Editing IPCC:AR6/SROCC/Chapter-2 (section)

== CCB.6 Glacier Projections in Polar and High Mountain Regions ==

<div id="section-2-2-3glaciers-block-1"></div>

Authors: Regine Hock (USA), Andrew Mackintosh (Australia/New Zealand), Ben Marzeion (Germany)

Century-scale projections for all glaciers on Earth including those around the periphery of Greenland and Antarctica are presented here. Projections of the Greenland and Antarctic ice sheets are presented in Chapter 4. Future changes in glacier mass have global implications through their contribution to sea level change (Chapter 4) and local implication, for example, by affecting freshwater resources (Section 2.3.1). Glacier decline can also lead to loss of palaeoclimate information contained in glacier ice (Thompson et al., 2017 <sup>[[#fn:r115|115]]</sup> ).

AR5 included projections of 21st century glacier evolution from four process-based global-scale glacier models (Slangen and Van De Wal, 2011 <sup>[[#fn:r116|116]]</sup> ; Marzeion et al., 2012 <sup>[[#fn:r117|117]]</sup> ; Giesen and Oerlemans, 2013 <sup>[[#fn:r118|118]]</sup> ; Radić et al., 2014). Results have since been updated (Bliss et al., 2014 <sup>[[#fn:r119|119]]</sup> ; Slangen et al., 2017 <sup>[[#fn:r120|120]]</sup> ; Hock et al., 2019 <sup>[[#fn:r121|121]]</sup> ) using new glacier inventory data and/or climate projections, and projections from two additional models have been presented (Hirabayashi et al., 2013 <sup>[[#fn:r122|122]]</sup> ; Huss and Hock, 2015 <sup>[[#fn:r123|123]]</sup> ). These six models were driven by climate projections from 8–21 General Circulation Models (GCMs) from the CMIP5 (Taylor et al., 2012 <sup>[[#fn:r124|124]]</sup> ) forced by various RCPs, and results are systematically compared in Hock et al. (2019) <sup>[[#fn:r125|125]]</sup> .

Based on these studies there is ''high confidence'' that glaciers in polar and high mountain regions will lose substantial mass by the end of the century. Results indicate global glacier mass losses by 2100 relative to 2015 of 18% ( ''likely'' range 11 – 25%) (mean of all projections with range referring to ± one standard deviation) for scenario RCP2.6 and 36% ( ''likely'' range 26 – 47%) for RCP8.5, but relative mass reductions vary greatly between regions (Figure CB6.1). Projected end-of-century mean mass losses relative to 2015 tend to be largest in mountain regions dominated by smaller glaciers and relatively little ice cover, exceeding on average 80%, for example, the European Alps, Pyrenees, Caucasus/Middle East, Low Latitudes and North Asia for RCP8.5 (see Figure 2.1 for region definitions). While these glaciers’ contribution to sea level is negligible their large relative mass losses have implications for streamflow (Section 2.3.1, FAQ 2.1).

The magnitude and timing of these projected mass losses is assigned ''medium confidence'' because the projections have been carried out using relatively simple models calibrated with limited observations in some regions and diverging initial glacier volumes. For example, mass loss by iceberg calving and subaqueous melt processes that can be particularly important components of glacier mass budgets in polar regions (McNabb et al., 2015 <sup>[[#fn:r126|126]]</sup> ) have only been included in one global-scale study (Huss and Hock, 2015 <sup>[[#fn:r127|127]]</sup> ). In addition instability mechanisms that can cause rapid glacier retreat and mass loss are not considered (Dunse et al., 2015 <sup>[[#fn:r128|128]]</sup> ; Sevestre et al., 2018 <sup>[[#fn:r129|129]]</sup> ; Willis et al., 2018 <sup>[[#fn:r130|130]]</sup> ).

The projected global-scale relative mass losses 2015 – 2100 correspond to a sea level contribution of 94 ( ''likely'' range 69 – 119) mm sea level equivalent (SLE) corresponding to an average rate of 1.1 ( ''likely'' range 0.8 – 1.4) mm SLE yr -1 for RCP2.6, and 200 ( ''likely'' range 156 to 240) mm SLE, a rate of 2.4 ( ''likely'' range 1.8 – 2.8) mm SLE yr -1 for RCP8.5, in addition to the sea level contribution from the Greenland and Antarctic ice sheets (Chapter 4). Averages refer to the mean and ranges to ± one standard deviation of all simulations. For RCP2.6, rates increase only slightly until approximately year 2040 with a steady decline thereafter, as glaciers retreat to higher elevations and reach new equilibrium. In contrast, for RCP8.5, the sea level contribution from glaciers increases steadily for most of the century, reaching an average maximum rate exceeding 3 mm SLE yr -1 (Hock et al., 2019 <sup>[[#fn:r131|131]]</sup> ). For both RCPs the polar regions are the largest contributors with projected mass reductions by 2100 relative to 2015 combined for the Antarctic periphery, Arctic Canada, the Greenland periphery, Iceland, Russian Arctic, Scandinavia and Svalbard ranging from 16% ( ''likely'' range 9 to 23%) for RCP2.6 to 33% ( ''likely'' range 22 to 44%) for RCP8.5. Due to extensive ice cover, these regions make up roughly 80% of the global sea level contribution from glaciers by 2100. The global projections are similar to those reported in AR5 for the period 2081 – 2100 relative to 1986 – 2005, if differences in period length and domain are accounted for (AR5’s glacier estimates excluded the Antarctic periphery). The eleven mountain regions covered in Chapter 2 are ''likely'' to lose 22 – 44% of their glacier mass by 2100 relative to 2015 for RCP2.6 and 37 – 57% for RCP8.5. Worldwide many glaciers are expected to disappear by 2100 regardless emission scenario, especially in regions with smaller glaciers ( ''very'' ''high confidence'' ) (Cullen et al. 2013 <sup>[[#fn:r132|132]]</sup> ; Rabatel et al., 2013 <sup>[[#fn:r133|133]]</sup> ; Huss and Fischer, 2016 <sup>[[#fn:r134|134]]</sup> ; Rabatel et al., 2017 <sup>[[#fn:r135|135]]</sup> ).

The global-scale projections (Figure CB6.1) are consistent with results from regional-scale studies using more sophisticated models. Kraaijenbrink et al. (2017) projected mass losses for all glaciers in High Mountain Asia of 64 ± 5% (RCP8.5) by the end of the century (2071 – 2100) compared to 1996 – 2015. A high-resolution regional glaciation model including ice dynamics indicated that by 2100 glacier volume in western Canada will shrink by ~70% (RCP2.6) to ~90% (RCP8.5) relative to 2005 (Clarke et al., 2015 <sup>[[#fn:r136|136]]</sup> ). Zekollari et al. (2019) <sup>[[#fn:r137|137]]</sup> projected that the glaciers in the European Alps will largely disappear by 2100 (94 ± 4% mass loss relative to 2017) for RCP8.5, while projected mass losses are 63 ± 11% for RCP2.6.

AR5 concluded with ''high confidence'' that due to a pronounced imbalance between current glacier mass and climate, glaciers are expected to further recede even in the absence of further climate change. Studies since AR5 agree and provide further evidence (Mernild et al., 2013 <sup>[[#fn:r139|139]]</sup> ; Marzeion et al., 2018 <sup>[[#fn:r139|139]]</sup> ).

<div id="section-2-2-3glaciers-block-2"></div>

<span id="figure-cb6.1"></span>
<!-- START IMG -->
<!-- IMG TITLE -->
'''Figure CB6.1'''

<span id="figure-cb6.1-projected-glacier-mass-evolution-between-2015-and-2100-relative-to-each-regions-glacier-mass-in-2015-100-based-on-three-representative-concentration-pathways-rcp-emission-scenarios-cross-chapter-box-1-in-chapter-1.-thick-lines-show-the-averages-of-46-to-88-model-projections-based-on-four-to-six-glacier-models-for-the"></span>
<!-- IMG CAPTION -->
'''Figure CB6.1 | Projected glacier mass evolution between 2015 and 2100 relative to each region’s glacier mass in 2015 (100%) based on three Representative Concentration Pathways (RCP) emission scenarios (Cross-Chapter Box 1 in Chapter 1). Thick lines show the averages of 46 to 88 model projections based on four to six glacier models for the […]'''
<!-- IMG FILE -->
[[File:ab4b15d55498916602753f775b21d72d IPCC-SROCC-CB_6_1.jpg]]

Figure CB6.1 | Projected glacier mass evolution between 2015 and 2100 relative to each region’s glacier mass in 2015 (100%) based on three Representative Concentration Pathways (RCP) emission scenarios (Cross-Chapter Box 1 in Chapter 1). Thick lines show the averages of 46 to 88 model projections based on four to six glacier models for the same RCP, and the shading marks ± 1 standard deviation (not shown for RCP4.5 for better readability). Global projections are shown excluding and including the Antarctic (A) and Greenland (G) periphery. Regional sea level contributions are given for three RCPs for all regions with &gt;0.5 mm sea level equivalent (SLE) between 2015–2100. The Low Latitudes region includes the glaciers in the tropical Andes, Mexico, eastern Africa and Indonesia. Region Alaska includes adjacent glaciers in the Yukon and British Columbia, Canada. Regions are sorted by glacier volume according to Farinotti et al. (2019). Data based on Marzeion et al. (2012); Giesen and Oerlemans (2013); Hirabayashi et al. (2013); Bliss et al. (2014); Huss and Hock (2015); Slangen et al. (2017). Modified from Hock et al. (2019).

<!-- END IMG -->
<span id="permafrost">
</span>
=== 2.2.4 Permafrost ===

<div id="section-2-2-4permafrost-block-1"></div>

This section assesses permafrost, but not seasonally frozen ground, in high mountain areas. As mountains also exist in polar areas, some overlap exists between this section and Chapter 3. Observations of permafrost are scarce (Tables 2.1 and 2.2, PERMOS, 2016; Bolch et al., 2018 <sup>[[#fn:r146|146]]</sup> ) and unevenly distributed among and within mountain regions. Unlike glaciers and snow, permafrost is a subsurface phenomenon that cannot easily be observed remotely. As a consequence, its distribution and change are less understood than for glaciers or snow, and in many mountain regions it can only be inferred (Gruber et al., 2017 <sup>[[#fn:r147|147]]</sup> ). Permafrost thaw and degradation impact people via runoff and water quality (Section 2.3.1), hazards and infrastructure (Section 2.3.2) and greenhouse gas emissions (Box 2.2).

AR5 and IPCC’s Special Report on ‘Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation’ (SREX) assessed permafrost change globally, but not separately for mountains. AR5 concluded that permafrost temperatures had increased in most regions since the early 1980s ( ''high confidence'' ), although warming rates varied regionally, and attributed this warming to increased air temperature and changes in snow cover ( ''high confidence'' ). The temperature increase for colder permafrost was generally greater than for warmer permafrost ( ''high confidence'' ). SREX found a ''likely'' warming of permafrost in recent decades and expressed ''high confidence'' that its temperatures will continue to increase. AR5 found decreases of northern high-latitude near surface permafrost for 2016–2035 to be ''very likely'' and a general retreat of permafrost extent for the end of the 21st century and beyond to be ''virtually certain'' . While some permafrost phenomena, methods of observation and scale issues in scenario simulations are specific to mountainous terrain, the basic mechanisms connecting climate and permafrost are the same in mountains and polar regions.

Between 3.6–5.2 million km 2 are underlain by permafrost in the eleven high mountain regions outlined in Figure 2.1 ( ''medium confidence'' ) based on data from two modelling studies (Gruber, 2012 <sup>[[#fn:r148|148]]</sup> ; Obu et al., 2019 <sup>[[#fn:r149|149]]</sup> ). For comparison, this is 14–21 times the area of glaciers (Section 2.2.3) in these regions (Figure 2.1) or 27–29% of the global permafrost area. The distribution of permafrost in mountains is spatially highly heterogeneous, as shown in detailed regional modelling studies (Boeckli et al., 2012 <sup>[[#fn:r150|150]]</sup> ; Bonnaventure et al., 2012 <sup>[[#fn:r151|151]]</sup> ; Westermann et al., 2015 <sup>[[#fn:r152|152]]</sup> ; Azócar et al., 2017 <sup>[[#fn:r153|153]]</sup> ; Zou et al., 2017 <sup>[[#fn:r154|154]]</sup> ).

Permafrost in the European Alps, Scandinavia, Canada, Mongolia, the Tien Shan and the Tibetan Plateau has warmed during recent decades and some observations reveal ground-ice loss and permafrost degradation ( ''high confidence'' ). The heterogeneity of mountain environments and scarcity of long-term observations challenge the quantification of representative regional or global warming rates. A recent analysis finds that permafrost at 28 mountain locations in the European Alps, Scandinavia, Canada as well as High Mountain Asia and North Asia, warmed on average by 0.19 ± 0.05 °C per decade between 2007–2016 (Biskaborn et al., 2019 <sup>[[#fn:r155|155]]</sup> ). Over longer periods, observations in the European Alps, Scandinavia, Mongolia, the Tien Shan and the Tibetan Plateau (see also Cao et al., 2018) show general warming (Table 2.1, Figure 2.5) and degradation of permafrost at individual sites (e.g., Phillips et al., 2009). Permafrost close to 0ºC warms at a lower rate than colder permafrost because ground-ice melt slows warming. Similarly, bedrock warms faster than debris or soil because of low ice content. For example, several European bedrock sites (Table 2.1) have warmed rapidly, by up to 1ºC per decade, during the past two decades. By contrast, total warming of 0.5ºC–0.8ºC has been inferred for the second half of the 20th century based on thermal gradients at depth in an ensemble of European bedrock sites (Isaksen et al., 2001 <sup>[[#fn:r156|156]]</sup> ; Harris et al., 2003 <sup>[[#fn:r157|157]]</sup> ). Warming has been shown to accelerate at sites in Scandinavia (Isaksen et al., 2007 <sup>[[#fn:r158|158]]</sup> ) and in mountains globally within the past decade (Biskaborn et al., 2019 <sup>[[#fn:r159|159]]</sup> ). During recent decades, rates of permafrost warming in the European Alps and Scandinavia exceeded values of the late 20th century ( ''limited evidence, high agreement'' ).

The observed thickness of the active layer (see Annex I: Glossary), the layer of ground above permafrost subject to annual thawing and freezing, increased in the European Alps, Scandinavia (Christiansen et al., 2010 <sup>[[#fn:r160|160]]</sup> ), and on the Tibetan Plateau during the past few decades (Table 2.2), indicating permafrost degradation. Geophysical monitoring in the European Alps during approximately the past 15 years revealed increasing subsurface liquid water content (Hilbich et al., 2008 <sup>[[#fn:r161|161]]</sup> ; Bodin et al., 2009 <sup>[[#fn:r162|162]]</sup> ; PERMOS, 2016 <sup>[[#fn:r163|163]]</sup> ), indicating gradual ground-ice loss.

During recent decades, the velocity of rock glaciers in the European Alps exceeded values of the late 20th century ( ''limited evidence, high agreement'' ). Some rock glaciers, that is, masses of ice-rich debris that show evidence of past or present movement, show increasing velocity as a transient response to warming and water input, although continued permafrost degradation would eventually inactivate them (Ikeda and Matsuoka, 2002 <sup>[[#fn:r164|164]]</sup> ). Rock glacier velocities observed in the European Alps in the 1990s were on the order of a few decimetres per year and during approximately the past 15 years they often were about 2–10 times higher (Bodin et al., 2009 <sup>[[#fn:r165|165]]</sup> ; Lugon and Stoffel, 2010 <sup>[[#fn:r166|166]]</sup> ; PERMOS, 2016 <sup>[[#fn:r167|167]]</sup> ). Destabilisation, including collapse and rapid acceleration, has been documented (Delaloye et al., 2010 <sup>[[#fn:r168|168]]</sup> ; Buchli et al., 2013 <sup>[[#fn:r169|169]]</sup> ; Bodin et al., 2016 <sup>[[#fn:r170|170]]</sup> ). One particularly long time series shows velocities around 1960 just slightly lower than during recent years (Hartl et al., 2016 <sup>[[#fn:r171|171]]</sup> ). In contrast to nearby glaciers, no clear change in rock glacier velocity or elevation was detected at a site in the Andes between 1955–1996 (Bodin et al., 2010 <sup>[[#fn:r172|172]]</sup> ). The majority of similar landforms investigated in the Alaska Brooks Range increased their velocity since the 1950s, while few others slowed down (Darrow et al., 2016 <sup>[[#fn:r173|173]]</sup> ).

Decadal-scale permafrost warming and degradation are driven by air temperature increase and additionally affected by changes in snow cover, vegetation and soil moisture. Bedrock locations, especially when steep and free of snow, produce the most direct signal of climate change on the ground thermal regime (Smith and Riseborough, 1996 <sup>[[#fn:r174|174]]</sup> ), increasing the confidence in attribution. Periods of cooling, one or few years long, have been observed and attributed to extraordinary low-snow conditions (PERMOS, 2016 <sup>[[#fn:r175|175]]</sup> ). Extreme increases of active-layer thickness often correspond with summer heat waves (PERMOS, 2016 <sup>[[#fn:r176|176]]</sup> ) and permafrost degradation can be accelerated by water percolation (Luethi et al., 2017 <sup>[[#fn:r177|177]]</sup> ). Similarity and synchronicity of interannual to decadal velocity changes of rock glaciers within the European Alps (Bodin et al., 2009 <sup>[[#fn:r178|178]]</sup> ; Delaloye et al., 2010 <sup>[[#fn:r179|179]]</sup> ) and the Tien Shan (Sorg et al., 2015 <sup>[[#fn:r180|180]]</sup> ), suggest common regional forcing such as summer air temperature or snow cover.

Because air temperature is the major driver of permafrost change, permafrost in high mountain regions is expected to undergo increasing thaw and degradation during the 21st century, with stronger consequences expected for higher greenhouse gas emission scenarios ( ''very high confidence'' ). Scenario simulations for the Tibetan Plateau until 2100 estimate permafrost area to be strongly reduced, for example by 22–64% for RCP2.6 and RCP8.5 and a spatial resolution of 0.5º (Lu et al., 2017 <sup>[[#fn:r181|181]]</sup> ). Such coarse-scale studies (Guo et al., 2012 <sup>[[#fn:r182|182]]</sup> ; Slater and Lawrence, 2013 <sup>[[#fn:r183|183]]</sup> ; Guo and Wang, 2016 <sup>[[#fn:r184|184]]</sup> ), however, are of limited use in quantifying changes and informing impact studies in steep terrain due to inadequate representation of topography (Fiddes and Gruber, 2012 <sup>[[#fn:r185|185]]</sup> ). Fine-scale simulations, on the other hand, are local or regional, limited in areal extent and differ widely in their representation of climate change and permafrost. They reveal regional and elevational differences of warming and degradation (Bonnaventure and Lewkowicz, 2011 <sup>[[#fn:r186|186]]</sup> ; Hipp et al., 2012 <sup>[[#fn:r187|187]]</sup> ; Farbrot et al., 2013 <sup>[[#fn:r188|188]]</sup> ) as well as warming rates that differ between locations (Marmy et al., 2016 <sup>[[#fn:r189|189]]</sup> ) and seasons (Marmy et al., 2013 <sup>[[#fn:r190|190]]</sup> ). While structural differences in simulations preclude a quantitative summary, these studies agree on increasing warming and thaw of permafrost for the 21st century and reveal increased loss of permafrost under stronger atmospheric warming (Chadburn et al., 2017 <sup>[[#fn:r191|191]]</sup> ). Permafrost thaw at depth is slow but can be accelerated by mountain peaks warming from multiple sides (Noetzli and Gruber, 2009 <sup>[[#fn:r192|192]]</sup> ) and deep percolation of water (Hasler et al., 2011 <sup>[[#fn:r193|193]]</sup> ). Near Mont Blanc in the European Alps, narrow peaks below 3,850 m a.s.l. may lose permafrost entirely under RCP8.5 by the end of the 21st century (Magnin et al., 2017 <sup>[[#fn:r194|194]]</sup> ). As ground-ice from permafrost usually melts slower than glacier ice, some mountain regions will transition from having abundant glaciers to having few and small glaciers but large areas of permafrost that is thawing (Haeberli et al., 2017 <sup>[[#fn:r195|195]]</sup> ).

<div id="section-2-2-4permafrost-block-2"></div>

<span id="table-2.1"></span>
<!-- START IMG -->
<!-- TABLE IMG -->
<!-- IMG TITLE -->
'''Table 2.1'''

Observed changes in permafrost mean annual ground temperature (MAGT) in mountain regions. Values are based on individual boreholes or ensembles of several boreholes. The MAGT refers to the last year in a period and is taken from a depth of 10–20 m unless the borehole is shallower. Region names refer to Figure 2.1. Numbers in brackets indicate how many sites are summarised for a particular surface type and area; the underscored value is an average. Elevation is metres above sea level (m a.s.l.)

<!-- IMG FILE -->
[[File:6620a2dfb66ec69295dd9bc6e55141a0 table2.1.png]]

<!-- END IMG -->
<div id="section-2-2-4permafrost-block-3">
</div>

<span id="table-2.2"></span>
<!-- START IMG -->
<!-- TABLE IMG -->
<!-- IMG TITLE -->
'''Table 2.2'''

Observed changes of active-layer thickness (ALT) in mountain regions. Numbers in brackets indicate how many sites are summarised for a particular surface type and area. Region names refer to Figure 2.1 . Elevation is metres above sea level (m a.s.l.).

<!-- IMG FILE -->
[[File:064eae0781b185825cf68925d98857bb table2.2.png]]

<!-- END IMG -->
<div id="section-2-2-4permafrost-block-4">
</div>

<span id="figure-2.5"></span>
<!-- START IMG -->
<!-- IMG TITLE -->
'''Figure 2.5'''

<span id="figure-2.5-mean-annual-ground-temperature-from-boreholes-in-debris-and-bedrock-in-the-european-alps-scandinavia-and-high-mountain-asia.-temperatures-differ-between-locations-and-warming-trends-can-be-interspersed-by-short-periods-of-cooling.-one-location-shows-degrading-of-permafrost.-overall-the-number-of-observed-boreholes-is-small-and-most-records-are"></span>
<!-- IMG CAPTION -->
'''Figure 2.5 | Mean annual ground temperature from boreholes in debris and bedrock in the European Alps, Scandinavia and High Mountain Asia. Temperatures differ between locations and warming trends can be interspersed by short periods of cooling. One location shows degrading of permafrost. Overall, the number of observed boreholes is small and most records are […]'''
<!-- IMG FILE -->
[[File:dd40b29ee885f4ec7ae85757776387e8 IPCC-SROCC-CH_2_5.jpg]]

Figure 2.5 | Mean annual ground temperature from boreholes in debris and bedrock in the European Alps, Scandinavia and High Mountain Asia. Temperatures differ between locations and warming trends can be interspersed by short periods of cooling. One location shows degrading of permafrost. Overall, the number of observed boreholes is small and most records are short. The depth of measurements is approximately 10 m, and years without sufficient data are omitted (Noetzli et al., 2018).

<!-- END IMG -->
<span id="lake-and-river-ice">
</span>
=== 2.2.5 Lake and River Ice ===

<div id="section-2-2-5lake-and-river-ice-block-1"></div>

Based on ''limited evidence'' , AR5 reported shorter seasonal ice cover duration during the past decades ( ''low confidence)'' , however, did not specifically address changes in mountain lakes and rivers. Observations of extent, timing, duration and thickness of lake and river ice rely mostly on ''in situ'' measurements (e.g., Sharma et al., 2019) and, increasingly on remote sensing (Duguay et al., 2014 <sup>[[#fn:r218|218]]</sup> ). Lake and river ice studies focusing specifically on mountain regions are rare but observations from lakes in the European Alps, Scandinavia and the Tibetan Plateau show highly variable trends in ice cover duration during the past decades.

For example, Cai et al. (2019) reported shorter ice cover duration for 40 lakes and longer duration for 18 lakes on the Tibetan Plateau during the period 2000–2017. Similarly, using microwave remote sensing, Du et al. (2017) found shorter ice cover duration for 43 out of 71 lakes &gt;50 km 2 including lakes on the Tibetan Plateau during 2002–2015, but only five of these had statistically significant trends ( ''p''   &lt;   0.05), due to large interannual variability. The variable trends in the duration of lake ice cover on the Tibetan Plateau between 2002–2015 corresponded to variable trends in surface water temperatures. Of 52 study lakes in this region, 31 lakes showed a mean warming rate of 0.055 ± 0.033 °C yr -1 , and 21 lakes showed a mean cooling rate of -0.053 ± 0.038 °C yr -1 during 2001–2012 (Zhang et al., 2014 <sup>[[#fn:r219|219]]</sup> ). Kainz et al. (2017) <sup>[[#fn:r220|220]]</sup> reported a significant ( ''p''   &lt;   0.05) increase in the interannual variability in ice cover duration for a subalpine lake in Austria during 1921–2015 in addition to a significant trend in later freeze on, earlier ice break up and shorter ice cover duration. A significant ( ''p'' &lt; 0.05) trend towards shorter ice cover duration was found for another Austrian alpine lake during 1972–2015 (Niedrist et al., 2018 <sup>[[#fn:r209|209]]</sup> ).

Highly variable trends were also found in the timing and magnitude of river ice jams during 1903–2015, as reported by Rokaya et al. (2018) for Canadian rivers, including rivers in the mountains. Most of the variability in river ice trends could be explained by variable water flow, in particular due to flow regulation.

There is ''high confidence'' that air temperature and solar radiation are the most important drivers to explain observed changes of lake ice dynamics (Sharma et al., 2019 <sup>[[#fn:r210|210]]</sup> ). In mountainous regions where the interannual variability in ice cover duration is high, additional drivers become important, for example, morphometry, wind exposure, salinity, and hydrology, in particular hydrological processes driven by glaciers (Kropácek et al., 2013 <sup>[[#fn:r211|211]]</sup> ; Song et al., 2014 <sup>[[#fn:r212|212]]</sup> ; Yao et al., 2016 <sup>[[#fn:r213|213]]</sup> ; Gou et al., 2017 <sup>[[#fn:r214|214]]</sup> ). Despite high spatial and temporal variability in lake and river ice cover dynamics in mountain regions there is ''limited evidence'' ( ''high agreement)'' that further air temperature increases will result in a general trend towards later freezing, earlier break-up, and shorter ice cover duration in the future (Gebre et al., 2014 <sup>[[#fn:r215|215]]</sup> ; Du et al., 2017 <sup>[[#fn:r216|216]]</sup> ).

Overall, there is only ''limited evidence'' on changes in lake and river ice specifically in the mountains, indicating a trend, but not universally, towards shorter lake ice cover duration consistent with increased water temperature.

<div id="section-2-2-5lake-and-river-ice-block-2" class="box"></div>

<span id="box-2.2-local-regional-and-global-climate-feedbacks-involving-the-mountain-cryosphere"></span>