Glacial lake outburst floods (GLOFs) pose a significant threat to downstream communities and infrastructure due to their potential to rapidly unleash stored lake water. The most common triggers of these GLOFs are mass movement entering the lake and/or the self-destruction of the terminal moraine due to hydrostatic pressures or a buried ice core. This study initially uses previous qualitative and quantitative assessments to understand the hazards associated with eight glacial lakes in the Nepal Himalaya that are widely considered to be highly dangerous. The previous assessments yield conflicting classifications with respect to each glacial lake, which spurred the development of a new holistic, reproducible, and objective approach based solely on remotely sensed data. This remote hazard assessment analyzes mass movement entering the lake, the stability of the moraine, and lake growth in conjunction with a geometric GLOF to determine the downstream impacts such that the present and future risk associated with each glacial lake may be quantified. The new approach is developed within a hazard, risk, and management action framework with the aim that this remote assessment may guide future field campaigns, modeling efforts, and ultimately risk-mitigation strategies. The remote assessment was found to provide valuable information regarding the hazards faced by each glacial lake and results were discussed within the context of the current state of knowledge to help guide future efforts.
Glacial lake outburst floods (GLOFs) unleash stored lake water, often causing enormous devastation downstream that can include high death tolls as well as the destruction of valuable lands and costly infrastructure. As the number and area of glacial lakes continue to increase (Bajracharya and Mool, 2009; Bolch et al., 2011; Gardelle et al., 2011; Carrivick and Tweed, 2013), assessing the risk associated with these potential GLOFs becomes increasingly important. This is especially true in the Himalaya, where a global assessment of the socio-economic impacts resulting from glacier outburst floods found that Nepal and Bhutan had the greatest economic consequences at the national level despite experiencing fewer floods than other parts of the world (Carrivick and Tweed, 2016). An assessment of previous GLOFs reveals the most common cause of failure in the Himalaya is mass movement (snow, ice, and/or rock) entering the lake (Richardson and Reynolds, 2000; Wang et al., 2012; Emmer and Cochachin, 2013) and subsequently overtopping and eroding the damming moraine. Other triggering mechanisms may include dam settlement and/or piping, the degradation of an ice-cored moraine, the rapid input of water from extreme events, and seismic events (Westoby et al., 2014). Events that occur over a relatively short time period, i.e., minutes to days, such as mass movement entering the lake, intensive rainfall, or intensive snowmelt, are referred to as dynamic events, while other events that occur over longer periods of time are referred to as self-destruction or long-term causes (Yamada, 1998; Emmer and Cochachin, 2013). Emmer and Cochachin (2013) highlight the complexity of these self-destructive events as they lump the failure from the degradation of buried ice, hydrostatic pressure, and/or the effects of time all together due to the difficulty of distinguishing the exact cause of failure. Similarly, Richardson and Reynolds (2000) were unable to identify the cause of over 23 % of the GLOFs in their study due to a lack of information. This uncertainty regarding other mechanisms of failure stresses the importance of taking a holistic approach towards assessing the hazard of these glacial lakes that accounts for the various triggering mechanisms and the stability of the moraine.
Methods have been developed to characterize the hazard and risk associated with glacial lakes in Cordillera Blanca (Reynolds, 2003; Hegglin and Huggel, 2008; Emmer and Vilímek, 2013, 2014), New Zealand (Allen et al., 2009), North America (Clague and Evans, 2000; O'Connor et al., 2001; McKillop and Clague, 2007a, b), the Swiss Alps (Huggel et al., 2004b; Nussbaumer et al., 2014), the Himalaya (Wang et al., 2008, 2012; ICIMOD, 2011; Fujita et al., 2013; Worni et al., 2013), the Tibetan Plateau (Wang et al., 2011), and other parts of high-mountain Asia (Bolch et al., 2011; Mergili and Schneider, 2011). These methods vary considerably based on the parameters considered, the level of importance placed upon each parameter, the amount and type of required input data, their ability to be transferred to other regions, and their levels of objectivity. Emmer and Vilímek (2013) applied a suite of existing hazard assessments to glacial lakes in Cordillera Blanca and found good agreement between the various methods despite the various studies using different parameters, various amounts of qualitative and quantitative information, and them being developed for specific regions. This good agreement suggests it may be feasible to accurately classify the hazard of a glacial lake and shows the parameters may be qualitative or quantitative as long as the developed analysis and thresholds are objective.
Unfortunately, the use of various approaches can also lead to different classifications of the hazard associated with an individual glacial lake. Imja Tsho (“Tsho” is Tibetan for lake), located in the Everest region of Nepal, provides an excellent example of these conflicting classifications. Some studies have stated that Imja Tsho is safe (Watanabe et al., 2009; Fujita et al., 2013), a very low risk (Hambrey et al., 2008), or a moderate risk (Budhathoki et al., 2010). Conversely, ICIMOD (2011) identified Imja Tsho as one of six glacial lakes in Nepal that is a high priority for further investigation, and currently remediation efforts to reduce its hazard are underway (UNDP, 2013). These remediation efforts and a similar project completed at Tsho Rolpa (Rana et al., 2000) are excellent steps forward for Nepal with respect to addressing the hazards associated with its glacial lakes. However, it is important that these efforts and resources are properly directed to ensure the development of the most cost-effective, Nepal-specific methods for successfully reducing the risk of glacial lakes as practiced by the Peruvians since the 1950s (Carey, 2010). Conflicting classifications of Imja Tsho suggest relief efforts may be misguided if the lake is indeed safe and also send mixed signals to the general public and downstream communities that these studies are meant to assist. Another good example of these conflicting reports is Chamlang South Tsho (also referred to as West Chamlang), where one study stated the lake was not particularly dangerous (Byers et al., 2013) and another stated the lake was potentially dangerous (Lamsal et al., 2016), despite having access to the same data sets.
These various assessments and conclusions highlight the importance of (1) developing a clear and straightforward holistic method for assessing the hazard associated with a glacial lake and (2) constructing a framework that guides management actions for a given glacial lake that accounts for the various levels of data collection, modeling, and field campaigns. An additional aspect that is important to incorporate is how the hazard associated with these glacial lakes will change as the glacier and glacial lakes continue to evolve. The negative mass balance regime across the central Himalaya (Kääb et al., 2012) will affect the continued growth of existing lakes and development of new lakes in overdeepened basins (Linsbauer et al., 2016). Anticipating this future development is especially important when one considers that successful remediation strategies on hazardous lakes often take more than a decade to secure funding and implement (Quincey et al., 2007).
This study assesses the performance of existing hazard assessment methods on potentially dangerous glacial lakes in Nepal to determine how well they agree with one another. These existing studies are used as the basis for developing a new holistic approach that is objective, repeatable, and uses readily available information such that it may be applied to any glacial lake. This approach will be developed within a management action framework such that it may be used to guide future field campaigns, research, and ultimately risk-mitigation strategies associated with hazardous glacial lakes.
This study assesses the hazard and risk of GLOFs for eight proglacial lakes in the Nepal Himalaya (Fig. 1). Six of these glacial lakes were identified by ICIMOD (2011) as being high priority for further investigation: Tsho Rolpa, Lower Barun Tsho, Imja Tsho, Lumding Tsho (also referred to as Tsho Og), Chamlang South Tsho (also referred to as West Chamlang), and Thulagi Tsho. The other two glacial lakes considered are Chamlang North Tsho (also referred to as Lake 464), which was found to be dangerous (Byers et al., 2013), and Dig Tsho, which experienced a GLOF in 1985 (Vuichard and Zimmermann, 1987). The level of investigation of each of lake is highly variable.
The three glacial lakes that have received the most attention are Imja Tsho
(86
Dig Tsho (86
The first method that was used to assess the hazard of these eight glacial lakes was a “shotgun approach” to determine how their classifications vary using previous qualitative and quantitative assessment methods. The shotgun approach uses the same studies as Emmer and Vilímek (2013) with the exception of Clague and Evans (2000) and Grabs and Hanisch (1993) as these required site-specific knowledge that was not possible to obtain from remote sensing. The use of solely remotely sensed data is one of the main goals of the new method and framework developed in this study. The qualitative methods used were O'Connor et al. (2001), Costa and Schuster (1988), and Wang et al. (2008). Wang et al. (2008) highlights 10 hazard parameters, but only gives thresholds for 8 of them; therefore, the 8 parameters with thresholds are used for this qualitative approach. As a specific hazard rating is unable to be determined from these qualitative approaches, the arithmetic mean of the three is used to rank the glacial lakes.
The locations of the eight glacial lakes assessed in this study in the Nepal Himalaya.
Previous qualitative hazard assessments applied to eight glacial lakes in Nepal. Fractions are the number of hazard parameters identified out of the total number of parameters used by each method. Details shown in Tables S.1–S.3.
Previous semi-quantitative and quantitative hazard assessments applied to eight glacial lakes in Nepal. Values and thresholds for classifications are specific to each study. Fractions are the number of scenarios for each lake that are considered to be highly dangerous. Details shown in Tables S.4–S.6.
The semi-quantitative method used in the shotgun approach was from Bolch et al. (2011), which was developed for glacial lakes in the Tien Shan using remotely sensed data. Bolch et al. (2011) use a term called lake area change based on a comparison of lake area to its initial area; however, the initial area of a lake is unclear, as they were all small melt ponds at one point in time. Therefore, the lake area change was simplified to give a value of 1 or 0 based on whether the lake has grown in the last decade or not, respectively. The quantitative approach of Wang et al. (2011) also used only remotely sensed data with specific thresholds determined from a statistical analysis of 78 lakes in the southeastern Tibetan Plateau. Wang et al. (2011) use the mean slope of the moraine based on a 100 m buffer around the lake; however, when this was applied to the glacial lakes in this study, the mean slope was zero or negative, indicating that within the first 100 m of the lake, the moraine is higher than the lake level. Therefore, the steep lakefront area (SLA) from Fujita et al. (2013) was used as a surrogate for the mean slope of the moraine so the method could still be applied and account for the slope of the moraine. A more detailed description of these studies may be found in Emmer and Vilímek (2013). Additionally, a new quantitative approach developed by Emmer and Vilímek (2014) for the Cordillera Blanca was used in this study. The approach assesses five GLOF scenarios: (1) moraine overtopping due to mass movement entering the lake, (2) overtopping from a flood upstream, (3) moraine failure due to mass movement entering the lake, (4) moraine failure from a flood upstream, and (5) failure resulting from a strong earthquake. Unfortunately, the locations of seepage points required for the last scenario were not identifiable from remote sensing, so the dam instability due to piping was not accounted for.
The qualitative hazard assessments show a good deal of variation between the three approaches (Table 1). The arithmetic mean reveals the most dangerous lakes based on these three qualitative approaches are Chamlang North Tsho, Chamlang South Tsho, and Tsho Rolpa, closely followed by the others with the exception of Imja Tsho and Thulagi Tsho. Imja Tsho has a lower value since these methods all use some form of mass movement entering the lake, which currently is not a threat at Imja Tsho. Thulagi Tsho is also not susceptible to ice avalanches and by some methods has a more stable moraine due to the presence of vegetation. The semi-quantitative and quantitative assessments give very different classifications of the hazard associated with each lake (Table 2). Bolch et al. (2011) emphasize the size of the lake and its ability to expand, so all the large glacial lakes that have expanded over the last decade are ranked as high danger (Imja Tsho, Lower Barun Tsho, Lumding Tsho, Thulagi Tsho, and Tsho Rolpa). The other glacial lakes that have already reached their maximum extent are classified as medium danger. Wang et al. (2011) is the complete opposite since the glacial lakes that have already reached their maximum extent are all high risk or very high risk, while the others are medium or low risk. This conflicting classification is due to the emphasis on parameters associated with mass movement entering the lake, i.e., both the distance and the slope between the lake and the glacier. The large glacial lakes that are still expanding have gentle slopes behind their calving fronts, which cause them to be classified as medium or low.
Emmer and Vilímek (2014) classify all the glacial lakes as highly dangerous as they are all susceptible to at least one GLOF scenario. Table 2 shows how many of the five scenarios are a potential threat to each glacial lake. The reason for this classification is that the parameters associated with the mass movement and overtopping scenario are dam freeboard and distance between the glacier and the lake. The eight lakes considered in this study all have outlet channels, so their freeboard by definition is zero. Furthermore, these eight glacial lakes are either in contact with or within 600 m of their mother glaciers, which results in this methodology considering them all to be susceptible to this dynamic failure. As Emmer and Vilímek (2014) discussed, the method was developed for scenarios related to Cordillera Blanca such that the framework could be transferred to other regions, but the exact scenarios or parameters used may not be representative of the main threats to other regions. Nonetheless, the assessment yields valuable insight showing Lumding Tsho, Lower Barun Tsho, and Tsho Rolpa have the greatest number of hazard scenarios due to the potential for a flood from a glacial lake upstream, which is important to consider.
Most frequently used parameters associated with previous studies (adapted from Emmer and Vilímek, 2013).
The shotgun approach shows that the hazard classification of each glacial lake varies greatly depending upon the selected method, which makes classifying the hazard associated with a particular glacial lake difficult. Fortunately, the shotgun approach is useful as it highlights the most commonly used parameters in previous studies. Table 3 shows the most frequently used parameters are mass movement entering the lake, the moraine width-to-height ratio, the presence of buried ice in the moraine, and the distance between the lake and the glacier. These naturally reflect the most common causes of GLOFs, i.e., mass movement entering the lake and the self-destruction of the moraine due to hydrostatic pressure and/or the degradation of buried ice (Richardson and Reynolds, 2000; Wang et al., 2012; Emmer and Cochachin, 2013). Furthermore, many of the other parameters are simply alternative forms of estimating the potential cause of failure; e.g., glacier snout steepness or the distance between the lake and the glacier are surrogate ways to estimate whether the lake is susceptible to an avalanche entering the lake. In this manner, the shotgun approach lends insight into the various parameters or methods that were used to estimate different triggering events. The variety of parameters and the frequency of their use highlight the parameters that are important to consider in addition to highlighting the importance of developing a holistic method that accounts for these various forms of failure.
Schematic of GLOF hazard parameters used in the new method: (1) snow/ice avalanche, (2) rockfall, (3) flood from upstream lake, (4) lake expansion, (5) hydrostatic pressure, (6) ice-cored moraine, and (7) downstream impact.
The conflicting hazard classifications from the shotgun approach cast uncertainty on the hazard of each glacial lake that can be confusing and misleading to the stakeholders these studies are meant to assist. Furthermore, they cast uncertainty on which glacial lakes should receive more attention through field campaigns and/or detailed analyses and the specific parameters that should be focused on. This study develops a new hazard and risk assessment framework that is holistic, objective, reproducible, and initially relies solely on remotely sensed data. Specifically, this framework focuses on two forms of glacial lake failure: dynamic and self-destructive. The term “self-destructive” failure (Yamada, 1998) is used here to avoid any confusion with long-term failures resulting from dynamic causes and lake growth; i.e., as a lake grows its expansion may make it susceptible to mass movement entering the lake from areas that could not previously reach the lake. Figure 2 shows the seven parameters used in this study are potential mass movement entering the lake from (1) a snow/ice avalanche, (2) a rockfall, or (3) an upstream flood, (4) the future expansion of the glacial lake, the stability of the moraine based on (5) the hydrostatic pressure and (6) the presence of buried ice, and (7) the downstream impact. These parameters are all estimated using simplistic models and globally available data sets. The approach is referred to as a “remote” hazard assessment. The integration of site-specific field data, high-resolution data sets, and more complex models to improve upon this remote assessment will be detailed in the discussion in conjunction with a brief description of the current state of knowledge for each glacial lake investigated.
The remote hazard assessment framework is intended to be used as a launching
point for assessing the hazards of glacial lakes. The parameters and models
used in this framework are all derived from globally available data sets. The
DEM utilized in this study is the ASTER GDEM, which is composed of
automatically generated DEMs from the Advanced Spaceborne Emission and
Reflection radiometer (ASTER) stereo scenes acquired from 2000 to 2008. The
ASTER GDEM V2 (hereon referred to as GDEM) has a horizontal resolution of
30 m and a vertical RMSE of
Mass movement entering a glacial lake is a highly hazardous situation for
glacial lakes as the entry may cause a tsunami-like displacement wave that
can trigger a GLOF. This study considers three types of dynamic failures:
snow/ice avalanches, rockfalls/landslides, and upstream GLOFs (Fig. 2). This
section describes the mass movement trajectories for avalanches and
rockfalls, while the upstream GLOFs are discussed later (Sect. 4.1.5).
Landsat imagery was used to automatically detect glacierized and
non-glacierized areas using a ratio of the NIR and SWIR 1 bands with a
threshold of 2.2 (Huggel et al., 2004a). In this simplified model, there is
no differentiation between snow and ice. Snow/ice avalanche-prone areas were
identified as any glacierized area with a slope greater than 45
The runout distance of the mass movement trajectories was computed using an
average slope threshold of 17 and 20
Lake growth is crucial to incorporate into hazard assessments as the expansion of a glacial lake may greatly alter the lake's proximity to potential hazards and increase the volume of water likely released in a GLOF. Mass movement entering the lake is the most common cause of a GLOF, so one must determine whether dynamic failure is both a current and/or future threat. Multi-spectral satellite imagery can be used to determine lake expansion rates semi-automatically using the normalized difference water index (NDWI) (McFeeters, 1996), which is a combination of the near-infrared (NIR) and blue bands. In the event that the blue band is not available or the contrast is not clear, the green (Bolch et al., 2008) and/or shortwave infrared (SWIR) bands (Somos-Valenzuela et al., 2014) may be used as a suitable alternative. Bolch et al. (2008) found the NDWI method yielded accurate estimates of lake area compared to manual delineations performed by Bajracharya et al. (2007).
One difficulty associated with the NDWI analysis is the objective selection of the threshold used to differentiate land and water. Bolch et al. (2011) found the threshold for Landsat images to range from 0.3 to 0.9 for glacial lakes in northern Tien Shan, but no clear instructions exist for selecting the threshold for each image and glacial lake. Thakuri et al. (2015) used the same technique at Imja Tsho and found the lake area to be constant between July and January each year. They suggested this was due to the lake level being constant, but measurements of lake level were not included. This study uses the same approach with Landsat imagery from 2000 to 2015 captured between September and January each year and assumes the width of the lake between the lateral moraines is constant based on the findings from Thakuri et al. (2015) and the assumption of a constant lake level. Two exceptions were made, one for Chamlang North Tsho, which used an image from May as there were fewer shadows during this time of year, and another for the supplementary image of Lower Barun Tsho in 2008, which used an image from the following April. Additionally, no clear-sky, non-banded Landsat imagery was available in 2003, 2004, and 2008 for Lower Barun Tsho, so Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) images between September and January were used for these years instead.
Thresholds were objectively selected such that the average width of the lake
was constant between images. This method forces any changes in lake area to
be the result of upglacier and/or downglacier expansion, which is the focus
of this study. In the event that clouds or the Landsat 7 stripping caused
portions of the lake to have gaps in the imagery, a second Landsat image was
used to fill in these missing areas using the same criteria. Additionally,
large debris-covered icebergs nearing the calving front that could cause
pixels to be misclassified as land instead of water were manually corrected
in post-processing. The maximum error in the lake area estimates was
calculated as the perimeter of the lake multiplied by half the pixel
resolution (
The expansion rate was estimated as the average rate of areal expansion for all the 10-year intervals available; i.e., if yearly delineations were available from 2000 to 2015, then the six values of the 10-year expansion were averaged to estimate the areal expansion rate. These expansion rates were used in conjunction with ice thickness estimates to determine future lake extents. The ice thickness data used in this study are from the GlabTop2 model (Frey et al., 2014), which estimates the ice thickness from a DEM and glacier outlines. The ice thickness of the glaciers in this study was upwards of 250 m thick behind the calving front of some lakes. The bed topography of the glaciers was computed as the ice thickness subtracted from the surface elevation, which was used to identify potential overdeepenings, i.e., locations in the bed topography that are sinks and allow the lake to expand.
Future lake projections were estimated using the average decadal areal expansion rates over the next 50 years such that the lake's risk of future dynamic failures could be assessed. Glacier flowlines were used to guide the direction of the expansion. The lake level was assumed to be constant over the next 50 years and was estimated as the average elevation of all the lake pixels based on the lake extent from 2000, as this elevation should be relatively constant over the time period when DEMs for the GDEM were being captured.
The exact cause of failure associated with a moraine that spontaneously
fails without any external influence is difficult to pinpoint, but is
commonly referred to as “self-destruction” (Yamada, 1998; Emmer and
Cochachin, 2013). One cause of self-destruction is when the hydrostatic
pressure, the pressure a column of water exerts on the moraine, exceeds the
structural capacity of the moraine. Many studies account for the moraine
stability using a ratio of the moraine width-to-height (Table 2), which is
subject to large errors if a global DEM is used (Fujita et al., 2008).
Fujita et al. (2013) developed a surrogate parameter known as the steep
lakefront area (SLA) using remotely sensed data that is not as susceptible
to the uncertainty associated with global DEMs. The SLA is the average slope
between the lake and any point within 1000 m of the moraine, which is
similar in concept to the mean slope of the moraine based on a 100 m buffer
that was used by Wang et al. (2011). The slope of any point within 1000 m of
the moraine is meant to capture the steepest slope between the lake and the
base of the terminal moraine. Fujita et al. (2013) examined moraines that
had previously failed and determined that lakes with a SLA less than
10
This study uses the SLA with a threshold of 10
The other main cause of failure associated with “self-destruction” is the melting of ice within a lake's terminal moraine, since a disintegrating ice core can undermine the structural integrity of the moraine (Richardson and Reynolds, 2000). This can have large implications for the hydrostatic pressure, piping/seepage, and reducing the height of the freeboard associated with the terminal moraine (Emmer and Cochachin, 2013). The importance of accounting for ice-cored moraines is apparent from previous studies (Table 2), but requires detailed information regarding the terminal moraine that is typically not available from remotely sensed data. Bolch et al. (2011) used permafrost as a surrogate parameter to suggest the potential of an ice-cored moraine. A similar approach was assessed in this study using permafrost maps (Gruber, 2012); however, a comparison between lakes with known ice cores and the permafrost maps revealed no correlation. McKillop and Clague (2007a) assessed the presence of an ice core according to the shape of the moraine using aerial imagery by assuming that a moraine with a rounded surface with arcuate ridges had an ice core, that a disproportionately large terminal moraine in front of a glacier was potentially ice-cored, and a narrow, sharp-crested moraine with an angular cross section was ice-free. Unfortunately, this approach is highly subjective and appears to fail for glacial lakes with ice cores in Nepal. For example, Tsho Rolpa has a narrow terminal and lateral moraine that would suggest its moraine is ice-free; however, it is known to have an ice core (Yamada, 1998; Richardson and Reynolds, 2000; ICIMOD, 2011).
In the field a common approach to determine whether a moraine is ice-cored is by observing ice cliffs or karst topography (Yamada, 1998; Richardson and Reynolds, 2000; ICIMOD, 2011). Another common way is to witness changes in the outlet channel over time, which has been observed at Imja Tsho (Watanabe, 1994), Tsho Rolpa (Yamada, 1998), and Thulagi Tsho (ICIMOD, 2011). This study takes a similar approach using satellite imagery and Google Earth to identify the presence of any water on the moraine or any changes in the outlet. If water is present on the moraine or any changes in the outlet are observed, the moraine is assumed to have an ice core. During the analysis, a combination of Google Earth and satellite imagery was found to help differentiate between shadows and water, but it is recommended to err on the side of caution when one is unsure. Unfortunately, the lack of clear thresholds for identifying water on the moraine or changes in the outlet adds a small amount of subjectiveness to this study. However, as this was the most effective and least subjective approach to identifying the presence of buried ice in the moraine, it was used for the remote assessment.
Flood models play a crucial role in a glacial lake hazard assessment as they identify areas at risk, which allows one to determine the downstream impact. Westoby et al. (2014) provides a thorough overview of the various types of floods and types of models that may be used to reconstruct a GLOF. In short, if the flood entrains enough sediment from the moraine and channel downstream, the flood may transform into a debris flow, which increases the momentum of the flood, thereby increasing the GLOF's extent and potential damage. The models used to reconstruct these debris flows or GLOFs range from simple computationally inexpensive GIS-based methods to computationally expensive, physically based numerical models. The GIS-based methods typically rely solely on the geometry of the downstream channel from a DEM. Numerical models have been used to resolve the flow of a GLOF in one or two dimensions. The benefit of two-dimensional models is their ability to capture more complex features and flow characteristics, e.g., multi-directional flows or super-elevation of flow around a channel bed (Westoby et al., 2014). The selection of a particular model or method typically depends on data availability and the desired model complexity.
This study explored the use of two computationally inexpensive flood models:
the Modified Single Flow direction (MSF) model developed by Huggel et
al. (2003) and the Monte Carlo Least Cost Path (MC-LCP) model developed by
Watson et al. (2015). The MSF model is a standard flow direction model that
“allows the flow to divert from the steepest descent direction up to
45
The lack of data related to previous GLOFs makes it difficult to assess the performance of different models. Watson et al. (2015) used the 1985 GLOF at Dig Tsho to compare the performance of the MSF and MC-LCP models to the actual flood extents. These models used the GDEM resampled to 15 m and were found to perform reasonably well, although the MC-LCP model had a larger inundation area and fewer artefacts. Watson et al. (2015) also observed that in high relief Himalayan catchments, the requirements for an artificially filled DEM by the MSF model created large linear inundation artefacts, whereas the MC-LCP model displayed improved flow routing and hence is more appropriate for assessing first-order socio-economic impacts of a potential GLOF. It is important to note that the GDEM reflects the post-GLOF terrain, which was severely altered by the GLOF (Vuichard and Zimmermann, 1987). Ideally, a pre-GLOF DEM would be used for model validation so modeled flood extents would not be affected by scouring and deposition in the main channel. Furthermore, the comparison reveals multiple areas where the MC-LCP model does not capture the actual flood extent, which would be highly problematic for a hazard assessment if these areas were populated.
As both the MSF and MC-LCP models have no physical basis, model selection was determined by the one that yielded the most reasonable, conservative estimate of inundation areas when compared to a two-dimensional debris-flow model, FLO-2D, from Imja Tsho (Somos-Valenzuela et al., 2015). Figure 3 shows the flood extent for the FLO-2D, MSF, and MC-LCP models along with the performance of the MC-LCP model for various DEMs (GDEM and SRTM) and resolutions (resampled to 15 or 30 m). The comparison revealed the MSF model (Fig. 3b) and the MC-LCP model resampled to a 15 m resolution for the SRTM DEM (Fig. 3c) and the GDEM (Fig. 3d) severely underestimated the flood extent. On the other hand, the MC-LCP model with the 90 m SRTM DEM yielded too conservative an estimate (Fig. 3g). The 30 m results for both the SRTM DEM (Fig. 3e) and GDEM (Fig. 3f) agreed well with the FLO2D results (Fig. 3a); however, a more detailed analysis revealed the GDEM tracked the main channel better. Therefore, the MC-LCP model with the 30m GDEM was used in this study to model the potential GLOFs from each lake. The GLOF from each lake was routed to the confluence of the Sun Kosi for seven of the eight glacial lakes, and approximately 60 km downstream for Thulagi Tsho based on the assumption that beyond this distance the downstream effects are minimal as the river is able to absorb the flood's energy (Vuichard and Zimmermann, 1987).
Inundated areas at Dingboche for a GLOF from Imja Tsho using
Descriptions of downstream impact classifications.
Buildings and agricultural land use data were extracted from the inundation extent of each MC-LCP GLOF scenario to provide a first-pass assessment of socio-economic implications. Buildings were downloaded from OpenStreetMap and validated in Google Earth, which was also used to update the data set, when necessary, using the most recent imagery. Areas of agricultural land were manually digitized in Google Earth and included all visibly managed land, i.e., land that appeared cleared, walled, farmed, or grazed, and would likely have detrimental socio-economic implications if flooded. The potential downstream impact was broken down into four categories: very high, high, moderate, and low (Table 4). Very high impact is defined as the potential loss of life with no warning (lodges/buildings) and the loss of costly projects or infrastructure (e.g., hydropower). High impact is defined as the potential loss of life with no warning (lodges/buildings) or costly projects (e.g., hydropower). Moderate impact is defined as any damage that is disruptive, which is meant to include damage to agricultural lands, bridges, trails, etc. Lastly, low impact is defined as having no impact on humans, infrastructure, or other projects. For the purpose of this study, buildings are assumed to have permanent occupants whose lives would be threatened in the event of a GLOF. Agricultural lands are considered to be moderate impact as their occupancy changes temporally depending on the season, and in the event that people are in the fields they may be able to hear and/or see an upstream flood and have an opportunity to move to safe ground. The definition of costly projects or infrastructure is fairly subjective, but is meant to refer to any hydropower system or similar project since the loss of a mature hydropower system can effect multiple generations and jeopardize the economic development of the country (Richardson and Reynolds, 2000).
Hazard classification flow chart for determining the hazard associated with a glacial lake (numbers refer to hazard parameters in Fig. 2).
The hazard elements described above are crucial for determining whether a glacial lake is susceptible to a dynamic or self-destructive failure. Figure 4 shows the workflow that is used to determine whether the lake is susceptible to failure and how the cause of failure translates into the hazard associated with the lake. The most dangerous situation is a glacial lake that is susceptible to both dynamic and self-destructive failures, which would classify the lake as a very high hazard. Susceptibility is defined as a hazard greater than low; i.e., a lake that is considered a moderate hazard for dynamic failure and a moderate hazard for self-destructive failure is still classified as very high hazard. The other scenario that classifies a lake as very high is a lake with a buried ice core that is susceptible to a snow/ice avalanche, as the ice core may alter the height of the moraine over time and/or the erosion and breach of the moraine. A lake that is susceptible to avalanches but does not have an ice core is classified as a high hazard. Snow/ice avalanches were given the highest hazard classification since they are the most frequent cause of failure in the Himalaya (Emmer and Cochachin, 2013). Additionally, any lake with a buried ice core that is susceptible to a rockfall or upstream GLOF, or has a steep SLA, is classified as a high hazard. These hazard ratings are meant to reflect self-destructive failures being the second most common cause of GLOFs followed by mass movement entering the lake from rock or liquid water (Emmer and Cochachin, 2013). As temperatures continue to increase, thereby promoting the formation of more glacial lakes and altering slope stability due to changes in permafrost (Haeberli et al., 2016), there is a possibility that failures from rockfalls and/or upstream GLOFs may become more common than they are in Cordillera Blanca (Emmer and Cochachin, 2013). A lake that does not have an ice core, but is susceptible to a rockfall or upstream GLOF, or has a steep SLA, is considered to be moderate hazard.
Risk management and action framework. Left table: GLOF risk is a function of the hazard and downstream impact. Right table: the recommended course of action for a given glacial lake based on the type of assessment.
Summary of the hazard parameters for the studied glacial lakes.
Summary of hazard, downstream impact, and risk for each glacial lake.
* Future hazard is very high and risk is high.
The risk of a GLOF is a function of a glacial lake's hazard and its potential downstream impact (Fig. 5). This study uses the same ranking scheme as Worni et al. (2013). Very high risk lakes are defined as any lake where the downstream impact is very high; i.e., lives and costly projects are threatened, and the hazard of the lake is very high. High risk is defined as a lake that has a high downstream impact and a high hazard, a very high hazard and high/moderate downstream impact, or vice versa. Moderate risk is defined as a lake that has a moderate hazard and moderate downstream impact, a moderate/low hazard and high downstream impact or vice versa, or a low hazard and very high downstream impact or vice versa. Lastly, low risk refers to the remaining scenarios that are less of a threat and is defined as lakes that have a low downstream impact and a low hazard, or a moderate hazard and low downstream impact or vice versa. It is important to note that any site that is at risk of a dynamic or self-destructive failure is valuable from an academic perspective as it may help improve the current state of knowledge of GLOF hazards.
The remote hazard assessment builds off of existing knowledge of glacial lakes in Nepal while integrating new approaches to developing a holistic understanding of their hazard and risks. The mass movement trajectories mark the first time these potential triggers have been modeled at a larger scale, which provides valuable information on the potential for mass to enter a lake in addition to identifying the locations of these avalanche- and rockfall-prone areas that should be further investigated. The stability of a moraine utilizes a previously developed approach, i.e., the SLA (Fujita et al., 2013), in combination with predictions of the presence of buried ice from satellite imagery such that self-destructive failures may be integrated into the hazard framework with the dynamic failures. Similarly, the downstream impacts use a previously developed GLOF model, i.e., the MC-LCP approach (Watson et al., 2015), to obtain a conservative estimate of potentially inundated areas downstream. For three of the eight glacial lakes considered in this study, this is the first time their potential GLOFs have been modeled. Furthermore, this study combined lake expansion rates from satellite imagery with ice thickness estimates to model the evolution of these glacial lakes to determine how their hazard may vary over the next 50 years.
Table 5 provides a brief summary of the hazard parameters for each glacial
lake. The potential for avalanches and rockfalls should not be surprising as
these glacial lakes have developed on avalanche-fed debris-covered glaciers,
so their surrounding slopes are commonly unstable. Three of the eight glacial
lakes are threatened by a potential upstream GLOF, which makes it important
to assess the hazards associated with these upstream lakes as well. The SLA
varies between 4.9
Area of glacial lakes derived from satellite imagery using the NDWI method from 2000 to 2015 (details in Table S.7).
The hazard of these eight glacial lakes varies from low to very high with
five of the eight lakes currently being very high hazard (Table 6). Four of
the lakes classified as very high hazard are susceptible to both dynamic and
self-destructive failures. Specifically, Chamlang South Tsho and Tsho Rolpa
are both very high hazard for dynamic failure and high hazard for
self-destructive failure. The MC-LCP GLOF modeling revealed potential to
damage buildings and large swathes of agricultural lands for all eight
glacial lakes and potential damage to hydropower projects for two of the
lakes. Therefore, the downstream impact for six of the lakes was high and
very high for the other two (Lower Barun Tsho and Thulagi Tsho). The amount
of buildings, agricultural land, and bridges affected varied greatly, which
was partly due to differences in the distance the GLOF was allowed to
propagate downstream, but also due to the amount of development below each
glacial lake. Therefore, the inundated buildings per km
Hazards and downstream impact for Imja Tsho:
Figure 6 shows that Lower Barun Tsho, Imja Tsho, and Lumding Tsho continue to expand rapidly, while Tsho Rolpa and Thulagi Tsho have stagnated in recent years. The other lakes (Chamlang North Tsho, Chamlang South Tsho, and Dig Tsho) have already detached from their glaciers and lack the ability to expand. These expansion estimates combined with the mass movement trajectories reveal that Imja Tsho is susceptible to a dynamic failure in the next 10–20 years (Table 5). These results have important implications for the hazard and risk associated with Imja Tsho as they indicate that in 10–20 years Imja Tsho may be a very high hazard and high risk (Table 6). Additionally, the expansion of Lower Barun Tsho and Lumding Tsho makes them susceptible to potential avalanches located further upstream; however, this does not alter their hazard ratings as they are presently at risk as well.
Details of downstream impacts from the MC-LCP GLOF models for each glacial lake.
Imja Tsho is one of the most well-studied glacial lakes in Nepal, yet the
remote hazard assessment still yields new relevant insight. The mass movement
trajectories (Fig. 7a, b) show that the lake is currently not at risk of a
dynamic failure. These trajectories are conservative estimates of avalanche
and rockfall runout, and for a worst-case scenario they remain at least 800
and 400 m from the calving front, respectively. The ice thickness behind the
calving front is greater than 200 m thick (Somos-Valenzuela et al., 2014;
Frey et al., 2014), which allows the lake to continue to expand (Fig. 7d). A
detailed analysis of the growth of Imja Tsho (Fig. 6) shows the lake is
expanding at an average rate of 0.032
Fortunately, at the present moment the lake is unlikely to have a
self-destructive failure as its SLA is fairly gentle at 6.8
Imja Tsho is currently classified as a moderate risk due to its low hazard and high downstream impact. These high downstream impacts highlight the importance of running hydrodynamic models with high-resolution imagery to improve the mapping of inundated areas and inform the local communities (Somos-Valenzuela et al., 2014). The expansion model reveals that if Imja Tsho continues to grow at its current pace, it will be a very high hazard and high risk in the next 10–20 years. Since the expansion is highly concerning, one important area of future work should be measuring the ice thickness upglacier of Imja-Lhotse Shar Glacier using geophysical techniques such that the potential future extent of Imja Tsho may be accurately quantified. Additionally, efforts should focus on modeling the GLOF process chain, i.e., mass movement entering the lake, the wave propagation, the possible breach of the moraine due to the overtopping, and the downstream impacts due to the future risk. The ice-cored moraine has already been well characterized (Hambrey et al., 2008), but is critical to avoid during any lake lowering activities such that a breach is not initiated. Furthermore, while self-destructive failure is not an immediate concern, repeat bathymetric studies on the outlet lakes would provide valuable information regarding the evolution of the moraine to understand how the hydrostatic pressures may change over time. Based on this remote assessment, the current risk mitigation actions to lower the level of the lake (UNDP, 2013), ideally by 20 m (Somos-Valenzuela et al., 2015), are well justified and hopefully will serve as a good example of how to mitigate the risk of a glacial lake before it becomes highly hazardous.
Lumding Tsho was classified as a high priority for further investigation
(ICIMOD, 2011), but has received very little attention beyond an analysis of
its expansion (Bajracharya and Mool, 2009) and SLA (Fujita et al., 2013). The
remote assessment confirms that Lumding Tsho is a very high hazard as it is
susceptible to both dynamic and self-destructive failures. The mass movement
trajectories reveal the lake is susceptible to avalanches and rockfalls from
the southern side slope (Fig. S5a, b). Furthermore, the lake is susceptible
to a GLOF from Lumding Tsho Teng located 600 m upstream of Lumding Tsho.
Lumding Tsho Teng (27
Bathymetric survey conducted on Lumding Tsho on 22–23 October 2015.
Lumding Tsho also has a SLA (10.5
An initial rapid reconnaissance was undertaken from 20 to 24 October 2015 to
assess the hazards associated with Lumding Tsho as directed by the management
action framework (Fig. 5). The short field campaign consisted of a
bathymetric study, streamflow measurements of the outlet, and visual
inspection of the terminal moraine and surrounding slopes. The bathymetric
survey was conducted over 2 days using an inflatable kayak and a Garmin
echoMAP 54dv to measure 4768 points of lake depth. The shoreline was
delineated using the NDWI methods as previously reported. The shoreline was
converted into point measurements that were used in conjunction with the
bathymetric survey to interpolate depth throughout the lake using the Topo to
Raster tool in ArcGIS (Somos-Valenzuela et al., 2014). The lake was found to
have an average depth of 51 m, a maximum depth of 114 m, and a total volume
of
Lumding Tsho and its surrounding slopes with areas of concern
highlighted in red showing
Streamflow from the outlet of the lake was measured to be
8.4 m
Visual inspection of the side slopes revealed the slopes were very steep and likely lacked the ability to generate a large rockfall (Fig. 9a, b). On the southern slope there was one large boulder that could be a potential concern (highlighted in Fig. 9b), but its limited elevation above the lake level would likely cause only a small surge wave. The snow and ice above the southern side slopes were also very steep and no hanging glaciers were apparent. A more detailed assessment of the stability of the side slopes would be invaluable in improving the likelihood of a rockfall or avalanche. Specifically, the three hanging glaciers located behind the calving front (Fig. 9a) are potential hazards as the lake grows. Physically based mass movement models would generate important information regarding the size and trajectories of these slope failures. Unfortunately, due to time constraints, a detailed analysis of Lumding Teng Tsho, the upstream glacial lake, was unable to be conducted. Similar knowledge of the stability of the slopes surrounding Lumding Teng Tsho would inform the likelihood of an upstream GLOF. A bathymetric study on Lumding Teng Tsho and an assessment of the stability of its moraine should also be a high priority. In the event of a GLOF, the northern lateral moraine of Lumding Tsho may or may not protect the lake from the upstream flood. A physically based GLOF model that accounts for erosion would yield insight into the potential breach of the lateral moraine and subsequent GLOF from Lumding Tsho. Based on the rapid field reconnaissance, the hazard associated with Lumding Tsho can be reduced from very high to high due to the apparent stability of the moraine. However, more detailed analyses of slope stability around Lumding Tsho and Lumding Teng Tsho in addition to modeling the potential GLOFs from both of these lakes should be a top priority.
Chamlang North Tsho has already reached its fullest extent, so its proximity
to hazards is not going to change over time (Fig. S1d). The southern and
eastern slopes reveal the lake is very susceptible to rockfalls and
avalanches to the extent that any rockfall or avalanche will enter the lake
(Fig. S1a, b). This potential for a dynamic failure is exacerbated by its
steep terminal moraine, which has a SLA of 18.8
Based on the remote assessment, the most important area of future investigation should be modeling the GLOF process chain. The avalanche-prone slopes were observed by Byers et al. (2013), who identified four overhanging glaciers on Chamlang North Tsho's southern slopes. The use of high-resolution imagery may help quantify the size of a potential avalanche. Byers et al. (2013) modeled a potential GLOF using the GDEM with the U.S. Army Corps of Engineers' Hydrologic Engineering Center River Analysis System (HEC-RAS). The ensuing flood model provides an improved estimate compared to the MC-LCP model, which estimates the water level will rise by 9 m at Bung; however, the downstream impacts were not detailed due to the lack of high-resolution imagery. The acquisition of high-resolution DEMs for this region and/or cross sections at critical locations should be a high priority. This high-resolution DEM should be used with two-dimensional flood models to more accurately estimate the downstream impacts. Additionally, a geotechnical survey including sedimentological sampling of the moraine should be performed such that erosion can be properly accounted for in breach scenarios (Westoby et al., 2015) and the stability of the moraine with respect to the current hydrostatic pressures may be quantified. Lastly, a bathymetric survey of Chamlang North Tsho is needed for the high-resolution physically based GLOF modeling.
Chamlang South Tsho has very similar hazard characteristics to Chamlang North
Tsho. Chamlang South Tsho is no longer expanding as it has already reached
its fullest extent (Fig. 6). Mass movement trajectories also show that the
lake is surrounded by unstable slopes such that any rockfall or avalanche
will enter the lake (Fig. S2a, b). The SLA exceeds the stability threshold
with a value of 10.5
The remote assessment of mass movement entering the lake from the
surrounding slopes was verified by field observations (Byers et al., 2013)
and satellite imagery (Lamsal et al., 2016). Sawagaki et al. (2012) also
performed a bathymetric survey on the lake in 2009 and estimated the total
volume to be
These conflicting studies, which had access to the same data
and had similar on-site observations, highlight the need to take an objective
approach towards understanding the risks associated with a glacial lake. The
main priority with respect to Chamlang South Tsho should be modeling the
entire GLOF process chain. This objective analysis would clarify the
conflicting views on the potential for a dynamic failure. As a first-pass
approach, the methods used by Heller and Hager (2010) were applied to
estimate the impulse wave height using one of the avalanche volumes and
tracks from the mass movement modeling. The modeled avalanche was
Additionally, the potential for a self-destructive failure needs to be explored in further detail as the remote assessment suggests the lake is unstable, which is supported by the SLA calculations in Fujita et al. (2013) and the observations and measurements from Lamsal et al. (2016). A chemical analysis of the seepage would lend insight as to the source of the water, i.e., whether the water is the melting of the ice core or lake water. Geophysical surveys of the terminal moraine should be used to determine the spatial extent and depth of the ice core and geotechnical surveys of the composition of the debris would be valuable for assessing the stability of the moraine in detail. Lastly, Lamsal et al. (2016) highlighted the potential downstream impact and Byers et al. (2013) noted that a GLOF from Chamlang South Tsho is a concern for these communities, so improved modeling using a physically based GLOF model should be a top priority. The combination of these modeling efforts and field measurements would definitively determine the hazard of the lake and the threat to downstream communities.
Dig Tsho is a prime example of why glacial lakes should not be prioritized
based on the size of the lake. In 1985, Dig Tsho was only 0.5 km
Field investigations should assess the current bathymetry of the lake to determine the amount of water that could be displaced by a GLOF. Most likely the maximum depth greatly diminished after the GLOF in conjunction with the reduction in the area of the lake, so the potential GLOF discharge would be smaller than the 1985 GLOF. A one-dimensional GLOF model that has been applied to other glacial lakes in Nepal (Byers et al., 2013; Khanal et al., 2015) would be beneficial in determining how the downstream impacts have changed based on the new bathymetry.
Lower Barun Tsho has received little attention despite being considered one
of the most dangerous glacial lakes in Nepal (ICIMOD, 2011). In this regard,
the remote assessment yields valuable information regarding its hazards and
can be used to guide future investigations of the lake. The mass movement
trajectories show the lake is very susceptible to rockfalls and avalanches
from its southern slope (Fig. S4a, b). Figure 6 also reveals that Lower Barun
Tsho has had the most rapid expansion rate of the eight lakes studied, with
an average growth of 0.054
Similar to Lumding Tsho, Lower Barun Tsho should be a main priority of future field campaigns in Nepal as the lake is a very high risk, but has received little attention. Field campaigns should focus on investigating the potential of mass movement entering the lake from the southern slopes of Lower Barun Tsho. This investigation should be coordinated with physically based modeling efforts of the GLOF process chain to determine how mass movement entering the lake will propagate across the lake and potentially breach the moraine. While the lake is unlikely to fail due to the hydrostatic pressures, a sedimentological survey of the composition of the moraine would greatly improve modeling a potential breach. Geophysical techniques should also be performed on the moraine to determine the presence and spatial extent of the potential ice core as this may have large implications on the breach of the GLOF. The expansion of Lower Barun Tsho is a large concern as it only increases its susceptibility to rockfall and avalanche-prone areas upglacier; therefore, geophysical techniques should be used on Barun Glacier to determine the maximum potential extent of the glacial lake. Additionally, bathymetric surveys should be performed on Lower Barun Tsho to aid modeling efforts of the GLOF process chain. Seto Pohkari also requires attention with regard to its bathymetry and modeling the process chain for avalanches and rockfalls from its surrounding slopes. Similar to Lumding Tsho, the northern lateral moraine of Lower Barun Tsho may protect the lake from the upstream GLOF. Therefore, a physically based flood model for both Lower Barun Tsho and Seto Pohkari would greatly improve the understanding of the risk faced by downstream communities.
Thulagi Tsho is one of the three glacial lakes where field campaigns were
performed to investigate the hazard of a GLOF by ICIMOD (2011). The results
from the remote assessment yield valuable information that may be used to
supplement these initial field campaigns. Mass movement trajectories reveal
the lake is susceptible to rockfalls from both side slopes but is not
susceptible to avalanches (Fig. S6a, b). Figure 6 shows that the lake growth
has stalled since 2010. The lake expansion model reveals the bed elevation of
the glacier behind the calving front is greater than the lake level,
indicating that the lake may have reached its maximum spatial extent
(Fig. S6d). An assessment of the terminal moraine shows there are ponds on
the terminal moraine and the outlet channel has changed in the last 15 years,
which suggests the moraine is ice-cored. Fortunately, the SLA of 7.1
As previously mentioned, the results of the remote assessment provide valuable information to supplement the work performed by ICIMOD (2011) and the one-dimensional GLOF modeling performed by Khanal et al. (2015). Specifically, the lake expansion model shows that Thulagi Tsho appears to have reached its maximum extent, which should be confirmed with a geophysical survey measuring ice thickness behind the calving front. Additionally, the mass movement trajectories are the first time any slope stability has been modeled at this site. These trajectories reveal the lake's vulnerability to mass movement entering the lake and should be the focus of future modeling efforts at this lake. High-resolution satellite imagery and field inspection should be used to determine the potential size of any rockfall such that these estimates may be applied to a physically based mass movement model. These mass movement models could be used in conjunction with the bathymetric survey by ICIMOD (2011) to model the wave propagation and breach of the moraine. A sedimentological survey to accompany the geophysical investigations performed by ICIMOD (2011) would allow the moraine stability and breach parameters to be quantified with greater accuracy. Furthermore, Khanal et al. (2015) also found the downstream impacts from a GLOF were very high; therefore, a two-dimensional physically based model should build off these results to more accurately quantify the risks and vulnerable areas, which may be used to inform the downstream communities.
Tsho Rolpa is arguably the most well-studied glacial lake in Nepal and
currently the only glacial lake that has been remediated (Richardson and
Reynolds, 2000). Richardson and Reynolds (2000) thoroughly discuss the
hazards associated with the glacial lake. Nonetheless, the remote assessment
yields valuable insight into the future development of the lake and potential
vulnerabilities that may guide future work that should be performed on the
lake. Similar to Thulagi Tsho, the lake expansion model shows that Tsho Rolpa
appears to have reached its maximum extent (Fig. S7d), which explains why the
lake area has been relatively constant over the last decade. A geophysical
survey behind the calving front would be beneficial to support or disprove
this finding. If the model is correct, this has large implications for the
hazard of the lake, as this will limit the magnitude of future avalanches
entering the lake; i.e., Richardson and Reynolds (2000) found that the
magnitude of avalanches was increasing as the lake grew. The mass movement
trajectories show that the lake is susceptible to avalanches from its
northern slope (Fig. S7a) and rockfalls from its surrounding side slopes
(Fig. S7b). The avalanche activity has been a major concern for Tsho Rolpa,
so the logical next step is to use physically based models to model the GLOF
process chain and determine how vulnerable the lake is to these threats.
Satellite imagery from the last decade also reveals changes in the islands
near the terminal moraine, which suggest the presence of an ice core
(Fig. S7c). The ice core has been confirmed and well documented (ICIMOD,
2011). Additionally, the SLA of 17.5
Additionally, there are three glacial lakes located upstream that are threats
to Tsho Rolpa. The first upstream glacial lake, Tsho Rolpa Upper 1
(27
The MC-LCP GLOF model of Tsho Rolpa (Fig. S7e, f) shows that 2787
buildings, 7.8 km
The remote assessment integrates the key hazard parameters in an objective manner that is repeatable and relies solely on globally available remotely sensed data. This study investigated eight glacial lakes in Nepal that are widely considered to be highly hazardous and was found to yield valuable insight with respect to each lake regardless of the amount of previous attention the lake had received. For Lumding Tsho and Lower Barun Tsho, this was the first time these lakes have been holistically studied since they were listed as a high priority of further investigation. For other glacial lakes that have already been studied extensively, e.g., Tsho Rolpa and Imja Tsho, the remote assessment yielded valuable information regarding their future expansions. This study is the first of its kind to incorporate detailed modeling of lake growth into a hazard assessment. For Imja Tsho this is particularly valuable as the assessment is able to identify future hazardous conditions before they occur and hopefully shows the benefit of implementing risk-mitigation strategies prior to the lake becoming highly hazardous.
The remote assessment is meant to be a simple tool for understanding the hazards and is meant to guide the focus of future modeling efforts and field campaigns. The difficulty associated with conducting fieldwork in these areas and the scarcity of site-specific field data required to adequately model the risk at each site, as discussed in this study, highlights the need for coordinated efforts amongst institutions and local agencies to address these knowledge gaps. This collaborative effort is crucial when one considers the variety of expertise that is required to conduct these field campaigns and effectively model the GLOF scenarios. Furthermore, despite the methods in this study only being applied to eight glacial lakes, the framework was developed such that future work may apply the remote assessment to all the glacial lakes in Nepal. In this manner, a holistic and objective understanding of the current and future states of GLOF hazards may be developed.
The land use and buildings used to quantify downstream impacts may be
downloaded at
The authors acknowledge the support of the NSF-CNH program (award no. 1516912), the USAID Climate Change Resilient Development (CCRD) project, and NASA Goddard Space and Flight Center/UMBC Maryland for the support of David Rounce. We also acknowledge the support of Dhananjay Regmi of Himalayan Research Expeditions for logistical support during fieldwork. The Landsat imagery used in this study was provided by the Land Processes Distributed Active Archive Center (LP DAAC). Edited by: W. Buytaert Reviewed by: two anonymous referees