulyana Nadia Horodyskyj

Science in the Wild, 40 S 35th St. Boulder, CO 80305, USA; [email protected]

Academic Editors: Daene C. McKinney and Alton C. Byers Received: 15 March 2017; Accepted: 15 May 2017; Published: 22 May 2017

Abstract: This paper investigates physical processes in the four sub-basins of Ngozumpa glacier’s terminal Spillway Lake for the period 2012–2014 in order to characterize lake deepening and mass transfer processes. Quantifying the growth and deepening of this terminal lake is important given its close vicinity to Sherpa villages down-valley. To this end, the following are examined: annual, daily and hourly temperature variations in the water column, vertical turbidity variations and water level changes and map lake floor sediment properties and lake floor structure using open water side-scan sonar transects. Roughness and hardness maps from sonar returns reveal lake floor substrates ranging from mud, to rocky debris and, in places, bare ice. Heat conduction equations using annual lake bottom temperatures and sediment properties are used to calculate bottom ice melt rates (lake floor deepening) for 0.01 to 1-m debris thicknesses. In areas of rapid deepening, where low mean bottom temperatures prevail, thin debris cover or bare ice is present. This finding is consistent with previously reported localized regions of lake deepening and is useful in predicting future deepening.

Keywords: glacier; lake; ﬂood; melting; Nepal; Himalaya; Sherpas

1. Introduction Since the 1950s, many debris-covered glaciers in the Nepalese Himalaya have developed large terminal moraine-dammed supraglacial lakes [1], which grow through expansion and deepening on the surface of a glacier [2–4]. As temperatures continue to rise [5] and lakes continue to grow in area and volume, they pose a ﬂooding risk to the Sherpa villages down-valley [6,7]. Two conditions suggested to promote the formation of supraglacial lakes are sloped surfaces <2◦ [8] and surface velocities <10 m/yr [9]. Hence, they occur in terminal “dead ice” zones. The Tsho Rolpa on northwest-flowing Trakarding glacier (Rowaling valley) and Tsho Imja on the west-flowing Imja-Lhotse Shar glacier (Khumbu valley) are two examples of moraine-dammed supraglacial lakes that have grown considerably in area and volume from 1950 to the present-day [10–14]. While there appears to be no immediate danger of lake drainage at Imja, continued monitoring is crucial to ensure the terminal ice dam, currently 500 m across, does not narrow and cause a catastrophic flood [15]. Field efforts in 2016 by the Nepalese Army have already lowered the level of the lake by 3 m in order to preemptively mitigate some of the hazard. The Tsho Rolpa differs in that its terminal ice dam already is much narrower and steeper [7]. In the event of ice collapse at the calving front, seiche waves could overtop the moraine and promote dam failure. In the late 1990s, a dam and spillway were constructed to lower the lake level and reduce the risk of a catastrophic drainage downstream [16]. Both lakes have large fetches of multiple kilometers and active calving fronts. Other than partially-ice-cored moraines at their termini, rocky (possibly ice-cored) lateral moraines bound them.

Water 2017, 9, 362; doi:10.3390/w9050362 www.mdpi.com/journal/water Water 2017, 9, 362 2 of 21

The Spillway is a base-level supraglacial lake on the south/southeast-ﬂowing Ngozumpa glacier Water 2017, 9, 362 2 of 23 in the Gokyo valley, between the Rowaling and Khumbu valleys, whose growth stages have been documentedThe Spillway since the is earlya base‐ 2000slevel supraglacial [4,17,18]. Spillwaylake on the has south/southeast the capacity‐flowing to grow Ngozumpa into a much glacier larger moraine-dammedin the Gokyo valley, lake (5–6between km inthe length) Rowaling similar and Khumbu to those valleys, on the whose Trakarding growth and stages Imja-Lhotse have been Shar glaciersdocumented (Figure1), since though the itearly could 2000s become [4,17,18]. much Spillway larger has in area the andcapacity volume to grow than into Tsho a much Rolpa larger and Tsho Imjamoraine given the‐dammed stagnation lake (5–6 of the km lower in length) 6.5 km similar of Ngozumpa to those on glacierthe Trakarding [4,9,18]. and Presently, Imja‐Lhotse the Shar Spillway glaciers (Figure 1), though it could become much larger in area and volume than Tsho Rolpa and consists of multiple sub-basins that are gradually merging into a larger lake. Over the past three Tsho Imja given the stagnation of the lower 6.5 km of Ngozumpa glacier [4,9,18]. Presently, the decades, Spillway has undergone cycles of rapid growth “spurts” alternating with periods of slower Spillway consists of multiple sub‐basins that are gradually merging into a larger lake. Over the past expansionthree decades, [16]. Deepening Spillway rateshas undergone vary locally cycles in theof rapid sub-basins growth from“spurts” less alternating than a meter with per periods year of to tens of metersslower per expansion year [4, 18[16].]. QuantifyingDeepening rates lake vary expansion locally in and the sub deepening‐basins from remains less than an importanta meter per objective,year givento the tens millions of meters of cubicper year meters [4,18]. of Quantifying water that canlake potentiallyexpansion and outburst deepening downstream, remains an affecting important Sherpa villages.objective, Towards given this the end, millions trainings of cubic of Sherpasmeters of via water a “Sherpa-Scientist that can potentially Initiative” outburst (SSI) downstream, were carried out duringaffecting the Sherpa course villages. of this Towards ﬁeld research, this end, to ensuretrainings local of Sherpas knowledge via a of“Sherpa the hazards,‐Scientist as Initiative” well as ways to counteract(SSI) were and carried mitigate out during hazards the shouldcourse of they this arise field inresearch, the future. to ensure As oflocal June knowledge 2016, one of weatherthe stationhazards, remains as well at the as terminusways to counteract of the Ngozumpa and mitigate glacier, hazards collecting should they meteorological arise in the future. data, As and of one June 2016, one weather station remains at the terminus of the Ngozumpa glacier, collecting pressure transducer remains in the Spillway Lake, collecting water level data. The data are periodically meteorological data, and one pressure transducer remains in the Spillway Lake, collecting water level downloaded by SSI-trained locals. Maintaining a research presence on the glacier is of importance as data. The data are periodically downloaded by SSI‐trained locals. Maintaining a research presence the Spillwayon the glacier Lake is continues of importance to expand as the Spillway and deepen. Lake continues to expand and deepen.

Figure 1. Overview map of the Khumbu region (April 2013 Landsat) showing Tsho Rolpa Lake Figure 1. Overview map of the Khumbu region (April 2013 Landsat) showing Tsho Rolpa Lake (NW‐flowing Trakarding glacier) in the Rowaling Valley; Spillway Lake (SE‐flowing Ngozumpa (NW-flowing Trakarding glacier) in the Rowaling Valley; Spillway Lake (SE-flowing Ngozumpa glacier) glacier) in the Gokyo Valley; and Imja Lake (W‐flowing Imja glacier) in the Khumbu Valley. in the Gokyo Valley; and Imja Lake (W-flowing Imja glacier) in the Khumbu Valley. While research has quantified rates of lake expansion [4], it is important also to understand Whileprocesses research controlling has quantifiedthe deepening rates rates of lakeof Spillway expansion Lake. [4 ],This it ispaper important investigates also tolake understand floor processesdeepening controlling in four the sub deepening‐basins of Spillway rates of SpillwayLake using Lake. measured This paperseasonal investigates bottom water lake temperatures floor deepening in fourand sub-basins numerical of modeling Spillway of Lake variable using thicknesses measured of seasonal debris cover bottom water temperatures and numerical modeling(0.01 ofto variable1 meter) to thicknesses predict lake of bottom debris coverice melt (0.01 rates. to 1The m) thickness to predict of lake debris bottom remains ice difficult melt rates. to The thicknessvalidate of debrisin the field, remains due to difficult its spatial to validateheterogeneity in the as field, seen at due the to surface its spatial of Ngozumpa heterogeneity glacier as [19]. seen at the surfaceHowever, of Ngozumpanumerically‐ glaciermodeled [19 melt]. However, rates using numerically-modeled temperature field data melt and rates calculated using temperaturethermal field dataconductivities and calculated provide thermal a framework conductivities with which provide to evaluate a framework previous depth with changes which in to Spillway evaluate Lake. previous Vertical water temperature and turbidity (suspended sediment concentration) measurements, depth changes in Spillway Lake. lake bottom temperatures, lake floor sediment properties and open water side‐scan sonar transects Verticalare presented water to temperature quantify rates and of ice turbidity melt beneath (suspended debris (sediment) sediment layers concentration) on lake floors measurements, through lake bottomheat conduction, temperatures, as well lake as to floor investigate sediment lake properties floor structure. and open The results water are side-scan used to sonar identify transects areas are presentedthat may to quantifypose the risk rates of rapid of ice deepening melt beneath and potential debris (sediment) loss of stored layers glacial on water lake in floors the future. through heat conduction, as well as to investigate lake floor structure. The results are used to identify areas that maypose the risk of rapid deepening and potential loss of stored glacial water in the future. Water 2017, 9, 362 3 of 21

2. Site Description

Ngozumpa is oneWater of 2017 Nepal’s, 9, 362 largest and longest glaciers. It ﬂows 18 km south-southeast3 of 23 from the ﬂanks of Cho Oyu (8188-m)2. Site Description and Gyachang Kang (7922-m) [17]. Near its terminus, at 4800-m, a large Ngozumpa is one of Nepal’s largest and longest glaciers. It flows 18 km south‐southeast from supraglacial lake has beenthe flanks growing of Cho Oyu (8188 in area‐m) and and Gyachang volume Kang (7922 since‐m) [17]. the Near 1980s its terminus, [4]. at 4800 Its‐ waterm, a level is controlled by the elevation of a spillwaylarge supraglacial along lake has the been western growing in area lateral-frontal and volume since the moraine, 1980s [4]. Its water as seen level is in Figure2. controlled by the elevation of a spillway along the western lateral‐frontal moraine, as seen in Figure 2.

Figure 2. Ngozumpa glacier’s terminal moraine and western outlet stream. Boulders along the outlet Figure 2. Ngozumpaare glacier’s a testament to terminal previous floods moraine in the area. and western outlet stream. Boulders along the outlet

are a testament to previousSince 2010, ﬂoods the development in the area.of a new connection between Spillway Lake and a stable pond called ‘Blue Lake’ (Figure 3 (feature labeled E)) may reroute water towards the eastern moraine [4], allowing for more rapid lake expansion, as well as a narrowing of the moraine dam. In 2009, an Since 2010, the developmentaverage lake depth of of 10 m a was new found, connection with a 26‐m maximum between depth in Spillway the north and a Lake stable region and a stable pond called ‘Blue Lake’ (Figure3 (featurein the south labeled(Figure 3 (feature E)) labeled may H)), reroute and a mean water depth of towards 14 m [4]. Spillway the easternmean depth is moraine [ 4], allowing shallower by an order of magnitude when comparing with the Tsho Rolpa and Imja supraglacial for more rapid lake expansion,lakes [10,20]. as well as a narrowing of the moraine dam. In 2009, an average lake depth of 10 m was found, with a 26-m maximum depth in the north and a stable region in the south (Figure3 (feature labeled H)), and a mean depth of 14 m [ 4]. Spillway mean depth is shallower by an orderWater of magnitude 2017, 9, 362 when comparing with the Tsho Rolpa and Imja supraglacial lakes [10,204 of]. 23

FigureFigure 3. The 3. The evolution evolution of of Spillway Spillway Lake’s Lake’s sub-basinssub‐basins since since 2001 2001 (A ()A and) and depth depth map map from from the original the original 2009/20102009/2010 survey survey work work (B ().B). Adapted Adapted fromfrom [4], [4], with with permission permission from from © Elsevier © Elsevier (Geomorphology (Geomorphology). ).

Rather than one large lake, Spillway consists of multiple sub‐basins separated by debris islands and surrounded by calving and melting ice walls. The sub‐basins are named Northwest (NW), Northeast (NE), Main (MB, for “Main Basin”) and Southwest (SW), with the north‐south fetch of the largest (MB) measuring at ~300 m (Figure 4). Comparatively, Tsho Rolpa and Imja have over 1‐km fetches, active calving fronts and bounds in the form of rocky, ice‐cored lateral moraines [10]. On Spillway, exposed ice walls facing south (near NW sub‐basin), north (near NE and Main sub‐basins), and east (located between NW and Main sub‐basins) contribute subaerial meltwater to the lake, keeping lake conditions cool during the melting (monsoon) season, despite higher air temperatures. The south‐facing walls surrounding the NW sub‐basin are prone to back‐wasting (given their <30° slopes) and coverage by thin debris. This led to a temporary stall in the northward expansion of the lake in recent years [18].

Water 2017, 9, 362 4 of 21

Rather than one large lake, Spillway consists of multiple sub-basins separated by debris islands and surrounded by calving and melting ice walls. The sub-basins are named Northwest (NW), Northeast (NE), Main (MB, for “Main Basin”) and Southwest (SW), with the north-south fetch of the largest (MB) measuring at ~300 m (Figure4). Comparatively, Tsho Rolpa and Imja have over 1-km fetches, active calving fronts and bounds in the form of rocky, ice-cored lateral moraines [10]. On Spillway, exposed ice walls facing south (near NW sub-basin), north (near NE and Main sub-basins), and east (located between NW and Main sub-basins) contribute subaerial meltwater to the lake, keeping lake conditions cool during the melting (monsoon) season, despite higher air temperatures. The south-facing walls surrounding the NW sub-basin are prone to back-wasting (given their <30◦ slopes) and coverage by thin debris. This ledWater to 2017 a, 9 temporary, 362 stall in the northward expansion of the lake in recent years5 of 23 [18].

Figure 4. 2016 GeoEye high‐resolution image of Ngozumpa glacier (A) with Spillway Lake close‐up (B), Figure 4. 2016showing GeoEye temperature high-resolution buoy locations image for ofthis Ngozumpa study in the glaciersub‐basins (A (yellow:) with Spillwaynorthwest; green: Lake close-up (B), showing temperaturenortheast; red: buoy main; locations blue: southwest), for this studyweather in station the sub-basins (white square) (yellow: and glacier northwest; outlet channel green: northeast; red: main; blue:(yellow southwest), arrow). weather station (white square) and glacier outlet channel (yellow arrow).

3. Field Methods 3. Field MethodsFour study regions were selected within the Northwest (NW), Northeast (NE), Main (MB) and Four studySouthwest regions (SW) weresub‐basins selected of Spillway within Lake the(Figure Northwest 4) based on area (NW), and depth Northeast maps, as (NE), seen in Main [4]. (MB) and The NW location was used to represent a shallowing region; NE and Main to represent deepening Southwest (SW)areas, sub-basins with the latter of further Spillway removed Lake from (Figureice walls; 4and) based SW to represent on area a and stable/slight depth shallowing maps, as seen in [ 4]. The NW locationregion. wasThe evolution used to of represent Spillway Lake’s a shallowing sub‐basins since region; 2001 reveals NE and that Mainin the NW to representsub‐basin, deepening areas, with thegrowth latter begins further in 2004 removed and expansion from continues ice walls; northward and from SW 2008–2010. to represent In the avicinity stable/slight of the NE shallowing sub‐basin buoy, expansion began in 2008 and continued through 2010. At the Main sub‐basin, region. Theexpansion evolution began of in Spillway 2001. In the Lake’svicinity of sub-basins the SW sub‐basin since buoy, 2001 the region reveals has seen that limited in the growth NW sub-basin, growth beginssince in its 2004 inception and in 2001. expansion ‘Blue Lake’ continues (Figure 3 (feature northward labeled E), from a shallow, 2008–2010. but optically In‐deep the lake, vicinity of the NE sub-basinhas buoy, remained expansion relatively stable, began with in some 2008 shoreline and continued degradation during through this time 2010. period. At Starting the Main in sub-basin, 2009, the development of a new channel routed some of its flow towards the south/southeast. expansion began in 2001. In the vicinity of the SW sub-basin buoy, the region has seen limited growth since its inception3.1. Temperature in 2001. and ‘Blue Water Lake’Level Buoys (Figure 3 (feature labeled E), a shallow, but optically-deep lake, has remained relativelyFor a week stable, in June with2014, instantaneous some shoreline vertical degradation temperature profiles during were this measured time period.using a Starting in 2009, the developmentweighted cable of and a newhandheld channel Oakton routed PC10 probe some with ofa resolution its ﬂow of towards 0.1 °C and theaccuracy south/southeast. of ±0.5 °C. For long‐term measurements (June 2013–June 2014), four buoys with three sensors on each line were constructed using surface and sub‐surface pool floats, anchored to the lake floor with sandbags 3.1. Temperaturecontaining and Water debris (mud Level and Buoys rock) dredged from the bottom. Depth was varied, in order to capture diversity within the sub‐basins. Onset temperature sensors (resolution 0.1 °C, accuracy ±0.2 °C) were For a weektied off in on June the Spectra 2014, fishing instantaneous cord at the verticalsurface, middle temperature and bottom proﬁles for each location. were measured Given using a weighted cablefluctuating and handheld water levels Oakton due to precipitation, PC10 probe ice with melt and a resolution inputs from of englacial 0.1 ◦C channels, and accuracy it was of ±0.5 ◦C. For long-termdifficult measurements to place sensors (June precisely 2013–June at the surface. 2014), As fourwater buoyslevel rose, with buoys three would sensors go underwater on each line were

Water 2017, 9, 362 5 of 21 constructed using surface and sub-surface pool floats, anchored to the lake floor with sandbags containing debris (mud and rock) dredged from the bottom. Depth was varied, in order to capture diversity within the sub-basins. Onset temperature sensors (resolution 0.1 ◦C, accuracy ±0.2 ◦C) were tied off on the Spectra fishing cord at the surface, middle and bottom for each location. Given fluctuating water levels due to precipitation, ice melt and inputs from englacial channels, it was difficult to place sensors precisely at the surface. As water level rose, buoys would go underwater during the monsoon season; and as lake levels dropped, the risk of exposure of the sensor to the air and solar radiation was higher. Thus, “surface” here is taken to represent 10 cm below the waterline. Given Secchi disk measurements of a 15–18-cm optical depth (and solar radiation penetration depth of twice this [21]), this seemed a reasonable compromise. Onset water depth loggers (pressure transducers) were placed at the bottom of the NW and SW sub-basins, to track water level changes.

3.2. Time-Lapse Imagery and Meteorological Data Commercially available time-lapse cameras were placed around the perimeter of Spillway Lake, in order to capture lake changes in real time. They were programmed to record a photo every hour for the duration of the monsoon from June–September 2013. During the winter months, with anticipated cold temperatures, they were reset to shoot every four hours. In April 2014, they were reset to record every hour, in order to capture spring thaw events. An Onset weather station was set up at the high point of a debris island, as seen in Figure4. It measured air temperatures (0.02 ◦C resolution), precipitation (0.2 mm resolution), wind speed (±1.1 m/s), wind direction (±5◦) and solar radiation (±5%).

3.3. Turbidity and Water Properties Turbidity with depth was measured in situ (June 2014) with a handheld turbidimeter, using a 90-degree scatter nephelometer on a 30-m weighted cable. The detected light intensity is directly proportional to the turbidity of the water. Water samples were collected and measured in the lab with units reported in nephelometric turbidity units (NTU). They were then ﬁltered, weighed and converted to suspended sediment concentration (SSC), as per the method in [10]. Other water properties, such as pH (±0.01) and conductivity (±1%), were measured with a handheld Oakton PC10 model. A black and white Secchi disk was used to determine optical depth, where an average of three measurements was taken on clear days, with the Sun overhead. Water chemistry was measured using an ICP-OES (major elements) and ICP-MS (trace elements) at the University of Colorado Boulder.

3.4. Sediment Properties Samples were collected from lake floors using an Ekman dredge with a trigger release messenger. In some regions, lake floor materials consisted of a thick mud packet, while in others, it included a mixture of rocks, pebbles and mud. Care was taken not to sample along shorelines where large rocks could damage the dredge. Samples were measured at the Colorado School of Mines using a KD-2 instrument to quantify heat capacity, thermal diffusivity and thermal conductivity, while particle density and porosity were measured separately. An Onset temperature sensor was buried in sandbags during the 2014 monsoon season, to compare bottom debris temperatures with bottom water temperatures.

3.5. Lake Floor Structure Open water side-scan sonar transects were collected in early June 2014, after the lake thaw and before the start of the monsoon. A remote-operated vehicle (ROV), developed by a team at the Milwaukee School of Engineering and Midwest ROV LLC, provided the platform for a commercial fish finder, transducer and GPS. The depths are not directly comparable with those reported in [4], who derived point-interpolated maps during the winter when the lake was frozen. Rather, these bathymetric results, with shorelines generated from collected ground waypoints (June 2014) and GeoEye (1 m/pixel) satellite imagery (December 2014), were used to determine 3D structure and lake floor characteristics. Using multiple sonar returns during the transects, relative roughness and Water 2017, 9, 362 6 of 21 hardness maps provide information on the presence of hard (rock) versus soft (mud) and compacted versus loose materials on the lake floors. A so-called “E1” layer (echo 1) is derived from the peak returnWater 2017 of, the 9, 362 first echo and represents the rugosity or roughness of the bottom. A peak Sv layer, which7 of 23 measures the strength of the sonar return as it reflects off the bottom, is used as a proxy for hardness, asas E2E2 layerslayers (secondary(secondary returnsreturns ofof thethe echo)echo) werewere notnot alwaysalways availableavailable oror reliable,reliable, givengiven thethe depthdepth rangerange neededneeded forfor thesethese typestypes ofof measurementsmeasurements (twice(twice thethe depthdepth ++ 1010 m)m) [[22,23].22,23].

4.4. Results

4.1.4.1. Sub-BasinSub‐Basin Vertical TemperaturesTemperatures and and Suspended Suspended Sediment Sediment Content Content TemperaturesTemperatures andand turbidityturbidity valuesvalues (converted(converted toto suspendedsuspended sedimentsediment concentrationconcentration (SSC),(SSC), asas seenseen inin FigureFigure S1)S1) werewere measuredmeasured inin thethe vicinityvicinity ofof thethe buoysbuoys forfor thethe sub-basinssub‐basins inin thethe morningsmornings andand afternoonsafternoons duringduring thethe ﬁrstfirst weekweek ofof JuneJune 2014.2014. FigureFigure5 5A,C,EA,C,E shows shows the the temperature temperature variations variations with with depthdepth for thethe NW,NW, NENE andand MBMB sub-basins,sub‐basins, respectively,respectively, whilewhile FigureFigure5 5B,D,FB,D,F shows shows the the suspended suspended sedimentsediment concentrationconcentration variations.variations. The lines represent morning and afternoon averages for the week.

FigureFigure 5. VerticalVertical temperature temperature and and turbidity turbidity variations variations in the in the NW NW (A,B (A), ,NEB), NE(C,D ()C and,D) andMain Main sub‐ sub-basinbasin (E,F ().E ,TheF). The blue blue line line is indicative is indicative of ofmorning morning averages, averages, while while the the black black dotted dotted line line represents afternoonafternoon averages.averages.

The general trend is warmer surface temperatures in the afternoon, given solar heating, but The general trend is warmer surface temperatures in the afternoon, given solar heating, but overall low temperatures (0.5–1.5 °C). In the NW sub‐basin (Figure 5A), temperatures decrease with overall low temperatures (0.5–1.5 ◦C). In the NW sub-basin (Figure5A), temperatures decrease with depth to 7 m and then slightly rise near the bottom. The NE sub‐basin (Figure 5C) is isothermal below a 2‐m depth, with a slight rise below 10 m to the bottom at 15 m. The MB sub‐basin (Figure 5E) is isothermal with depth until reaching 0.1 °C at the bottom, which is interpreted to be either bare ice or a very thin layer of debris covering the ice. A general trend of an increase in suspended sediment concentration (from ~0.25–0.4 g/L) with depth is seen for the sub‐basins, except for the NE, which shows a drop in SSC from a 10–12‐m depth.

Water 2017, 9, 362 7 of 21 depth to 7 m and then slightly rise near the bottom. The NE sub-basin (Figure5C) is isothermal below a 2-m depth, with a slight rise below 10 m to the bottom at 15 m. The MB sub-basin (Figure5E) is isothermal with depth until reaching 0.1 ◦C at the bottom, which is interpreted to be either bare ice or a very thin layer of debris covering the ice. A general trend of an increase in suspended sediment concentration (from ~0.25–0.4 g/L) with depth is seen for the sub-basins, except for the NE, which shows a drop in SSC from a 10–12-m depth. Lower SSC at the surface is attributed to wind-mixing (see Figure S2). This effect can be seen in the NW sub-basin (Figure5B) when comparing the morning and afternoon SSC levels. The decrease in SSC at depth in the NE sub-basin could be an indication of horizontal clean water intrusions from an englacial channel and/or from the nearby north-facing ice walls, although temperatures slightly rise at this depth. There are no discernible pycnoclines or thermoclines in the sub-basins. Substituting water density for temperature in Figure5 reveals that higher density water overlies lower density. For pure water, this condition will cause overturning. Therefore, it must be that in these glacial lakes, the suspended sediment content governs the water density, not temperature. This is similar to the relationship between density and temperature found in the nearby Tsho Rolpa glacial lake [10]. Sieving of dredge samples revealed the finest fraction to be d < 44 µm (clay and silt). Water samples measured from Tsho Rolpa lake had similar grain sizes and increases in SSC with depth, although at twice the concentration (up to 0.8 g/L), attributed to feeder sediment underflows from the large ice calving front [10]. Calmer areas of Tsho Rolpa showed ranges in SSC with depth similar to that recorded here. Applying Stoke’s law, the settling velocity of clay and silt-sized particles is found to be quite low and results in a settling time range of 2 days–3 weeks per every 10 m in the lake. Given the sub-basin suspended sediment concentration profiles with depth and known settling velocity for particle size, diffusivity can be calculated using: wsC = εs(∆C/∆z) where ws is the settling velocity of the mean grain sizes (m/s), εs is the diffusivity for suspended sediment and (∆C/∆z) is the mean SSC gradient of a size fraction, as seen in [11]. The values here range much higher than molecular thermal diffusivity at 1 × 10−7 m2/s, suggesting that part of the mixing is due to turbulence, presumably induced by wind and/or calving at the surface and/or thermal currents at depth. NW and Main sub-basin show similar diffusivities and thus likely have similar mixing intensities, while NE and SW (Figure S3) have the highest diffusivities, indicating higher mixing intensity likely due to the proximity of subaerial ice walls for the former and strong currents (0.8–1 m/s) at the latter. The mostly isothermal nature of these sub-basins, in combination with SSC values, suggests an overall well-mixed system during the monsoon. Flocculation of suspended sediment was found to occur in collected water samples, providing a mechanism for increasing settling velocity and, thus, accumulation of these small particles at depth when surface and subsurface melting activity (mixing) slows during the dry season.

4.2. Water Properties Major elements (Si, Fe, Mg, Ca, Al and Na, in ppm), the trace element arsenic (As, in ppb) and suspended sediment concentration (g/L) were measured in the lab for the sub-basins and surroundings of interest (Table1), including inflow channels, an ice wall, Spillway’s outflow and a nearby stable blue lake. NE is the least turbid of the sub-basins, but with comparable chemistry. This correlates well with the in situ field measurements. As expected, the ice wall sample has the lowest concentration of these elements, due to minimal mixing with surrounding debris and is characterized by having the clearest water. Water 2017, 9, 362 8 of 21

Table 1. Major and trace (As) elements and suspended sediment concentration (SSC) values from locations in and around Spillway Lake.

Si Fe Mg Ca Al Na As SSC Samples (ppm) (ppm) (ppm) (ppm) (ppm) (ppm) (ppm) (g/L) NW basin 14.88 7.00 2.80 10.43 8.00 12.10 13.465 0.43 NW inflow 18.64 0.15 0.24 12.93 0.31 0.61 10.873 0.22 Ice wall 4.32 0.12 0.08 0.64 0.20 0.23 DL 0.07 NE basin 20.81 5.62 1.93 10.73 10.03 16.23 13.763 0.21 NE inflow 12.93 0.02 0.33 15.73 0.04 0.34 0.880 0.06 Main basin 16.57 4.69 1.71 10.00 8.00 25.13 16.363 0.40 SW basin 22.18 6.18 2.14 11.15 10.99 19.77 14.393 0.40 Outflow 15.96 6.00 3.00 11.08 9.00 13.60 14.681 0.46 Blue lake 2.92 0.19 0.27 20.06 0.40 19.73 45.456 0.04

Inflow from exposed englacial channels found near NW and NE differ, with the latter possessing significantly less concentrated suspended sediment, at 0.06 g/L; and the former with higher SSC, at 0.22 g/L. The cleaner NE inflow, one of many channels that exist beneath the water’s surface, may help explain the overall low turbidity in the NE sub-basin. High amounts of Si, Fe and Al in the sub-basins are likely due to sediment inputs, while high values of Ca and As in the stable blue lake are likely from weathering of the surrounding leucogranitic rocks, which contain garnet, tourmaline and chalcopyrite. The blue lake’s SSC is the lowest; it has the largest optical depth and the highest As (45 ppb), nearly four-times that found in Spillway Lake. The U.S. Environmental Protection Agency (EPA) sets the safe drinking water limit at 10 ppb of As. Given the proximity of other such blue lakes to hiking trails, care should be taken when drinking this water, although the situation is not as dire as in Bangladesh’s drinking water, where values reach 500 ppb [24]. Sub-basins show similar (basic) pH values and very low conductivity, as expected for these freshwater lakes (Table2). Optical depth, determined with a Secchi disk, shows equivalent depths for NW, MB and SW. NE’s higher optical depth correlates with its lower turbidity, as measured in the ﬁeld.

Table 2. Water properties for the sub-basins of Spillway Lake. Values of pH and conductivity were averaged with depth.

Sub-basin pH Conductivity (µS) Optical Depth (cm) NW basin 8.59 68.87 18 NE basin 8.39 70.37 32 Main basin 8.21 69.23 17.5 SW basin 8.32 68.47 15

4.3. Annual Temperature Variations Surface, middle and bottom temperatures were measured for the NW (10-m buoy), NE (15-m buoy), Main (17-m buoy) and SW (3-m buoy) sub-basins from June 2013–June 2014. Figure6 shows the annual trends for each. NW buoy (A) has the lowest average annual temperatures, at −0.187 ◦C (surface), due to a long period of ice cover; 0.388 ◦C (middle); and 0.324 ◦C (bottom). Its bottom temperatures are the lowest of all of the sub-basins, and it was the last basin to completely thaw. NE buoy (B) is warmer at 0.651 ◦C (surface); 0.738 ◦C (middle); and 0.865 ◦C (bottom); with the bottom temperatures as second warmest, despite the 15-m depth. The Main buoy has annual average temperatures of 0.136 ◦C (surface), due to long ice cover; 0.706 ◦C (middle); and 0.747 ◦C (bottom). The SW buoy is the warmest with annual average surface temperatures of 0.75 ◦C; 1.02 ◦C (middle) and 1 ◦C (bottom). Given the shallow optical depth of these sub-basins, shortwave radiation is unlikely to reach the bottom, as noted on smaller supraglacial ponds on the Khumbu glacier [13,25]. The warmer bottom temperatures at SW are likely due to the shallower depth, while for MB and NE, the presence of an insulating debris layer keeps bottom temperatures warmer. Water 2017, 9, 362 9 of 21

Close-ups of seasonal temperature variations can be seen in Figures7 and8 (Figures S4 and S5 for NW sub-basin and NE sub-basin, respectively), showing the deepest (MB) and shallowest (SW) buoy locations. During the summer, bottom temperatures at the MB site reach a maximum of 1.2 ◦C, while averaging ◦ ◦ ◦ 0.7 C.Water Comparatively, 2017, 9, 362 SW summer bottom temperatures reach a maximum of 3.26 C, averaging10 of 1.45 23 C. Water 2017, 9, 362 10 of 23

Figure 6. Annual temperatures at the surface (in blue), middle (in green), and bottom (in red) for each FigureFigure 6. Annual 6. Annual temperatures temperatures at theat the surface surface (in (in blue), blue), middle (in (in green), green), and and bottom bottom (in (inred) red) for each for each sub‐basin from 01 June 2013 to 01 June 2014. (A) NW buoy; (B) NE buoy; (C) MB (Main basin) buoy; sub‐basin from 01 June 2013 to 01 June 2014. A(A) NW buoy; (BB) NE buoy; (C)C MB (Main basin) buoy; sub-basinand ( fromD) SW 01 buoy. June The 2013 arrows to 01 in June each 2014. show (the) arrival NW buoy; of Cyclone ( ) NE Phailin buoy; (October ( ) MB 2013). (Main basin) buoy; and (Dand) SW (D) buoy. SW buoy. The The arrows arrows in eachin each show show the the arrival arrival of of CycloneCyclone Phailin Phailin (October (October 2013). 2013).

Figure 7. Seasonal temperature variations in the Main sub‐basin buoy location for the Figure 7. Seasonal temperature variations in the Main sub‐basin buoy location for the Figuresummer/monsoon 7. Seasonal (A temperature), fall (B), winter variations (C) and spring in the (D). Surface Main sub-basinin blue; middle buoy in green; location bottom for the summer/monsoon (A), fall (B), winter (C) and spring (D). Surface in blue; middle in green; bottom summer/monsoonin red. (A), fall (B), winter (C) and spring (D). Surface in blue; middle in green; bottom in red.in red.

Water 2017, 9, 362 10 of 21 Water 2017, 9, 362 11 of 23

FigureFigure 8. Seasonal8. Seasonal temperature temperature variations variations in the SW in sub-basin the SW buoy sub location‐basin forbuoy the summer/monsoonlocation for the (Asummer/monsoon), fall (B), winter ( C(A)), and fall spring (B), winter (D). Surface (C) and in spring blue; ( middleD). Surface in green; in blue; bottom middle in red. in green; bottom in red. Both sub-basins reveal similar temperatures from top to bottom, indicating that during the Both sub‐basins reveal similar temperatures from top to bottom, indicating that during the monsoon, the lake is mostly isothermal and well mixed. Inputs include subaerial meltwater from monsoon, the lake is mostly isothermal and well mixed. Inputs include subaerial meltwater from surrounding ice walls, englacial/subglacial sources and precipitation. During the fall, average bottom surrounding ice walls, englacial/subglacial sources and precipitation. During the fall, average bottom temperatures increase to 1 ◦C for MB (max of 1.59 ◦C) and drop to 1.22 ◦C (max of 2.5 ◦C) for the temperatures increase to 1 °C for MB (max of 1.59 °C) and drop to 1.22 °C (max of 2.5 °C) for the SW SW site. During this time, due to decreasing air temperatures, subaerial and englacial melting inputs site. During this time, due to decreasing air temperatures, subaerial and englacial melting inputs decrease, compared with the summer. SW is able to cool more quickly presumably due to its shallower decrease, compared with the summer. SW is able to cool more quickly presumably due to its depth and larger diurnal temperature amplitudes. shallower depth and larger diurnal temperature amplitudes. Cyclone Phailin dropped one meter of snow in the area in mid-October (2013), leading to a Cyclone Phailin dropped one meter of snow in the area in mid‐October (2013), leading to a decrease in surface and bottom temperatures, from which point the latter did not recover to pre-cyclone decrease in surface and bottom temperatures, from which point the latter did not recover to values. During the winter freeze, bottom temperatures remain above zero for both sites (0.57 ◦C for MB; pre‐cyclone values. During the winter freeze, bottom temperatures remain above zero for both sites 0.43 ◦C for SW). The delay in surface freezing is due to strong (0.8 m/s) surface currents in the vicinity (0.57 °C for MB; 0.43 °C for SW). The delay in surface freezing is due to strong (0.8 m/s) surface of both buoy locations; these have been documented in the ﬁeld to persist through late December. currents in the vicinity of both buoy locations; these have been documented in the field to persist During the spring thaw, the surfaces undergo freeze-thaw cycles for most of April, until total thaw of through late December. During the spring thaw, the surfaces undergo freeze‐thaw cycles for most of the lakes occurs in early May for MB (see Figure S6 for the time-lapse sequence), while for SW, the April, until total thaw of the lakes occurs in early May for MB (see Figure S6 for the time‐lapse thaw occurs a few weeks earlier. Bottom temperatures peak out at 2.45 ◦C for MB and 3.26 ◦C for SW, sequence), while for SW, the thaw occurs a few weeks earlier. Bottom temperatures peak out at before the water column mixes again at the end of the month. A late season snowstorm accounts for 2.45 °C for MB and 3.26 °C for SW, before the water column mixes again at the end of the month. A the anomaly seen in the data on 26 May. late season snowstorm accounts for the anomaly seen in the data on 26 May. A close-up of May 2014 is seen in Figure9 for the NE and Main sub-basin buoy locations, along A close‐up of May 2014 is seen in Figure 9 for the NE and Main sub‐basin buoy locations, along with meteorological data, in order to understand the event that occurred before the 26 May snowstorm with meteorological data, in order to understand the event that occurred before the 26 May (Figure9C) that led to cold temperatures and mixing throughout the water columns (Figure9B,D). snowstorm (Figure 9C) that led to cold temperatures and mixing throughout the water columns No anomalous air temperatures were recorded for 23 May (Figure9A), although NE reached a peak (Figure 9B,D). No anomalous air temperatures were recorded for 23 May (Figure 9A), although NE surface temperature of 4.5 ◦C, and MB saw a comparatively high 3.2 ◦C. In the former case, this would reached a peak surface temperature of 4.5 °C, and MB saw a comparatively high 3.2 °C. In the former lead to stratiﬁcation of the water column, given lower buoyancy of the warm surface water. In the case, this would lead to stratification of the water column, given lower buoyancy of the warm surface latter case, bottom temperatures of 0.2 ◦C would be less dense and would want to rise to replace water. In the latter case, bottom temperatures of 0.2 °C would be less dense and would want to rise the surface water. The snowstorm interferes with the mixing ‘signal’ in the water column, following to replace the surface water. The snowstorm interferes with the mixing ‘signal’ in the water column, this event. However, on 23 May, the middle temperature in the MB sub-basin tracks with the bottom following this event. However, on 23 May, the middle temperature in the MB sub‐basin tracks with temperature of 0.2 ◦C, indicating that this may be an englacial/subglacial emergence of meltwater the bottom temperature of 0.2 °C, indicating that this may be an englacial/subglacial emergence of (Figure9D). The NE sub-basin records a similar 0.2 ◦C bottom temperature drop, though the middle meltwater (Figure 9D). The NE sub‐basin records a similar 0.2 °C bottom temperature drop, though the middle does not track as clearly as in the MB location. A similar event in NE a week prior shows

Water 2017, 9, 362 11 of 21

Water 2017, 9, 362 12 of 23 doesWater not track2017, 9, as362 clearly as in the MB location. A similar event in NE a week prior shows12 of bottom 23 temperaturesbottom temperatures again dropping again to dropping low values, to low although values, notalthough as low not as as the low 23 as May the event,23 May while event, the while surface remainsbottomthe unchanged.surface temperatures remains The unchanged. again distance dropping between The distance to low these values, between sub-basins although these sub is not approximately‐basins as low is as approximately the 23 200 May m. event, 200 m. while the surface remains unchanged. The distance between these sub‐basins is approximately 200 m.

FigureFigure 9. Air 9. temperatures Air temperatures (A )(A and) and solar solar radiation radiation ((CC) during May May 2014. 2014. Temperatures Temperatures do not do reach not reach their peak until later in the month. Solar radiation remains high (>1000 W/m2) during the month, but theirFigure peak until 9. Air later temperatures in the month. (A) and Solar solar radiation radiation remains (C) during high May (>1000 2014. W/mTemperatures2) during do the not month, reach but theirthe 26 peak May until snowstorm later in thedrops month. this toSolar <400 radiation W/m2. Zoom remains‐ins highof the (>1000 NE buoy W/m location2) during (B the) and month, the Main but the 26 May snowstorm drops this to <400 W/m2. Zoom-ins of the NE buoy location (B) and the Main thebuoy 26 locationMay snowstorm (D) for May drops 2014 this show to <400 anomalous W/m2. Zoom bottom‐ins water of the temperatures NE buoy location on 23 (May.B) and the Main buoy location (D) for May 2014 show anomalous bottom water temperatures on 23 May. buoy location (D) for May 2014 show anomalous bottom water temperatures on 23 May. Another close‐up of May 2014 is shown in Figure 10, in which water temperature and pressure AnothertransducerAnother close-up data close provide‐up of Mayof Maymore 2014 2014 insight is is shown shown to the in in 23 FigureFigure May anomaly 10,10 ,in in which which from water waterthe othertemperature temperature sub‐basins. and andpressureThe pressureNW transducertransducerbuoy location data data provide shows provide more a more more insight chaotic insight to the gainto the 23 in May 23water May anomaly level anomaly during from from the the the spring other other sub-basins.thaw sub ‐(Figurebasins. The The10A). NW NW Its buoy locationbuoyproximity shows location to a moredebris shows chaotic‐covered a more gain icechaotic walls in water gaincould in level cause water during occasional level theduring springdisruptions, the thawspring yet (Figure thaw the overall(Figure 10A). increasing Its10A). proximity Its water level pattern continues from 11 May to 23 May. The same gradual rise is seen in the SW buoy to debris-coveredproximity to debris ice walls‐covered could ice cause walls occasionalcould cause disruptions,occasional disruptions, yet the overall yet the increasing overall increasing water level waterlocation level (Figure pattern 10B). continues A gain fromof 0.5 11 m Mayoccurs to during23 May. the The time same of gradualthe anomaly, rise is though seen in it the starts SW inbuoy the pattern continues from 11 May to 23 May. The same gradual rise is seen in the SW buoy location locationNW location (Figure at 10B).7 a.m. A ongain 23 of May. 0.5 m The occurs gain during in SW the is timemore of gradual, the anomaly, peaking though out ait daystarts later. in the A (Figure 10B). A gain of 0.5 m occurs during the time of the anomaly, though it starts in the NW location NWsnowstorm location keeps at 7 thea.m. level on high23 May. through The 26gain May, in andSW byis themore end gradual, of the month, peaking both out sub a‐ basinsday later. return A at 7 a.m.snowstormto their on 23pre May. ‐keepsgain Thelevels. the level gain high in SW through is more 26 gradual,May, and peakingby the end out of the a day month, later. both A snowstorm sub‐basins return keeps the levelto high theirInto through pre the‐gain first 26 levels. week May, of and June, by the the gradual end of rise the in month, water level both continues, sub-basins due return to the to start their of the pre-gain melting levels. Intoseason. theInto While ﬁrst the first week the week NE of and June,of June, MB the thesub gradual gradual‐basins risesaw rise inina signal water oflevel level englacial/subglacial continues, continues, due due to the toactivity the start start (Figureof the of melting the 9B,D), melting season.season.it is While not While as the evident NEthe NE and in NW,and MB MB given sub-basins sub very‐basins low saw saw bottom a a signal signal temperatures of of englacial/subglacial englacial/subglacial to begin with. activity activitySW is a(Figure shallow (Figure 9B,D), 9subB,D),‐ it is notitbasin, as is evidentnot with as evident no in ice NW, surroundings. in given NW, given very lowItvery does bottom low not bottom see temperatures a sudden temperatures drop to in begin to bottom begin with. with.temperature SW SW is is a a shallow on shallow 23 May. sub-basin, sub ‐ with nobasin, ice with surroundings. no ice surroundings. It does not It does see anot sudden see a sudden drop in drop bottom in bottom temperature temperature on 23on May.23 May.

Figure 10. Cont. Water 2017, 9, 362 12 of 21 Water 2017, 9, 362 13 of 23

FigureFigure 10. Surface 10. Surface (blue), (blue), middle middle (green) (green) and and bottom bottom (red) water water temperatures temperatures for for NW NW (A) ( Aand) and SW (B) SW (B) buoy locations. Pressure transducer data reveal a steady rise in water level at the SW site buoy locations. Pressure transducer data reveal a steady rise in water level at the SW site during the during the spring thaw, with a more chaotic gain seen in NW. spring thaw, with a more chaotic gain seen in NW. 4.4. Sediment Properties and Ice Melt Rates 4.4. Sediment Properties and Ice Melt Rates An Ekman dredge was used to collect samples from the sub‐basins, to measure material propertiesAn Ekman in dredge the lab was including used to grain collect size, samples particle fromdensity, the thermal sub-basins, conductivity, to measure heat material capacity properties and in thediffusivity. lab including Consistently grain size, at the particle buoy locations, density, dredged thermal materials conductivity, included heat a surface capacity of andcompacted diffusivity. Consistentlyfine mud, at with the buoysilt and locations, sand‐sized dredged particles materials and pebbles included mixed ain surface with depth. of compacted Quantifying ﬁne thermal mud, with silt andconductivity sand-sized of sediment particles is and important pebbles when mixed attempting in with to depth. model Quantifyingice melt below thermal debris. Underwater, conductivity of sedimentthere is no important need to account when attemptingfor meteorological to model variables, ice melt thereby below reducing debris. the Underwater, complexity. Conductive there is no need heat flux can be calculated using: to account for meteorological variables, thereby reducing the complexity. Conductive heat ﬂux can be calculated using: Qc = −k dT/dz − where k is the thermal conductivity of the materialQc = k (W/m dT/dz‐K), T is the temperature within the sediment wherelayer k is at the some thermal point conductivity (z) within that of layer the material (m) and (W/m-K),Qc is the downward T is the temperature conductive heat within flux. the As sediment per [26] and [27], simplifying the temperature gradient to be linear within the debris layer, the equation layer at some point (z) within that layer (m) and Qc is the downward conductive heat flux. As per [26] becomes: and [27], simplifying the temperature gradient to be linear within the debris layer, the equation becomes: Qc = k(Ts − Ti)/hd Qc = k(Ts − Ti)/hd where Ts is the temperature at the surface of the debris and Ti is the temperature of the underlying ice, taken to be 0 °C. where T is the temperature at the surface of the debris and T is the temperature of the underlying ice, s The melt rate can be calculated using: i taken to be 0 ◦C. The melt rate can be calculated using:M = 3.1536 × 107 (Qc/Lρ)

where the constant converts time from seconds to years, Qc is the calculated heat flux; ρ is the density 7 of ice; and L is the latent heat of melting.M = 3.1536Debris ×samples10 (Q werec/Lρ measured) in the lab for their thermal properties (Table S1). The thermal conductivity is ~1.8 W/m‐K, close to that found previously where(1.7 the W/m constant‐K) for debris converts on the time surface from of Ngozumpa seconds to in years, [19]. Qc is the calculated heat flux; ρ is the density ofCalculated ice; and Lice is melt the rates latent can heat be found of melting. in Table Debris 3, which samples compares were seasonal measured rates for in the the lab for their thermalsub‐ propertiesbasins for sediment (Table S1). thicknesses The thermal varying conductivity from 0.1–1 is m. ~1.8 The W/m-K, thinner the close debris, to that the foundfaster the previously ice (1.7 W/m-K)will melt. for The debris rates are on upwards the surface of 10 of m/yr Ngozumpa in all sub in‐basins, [19]. except for NW, which was the coldest overallCalculated basin. ice At melt only rates 10 cm can of besediment, found inice Table melt 3rates, which are reduced compares significantly, seasonal rates to less for than the 2 sub-basins m/y. for sedimentWith a 1‐ thicknessesm thickness varyingof sediment from assumed 0.1–1 m. to The line thinnerthe bottom the of debris, Spillway the fasterLake [4], the the ice melting will melt. is The ratesnegligible are upwards in all ofsub 10‐basins, m/yr even in all the sub-basins, warm SW location, except forfor observed NW, which annual was bottom the coldest temperatures. overall basin. Figure 11 compares the sub‐basins across the seasons, using only one debris thickness (1 cm) as At only 10 cm of sediment, ice melt rates are reduced significantly, to less than 2 m/yr. With a 1-m an example. In NW, NE and MB, ice melt rates are highest during the fall, while for SW, summer thicknessrates are of sediment higher. High assumed rates in to the line SW the basin bottom can be of attributed Spillway to Lake its shallower [4], the melting depth and is negligible significant in all sub-basins,distance even away the from warm actively SW calving location, and/or for observed melting ice annual walls. bottomThe decline temperatures. of surface melting in the Figure 11 compares the sub-basins across the seasons, using only one debris thickness (1 cm) as an example. In NW, NE and MB, ice melt rates are highest during the fall, while for SW, summer rates are higher. High rates in the SW basin can be attributed to its shallower depth and significant distance away from actively calving and/or melting ice walls. The decline of surface melting in the fall in the Water 2017, 9, 362 13 of 21 other basins aids in heat storage. Winter and spring rates are low in the NW buoy site, given its very cold temperatures throughout. The rates are comparable for the NE and MB. With the addition of an ice cover, heat can be dedicated towards bottom melting. This has been documented by [25] for small supraglacial ponds on the Khumbu glacier.

Table 3. Seasonal ice melt rates (m/yr) for the sub-basins, with simulated debris thicknesses.

Seasonal Bottom Ice Melt (in Meters) Sub-basin Debris Thickness (m) Summer (m) Fall (m) Winter (m) Spring (m) 0.01 2.1 2.92 0.51 0.37 0.02 1.05 1.46 0.26 0.19 0.05 0.42 0.58 0.1 0.07 0.1 0.21 0.29 0.05 0.04 NW sub-basin 0.15 0.14 0.19 0.03 0.02 0.3 0.07 0.1 0.02 0.01 0.5 0.04 0.06 0.01 0.01 1 0.02 0.03 0.01 0 0.01 3.23 4.53 2.66 3.02 0.02 1.6 2.25 1.31 1.51 0.05 0.64 0.9 0.52 0.52 0.1 0.32 0.45 0.26 0.3 Main sub-basin 0.15 0.21 0.3 0.17 0.2 0.3 0.11 0.15 0.09 0.1 0.5 0.06 0.09 0.05 0.06 1 0.03 0.04 0.03 0.03 0.01 4.42 5.39 2.84 3.18 0.02 2.21 2.67 1.4 1.58 0.05 0.88 1.07 0.56 0.63 0.1 0.44 0.53 0.28 0.32 NE sub-basin 0.15 0.29 0.36 0.19 0.21 0.3 0.15 0.18 0.09 0.11 0.5 0.09 0.11 0.06 0.06 1 0.04 0.05 0.03 0.03 0.01 6.58 5.71 2.02 3.37 0.02 3.29 2.85 1.01 1.68 0.05 1.32 1.14 0.4 0.67 0.1 0.66 0.57 0.2 0.34 SW sub-basin 0.15 0.44 0.38 0.13 0.22 0.3 0.22 0.19 0.07 0.11 0.5 0.13 0.11 0.04 0.07 1 0.07 0.06 0.02 0.03

Water 2017, 9, 362 15 of 23

Figure 11. IceFigure melt 11. (meters) Ice melt (meters) for the for sub-basins the sub‐basins during during thethe summer summer (blue), (blue), fall (orange), fall (orange), winter (black) winter (black) and springand (green), spring using (green), 1 using cm of 1 cm debris of debris thickness. thickness.

Figure 12 compares June–October 2014 surface and bottom water temperatures in the Main sub‐ basin, with debris temperature from a sensor buried in a sandbag filled with sediment that had been dredged from the bottom. This buoy was placed within 10 m of a west‐facing ice wall. From mid‐ June–mid‐August, temperatures drop from 1.8 down to 0.2 °C, due to the meltwater inputs. No snow was seen in time‐lapse imagery in mid‐August, when debris and bottom temperatures dropped to 0.2 °C. This indicates an upwelling of englacial/subglacial meltwater. The temperatures begin to rise again from mid‐August–mid‐October. During this time, air temperatures remained high until August and then started gradually dropping until October. Cyclone Hudhud dropped 30 cm of snow in mid‐ October 2014, explaining the sudden temperature drop in the water column. The data show an average of 0.75 °C for bottom water temperature and 0.62 °C debris temperature during this timeframe. The 0.15 °C difference implies an overestimate of ice melt using June 2013–June 2014 data.

Water 2017, 9, 362 15 of 23

Figure 11. Ice melt (meters) for the sub‐basins during the summer (blue), fall (orange), winter (black) Water 2017and, 9spring, 362 (green), using 1 cm of debris thickness. 14 of 21

Figure 12 compares June–October 2014 surface and bottom water temperatures in the Main sub‐ Figure 12 compares June–October 2014 surface and bottom water temperatures in the Main basin, with debris temperature from a sensor buried in a sandbag filled with sediment that had been sub-basin, with debris temperature from a sensor buried in a sandbag ﬁlled with sediment that dredged from the bottom. This buoy was placed within 10 m of a west‐facing ice wall. From mid‐ had been dredged from the bottom. This buoy was placed within 10 m of a west-facing ice wall. June–mid‐August, temperatures drop from 1.8 down to 0.2 °C, due to the meltwater inputs. No snow From mid-June–mid-August, temperatures drop from 1.8 down to 0.2 ◦C, due to the meltwater inputs. was seen in time‐lapse imagery in mid‐August, when debris and bottom temperatures dropped to No snow was seen in time-lapse imagery in mid-August, when debris and bottom temperatures 0.2 °C. This indicates an upwelling of englacial/subglacial meltwater. The temperatures begin to rise dropped to 0.2 ◦C. This indicates an upwelling of englacial/subglacial meltwater. The temperatures again from mid‐August–mid‐October. During this time, air temperatures remained high until August begin to rise again from mid-August–mid-October. During this time, air temperatures remained high and then started gradually dropping until October. Cyclone Hudhud dropped 30 cm of snow in mid‐ until August and then started gradually dropping until October. Cyclone Hudhud dropped 30 cm of October 2014, explaining the sudden temperature drop in the water column. The data show an snow in mid-October 2014, explaining the sudden temperature drop in the water column. The data average of 0.75 °C for bottom water temperature and 0.62 °C debris temperature during this show an average of 0.75 ◦C for bottom water temperature and 0.62 ◦C debris temperature during this timeframe. The 0.15 °C difference implies an overestimate of ice melt using June 2013–June 2014 data. timeframe. The 0.15 ◦C difference implies an overestimate of ice melt using June 2013–June 2014 data.

Water 2017, 9, 362 16 of 23

FigureFigure 12. 12.Air Air temperatures temperatures (A) and(A) surfaceand surface (blue), (blue), bottom bottom (red) water (red) temperatures water temperatures and debris and (black) debris temperatures(black) temperatures (B) were measured(B) were measured in the Main in sub-basinthe Main sub from‐basin June–October from June–October 2014. 2014.

4.5.4.5. Sonar Sonar and and Structure Structure AnAn open open water water bathymetric bathymetric survey survey was was conducted conducted on on the the sub-basins sub‐basins (and (and surrounding surrounding smaller smaller ponds)ponds) of of Spillway Spillway Lake Lake in in the the first first week week of of June June 2014, 2014, using using a 200-kHza 200‐kHz dual dual down down and and side side imaging imaging commercialcommercial fish fish finder, finder, with with±1 ±1 m m vertical vertical accuracy. accuracy. A transducer, A transducer, GPS GPS and and handheld handheld unit, unit, powered powered by a 20-amp-hby a 20‐amp battery‐h battery in a waterproof in a waterproof casing, casing, were mounted were mounted on a boat on and a boat towed and by towed a sea kayak.by a sea Lateral kayak. positionLateral accuracy position is accuracy estimated is asestimated±0.5 m. as The ±0.5 3D m. modeling The 3D results modeling from results the four from sub-basins the four are sub seen‐basins in Figureare seen S7. in The Figure NW sub-basinS7. The NW (A) sub has‐basin a maximum (A) has a depth maximum of 21 depth m, with of a21 ridge-like m, with a structure ridge‐like trending structure west-east,trending awest relict‐east, of the a relict hummocky of the hummocky surface prior surface to water prior infill to water from infill lake from expansion lake expansion in 2004–2008. in Similar2004–2008. relict Similar hummocky relict terrain hummocky is seen terrain in the is Main seen sub-basinin the Main (C, sub maximum‐basin (C, depth maximum of 22 depth m) as lakeof 22 m) as lake expansion began in that area in 2004. The NE and SW sub‐basins are shallower, at a 15‐m and 14‐m depth, respectively. Lake floor roughness can be estimated by acoustic backscatter in the primary echo region (E1 layer). To the side‐scan sonar, smooth bottoms, such as compacted muds and sands, act as acoustic mirrors and reflect most of the returning echo away from the receiver. The result is an echo amplitude waveform with a sharp rise and rapid deterioration [22,23]. In contrast, irregular bottom surfaces, consisting of rocks and boulders, reflect sound with multi‐path back to the receiver. Thus, these kinds of surfaces will increase the proportion of the total transmitted sound that is reflected back to the transducer. In this case, the waveform rises quickly, but deteriorates slowly. Figure 13 provides an example of open‐water data collected in the Main sub‐basin of Spillway Lake.

Water 2017, 9, 362 15 of 21

expansion began in that area in 2004. The NE and SW sub-basins are shallower, at a 15-m and 14-m depth, respectively. Lake floor roughness can be estimated by acoustic backscatter in the primary echo region (E1 layer). To the side-scan sonar, smooth bottoms, such as compacted muds and sands, act as acoustic mirrors and reflect most of the returning echo away from the receiver. The result is an echo amplitude waveform with a sharp rise and rapid deterioration [22,23]. In contrast, irregular bottom surfaces, consisting of rocks and boulders, reflect sound with multi-path back to the receiver. Thus, these kinds of surfaces will increase the proportion of the total transmitted sound that is reflected back to the transducer. In this case, the waveform rises quickly, but deteriorates slowly. Figure 13 provides an example of Wateropen-water 2017, 9, 362 data collected in the Main sub-basin of Spillway Lake. 17 of 23

FigureFigure 13. 13. SampleSample sonar sonar image image from from the the Main Main sub sub-basin‐basin of ofSpillway Spillway Lake, Lake, showing showing the the primary primary echo echo (E1(E1 layer), layer), which which measures measures roughness roughness (the (the signal signal can can be be mirrored mirrored back back or or take take a multi a multi-path),‐path), and and the the secondarysecondary (peak (peak Sv Sv layer) layer) echo, echo, which which measures measures hardness hardness based based on on the the material’s material’s acoustic acoustic absorption. absorption. TheThe former former shows shows sonar sonar returns returns corresponding corresponding with with color color-coded‐coded dots. dots. The The darker, darker, the the rougher, rougher, in ina a relativerelative scale. scale.

To determine bottom hardness, the secondary echo region (E2) can be used. This is a measure of To determine bottom hardness, the secondary echo region (E2) can be used. This is a measure of the acoustic absorption or impedance of the bottom. Mud and materials with high water content will the acoustic absorption or impedance of the bottom. Mud and materials with high water content will absorb a large proportion of the transmitted acoustic energy, reducing the echo amplitude, whereas absorb a large proportion of the transmitted acoustic energy, reducing the echo amplitude, whereas hard substrates will reflect a greater proportion of the transmitted energy, resulting in high amplitude hard substrates will reflect a greater proportion of the transmitted energy, resulting in high amplitude secondary echoes [22,23]. In cases where E2 returns are not picked up, due to insufficient depth, peak secondary echoes [22,23]. In cases where E2 returns are not picked up, due to insufficient depth, peak Sv is used instead. This is calculated from the primary echo and defined as the observed maximum Sv is used instead. This is calculated from the primary echo and defined as the observed maximum volumetric backscatter (Sv) of all of the acoustic samples within the layer between the sounder volumetric backscatter (Sv) of all of the acoustic samples within the layer between the sounder detected detected bottom and the bottom of the E1 layer. bottom and the bottom of the E1 layer. It is important to note that E1, E2 and peak Sv provide only a relative measure of substrate It is important to note that E1, E2 and peak Sv provide only a relative measure of substrate roughness and hardness. The measurements are highly influenced by the acoustic footprint size, thus roughness and hardness. The measurements are highly influenced by the acoustic footprint size, thus limiting the ability to resolve changes with increasing water depth. E2 measurements rely on sound limiting the ability to resolve changes with increasing water depth. E2 measurements rely on sound reflections from the lake surface and so are influenced by wind and wave conditions. For this reason, reflections from the lake surface and so are influenced by wind and wave conditions. For this reason, the four sub‐basins cannot be compared with each other (rockier, harder, etc.). the four sub-basins cannot be compared with each other (rockier, harder, etc.). Figures 14 and 15 show NE and Main sub‐basin E1 and peak Sv maps (Figures S8 and S9 for NW and SW sub‐basins). In both cases, (A) reveals smooth substrate from the E1 return, while (C) shows soft substrate from the calculated peak Sv. (B) and (D) show rocky (E1) and hard (peak Sv), respectively. Rocky and hard returns in Figure 14B,D are seen in the eastern sector and correlate well with recent shallowing.

Water 2017, 9, 362 16 of 21

Figures 14 and 15 show NE and Main sub-basin E1 and peak Sv maps (Figures S8 and S9 for NW and SW sub-basins). In both cases, (A) reveals smooth substrate from the E1 return, while (C) shows soft substrate from the calculated peak Sv. (B) and (D) show rocky (E1) and hard (peak Sv), respectively. Rocky and hard returns in Figure 14B,D are seen in the eastern sector and correlate well with recent shallowing. Water 2017, 9, 362 18 of 23

Water 2017, 9, 362 18 of 23

Figure 14. NE sub‐basin roughness and hardness maps, generated from E1 and peak Sv acoustic Figurebackscatter 14. NE sub-basin values. Scaled roughness smooth (A and), rough hardness (B), soft maps, (C) and generated hard (D) are from shown, E1 with and boxes peak overlain Sv acoustic Figure 14. NE sub‐basin roughness and hardness maps, generated from E1 and peak Sv acoustic backscatterto show values. smooth Scaled‐hard returns, smooth interpreted (A), rough to ( Bbe), bare soft ice. (C) and hard (D) are shown, with boxes overlain backscatter values. Scaled smooth (A), rough (B), soft (C) and hard (D) are shown, with boxes overlain to show smooth-hard returns, interpreted to be bare ice. to show smooth‐hard returns, interpreted to be bare ice.

Figure 15. Cont.

Water 2017, 9, 362 17 of 21 Water 2017, 9, 362 19 of 23

Figure 15. Main sub‐basin roughness and hardness maps, generated from E1 and peak Sv acoustic Figure 15. backscatterMain sub-basin values. Smooth roughness (A), rough and (B), hardness soft (C) and maps, hard generated(D) are shown, from with E1 boxes and overlain peak Sv to acoustic backscattershow values. smooth‐hard Smooth returns (A (A,D),), rough interpreted (B), soft to (beC )bare and ice. hard (D) are shown, with boxes overlain to show smooth-hard returns (A,D), interpreted to be bare ice. Rocky and hard returns in MB (Figure 15B,D) are seen along the shore and on the hummocky shallower terrain in the middle of the basin. Overlain black boxes highlight areas with smooth (A) Rockyand hard and (D) hard returns, returns which in MBreveal (Figure areas of 15 bareB,D) or are nearly seen bare along ice. More the shore field data and are on needed the hummocky to shallowervalidate terrain these in results the middle using seismic of the and/or basin. ground Overlain‐penetrating black boxes radar. highlight However, in areas the case with of smooth the (A) and hardMain (D) sub returns,‐basin, vertical which temperature reveal areas measurements of bare or (Figure nearly 5) bareshowed ice. 0.1 More°C at the field bottom, data while are needed to validatefield thesephotos results (Figure using S10) show seismic upwelling and/or activity ground-penetrating in the region, likely radar. pointing However, to an inactive the case of the Mainenglacial/subglacial sub-basin, vertical input and temperature near‐bare ice measurements in the region. (Figure5) showed 0.1 ◦C at the bottom, while field5. Discussion photos (Figure S10) show upwelling activity in the region, likely pointing to an active englacial/subglacial input and near-bare ice in the region. 5.1. Sub‐Basins of Spillway 5. Discussion The NW buoy location, at a 10‐m depth, recorded the coldest annual temperatures from top to bottom. This cold, turbid sub‐basin emerged in 2004 and began expanding towards the north 5.1. Sub-Basins of Spillway beginning in 2008. South‐facing ice walls were actively calving in 2009/2010, during initial survey Thework NW as buoyseen in location, [4]. Upon at return a 10-m to the depth, field site recorded in 2012, the walls coldest were annual found to temperatures be covered with from top to bottom.thick Thisdebris. cold, Extensive turbid and sub-basin high amounts emerged of shallowing in 2004 in the and sub began‐basin from expanding towards the north 2009–2012 are interpreted to represent the fall of debris (rocks) into the water. The area of deepening beginning in 2008. South-facing ice walls were actively calving in 2009/2010, during initial survey towards the west did so at rates of ~1.5–5 m/y. Using bottom temperatures for NW, this can be work asachieved seen in with [4]. Upon1–5 cm return of sediment to the cover. field This site is in plausible, 2012, the given walls the were relatively found young to be age covered (five years) with thick debris.of Extensive the sub‐basin. and highEnhanced amounts water ofmixing shallowing due to calving in the sub-basinand debris fromfalls can 2009–2012 also contribute are interpreted to to representpotentially the fall warmer of debris temperatures (rocks) at into depth, the although water. Thea sensor area near of deepeninga melting/calving towards ice wall the in west the did so at ratesMain of ~1.5–5 sub‐basin m/yr. (June–October Using bottom 2014; Figure temperatures 12B) recorded for very NW, low this surface can and be bottom achieved temperatures with 1–5 cm of sedimentduring cover. the This melting is plausible, season. given the relatively young age (five years) of the sub-basin. Enhanced The NE buoy location, at a 15‐m depth, was the second warmest sub‐basin. It was optically the water mixingclearest, due as well to calving (Table 2). and The debris emergence falls of can the main also contributebody of this tosub potentially‐basin was in warmer 2008, with temperatures growth at depth, althoughof the eastern a sensor sector documented near a melting/calving in 2009 and west ice‐facing wall ice in walls the Mainobserved sub-basin in 2010. The (June–October ice walls in 2014; Figure 12theB) eastern recorded sector very have low since surface been covered and bottom with debris. temperatures Rough and hard during sonar the returns melting in the season. vicinity The NE buoy location, at a 15-m depth, was the second warmest sub-basin. It was optically the clearest, as well (Table2). The emergence of the main body of this sub-basin was in 2008, with growth of the eastern sector documented in 2009 and west-facing ice walls observed in 2010. The ice walls in the eastern sector have since been covered with debris. Rough and hard sonar returns in the vicinity (Figure 14B,D) possibly signify debris collapse into the basin. The deepening in certain locations here is on the order of 3–5 m/yr. Using the bottom water temperatures, achieving this rate requires <5 Water 2017, 9, 362 18 of 21 cm of debris, which is possible given the relatively young age (five years) of the basin, in certain spot locations. The Main buoy location was in 17 m of water. The emergence of the southern sector started in 2001, with northward expansion beginning in 2004. Northwestern expansion began in 2008. Thus, the oldest part of the sub-basin is in the south, while the youngest is towards the NW. The sonar returns, as seen in Figure 15, reveal smooth and soft surfaces in the deep, older areas of the basin, here interpreted to be mud. Sonar returns and substrate classification in the younger parts of the sub-basin point to hard and smooth surfaces, and more research is needed to determine whether or not this is bare ice. However, this would correlate well with the high (5–7 m/yr) deepening rates in the region. Using the recorded bottom water temperatures, achieving these deepening rates in the older region of the basin would requires <5 cm of debris. However, given the age of the basin (13 years), there should have been sufficient time to accumulate thick debris. Documented flow currents of >1 m/s, even late into the dry season in the narrow region connecting Main to SW, can potentially extend the settling time of particles in this basin and/or bottom currents can potentially lead to lake floor scouring of debris. The SW buoy location was in 3 m of water and the warmest overall location. This sub-basin is the oldest, at 13 years, and has remained a shallow/stable region during this time [4]. While the bottom temperatures recorded here can lead to the highest ice melt rates, sediment thickness is likely very high. A 0.3 m/yr sedimentation rate was documented in this region [4]. The bottom temperatures at a shallow depth can still be used as a proxy for other shallow regions in the vicinity of the SW basin and can help explain (limited) shoreline degradation in the region since 2001.

5.2. Tsho Rolpa, Imja and Spillway Investigations at Tsho Rolpa revealed vigorous circulation of surface water due to wind-driven currents across a 3-km fetch, as well as an established thermocline at 25 m and weak pycnocline [10]. This allows for heat transfer between water and ice along the calving front and on the thinly sediment-covered lake ice ﬂoor. Warm mean annual water temperatures (2–3 ◦C) and heat conduction through sediments can lead to melting and deepening of the ice ﬂoor on Tsho Rolpa at much faster rates than Spillway, given these higher temperatures. No thermocline was observed at Imja, indicating that wind mixing is not as prevalent. It is an isothermal lake, at 3 ◦C, which may encourage higher rates of dead-ice melt through a sediment layer. Despite the capacity for mixing, given strong valley winds, the presence of extensive dead ice near the terminus, as well as elevated debris islands results in roughness and turbulence, thereby lessening mixing in Imja given these structures can behave as energy dissipaters, shielding the lake from wind [11,20]. Spillway Lake is comparatively shallow (Tsho Rolpa’s thermocline is at Spillway’s maximum depth), with a much smaller fetch (300 m compared with 3 km) and no established thermocline or pycnocline. Rather than a calving front, the lake is surrounded by ice walls and is fed by multiple supraglacial, englacial and possibly subglacial channel inputs. Although comparably strong valley winds blow during the summer, hummocky terrain and multiple debris islands dotted between the sub-basins, especially between Main and NW sub-basins, likely result in high roughness and turbulence, like that seen on Imja, minimizing the wind effect. Subaqueous calving events, as readily observed on the debris-covered Tasman glacier (New Zealand), can lead to rapid deepening and destabilization [28]. It is unknown how frequently this occurs on Himalayan debris-covered glaciers [16]. No instances of subaqueous calving were observed in our time-lapse cameras during 2013–2014, although that does not mean it does not occur.

6. Conclusions Precise 4D measurements of temperature and associated water opacity, sediment concentration and water depth over two seasons of snow accumulation and melt have elucidated several features that appear important in governing the depth and areal expansion of supraglacial lakes. Spillway Lake Water 2017, 9, 362 19 of 21 on Ngozumpa glacier affords a view of contrasting processes from which we may forecast the future development of Spillway and other similar lakes. In areas of very thick sediment cover, like the older southern region of Spillway, bottom deepening is expected to be minimal, although expansion can continue towards the north and west, exploiting weaknesses (cracks) in the terrain. In younger parts of the lake, especially in the NE sub-basin with actively melting and calving north-facing ice walls, less time to accumulate debris and higher ice melt rates under this thinner debris means that deepening rates are expected to continue to increase. In the Main sub-basin, despite older and younger regions, deepening is continuing at the highest rate in Spillway Lake. More information is needed, in the form of seismic or radar survey, to better quantify debris thickness and to detect the presence of bare ice in this region, while current-meters set at depth can be used to quantify the ﬂow associated with englacial and subglacial inputs. Higher resolution multi-beam sonar surveys could prove useful in mapping subaqueous ice cliffs in the vicinity of calving ice walls. In view of the danger from large waves associated with calving during the measurements, unmanned remotely steerable sensor platforms offer advantages for future survey work.

Supplementary Materials: The following are available online at www.mdpi.com/2073-4441/9/5/362, Figure S1: Conversion of field and lab-measured turbidity values (nephelometric turbidity units (NTU)) to suspended sediment concentration (SSC) in g/L; Figure S2. Wind direction (red) and wind speed (black) data subset from June 2014. General trends include highest wind speeds (>5 m/s) in the afternoon, starting at 12 p.m. During the day, wind direction (~130◦) comes via valley winds from the southeast. At night, wind direction (~250–300◦), and speed (<2 m/s) switch to weaker northwesterly mountain downdrafts; Figure S3. SW sub-basin temperature (A) and suspended sediment concentration (B) with depth. Shallow depth leads to much higher overall temperatures from top to bottom. Similar to the other sub-basins, SSC increases with depth, indicating a stable density structure; Figure S4. Seasonal temperature variations in NW sub-basin buoy location for the monsoon (A), fall (B), winter (C) and spring (D). Surface in blue; middle in green; bottom in red; Figure S5. Seasonal temperature variations in NE sub-basin buoy location for the monsoon (A), fall (B), winter (C) and spring (D). Surface in blue; middle in green; bottom in red; Figure S6. Time-lapse sequence of the Main sub-basin thaw from 11 May–19 May 2014; Figure S7. Derived 3D models of the sub-basins, with 2–3 × vertical exaggeration. NW sub-basin (A); NE sub-basin (B); Main sub-basin (C) and SW sub-basin with surroundings (D); Figure S8. NW sub-basin roughness and hardness maps, generated from E1 and peak Sv acoustic backscatter values. Smooth (A), rough (B), soft (C) and hard (D) are shown; Figure S9. Zoom-in of SW sub-basin roughness and hardness maps, generated from E1 and peak Sv acoustic backscatter values. Smooth (A) and soft (B) are shown, showing good overlap and consistent with high sedimentation rates in this region; Figure S10. Frozen-over upwelling (December 2013) in the northeastern part of Main basin, Spillway Lake; Table S1: Six samples (one from each sub-basin and two extra within the Main sub-basin) and the average values for thermal conductivity (k), thermal resistivity (rho), volumetric heat capacity (C) and thermal diffusivity (D), with error. Acknowledgments: The author wishes to thank the U.S. Fulbright Program; a USAID HiMAP climber-scientist grant; The Geological Society of America; the National Snow and Ice Data Center (Richard Armstrong); and crowd fundraising supporters, namely Adrian Sheremeta; Arthur Few; Benjamin Diedrich; Jeff Daulton; and John Cassese (www.petridish.org (no longer in operation)) for providing the funds to complete this research. The author is indebted to Patrick Rowe (Midwest ROV LLC) and the Milwaukee School of Engineering; Roger Bilham (University of Colorado Boulder); and David Breashears (GlacierWorks) for their technical and field assistance. Thanks to Emma Marcucci, Daniel Zietlow, Cecil Goodson, Sam Ecenia and David Byrne for assistance with collection of water and mud samples; deployment of temperature buoys and weather stations; and assistance with rowing of a hand-powered kayak for initial open-water sonar transect data collection. Thank you to Lhakpa Nuru Sherpa for hospitality at the Cho La Pass Resort in Tangnak, Nepal during the field work; Amrit Thapa (Epic Himalaya) for logistical support; Benjamin Pothier (Paris, France) for filming on-location; The Crowd & The Cloud PBS citizen science series for featuring our work with the Sherpa-Scientist Initiative; and laboratory facilities at the Colorado School of Mines and University of Colorado Boulder (Fred Luiszer). Helpful conversations with Doug Benn and Sarah Thompson proved invaluable for the completion of this study. Conflicts of Interest: The authors declare no conflict of interest.

References

1. Yamada, T. Glacier Lake and Its Outburst Flood in the Nepal Himalaya. Monograph No. 1. Data Centre for Glacier Research; Japanese Society of Snow and Ice: Tokyo, Japan, 1998; p. 96. 2. Komori, J. Recent expansion of glacial lakes in Bhutan Himalayas. Quat. Int. 2008, 184, 177–186. [CrossRef] 3. Sakai, A.; Nishimura, K.; Kadota, T.; Takeuchi, T. Onset of calving at supraglacial lakes on debris-covered glaciers of the Nepal Himalaya. J. Glaciol. 2009, 55, 909–917. [CrossRef] Water 2017, 9, 362 20 of 21

4. Thompson, S.; Benn, D.I.; Dennis, K.; Luckman, A. A rapidly growing moraine dammed glacial lake on Ngozumpa Glacier, Nepal. Geomorphology 2012, 75, 266–280. [CrossRef] 5. Shrestha, A.B.; Wake, C.P.; Mayewski, P.A.; Dibb, J.E. Maximum temperature trends in the Himalaya and vicinity: An analysis based on temperature records from Nepal for the period 1971–94. J. Clim. 1999, 12, 2775–2786. [CrossRef] 6. Kadota, T. Report for the Field Investigation on the Tsho Rolpa Glacier Lake, Rolwaling Valley, February 1993–June 1994; WECS: Kathmandu, Nepal, 1994. 7. Richardson, S.D.; Reynolds, J.M. An overview of glacial hazards in the Himalayas. Quat. Int. 2000, 65, 31–47. [CrossRef] 8. Reynolds, J.M. On the Formation of Supraglacial Lakes on Debris-Covered Glaciers. In Proceedings of Symposium at Seattle 2000—Debris Covered Glaciers, Seattle, WA, USA, September 2000; International Association of Hydrological Sciences Publication: Seattle, WA, USA, 2000; Volume 264, pp. 153–161. 9. Quincey, D.J.; Richardson, S.D.; Luckman, A.; Lucas, R.M.; Reynolds, J.M.; Hambrey, M.J.; Glasser, N.F. Early recognition of glacial lake hazards in the Himalaya using remote sensing datasets. Glob. Planet. Chang. 2007, 56, 137–152. [CrossRef] 10. Chikita, K.A.; Jha, J.; Yamada, T. Hydrodynamics of a supraglacial lake and its effect on the basin expansion: Tsho Rolpa, Rolwaling Valley, Nepal Himalaya. Arct. Antarct. Alp. Res. 1999, 31, 58–70. [CrossRef] 11. Chikita, K.A.; Jha, J.; Yamada, T. Sedimentary effects on the expansion of a Himalayan supraglacial lake. Glob. Planet. Chang. 2001, 28, 23–34. [CrossRef] 12. Watanabe, T.; Kameyama, S.; Sato, T. Imja glacier dead-ice melt rates and changes in a supraglacial lake, 1989–1994, Khumbu Himal, Nepal: Danger of lake drainage. Mt. Res. Dev. 1995, 15, 293–300. [CrossRef] 13. Sakai, A.; Chikita, K.; Yamada, T. Expansion of a moraine-dammed lake, Tsho Rolpa, in Rolwaling Himal, Nepal Himalaya. Limnol. Oceanogr. 2000, 45, 1401–1408. [CrossRef] 14. Fujita, K.; Sakai, A.; Nuimura, T.; Yamaguchi, S.; Sharma, R.R. Recent changes in Imja glacial lake and its damming moraine in the Nepal Himalaya revealed by in situ surveys and multi-temporal ASTER imagery. Environ. Res. Lett. 2009, 4, 045205. [CrossRef] 15. Hambrey, M.J.; Quincey, D.J.; Glasser, N.F.; Reynolds, J.M.; Richardson, S.D.; Clemmens, S. Sedimentological, geomorphological and dynamic context of debris mantled glaciers, Mount Everest region, Nepal. Quat. Sci. Rev. 2008, 27, 2361–2389. [CrossRef] 16. Benn, D.I.; Bolch, T.; Hands, K.; Gulley, J.; Luckman, A.; Nicholson, L.I.; Quincey, D.; Thompson, S.; Toumi, R.; Wiseman, S. Response of debris-covered glaciers in the Mt. Everest region to recent warming, and implications for outburst ﬂood hazards. Earth-Sci. Rev. 2012, 114, 156–174. 17. Benn, D.I.; Wiseman, S.; Hands, K.A. Growth and drainage of supraglacial lakes on the debris-mantled Ngozumpa Glacier, Khumbu Himal, Nepal. J. Glaciol. 2001, 47, 626–638. [CrossRef] 18. Thompson, S.; Benn, D.I.; Mertes, J.; Luckman, A. Stagnation and mass loss on a Himalayan debris-covered glacier: Processes, patterns and rates. J. Glaciol. 2016, 62, 467–485. [CrossRef] 19. Nicholson, L.A.; Benn, D.I. Calculating ice melt beneath a debris layer using meteorological data. J. Glaciol. 2006, 52, 463–470. [CrossRef] 20. Chikita, K.A.; Yamada, T. The expansion of Himalayan glacial lakes due to global warming: Field observations and numerical simulation. In Regional Hydrological Impacts of Climatic Change – Impact Assessment and Decision Making in Seventh IAHS Scientiﬁc Assembly; IAHS Publication 295: Foz do Iguacu, Brazil, 2005. 21. Preisendorfer, R.W. Secchi disk science: Visual optics of natural waters. Limnol. Oceanogr. 1986, 31, 909–926. [CrossRef] 22. Chivers, R.C.; Emerson, N.; Burns, D.R. New acoustic processing for underway surveying. Hydrogr. J. 1990, 56, 8–17. 23. Kloser, R.J.; Bax, N.J.; Ryan, T.; Williams, A.; Barker, B.A. Remote sensing of seabed types in the Australian Southeast Fishery; development and application of normal-incident acoustic techniques and associated ground truthing. Mar. Freshw. Res. 2001, 52, 475–489. [CrossRef] 24. Smith, A.H.; Lingas, E.O.; Rahman, M. Contamination of drinking water by arsenic in Bangladesh: A public health emergency. Bull. World Health Organ. 2000, 78, 1093–1103. [PubMed] Water 2017, 9, 362 21 of 21

25. Sakai, A.; Takeuchi, N.; Fujita, K.; Nakawo, M. Role of Supraglacial Ponds in the Ablation Process of a Debris-Covered Glacier in the Nepal Himalayas. In Proceedings of Symposium at Seattle 2000—Debris Covered Glaciers, Seattle, WA, USA, September 2000; International Association of Hydrological Sciences: Seattle, WA, USA, 2000; Volume 264, pp. 119–130. 26. Kraus, H. An Energy Balance Model for Ablation in Mountainous Areas. In Proceedings of Snow and Ice Symposium, Moscow, Russia, August 1971; IAHS Publication: Moscow, Russia, 1975; pp. 74–82. 27. Nakawo, M.; Young, G.J. Field experiments to determine the effect of a debris layer on ablation of glacier ice. Ann. Glaciol. 1981, 2, 85–91. [CrossRef] 28. Röhl, K. Characteristics and evolution of supraglacial ponds on debris-covered Tasman Glacier, New Zealand. J. Glaciol. 2008, 54, 867–880. [CrossRef]

© 2017 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).