2. CLIMATE DYNAMICS AND GLACIER CHARACTERISTICS

The hydrological cycle of the HK region is complex because of the impact of two circulation systems, the Asian Monsoon (AM) and Western Disturbances (WD) (Bookhagen and Burbank, Reference Bookhagen and Burbank2010). Most glaciers in the eastern and central Himalaya experience maximum accumulation in the summer due to high monsoonal precipitation and high elevations, where periods of summertime ablation punctuate overall summer-long snow accumulation (Ageta and Higuchi, Reference Ageta and Higuchi1984). The summer accumulation in the western Himalaya is weak while the AM barely reaches to the Karakoram Range (Bookhagen and Burbank, Reference Bookhagen and Burbank2010). In the Karakoram, WD is the important source of moisture providing maximum precipitation during winter due to strong storms (Lang and Barros, Reference Lang and Barros2004) with generally short life spans of 2–4 days (Dimri and Mohanty, Reference Dimri and Mohanty2009). The western Himalaya is a transition region receiving precipitation from both the AM and WD (Azam and others, Reference Azam2016). Further, there are strong orographic differences in precipitation from south to north across the HK (Shrestha and others, Reference Shrestha, Wake, Mayewski and Dibb1999). The HK is a barrier to monsoon winds, causing maximum precipitation on southern slopes with a regional east to west decrease in the monsoon intensity (Shrestha and others, Reference Shrestha, Wake, Mayewski and Dibb1999) and large local orographic controls on climate. For instance, AM provided low precipitation (21% of the annual sum) on the leeward side of the orographic barrier at Chhota Shigri Glacier (western Himalaya) and high precipitation (51% of annual total) on the windward side at Bhuntar city (~50 km south from Chhota Shigri) (Azam and others, Reference Azam2014b). Therefore, depending on their geographical position and regional orography, the glaciers in the HK region are subjected to different climates. This variability of precipitation regimes along the HK region begets varying types and behaviors of glaciers over short distances (Maussion and others, Reference Maussion2014). Five classes of glaciers were defined (Maussion and others, Reference Maussion2014): two dominant classes with winter (DJF) and summer (JJA) accumulation type, a class with maximum precipitation in pre-monsoon (MAM) months and two intermediate classes which tend to receive either winter (DJF/MAM) or summer (MAM/JJA) precipitation but with less pronounced centers (precipitation amounts) in winter or summer seasons, respectively.

3. GLACIER FLUCTUATIONS AND AREA CHANGE RECORDS

We present the maximum possible compilation of glaciological studies including many Geological Survey of India reports, which are not available easily, but scanned for this study. Compared with snout fluctuation records for almost 100 glaciers/basins in a previous review (Bolch and others, Reference Bolch2012), our compilation includes historical records of 154 glaciers/basins, some available since the 1840s (Fig. 1a). We carefully assess the quality of satellite imageries or topographic maps used to estimate the length changes and assign the caution flags to each length change record (Supplementary Table S2). The longest historical records are for Milam, Gangotri, Pindari, Siachen and Biafo glaciers. Between the mid-19th and mid-20th century, glacier fluctuation records (Fig. 2a) are at multi-decadal scale, extracted from field photographs or maps. After that, the records are usually available at decadal scale, and for some glaciers like Chorabari, Pindari and Raikot records are available annually for recent years. There are 25 records available per year between the mid-19th and the mid-20th century (Fig. 2b). An abrupt increase of records occurred in the early 1960s, with a peak exceeding 125 records per year in the 1975–2000 period (Fig. 2b). This is because Survey of India maps, available since the 1960s, were combined with recent satellite images or field surveys to estimate the fluctuations (Supplementary material).

Fig. 1.

Spatial glacier/basin behaviors over the HK region (Note: the observation time is different and given in corresponding supplementary tables). The regions are defined following (Bolch and others, Reference Bolch2012). Symbology: Circles represent the glacier scale observations while squares represent basin/regional scale observations. Red (or blue) color represents negative (or positive) changes in length, area or mass balance. The abbreviations are given in corresponding supplementary tables. (a) Glacier snout fluctuations for 152 glaciers and 2 basins (Supplementary Table S2). (b) Area changes for 24 glaciers and 47 basins (Supplementary Table S3). (c) Glaciological mass balances for 24 glaciers (Supplementary Table S4 and S6) and (d) geodetic mass balances for 10 glaciers and 24 basins/regions (Supplementary Table S8).

Fig. 2.

(a) Length change of selected glaciers in the HK region over the last 170 years (b) Number of data records (c) Area change rates for HK region. The rates were calculated in percent change per year with respect to the initial observed area. Note that the unweighted mean area rates are calculated for 5-year period from a varying number of values depending on the period. The black and orange boxes represent the ±1 Std dev. envelope for each 5-year mean area change rate and calculated from the area change rates available for the corresponding period. Data and references used in the figure are listed in Supplementary Table S2 and S3.

Since the mid-19th century, a majority of Himalayan glaciers retreated with rates varying regionally and from glacier to glacier, while in the Karakoram, the glaciers snouts have been retreating, stable or advancing (Fig. 2a; Supplementary Table S2). The interpretation of the length record in the Karakoram is complicated by the occurrence of surges. For example, in upper Shyok Valley in the Karakoram, 18 glaciers (1 004 km²) out of 2 123 glaciers (2 978 km²) showed surge-type behavior (Bhambri and others, Reference Bhambri2013).

Glacier area change studies were performed generally at basin-wide scale with few individual glacier estimates (Fig. 1b; Supplementary Table S3). We assign the caution flags to each area change record based on the quality of satellite imageries and topographic maps used in the studies (Supplementary Table S3). Along the Himalayan Range shrinkage is common over the last 5–6 decades (Fig. 1b) with high variability in rates ranging from –0.07% a⁻¹ for Kimoshung Glacier over 1974–2015 to –1.38% a⁻¹ for Ganju La Glacier over 2004–14 (Fig. 2c). Conversely, glaciers in the Karakoram Range showed a slight shrinkage or stable area since the mid-19th century (Supplementary Table S3).

The example of the Khumbu Region (Everest) is striking to illustrate how difficult it can be to compare different area change estimates. A small fraction of the glaciers in this region (~92 km²) showed an area reduction of –0.12% a⁻¹ over 1965–2005 (Bolch and others, Reference Bolch, Buchroithner, Pieczonka and Kunert2008). Another study (Salerno and others, Reference Salerno, Buraschi, Bruccoleri, Tartari and Smiraglia2008), using historical maps (1 : 50 000 scale), reported similar reduction rate of –0.14% a⁻¹ over a 404 km² glacierized area in the same region for 1956–90. A study covering 4000 km² over the Koshi Basin (including Khumbu Region) estimated the much more negative rate of area changes of –0.59 ± 0.17% a⁻¹ over 1976–2009 (Shangguan and others, Reference Shangguan2014). The large discrepancies in those estimates can be attributed to the differing extents of the glacierized area, difficulties in mapping debris-covered glaciers, different observation periods, methodologies, data used and related uncertainties.

The glacier area shrinkage causes fragmentation of large glaciers; consequently, the number of glaciers increased in the Himalaya over the past 5–6 decades (Kulkarni and others, Reference Kulkarni2007; Ojha and others, Reference Ojha2016). The area shrinkage of clean-ice areas, and perhaps reducing glacier flow speeds and possibly accelerating mass wasting from deglaciated surfaces also resulted in increasing supraglacial debris-covered area in the Himalaya (Scherler and others, Reference Scherler, Bookhagen and Strecker2011; Nuimura and others, Reference Nuimura, Fujita, Yamaguchi and Sharma2012; Thakuri and others, Reference Thakuri2014). Small and low elevation glaciers were found to be shrinking faster than larger ones (Thakuri and others, Reference Thakuri2014; Ojha and others, Reference Ojha2016). Studies with multiple observation periods revealed specific trends of area shrinkage rates from different regions of the HK. Steady trends in the eastern Himalaya (Racoviteanu and others, Reference Racoviteanu, Arnaud, Williams and Manley2015), accelerated shrinkage in the central Himalaya (Bolch and others, Reference Bolch, Buchroithner, Pieczonka and Kunert2008; Thakuri and others, Reference Thakuri2014) and decreasing shrinkage in Zanskar and Ravi basins of the western Himalaya (Schmidt and Nüsser, Reference Schmidt and Nüsser2012; Chand and Sharma, Reference Chand and Sharma2015) were found over the last 5–6 decades.

The mean shrinkage rates were computed for the Himalayan glaciers using 60 data series (excluding highly uncertain data; Supplementary Table S3). The shrinkage rates were available since the 1950s to present, but a time window from 1960 to 2010 was selected to avoid the sparse data outside this window. Despite the regional variation in shrinkage rates, Himalayan glaciers showed a continuous mean area shrinkage over the last 5 decades which is slightly higher (more negative) between 1980 and 2000 (Fig. 2c). The calculated unweighted mean area shrinkage for Himalayan glaciers is –0.36% a⁻¹ for 1960–2010. This rate is less than the unweighted mean shrinkage rate (–0.57% a⁻¹) and very close to the weighted mean shrinkage rates (–0.40% a⁻¹) calculated for the whole of High Mountain Asia (HMA) (Cogley, Reference Cogley2016) for the same period. Only four area change studies are available from the Karakoram Range covering different regions (Supplementary Table S3). A time window of 1980–2010 was selected to get at least three estimates for the mean rate of area change. The unweighted mean rate of area change for the Karakoram glaciers varies from ~ –0.06% a⁻¹ during the 1980s to almost 0% a⁻¹ in 2000s (Fig. 2c), but the temporal trend is hard to interpret and probably not significant given the scarcity of the measurements. Overall, these close to 0% a⁻¹ area changes rates are in line with the Karakoram Anomaly (Hewitt, Reference Hewitt2005; Gardelle and others, Reference Gardelle, Berthier and Arnaud2012; Kääb and others, Reference Kääb, Berthier, Nuth, Gardelle and Arnaud2012). Importantly, due to the unweighted nature of these mean rates, and the fact that small glaciers tend to lose area faster than big ones when expressed in % change, these means are different than the mean rate of change of total glacierized area.

4. GLACIER MASS CHANGES

Although the databases of the glacier changes in the HK have greatly improved recently, field observations about mass balance are still scarce. In this review, we include results of mass-balance changes at both glacier scale (from the glaciological method, geodetic method and model results) and regional scale (geodetic method).

The glaciological mass balance is an undelayed, direct response to meteorological changes (Oerlemans, Reference Oerlemans2001) and thus mass-balance observations are needed to study the climate change especially in remote areas such as the HK, where our knowledge of climate–glacier relationship is still partial. Regrettably only 24 glaciers, covering an area of ~112 km² (~0.5% of the total Himalayan glacierized area (Bolch and others, Reference Bolch2012)), have been surveyed in the Himalayan Range using the glaciological method (Supplementary Tables S4–S5), which is sparse in comparison with most other large glacierized regions of the world (Zemp and others, Reference Zemp2015). Glaciological mass-balance measurements in the Himalayan Range are often challenging because of the vast glacierized area, high altitude, rugged terrain, extreme climate, and political and cultural boundaries. In the Karakoram, where the challenges are the toughest, the glaciological measurements were performed only on the ablation area of Baltoro Glacier (Mayer and others, Reference Mayer, Lambrecht, Belo, Smiraglia and Diolaiuti2006). Since the first mass-balance observation on Gara Glacier in 1974/75, the annual mass balances in the Himalaya have mostly been negative, with only 16 positive annual mass-balance observations out of a total of 142 observations (see the unweighted mean mass-balance series in Fig. 3a and Supplementary Table S6). The longest continuous series is just 12 years for Chhota Shigri Glacier (–0.56 ± 0.40 m w.e. a⁻¹ over 2002–14) (Azam and others, Reference Azam2016).

Fig. 3.

Mass balances in the HK region. (a) Annual glaciological mass balances for all 24 glaciers. Red thick line is the mean mass balances for the Himalayan Range calculated using 24 glaciers’ data, black thick line is the mean mass balances for the Himalayan Range calculated using 20 screened glaciers, and the blue thick line represents the global mean mass balances between 1975 and 2014 calculated from 37 reference glaciers of the World Glacier Monitoring Service (Zemp and others, Reference Zemp2012, WGMS Reference Zemp, Nussbaumer, Naegeli, Gärtner-Roer, Paul, Hoelzle and Haeberli2013). (b) Number of data points available each year. (c) Geodetic mass balances. Black thick line is the mean geodetic mass balances for the Himalayan Range while green thick line is the mean geodetic mass balances for the Karakoram Range and (d) Annual modeled/hydrological mass balances. Black thick line is the mean modeled mass balances. Abbreviations in different panels are glacier/region names, available in Supplementary Tables S4, S8, and S9. Note that the mean mass balances are unweighted and calculated for the 5-year period from a varying number of mass balance values available for each period. The black boxes represent the ±1 Std dev. envelope for each 5-year mean mass balance and calculated from the mass-balance values available for the corresponding period.

Previous compilations of field-based mass balances (Cogley, Reference Cogley2011; Bolch and others, Reference Bolch2012) yielded regional mean mass balances for the Himalaya using all available glaciological mass-balance data without quality checks. As already highlighted (Gardner and others, Reference Gardner2013), this problem, combined with bias and lack of representativeness due to benchmark glacier selection, results in regional glaciological mass balances that are more negative compared with the values derived from remote-sensing data, and might lead to an overestimation of the sea-level rise contribution of Himalayan glaciers. The surveyed glaciers are generally chosen for their easy access, low altitudes, small sizes (mean area of surveyed glaciers is 4.6 km²; Supplementary Table S4) and low coverage by debris. Selection criteria introduce a strong bias toward rapid response type glaciers, or at least does not represent the population as a whole; representation is especially lacking for the large, complex glaciers, which dominate the total ice mass in the HK region. Further, almost half of the glaciological studies were conducted by the Geological Survey of India and are published in internal reports (Supplementary Table S4) and often lack the details of mass-balance observations (Supplementary Table S5). These problems make difficult the identification of the reliable mass-balance data series and extension of the data to represent the Himalayan region as a whole. The mass balances of the avalanched-fed Hamtah Glacier (–1.43 m w.e. a⁻¹ between 2000 and 2012) were already found to be negatively biased (Vincent and others, Reference Vincent2013). Therefore, we performed a detailed systematic check of the reliability of each mass-balance time series using several criteria, such as density of point measurement network, stake material, availability of snow density field measurements, error analysis in mass-balance estimates, map quality, debris-cover extent, avalanche contribution, verification of glaciological mass balance with geodetic mass balance and relationships of mass balances with equilibrium line altitude, temperature and precipitation, etc. and classify the mass-balance time series in four categories (excellent, good, fair and dubious; Supplementary Table S7) (details are in the Supplementary material).

Mass balances of six glaciers are dubious (Supplementary Table S7) and might have biases. Except for Kangwure Glacier, these glaciers either receive their accumulation through avalanches (Changmekhangpu, Dunagiri, Hamtah and Kolahoi glaciers) or are highly debris covered (50–80% debris-covered area on Changmekhangpu, Chorabari, Dunagiri, Hamtah glaciers) (Supplementary material). Due to steep topography, many HK glaciers receive a large part of their accumulation from avalanches (Racoviteanu and others, Reference Racoviteanu, Bahuguna, Kargel, Leonard, Bishop, Kääb and Raup2014), a mass input that has not been quantified yet (Laha and others, Reference Laha2017). Avalanches sometimes destroy the ablation/accumulation stakes. Most accumulation is close to the glacier headwalls and cannot be safely monitored with the glaciological method. Full representation of all glacier types should include avalanche-fed glaciers, but for the preceding reasons, these glaciers themselves cannot be fully and correctly surveyed. The debris distribution on highly debris-covered glaciers is, generally, heterogeneous across the surface (often the stakes are installed over a few decimeter thick debris cover); hence the ablation measurement with stakes is location specific. Further, the stakes are generally installed at locations easy to access, and so the effects of supraglacial ponds or ice cliffs, known to be melt ‘hot spot’ (Sakai and others, Reference Sakai, Nakawo and Fujita2002), is not included in the mass-balance estimates. In support of previous studies (Buri and others, Reference Buri, Pellicciotti, Steiner, Miles and Immerzeel2016; Miles and others, Reference Miles2016; Vincent and others, Reference Vincent2016), we find that conventional glaciological mass-balance methods applied to highly debris-covered and avalanche-fed glaciers are error prone, but the selection of simple, safe, clean-ice glaciers might introduce a bias toward more negative regional assessments. Therefore, glacier-wide mass balance on highly avalanche-fed and debris-covered glaciers should be estimated by remote-sensing methods. Despite errors and biases, in situ glaciological data on avalanche-fed and debris-covered glacier are needed for ground truthing of remotely sensed data, modeling and process understanding (Banerjee and Shankar, Reference Banerjee and Shankar2013; Vincent and others, Reference Vincent2016). It is thus necessary to recognize and explain the biases pertaining to the benchmark glacier sample of the whole glacier population, versus biases which may render an individual glacier's data misleading and thus favoring exclusion.

Excluding these six dubious mass-balance series (total 37 annual mass-balance data points), the mean mass balances for the Himalayan range were less negative (Fig. 3a). The revised mean mass balance during 1975–2015 for 18 screened glaciers is –0.49 m w.e. a⁻¹ versus –0.59 m w.e. a⁻¹ for all 24 glaciers. These revised mean mass balances show a moderate wastage since 1975 that generally follows the global trend before 2000 (Fig. 3a). After 2000, a positive deviation of Himalayan mean mass balances from the increasingly negative global mean seems to be consistent with the regional satellite-based mass changes estimated over recent years that suggest two or three times less negative mass balances for the Pamir-Karakoram-Himalaya than for the global mean (Kääb and others, Reference Kääb, Berthier, Nuth, Gardelle and Arnaud2012; Gardelle and others, Reference Gardelle, Berthier, Arnaud and Kääb2013). We note that our inference of reduced Himalayan glacier mass loss after 2000 is based on a changing sample of field-monitored glaciers and thus need to be confirmed in the future. In particular, rare remote-sensing mass-balance estimates for several periods do not show such a trend (Ragettli and others, Reference Ragettli, Bolch and Pellicciotti2016). However, remote-sensing estimates themselves carry their own uncertainty, as illustrated by the fact that the glacier-wide mass balance in the Langtang area (Nepal) had to be revisited by Ragettli and others (Reference Ragettli, Bolch and Pellicciotti2016) compared with a similar earlier assessment by Pellicciotti and others (Reference Pellicciotti, Stephan, Miles, Immerzeel and Bolch2015). Two major sources of uncertainty in DEM-based geodetic mass-balance estimates are the poor quality of DEM in the texture-less accumulation areas (Ragettli and others, Reference Ragettli, Bolch and Pellicciotti2016) and the unknown penetration of the SRTM C-band and X-band radar signal into dry snow and firn (Barandun and others, Reference Barandun2015; Kääb and others, Reference Kääb, Treichler, Nuth and Berthier2015; Dehecq and others, Reference Dehecq, Millan, Berthier, Gourmelen and Trouve2016; Round and others, Reference Round, Leinss, Huss, Haemmig and Hajnsek2017).

With recent progress in satellite data acquisitions and processing, and the availability of declassified stereo images from spy satellites, several estimates provided the geodetic mass changes at glacier- or region-wide scale (Fig. 3c; Supplementary Table S8). Several comprehensive assessments of glacier mass changes in the HK have been obtained from 2003 to 2009 using ICESat laser altimetry (Kääb and others, Reference Kääb, Berthier, Nuth, Gardelle and Arnaud2012, Reference Kääb, Treichler, Nuth and Berthier2015; Gardner and others, Reference Gardner2013; Neckel and others, Reference Neckel, Bolch and Hochschild2014) and, despite differences, revealed a contrasted pattern of mass change in the HK with strong thinning in the south-east Tibetan Plateau and in the western Himalaya and no significant elevation change in the Karakoram and over the Western part of the Tibetan plateau. These ICESat-based mass-balance estimates tended to be more negative than the ones derived by comparing SPOT5 and SRTM DEMs for nine sub-regions of KH (Gardelle and others, Reference Gardelle, Berthier, Arnaud and Kääb2013). The differences in these estimates are difficult to investigate as the study periods and sample regions are mismatched (Supplementary Table S8). However, the difficulty of accounting for the penetration of the SRTM radar signal into snow and ice may explain some of these differences (Kääb and others, Reference Kääb, Treichler, Nuth and Berthier2015) and the varying treatment of errors (not all studies account for systematic errors), but we also note that at the error limits, the estimates overlap. Hence, we find an encouraging approach toward consistency. The contrasted pattern of change has recently been confirmed for a longer time period (2000–16) and at the scale of individual glaciers using multi-temporal analysis of 50 000 ASTER DEMs (Brun and others, Reference Brun, Berthier, Wagnon, Kääb and Treichler2017).

Measurement of the time-varying gravity fields by the GRACE (Gravity Recovery and Climate Experiment) satellites have led to varying estimates of glacier mass changes in our study region (Jacob and others, Reference Jacob, Wahr, Pfeffer and Swenson2012; Gardner and others, Reference Gardner2013). However, they remain difficult to interpret glaciologically due to their coarse resolution, leakage of the strong signal of groundwater depletion from north India (Rodell and others, Reference Rodell, Velicogna and Famiglietti2009), the fact that glacier meltwater may be stored in nearby glacier lakes (Song and others, Reference Song, Huang, Richards, Ke and Hien Phan2014) and the apparent large positive mass change anomaly over the Tibetan Plateau (Yi and Sun, Reference Yi and Sun2014), probably due to increased precipitation (Zhang and others, Reference Zhang2017). As an example, in their sea-level budget assessment during 2002–14, Reager and others (Reference Reager2016) were cautious not to use GRACE data to update glacier mass loss in High Mountain Asia, relying instead on the 2003–09 ICESat mass change estimates.

The individual geodetic mass balances are available at multiannual scale (Supplementary Table S8). The mean geodetic mass balance for the Himalayan Range was –0.37 m w.e. a⁻¹ between 1962 and 2015. Conversely, the Karakoram Range exhibited balanced mass budget with –0.01 m w.e. a⁻¹ between 1975 and 2010 against the mean mass balance of –0.37 m w.e. a⁻¹ for the Himalayan region over the same period. Therefore, the ‘Karakoram Anomaly’ can be extended back at least to the mid-1970s (Bolch and others, Reference Bolch, Pieczonka, Mukherjee and Shea2017; Zhou and others, Reference Zhou, Li and Li2017). The mean balanced mass over the Karakoram Range is consistent with recent glacier snout stability (Fig. 2a), whereas the continuing negative balances over the Himalaya are consistent with glacial lake growth since the early 1960s (Bajracharya and Mool, Reference Bajracharya and Mool2009).

Some discrepancies in glaciological and geodetic mean mass-balance estimates are obvious because of (i) selection bias and lack of representativeness of glaciers used for glaciological measurements (ii) different satellite data types and methodologies for geodetic mass balance and (iii) the larger area covered by geodetic estimates. Compared with mean mass balance of –0.59 m w.e. a⁻¹ for all 24 observed glaciers between 1975 and 2015, the screened mean mass balance of –0.49 m w.e. a⁻¹ over the same period is closer to the agreement with the geodetic mean mass balance of –0.37 m w.e. a⁻¹ for the Himalayan Range over 1975–2015. However, these screened glaciological mass balances remain too sparse in time and space to obtain a robust regional average and an unambiguous temporal trend. Our analysis highlights the sensitivity of the regional average to the addition/subtraction of just a few glaciological measurements. Glaciological measurements should be retained for the understanding of physical processes, validation of remotely sensed measurements, calibration/validation of glacio-hydrological models and development of process-based models for future glacier changes. Our recommendation is that glaciological mass balances should not be used for the computation of regional mass balance (Sherpa and others, Reference Sherpa2017) and for sea-level rise contribution from the Himalayan range. This recommendation is paired with suggested improvements in the benchmark glacier network.

Mass-balance modeling is becoming widely used in the HK region with growing satellite and recent in situ meteorological data availability. A few studies estimated mass balances using different models such as the hydrological model for Siachen Glacier; temperature index model for Chhota Shigri, Langtang and Mera glaciers; albedo model for Chhota Shigri and regression (mass balance-meteorological parameters) model for Kangwure Glacier (Fig. 3d; Supplementary Table S9). Modeling of mass balance over the historic period of observations using both in situ and satellite measurements may finally give high spatial and temporal resolution, long-term continuity and geographic completeness, thus filling data gaps and addressing current inhomogeneities (Supplementary Tables S4 and S8).

Based on in situ field measurements, Vincent and others (Reference Vincent2013) showed that Chhota Shigri Glacier was near balanced conditions during the1990s. The reconstructed mass change of Chhota Shigri Glacier corresponded to a moderate mass loss of –0.30 ± 0.36 m w.e. a^–1 between 1969 and 2012, with, interestingly, no significant mass change between 1986 and 2000 (–0.01 ± 0.36 m w.e. a^–1) (Azam and others, Reference Azam2014a). Further, Mera Glacier mass-balance reconstruction over 1961–2007 showed strong similarities with the Chhota Shigri reconstruction with balanced conditions from the late 1980s to the early 1990s (Shea and others, Reference Shea, Immerzeel, Wagnon, Vincent and Bajracharya2015) (Fig. 3d, Table S10). Our glaciological mean mass balances for the Himalayan glaciers, do not show balanced conditions over the mid-1980s to the late 1990s (Fig. 3a). Few measurements are available for this period (Fig. 3b) when the glaciers may have been in balance. Supporting the modeling studies, only Tipra bank glaciological mass-balance series was close to steady state with mean mass balance of –0.14 m w.e. a^–1 between 1981 and 1988. The appearance of nearly balanced conditions on Chhota Shigri, Mera and Tipra Bank glaciers could be because of regional orography and glacier-specific dynamics. For instance, Mera Glacier showed almost no mass change (–0.02 m w.e. a^–1) between 2010 and 2015, while Pokalde and Changri Nup glaciers, in the nearby interior of the range, showed a rapid wastage of –0.63 and –1.24 m w.e. a^–1, respectively, over the same period (Supplementary Table S6, Sherpa and others, Reference Sherpa2017). Given the limited number of modeled mass-balance series for mean mass-balances computation, we believe these means should not be considered as regional representative and must be handled cautiously.

5. GLACIER WASTAGE AND DEBRIS COVER

Debris-covered glaciers are widespread in the HK region (Scherler and others, Reference Scherler, Bookhagen and Strecker2011). Thermally insulating debris due to rugged topography and strong avalanche activity might slow the glacier mass wastage, thus resulting in longer timescales of mass loss (Rowan and others, Reference Rowan, Egholm, Quincey and Glasser2015; Banerjee, Reference Banerjee2017). This insulating effect has been recently quantified on Changri Nup Glacier (between 5240 and 5525 m a.s.l.) (Vincent and others, Reference Vincent2016) where the area-averaged ablation of the entire debris-covered area is reduced by 1.8 m w.e. a^–1. Fine-grained thick and intact debris cover (0.5 m or more) nearly stop the surface ablation (Potter and others, Reference Potter1998; Konrad and others, Reference Konrad1999). Further, a large number of glaciers in the HK region are avalanche-fed and accumulating debris continuously. Indeed such phenomena exert strong effects on glacier dynamics; the effects that are very poorly understood in the HK region (see Section 6). The role of supra-glacial debris cover and thus melt beneath debris cover has been investigated (Lejeune and others, Reference Lejeune, Bertrand, Wagnon and Morin2013; Collier and others, Reference Collier2015). A few recent studies also addressed the role of backwasting of supraglacial ice cliffs (Buri and others, Reference Buri, Pellicciotti, Steiner, Miles and Immerzeel2016) and supraglacial ponds on melting of debris-covered glaciers (Miles and others, Reference Miles2016; Watson and others, Reference Watson, Quincey, Carrivick and Smith2016). Furthermore, internal ablation (enlargement of englacial conduits) has both a direct and indirect effect on mass loss, through melting and collapse of ice surfaces (Thompson and others, Reference Thompson, Benn, Mertes and Luckman2016; Benn and others, Reference Benn2017). Recent satellite observations of glacier dynamics (Scherler and others, Reference Scherler, Bookhagen and Strecker2011), supported by simplified models (Banerjee and Shankar, Reference Banerjee and Shankar2013), have shown that debris-covered glacier losses occur mostly by thinning without significant retreat in response to climatic warming (Rowan and others, Reference Rowan, Egholm, Quincey and Glasser2015; Banerjee, Reference Banerjee2017). Further, regional-scale studies have shown that the thinning rates of debris-free and debris-covered ice are not different (Kääb and others, Reference Kääb, Berthier, Nuth, Gardelle and Arnaud2012; Nuimura and others, Reference Nuimura, Fujita, Yamaguchi and Sharma2012); this is ascribed to dynamics (Banerjee and Shankar, Reference Banerjee and Shankar2013; Banerjee, Reference Banerjee2017) and strongly enhanced wastage at thermokarst features like supraglacial ponds, ice cliffs and pro-glacial lakes (Sakai and others, Reference Sakai, Nakawo and Fujita2002; Buri and others, Reference Buri, Pellicciotti, Steiner, Miles and Immerzeel2016; Miles and others, Reference Miles2016; Watson and others, Reference Watson, Quincey, Carrivick and Smith2016), which counteract the effect of insulation by debris. Future investigations are needed to quantify the role of debris cover on HK glaciers.

As a result of mass wastage (Figs 3a, b), increasing supraglacial debris-covered area was reported in the Himalaya (Scherler and others, Reference Scherler, Bookhagen and Strecker2011; Nuimura and others, Reference Nuimura, Fujita, Yamaguchi and Sharma2012; Thakuri and others, Reference Thakuri2014) over the last 5–6 decades. Conversely, in the Karakoram Range, the nearly balanced mass budget since the 1970s (Fig. 3b) are accompanied with nearly unchanged supraglacial debris-covered area between 1977 and 2014 (Herreid and others, Reference Herreid2015). Based on glaciers’ response times, a study (Rankl and others, Reference Rankl, Kienholz and Braun2014) inferred a shift of the Karakoram glaciers from negative to balanced/positive mass budgets in the 1980s or 1990s, but recent findings (Bolch and others, Reference Bolch, Pieczonka, Mukherjee and Shea2017; Zhou and others, Reference Zhou, Li and Li2017) of steady-state mass balances since the 1970s suggest this shift to be during or preceding the 1970s.

6. ADJUSTING GLACIER DYNAMICS

Changes in mass balance (Figs 3c, d) influence the glacier dynamics (Cuffey and Paterson, Reference Cuffey and Paterson2010); hence, an adjustment in HK glacier flow is expected. There has been recent progress in regional satellite-based glacier velocity mapping (Dehecq and others, Reference Dehecq, Gourmelen and Trouvé2015; Bhattacharya and others, Reference Bhattacharya2016). However, ice thickness data – also scarce in the HK – are needed with velocity measurements to study the glacier dynamics. On Chhota Shigri Glacier, field-based surface velocities and ice thickness were found to be reducing since 2003 (Supplementary material), which suggest that the glacier is adjusting its dynamics in response to its negative mass balances (Azam and others, Reference Azam2012). Since the Little Ice Age, the mean ice thickness and surface velocities at Khumbu Glacier were also found to be decreasing, suggesting that Khumbu Glacier is out of balance with climate (Rowan and others, Reference Rowan, Egholm, Quincey and Glasser2015), consistent with a lengthy response time (Jóhannesson and others, Reference Jóhannesson, Raymond and Waddington1989). The mean velocity on Gangotri Glacier decreased by ~6.7% between 2006–14 and 1993–2006, a likely response to negative mass budget (Bhattacharya and others, Reference Bhattacharya2016). To understand the impact of climate change on the HK glacier dynamics and to confirm these glacier-scale findings, more region-scale dynamics studies are urgently needed. Hitherto these handful studies on Chhota Shigri, Khumbu and Gangotri glaciers support the reduced glacier velocities found in response to the negative glacier mass balances in several regions of the world (Heid and Kääb, Reference Heid and Kääb2012). Recent accelerations of Baltoro Glacier (Karakoram, Pakistan) (Quincey and others, Reference Quincey2009) and other Karakoram glaciers may be linked with positive mass balances, but the surging phenomenon and a wide range of flow instabilities (Scherler and Strecker, Reference Scherler and Strecker2012; Bhambri and others, Reference Bhambri, Hewitt, Kawishwar and Pratap2017) complicate this relationship (Heid and Kääb, Reference Heid and Kääb2012).

7. GLACIER/CLIMATE RELATIONSHIP

Many glacial processes such as glacier surface mass balance and glacier runoff respond simultaneously with changes in climate; other responses such as length, area, ice velocity, ice thickness profiles are delayed. For example, a step change in climate may take a century or longer to manifest in an approach toward a new equilibrium glacier length, area and thickness profile (Jóhannesson and others, Reference Jóhannesson, Raymond and Waddington1989). The glaciers in the region of low annual temperature range (10 < ΔT < 20°C) such as in the Himalaya were found to have higher mass-balance sensitivity to climate (temperature and precipitation) change, while glaciers having a higher annual temperature range (20 < ΔT < 30°C) such as in the Karakoram have lower climate sensitivity (Sakai and others, Reference Sakai2015). Accordingly, eastern and central Himalaya have a higher sensitivity than western Himalaya and Karakoram (Fujita, Reference Fujita2008; Sakai and others, Reference Sakai2015), a feature that alone explains a larger part of the contrasted pattern of mass loss measured using laser altimetry (Sakai and Fujita, Reference Sakai and Fujita2017). Some studies attempted to understand the mass balances with local meteorological data (Azam and others, Reference Azam2014a; Sherpa and others, Reference Sherpa2017). The glacier wastage in the Himalayan Range is consistent with increasing temperature (Shrestha and others, Reference Shrestha, Wake, Mayewski and Dibb1999; Dash and others, Reference Dash, Jenamani, Kalsi and Panda2007; Dimri and Dash, Reference Dimri and Dash2012; Banerjee and Azam, Reference Banerjee and Azam2016) and decreasing precipitation (Bhutiyani and others, Reference Bhutiyani, Kale and Pawar2010; Dimri and Dash, Reference Dimri and Dash2012). For instance, on the East Rongbuk Glacier (Everest area) the decrease in snow accumulation from 1970 to 2000 (Kaspari and others, Reference Kaspari2008) might be related to the weakening of the AM (Bingyi, Reference Bingyi2005). An extremely ambitious global temperature rise of 1.5°C would lead to a warming of 2.1  ±  0.1°C in HMA (including HK region) and that 64  ±  7% of the present-day ice mass stored in the HMA glaciers will remain by the end of the century (Kraaijenbrink and others, Reference Kraaijenbrink, Bierkens, Lutz and Immerzeel2017). Avalanche-fed glaciers in the HK region are sensitive to rising temperature not only through increased melting, but also through a rise in rain/snow transition elevation during the monsoon; this especially impacts avalanche-fed glaciers because their accumulation zones are relatively low due to the downward transfer of snow into avalanche cones (Benn and others, Reference Benn2012). The conditions which make Karakoram and Himalayan glaciers different could be attributed to increasing winter precipitation in the former (Fowler and Archer, Reference Fowler and Archer2005) or the weaker sensitivity of their winter accumulation to warming (Kapnick and others, Reference Kapnick, Delworth, Ashfaq, Malyshev and Milly2014; Sakai and Fujita, Reference Sakai and Fujita2017). Moreover, cooler summers, greater summer cloudiness and snow cover, and decreasing maximum and minimum temperatures (Fowler and Archer, Reference Fowler and Archer2005; Shekhar and others, Reference Shekhar, Chand, Kumar, Srinivasan and Ganju2010; Bashir and others, Reference Bashir, Zeng, Gupta and Hazenberg2017; Forsythe and others, Reference Forsythe, Fowler, Li, Blenkinsop and Pritchard2017) reduce the average ablation rates or the duration of the ablation season (Hewitt, Reference Hewitt2005) thereby resulting in quasi-stable mass balance.

The physical basis of glacier/climate relationships can be understood by studying the glacier surface energy balance (SEB). In the HK, only a few SEB studies are available; therefore, we spread our area of interest to the whole HMA region. Generally, the SEB studies are restricted to understand the melt processes at point scale over clean glaciers during the summer-monsoon months (Supplementary Table S11). Some intrinsic discrepancies are evident in the comparison because of different models/methods for SEB calculations, time periods, or climatic conditions. Yet, similar to glaciers worldwide (Favier and others, Reference Favier, Wagnon, Chazarin, Maisincho and Coudrain2004; Andreassen and others, Reference Andreassen, Van Den Broeke, Giesen and Oerlemans2008), net short wave radiation flux is the largest source of energy on glacier surfaces in the HMA region and mainly controls the temporal variability of melting, whereas net longwave radiation flux is the greatest energy sink. The net all-wave radiation flux provides the maximum energy flux with >80% contribution to the glacier surface during the summer for the observed HMA glaciers except for Guxiang No. 3 (65%). Sensible turbulent heat flux, always positive over debris-free areas, complements the net radiation flux. Latent heat flux also brings some energy, at least during the core summer-monsoon period, in the form of re-sublimation/condensation of moisture on the glaciers directly affected by monsoonal activity like Chhota Shigri, AX010, Parlung No. 4 and Guxiang No. 3 (Supplementary Table S11). However, depending on the monsoon intensity, the duration of the re-sublimation period can vary from a few weeks (e.g., on Chhota Shigri (Azam and others, Reference Azam2014b)) to a few days (on Parlung No. 4, where re-sublimation occurs on rare days (Zhu and others, Reference Zhu2015)). With continuous negative latent heat flux, sublimation prevails in the summer over the ablation zones of the glaciers less affected by the monsoon and more affected by drier conditions (e.g., Zhadang Glacier, central Tibetan Plateau (Zhu and others, Reference Zhu2015) and Baltoro Glacier in the Karakoram (Collier and others, Reference Collier2013)). Therefore, dry climate conditions over the central Tibetan Plateau, Karakoram and northwest Himalaya (Ladakh, Zanskar regions) point toward mass loss through sublimation. For instance, on Puruogangri ice cap (north-central Tibetan Plateau), sublimation accounted for 66% of its total mass loss from October 2001 to September 2011 (Huintjes and others, Reference Huintjes, Neckel, Hochschild and Schneider2015). The conductive heat flux or heat flux from precipitation is normally small compared with other terms of the SEB. Therefore, the glaciers under drier conditions seem to lose a significant mass fraction through sublimation, while condensation/re-sublimation dominates over glaciers directly influenced by the monsoon. However, we stress that these conclusions are based on a few sporadic studies and need to be confirmed in near future by developing more glacier-scale as well as region-scale SEB studies.

In the western Himalaya, SEB analysis on Chhota Shigri Glacier suggests a clear control of the summer monsoon on annual mass balance through surface albedo change (reduced absorption of solar radiation when monsoonal snow falls occur) (Azam and others, Reference Azam2014b). Handling large-scale circulation analysis over HMA, another study (Mölg and others, Reference Mölg, Maussion and Scherer2014) suggested that mass balance is mainly determined by the precipitation amounts in May–June and shaped by the intensity of summer-monsoon onset and WD dynamics.

8. HYDROLOGICAL REGIMES OF HK RIVERS

Glaciers in a basin can alter the river discharge characteristics at different temporal scales from daily to multi-century (Jansson and others, Reference Jansson, Hock and Schneider2003). Glacier melt contribution to total river discharge depends on the percentage of glacierized area at any given basin outlet hence, going up in the basin, glacier melt contribution increases (Kaser and others, Reference Kaser, Großhauser and Marzeion2010). Glacier runoff contribution in the HK has strong seasonality and follows the seasonality of precipitation/glacier melt in the basins (Bookhagen and Burbank, Reference Bookhagen and Burbank2010; Kaser and others, Reference Kaser, Großhauser and Marzeion2010). Total discharge at any basin outlet is the sum of rain runoff, snow melt, glacier melt and base flow (Lutz and others, Reference Lutz, Immerzeel, Shrestha and Bierkens2014). A variety of methods with different complexity such as empirical relationships between precipitation and discharge (Thayyen and Gergan, Reference Thayyen and Gergan2010), ice ablation models (Racoviteanu and others, Reference Racoviteanu, Armstrong and Williams2013), the hydrograph separation method (Mukhopadhyay and Khan, Reference Mukhopadhyay and Khan2014a, Reference Mukhopadhyay and Khan2015), chemical tracer methods (Racoviteanu and others, Reference Racoviteanu, Armstrong and Williams2013) and distributed glacio-hydrological models (Lutz and others, Reference Lutz, Immerzeel, Shrestha and Bierkens2014; Ragettli and others, Reference Ragettli2015) are used to understand the discharge composition in the HK. Generally, the meteorological stations are installed at valley bottoms; knowledge of the precipitation distribution at glacier altitudes is nearly lacking, which makes it difficult to develop hydrological models. Few studies (Immerzeel and others, Reference Immerzeel, Wanders, Lutz, Shea and Bierkens2015; Sakai and others, Reference Sakai2015) used the glacier mass balances to inversely infer the precipitation at glacier altitudes. They generally suggest that the amount of precipitation required to sustain the observed mass balances is far beyond what is observed at valley stations or estimated by gridded precipitation data (Immerzeel and others, Reference Immerzeel, Wanders, Lutz, Shea and Bierkens2015). Another challenge in runoff modeling of the HK glaciers is the presence of debris cover. Detailed glacio-hydrological studies explaining the physical basis of discharge generation from debris-cover glaciers are still regionally sparse (Fujita and Sakai, Reference Fujita and Sakai2014; Ragettli and others, Reference Ragettli2015). Further, most models do not include sublimation and wind erosion in these methods/models, yet sublimation may be a vital share of the glacier mass wastage in the dry conditions of the Tibetan Plateau and parts of the Karakoram (see Section 7) or even on wind-exposed high-elevation slopes in the Himalaya due to strong winds mostly in winter (Wagnon and others, Reference Wagnon2013).

The HK glaciers play a significant role with varying contributions of glacier and snow melt to the total discharge of HK rivers (Lutz and others, Reference Lutz, Immerzeel, Shrestha and Bierkens2014; Pritchard, Reference Pritchard2017). For instance, over 1998–2007, in the upper Indus Basin (whole basin excluding Indo-Gangetic plains) stream flow was dominated by glacier meltwater, contributing almost 41% of the total discharge, while in the upper Ganges and upper Brahmaputra basins (whole basins excluding Indo-Gangetic plains) contribution was much lower, i.e., ~12 and 16%, respectively (Lutz and others, Reference Lutz, Immerzeel, Shrestha and Bierkens2014). In general, glacial melt dominates snow melt in all these basins (Lutz and others, Reference Lutz, Immerzeel, Shrestha and Bierkens2014) but it varies intrabasinally. In the upper Indus Basin, glacial melt dominates in the Karakoram, while in the western Himalaya, snow melt contributes more than glacial melt to the total discharge (Mukhopadhyay and Khan, Reference Mukhopadhyay and Khan2015). Initially, the positive mass budgets of the central Karakoram (Gardelle and others, Reference Gardelle, Berthier and Arnaud2012) were linked with decreased river flow (Fowler and Archer, Reference Fowler and Archer2006). Recently, this relationship was questioned with the finding of increasing river flows in the central Karakoram during the melt season from 1985 to 2010 (Mukhopadhyay and Khan, Reference Mukhopadhyay and Khan2014a) and the nearly balanced mass budgets (Kääb and others, Reference Kääb, Treichler, Nuth and Berthier2015). The increasing river flows are now thought to be associated with increasing mass turnover as a result of increased temperature and precipitation, but under near-neutral mass balance (Mukhopadhyay and Khan, Reference Mukhopadhyay and Khan2014b). Mass wastage over the Himalayan region is expected to modify the runoff regimes. The glacier-wastage contribution (net water withdrawal from glacier storage) is a moderate fraction of the total annual glacial meltwater (Kääb and others, Reference Kääb, Berthier, Nuth, Gardelle and Arnaud2012; Gardelle and others, Reference Gardelle, Berthier, Arnaud and Kääb2013), hence, discharge composition and its seasonal variation will shift in the future due to changing climate, partly but not only because of changing glaciers.

Future expected temperature increase over the HK region will affect river hydrology in three ways: (i) phase change of precipitation from snow to rain directly contributes to the discharge, (ii) the reduced albedo because of less snow imparts more absorption of solar radiation and thus enhances melting, and (iii) earlier seasonal onset and later end of snow/ice melting. In the Shigar catchment (upper Indus basin), snowmelt is projected to occur earlier in the melting season (Soncini and others, Reference Soncini2015). In the HK region, the glacier melt contribution is projected to increase until 2050 and then decrease (Immerzeel and others, Reference Immerzeel, Pellicciotti and Bierkens2013; Soncini and others, Reference Soncini2015; Shea and Immerzeel, Reference Shea and Immerzeel2016). Discharge is projected to increase at catchment (Immerzeel and others, Reference Immerzeel, Pellicciotti and Bierkens2013; Soncini and others, Reference Soncini2015) and basin scales (Lutz and others, Reference Lutz, Immerzeel, Shrestha and Bierkens2014) up to at least 2050 in HK rivers. The projected increase in discharge is mainly due to the enhanced melt for the Indus Basin and increase in precipitation for the Ganges and Brahmaputra basins (Lutz and others, Reference Lutz, Immerzeel, Shrestha and Bierkens2014; Tahir and others, Reference Tahir, Adamowski, Chevallier, Haq and Terzago2016). However, partly due to the large differences in glacier area/volume estimates, the future predictions inherit large uncertainties.

9. CONCLUSIONS AND OUTLOOK

With increasing recent attention of the scientific community, the understanding of the HK glaciers has grown swiftly, yet this understanding remains weak in spatial and temporal coverage compared with many other mountain ranges of the world. Most glaciers are retreating, shrinking and losing mass with variable rates along the Himalayan Range. These trends are generally consistent with climate warming and decreasing precipitation. The supraglacial debris-covered area in the Himalaya has increased due to glacier shrinkage and debris accumulation. Since the 1970s, the mass budget of Karakoram glaciers has been almost balanced, a global exception initially referred as the ‘Karakoram Anomaly’ and now known to extend to the west of the Tibetan Plateau (western Kunlun and eastern Pamir). This multi-decadal mass stability is in line with nearly unchanged debris-cover. The anomalous behavior of Karakoram glaciers can be linked with increased precipitation and cooler summers as well as their lower sensitivity to temperature change. Often, the climatic response of HK glaciers has been analyzed using temperature and precipitation data generally from meteorological stations at low altitude. Glacier SEB studies, explaining the physical processes of mass change, are still sparse. We suggest that glaciers in dry regions lose a significant amount of mass through sublimation, while condensation/re-sublimation is dominant over glaciers more directly influenced by monsoons. Thus, sublimation should be included in hydrological modeling at least over dry regions, such as the northwest Himalaya and Karakoram, especially on the cold, dry Tibetan side.

Some glaciological measurements are likely errant and together they contribute to regional glaciological mass balances that are more negative than geodetically based measurements. Once the dubious glaciological data are excluded, the mean regional mass balance has better agreement with the geodetic mass balance in the Himalaya. Considering the deficient statistical representation of glaciers in glaciological mass-balance data, this improved agreement with geodetic mass balances could be somewhat fortuitous. Our screening of all 24 glaciological mass-balance series from the Himalaya identified possible biased series and the evidence suggests that the glaciological method is unsuitable if the glacier is highly debris-covered or surrounded by steep valley walls that induce avalanches. Yet, point mass-balance measurements on such glaciers are of great importance for process understanding and modeling. While most Himalayan glaciers have lost mass over the last 5–6 decades, a few showed steady-state mass-balance episodes. Temporal and spatial variability of mass balances relate to heterogeneous climatic conditions, which vary among or within mountain ranges, or even within the same valley. These findings pose vexing questions about the representativeness of benchmark glaciers and highlight the importance of selection of glaciers for field measurement.

Bolch and others (Reference Bolch2012) tracked mass wastage from 1963 to 2010 and saw accelerating ice loss, especially after 1995, with the mean Himalayan glacier wastage trend generally following that of the global mean. Our unscreened dataset, covering a similar but updated time span, leads to a similar conclusion as that of Bolch and others (Reference Bolch2012), but our screened data indicate either nearly constant magnitudes of wastage rate or at least less acceleration, in agreement with geodetic mass-balance estimates. Due to the scarcity and biases of glaciological measurements, remote-sensing methods should be preferred for the computation of sea-level rise contribution from the HK region.

Studies have shown decreasing glacier thickness and velocities due to mass wastage in the Himalaya, while in the Karakoram velocities are temporally more variable, but not much is known about local responses of glaciers in the HK range. Though melt contributions from the HK glaciers are projected to increase until 2050 and then decrease (Immerzeel and others, Reference Immerzeel, Pellicciotti and Bierkens2013; Soncini and others, Reference Soncini2015; Shea and Immerzeel, Reference Shea and Immerzeel2016), the wide range of area and volume estimates (glacier inventories still have ~30% range in estimates) introduces big uncertainties in runoff. Accurate glacier inventory and additional in situ measurements of glacier volumes are needed to validate the area-volume parameterizations and volume distribution models (Frey and others, Reference Frey2014; Farinotti and others, Reference Farinotti2017) used in hydrological modeling. These models are hampered by poor information on the amount, spatial distribution and phase state of high-elevation precipitation, and of permafrost in the HK.

The projected increase in discharge if coupled with extreme rainfall in the future may result in floods that may further induce rapid erosion, landslides, glacial lake outburst floods (GLOFs), etc. Such an event occurred in Kedarnath (Uttarakhand, India) during 15–17 June 2013 when extreme 1-day rainfall of 325 mm was recorded; this extreme event is believed to be a result of the summer monsoon and WD convergence (Bhambri and others, Reference Bhambri2015). This rainfall event also caused the collapse of the moraine-dammed Chorabari Glacier Lake. The resulting flood devastated the Kedarnath valley and downstream, and killed hundreds of people (Dobhal and others, Reference Dobhal, Gupta, Mehta and Khandelwal2013). Several glacial lakes were found potentially dangerous in the HK (Bajracharya and Mool, Reference Bajracharya and Mool2009; Fujita and others, Reference Fujita2013), but the prediction of extreme rainfall events, the growth of possible future lakes (Linsbauer and others, Reference Linsbauer2016), related risks or GLOFs is still difficult because of limited regional forecasting abilities (Bookhagen, Reference Bookhagen2010). Sometimes glacial surges develop a natural dam and block the river streams that further result in outburst floods (Hewitt and Liu, Reference Hewitt and Liu2010; Round and others, Reference Round, Leinss, Huss, Haemmig and Hajnsek2017).

Our mass-balance reliability assessment using scoring method highlights the weaknesses in available mass-balance series. We recommend long-term continuation and expansion of ongoing in situ glaciological observations in the HK so that these glaciers can be used for climate change and applications studies. Further, for ongoing or future observations, we suggest to systematically include uncertainty estimates, use an optimized density of ablation stakes, perform systematic accumulation measurements with appropriate methods (e.g., use of artificial layer to mark the glacier surface), and verify in situ glaciological mass balances with geodetic mass balances.

We also propose the establishment of a network of high altitude meteorological and discharge stations covering more glaciers/watersheds in the HK region, adoption of a holistic approach to understand mass-balance-climate/hydrology relationships, and development of data sharing policies among the HK countries so that the large-scale modeling of the future glacier and runoff evolution can be done with improved accuracy.

The pace of glacier change is fast enough that many applications and interests should consider climate-change induced glacier responses and associated hydrological changes, but is slow enough that well-planned and locally tailored approaches to adaptation are both possible and needed.