Understanding the current status, recent changes, and likely future evolution of glaciers in High Mountain Asia (HMA) is important for a number of reasons, including evaluating the status of glacial water resources and how these may evolve under climate change (Immerzeel et al., 2010; Rasul, 2016; Lalande et al., 2021) and how glacier-related changes affect glacial hazards (e.g. Harrison et al., 2018; Shugar et al., 2021). Given the likely warming by the end of the 21st century in large parts of HMA, this understanding becomes a critical issue, as cryosphere-derived water supply affects the livelihoods of hundreds of millions of people and the stability of ecosystems downstream. A total of 8 of the 27 low-income and lower-middle-income economies identified by the United Nations (UN) Development Programme in Asia are currently affected by climate-driven water supply issues in HMA (Harrison et al., 2021). The Sustainable Development Goals (SDGs), adopted by all UN and Asian governments, aim to substantially increase the water-use efficiency across all sectors, to ensure sustainable withdrawals and supply of freshwater, whilst also reducing the number of people experiencing water scarcity (e.g. Bhaduri et al., 2016) by 2030. There are also concerns about the impact of future glacier ice loss on global sea level change (e.g. Marzeion et al., 2020) and on glacier-related hazards, such as glacier lake outburst floods, rock slides and falls, and rapid changes in slope and catchment sediment yield (e.g. Li et al., 2022).

As a consequence of these concerns, there has been long-standing scientific and policy focus on modelling changes in glacier mass balance and understanding the implications of climate change for mountain water supplies (e.g. Nie et al., 2021). Numerous modelling studies have projected the impacts of climate change on glacier mass loss in HMA (e.g. Kraaijenbrink et al., 2017; Hock et al., 2019a; Hugonnet et al., 2021; Rounce et al., 2020, 2023) (Table 1) and downstream river runoff (e.g. Sorg et al., 2012; Lutz et al., 2014; Huss and Hock, 2015). Since 2013, these have tended to use the outputs from the CMIP5 set of model runs; while the latest CMIP6 model runs are now available, few projections from this have been employed so far.

Although existing modelling approaches are useful to assess pathways for future ice loss from the region (Table 1), these generally assume that the different ways in which mountain glaciers will evolve under future climate change have been accurately captured. We argue that this is not necessarily the case (Harrison et al., 2021). The common view is that the expected rise in air temperature over this century is expected to lead to the almost complete melting of glaciers in HMA by 2100 (e.g. Rounce et al., 2023; Chen et al., 2023a, b); thus there is a simple relationship between temperature forcing and glacier mass balance response. Current modelling studies support substantial but incomplete melting, for example, 60 %–98 % reduction in glacier mass under RCP8.5 by the end of the century according to Shannon et al. (2019). This outcome arises from a combination of the reduction in accumulation as more precipitation falls in the form of rain and the enhanced melting associated with rising temperatures (see Table 1).

However, there are regional differences in mass loss projections across HMA which partly reflect model uncertainty at fine spatial scales (Chen et al., 2023a). Kraaijenbrink et al. (2017) used a global ensemble of 110 GCM runs from CMIP5 to assess the glacial response driven by emissions under RCP2.6 scenarios and a consequent increase in Global Mean Surface Temperature (GMST) of 1.5 °C above pre-industrial conditions. This result suggests a probable warming of 2.1±0.1 °C for HMA by 2100, even at this low-emission scenario. They also assessed likely regional changes and argued that parts of the western Pamir and the Qilian Shan of northern China will lose most of their glacier mass compared to the present day by 2100, with only 32±14 % and 30±5 % ice mass remaining, respectively. In this study, the Karakoram region shows more resilience to climate warming, with a projected 80±7 % of ice volume remaining by 2100. This anomaly is attributed partly to the role of supraglacial debris cover in maintaining ice mass and to the role of winter precipitation in maintaining accumulation.

Whilst such modelling experiments suggest varying glacier volume loss in HMA by 2100, the physical response of glaciers to climate change varies enormously across the Himalayas and the wider HMA region, and this is caused by varying exposure to monsoonal and westerly atmospheric flows (e.g. Mölg et al., 2014); changes in surface albedo; and the variability in local catchment characteristics such as local topography, aspect, and geology (e.g. Fugger et al., 2022). Glaciers experiencing accumulation in the summer months also undergo ablation at this time (e.g. Fujita and Ageta, 2000). It is also known that the timing and amount of monsoon snowfall can impact on mass balance in different regions of the Himalayas and for subsequent seasons (Mölg et al., 2014; Bonekamp et al., 2019), and these factors are not fully captured in existing climate models.

Supraglacial debris is now recognized as an important factor that may variably amplify or buffer glacier mass balance response to temperature forcing (e.g. Herreid and Pellicciotti 2020; Shrestha et al., 2020; Chen et al., 2023a; Pratap et al., 2023). Kraaijenbrink et al. (2017) were among the first to model the impact of debris cover on glacier melt in HMA under different climate projections, and they showed that, under RCP4.5, RCP6.0, and RCP8.5, glacier mass losses would be 49±7 %, 51±6 %, and 64±5 %, respectively, by 2100 compared with the present day. More recently, Compagno et al. (2022a) used the five Shared Socioeconomic Pathways (SSP119, SSP126, SSP245, SSP370, and SSP585) from CMIP6 to assess the future evolution of debris cover and its impact on glacier dynamics for all HMA glaciers. They showed a general increase in glacier debris cover with increasing radiative forcing and local increases in debris thickness on individual glaciers (see also Scherler et al., 2018; Mölg et al., 2020). At a smaller scale, Rowan et al. (2015) applied a numerical model to estimate the evolution of the debris-covered Khumbu Glacier and predicted a decrease in glacier volume of 8 %–10 % by 2100.

Despite this body of research, we argue that more work needs to focus on likely geomorphic responses to glacier mass loss across and within HMA if we are to better understand landscape evolution during future deglaciation and any hydrological implications that follow. The landscape responses that might increase in scale and spatial impact include rock surface weathering, slope sediment supply and downslope sediment yield, mass movements such as rock slope failures and debris flows, and ecological succession and slope greening, along with their biophysical feedbacks (e.g. Knight, 2024). All such geomorphic processes can deliver debris to glacier surfaces and surrounding slopes and valley floors, thereby contributing to reduced melting through insulation of the ice beneath, alongside downstream changes in sediment supply and changes in river transport capacity.

In this Perspective Article, we explore and highlight some of the implications of these modelling exercises, the majority of which project sustained reduction in glacier mass balance (e.g. Edwards et al., 2021), and highlight some plausible alternative scenarios for how glacier systems in HMA might evolve to 2100 and beyond. We consider two broad scenarios of how mountain glaciers might respond to climate change: the Major Ice Loss (MIL) view and the Paraglacial Transition (PT) view. Both of these end-member evolutionary pathways necessarily represent simplifications of future glacier behaviour, yet they can usefully explore how HMA glaciers could evolve over future decades. The two pathways highlight the contingency of glacier evolution to the geomorphological, hypsometric, geological, and climatic variations that exist over HMA and that conventional climate model outcomes do not successfully capture. It is already known, for instance, that the glacier responses to recent climate changes have been spatially and temporally variable across HMA (e.g. Rounce et al., 2020). Thus, our more generalized approach is grounded in an understanding of known glacier behaviour and the properties of HMA glaciers. Given these caveats, we end by discussing some of these regional differences.

An important distinguishing feature between the MIL and PT pathways is that direct ice melt is only a factor while the ice mass still exists, ceasing when the ice mass is gone, and tends to affect local areas only. By contrast, paraglaciation affects wider geographical areas outside of the ice mass and can extend for decades to millennia after full ice mass loss (Ballantyne, 2002). Understanding which pathway of evolutionary development is followed by any ice mass at any point in time has implications for predictability, hazard risk, and sustainable development. MIL and PT pathways represent end-members along an evolutionary trajectory. It is likely that the development of any one glacier is dependent on their initial conditions and on changes in ice mass volume or surface debris. However, both of these elements are more than a simple volumetric analysis because this does not account for the detailed dynamics or spatial/temporal patterns of ice or debris. These different pathways are likely to display important regional variation in the response of clean-ice glaciers, debris-covered glaciers, and rock glaciers to future climate change in HMA. These responses will not only be driven by variations in regional temperatures, but also by changes in the behaviour of the Indian and East Asian summer monsoons and the Western Disturbance (e.g. Fugger et al., 2022). Whilst regional climatic differences remain largely unexplored (however, see Brun et al., 2019), we can hypothesize that the areas that will undergo a transition from ice glaciers to rock glaciers and other ice debris landforms most readily will be those where debris-covered glaciers are already most common because the debris supply in those areas is high. Furthermore, as high-elevation clean ice thins, and as freeze–thaw and frost-shattering environments move to higher altitudes, debris supply will increase in some high-elevation areas that presently produce little debris. Where these environmental domains move into high-elevation cirques, especially north-aspect cirques, “periglacial” rock glacier development associated with permafrost creep will be favoured (cf. Haeberli et al., 2006). Because of the exceptionally high relief of the Himalayas and many other parts of HMA, rock glacier development may proceed on regional scales greater than seen in other mountains globally today, though it will take centuries for rock glaciers to develop. Improved understanding of debris supply processes to valley bottoms is restricted by the relative lack of modelling at regional scales that specifically considers the role of debris cover on glacier dynamics and mass balance (e.g. Racoviteanu et al., 2022).

We can see, then, that the rock glacier response to deglaciation envisioned by the PT scenario is likely to be highly complex, with the full suite of rock glaciers (“periglacial” and “glacier-derived”) and other ice debris landforms developing in different locations and regions and over different timescales. This complex response will be driven by climate change as a first-order control and by debris supply and glaciological factors as secondary factors. Separate from rock glaciers, the evolution of undifferentiated ice debris landforms during glacier recession has hardly been discussed in the cryosphere literature from HMA (however, see Bolch et al., 2019). As a result, there is uncertainty in evaluating precisely how the PT scenario would develop and how quickly and what form the equilibrium landscape might present.

How do we assess which of the MIL and PT scenarios are more likely and what their probable future spatial distributions are? A first-order understanding might be gained by a simple evaluation of how glaciers have behaved in the past in response to known climate forcings; from this, we can suggest whether these glaciers have shown high or low sensitivity to past or recent climate change (Harrison, 2009). Although this is a uniformitarian view, if glaciers have demonstrated a high sensitivity to past climate change, then this tends to support the MIL scenario for the future responses of glaciers to climate change. Alternatively, if the glaciers have shown low sensitivity or delayed response to past climate change, then this might make the PT more likely, even if the forcings are different at different times and glaciers today are in different states than they were in the past. To achieve this, we need to establish the extent to which glaciers have responded to the warming since the regional Last Glacial Maximum or the late Holocene Neoglacial Maximum (often seen as equivalent to the European Little Ice Age).

Therefore, we compiled published studies that have dated Himalayan moraine sequences from the western, central, and eastern Himalayas (Fig. 2 and 10.5281/zenodo.15828526; Harrison, 2025). We analysed moraine ages from three time periods: the regional Last Glacial Maximum from 18 to 24 ka (Owen et al., 2002), a period covering the regional Younger Dryas from 12 880 and 11 640 ka (Rawat et al., 2012), and the regional Little Ice Age between 1300–1600 AD (Rowan, 2017) (Fig. 2). While there is evidence that glaciers in several areas reached their late Pleistocene limits earlier than Marine Isotope Stage (MIS) 2 (Benn and Owen, 1998; Owen et al., 2002), overall, these data show that glaciers in the Himalayas have not receded much further behind dated glacier limits over these time periods. Figure 2 shows that, as expected, glaciers in different regions of the Himalayas have responded differently to past climate change. The results are averaged by region and show considerable local variability. However, overall, glaciers in the western and central regions of HMA have retreated less over these time periods than those in the eastern part of the Himalayas; this spatial pattern might reflect the reduction in the monsoon and intensification of westerly influences on glacier mass balance towards the western Himalayas (Kumar et al., 2020; Hunt, 2024).

One current limitation with this approach is that few moraines have been dated in HMA and that most that have been are dated to the regional Neoglacial (Rowan, 2017). Furthermore, most dated moraines are associated with fluctuations in large valley glaciers and therefore might not reflect the behaviour of the more numerous and climatically sensitive smaller glaciers.

However, another way to assess glacier response to climate change in general (and temperature rise in particular), and therefore the relative likelihood of the MIL or PT models being dominant, is by monitoring their Equilibrium Line Altitude (ELA), i.e. the altitude on the glacier surface where the theoretical mass balance is zero at a given point in time (Zemp et al., 2006; Cogley, 2011). For instance, in the Khumbu region in the central Himalayas, Marine Isotope Stage (MIS) 2 moraines (equivalent in age to the global Last Glacial Maximum (LGM)) are located just 5 km or so from modern ice limits and in many cases reflect a 200–300 m reduction in glacier ELA at this time compared with current glacial ELAs in the region (Richards et al., 2000) and also reduced insolation driving a weakened monsoon (Owen et al., 2002). At the western end of the Himalayas, much research has concentrated on the Nanga Parbat massif to the south of the Karakoram. Here, work has shown that glaciers draining Nanga Parbat do not show an MIS 2 maximum, although moraines of MIS 3 are present downvalley (Phillips et al., 2000). The absence of evidence for an LGM-age advance of the glacier may reflect aridity during MIS 2 in this region and therefore low glacier sensitivity to atmospheric temperatures at this time (e.g. Yan et al., 2021), i.e. that glacier dynamics here are precipitation-controlled rather than temperature-controlled.

Glacier recession since the regional Neoglacial maximum of the late 18th century also supports the present low sensitivity of many Himalayan glaciers to climate change (Rowan, 2017). For instance, at Ama Dablam and Lhotse in the Khumbu region of Nepal, present glacier margins have retreated by around 1 km from Neoglacial moraine limits. Similarly, in the monsoon transition zone of the Indian Himalayas, the debris-covered Bara Shigri Glacier has retreated less than 3 km since the 1850s (Chand et al., 2017). Overall, Fig. 2 demonstrates that glacier termini in the western and central parts of the Himalayas have retreated less than 2 km since the end of the Neoglacial maximum. Compared with considerable volumetric ice loss since this time (Lee et al., 2021), if future glacier response mirrors that of the past, then this supports our contention that future glacier loss will dominantly involve downwasting of glacier surfaces rather than terminus retreat. We argue that this favours the development of stagnant glacier tongues and enhances the likelihood of further transition to rock glaciers (Jones et al., 2019) and other ice debris landforms (e.g. Johnson, 1980; Monnier and Kinnard, 2017; Anderson et al., 2018) and thus the PT rather than the MIL pathway. From this albeit limited and incomplete data set, we suggest that glaciers from the western, eastern, and northern Himalayas displayed low sensitivity to climate change during the regional LGM and the Neoglacial, and this supports the PT scenario of glacier responses to future climate change.

It is also likely that different glaciers in HMA are at different stages of deglacial evolution, depending on their size, location, and mass balance. This means that some show a MIL response, some show a PT response, and some may be changing from the former to the latter (Fig. 3). It is, however, uncertain as to which clean-ice or debris-covered glaciers will transition to rock glaciers and other ice debris landforms and which will melt significantly in response to climate warming. However, we can say, based on current observations, that small glaciers located below the regional ELA are projected to disappear, as they will not be able to adapt to future climate (ICIMOD, 2023), as is the case in the Andes (Ramírez et al., 2001) and the European Alps (e.g. Zemp et al., 2006; Žebre et al., 2021). Some of these features may completely disappear if debris supply is low, and others may undergo transitions to rock glaciers that may be stabilized by a combination of high snowfall and high debris production. We hypothesize that the transition process is dominated by debris fluxes from mountain slopes and the connectivity between these sites and downwasting glacier surfaces below the ELA (Fig. 3). Therefore, the transition between MIL and PT pathways is most efficient in areas where high mountain slopes are producing rock slope failures, rock falls, and debris flows and where lateral moraines are absent or poorly developed and therefore allow sediment access from valley sides to the glacier surface. Climate change will therefore represent a first-order control on glacier behaviour, but glacial processes creating lateral moraines will play a significant second-order control.

Finally, the consequences of having more persistent glacial ice in HMA would be profound. This outcome would mean that there is more time to generate climate change mitigation and adaptation schemes in the wider region and to develop technological fixes to the challenges of changing hydrological resources (including total volumes and seasonality). In addition, several of the UN Sustainable Development Goals, such as Clean Water and Sanitation, might be achieved more easily than previously expected if cryospheric water sources persist longer.

We have argued that there is a general consensus from climate modelling that Himalayan and wider HMA glaciers will reduce their volume significantly (perhaps by up to 90 % by 2100) in response to projected climate warming (see Shannon et al., 2019) and that most small glaciers will completely disappear by this time. However, we have presented an alternative Paraglacial Transition scenario where many glaciers transform into rock glaciers and other ice debris landforms as climate change progresses. This serves to inhibit ice melt and increase the resilience of the Himalayan glacial system to future climate change by increasing the longevity of ice bodies in the landscape. The Major Ice Loss (MIL) and Paraglacial Transition (PT) scenarios discussed here represent end-members of possible glacial system responses to future climate change (Fig. 3).

While the MIL scenario will also lead to a range of paraglacial responses from deglaciating catchments, we argue that this will not necessarily change the future evolution of individual glaciers which will continue to melt in response to ongoing climate warming (although increased snowfall associated with a warming atmosphere might reduce net mass loss). This MIL scenario continues to dominate the literature based on climate model assessments of glacier melt. However, there are relatively few published studies on the development of rock glaciers and their importance in HMA (see Harrison et al., 2021, for a discussion of this). More research needs to be conducted on the different ice masses and rock glaciers of HMA and the paraglaciation of the region if the PT view is to be properly assessed. Such future work is made challenging by the difficulty of assessing ice content in rock glaciers and other debris-covered landforms such as lateral and terminal moraines, especially in remote, high-altitude settings. How many rock glaciers have derived from the downwasting of glacial ice (Knight et al., 2019) and how many are derived from the creep of ice-rich permafrost (e.g. Haeberli et al., 2024) is also unknown. How rock glaciers respond to climate change in HMA is also hardly known, particularly given their likely long response times.

Critical research is needed in order to evaluate the operations and outcomes of the MIL and PT scenarios and their possible interactions on individual glaciers. Future research imperatives therefore include (1) determination of debris fluxes throughout the region for the full range of geological materials, slopes, microclimates, and glacier types; (2) long-term monitoring of glacier mass balance across the region in order to evaluate cryospheric sensitivity to climate forcing; (3) measurement of contemporary debris fluxes and distributions on different glacier types; (4) present and past climate modelling at high resolution to include changing debris cover; and (5) projections into the future for the full range of climate scenarios. Development of ultra-downscaled climate modelling that is responsive to the full range of HMA relief and slopes with resolutions enough to resolve individual cirques is also needed. This may currently be possible for small local geographic domains sufficient to sample different parts of HMA.