Editing IPCC:AR6/SROCC/Chapter-3 (section)

== 3.4 Arctic Snow, Freshwater Ice and Permafrost: Changes, Consequences and Impacts ==

<span id="observations"></span>
=== 3.4.1 Observations ===

<div id="section-3-4-1-1seasonal-snow-cover"></div>

<span id="seasonal-snow-cover"></span>
==== 3.4.1.1 Seasonal Snow Cover ====

<div id="section-3-4-1-1seasonal-snow-cover-block-1"></div>

Terrestrial snow cover is a defining characteristic of the Arctic land surface for up to nine months each year, with changes influencing the surface energy budget, ground thermal regime and freshwater budget. Snow cover also interacts with vegetation, influences biogeochemical activity and affects habitats and species, with consequences for ecosystem services. Arctic land areas are almost always completely snow covered in winter, so the transition seasons of autumn and spring are key when characterising variability and change.

<div id="section-3-4-1-1seasonal-snow-cover-block-2"></div>

<span id="extent-and-duration"></span>
===== 3.4.1.1.1 Extent and duration =====

Dramatic reductions in Arctic (land areas north of 60°N) spring snow cover extent have occurred since satellite charting began in 1967 (Estilow et al., 2015). Declines in May and June of –3.5% (± 1.9%) and –13.4% respectively per decade (± 5.4%) between 1967 and 2018 (relative to the 1981–2010 mean) were determined from multiple datasets based on the methodology of (Mudryk et al., 2017 <sup>[[#fn:r1351|1351]]</sup> ) (Figure 3.10) ( ''high confidence'' ). The loss of spring snow extent is reflected in shorter snow cover duration estimated from surface observations (Bulygina et al., 2011 <sup>[[#fn:r1352|1352]]</sup> ; Brown et al., 2017 <sup>[[#fn:r1353|1353]]</sup> ), satellite data (Wang et al., 2013 <sup>[[#fn:r1354|1354]]</sup> ; Estilow et al., 2015 <sup>[[#fn:r1355|1355]]</sup> ; Anttila et al., 2018 <sup>[[#fn:r1356|1356]]</sup> ), and model-based analyses (Liston and Hiemstra, 2011 <sup>[[#fn:r1357|1357]]</sup> ) ( ''high confidence'' ). These trends range between –0.7 and –3.9 days per decade depending on region and time period, but all spring snow cover duration trends from all datasets are negative (Brown et al., 2017 <sup>[[#fn:r1358|1358]]</sup> ). These same multi-source datasets also identify reductions in autumn snow extent and duration (-0.6 to -1.4 days per decade; summarized in Brown et al., 2017) ( ''high confidence'' ). There is ''low confidence'' in positive October and November snow cover extent trends apparent in a single dataset (Hernández-Henríquez et al., 2015 <sup>[[#fn:r1359|1359]]</sup> ) because they are not replicated in other surface, satellite and model datasets (Brown and Derksen, 2013 <sup>[[#fn:r1360|1360]]</sup> ; Mudryk et al., 2017 <sup>[[#fn:r1361|1361]]</sup> ).

<div id="section-3-4-1-1seasonal-snow-cover-block-3"></div>

<span id="depth-and-water-equivalent"></span>
===== 3.4.1.1.2 Depth and water equivalent =====

Weather station observations across the Russian Arctic identify negative trends in the maximum snow depth between 1966 and 2014 (Bulygina et al., 201 <sup>[[#fn:r1362|1362]]</sup> ; Osokin and Sosnovsky, 2014 <sup>[[#fn:r1363|1363]]</sup> ). There is ''medium confidence'' in this trend because the pointwise nature of these measurements does not capture prevailing conditions across the landscape. Seasonal maximum snow depth trends over the North American Arctic are mixed and largely statistically insignificant (Vincent et al., 2015 <sup>[[#fn:r1364|1364]]</sup> ; Brown et al., 2017 <sup>[[#fn:r1365|1365]]</sup> ). The timing of maximum snow depth has shifted earlier by 2.7 days per decade for the North American Arctic (Brown et al., 2017 <sup>[[#fn:r1366|1366]]</sup> ); comparable analysis is not available for Eurasia. Gridded products from remote sensing and land surface models identify negative trends in snow water equivalent between 1981 and 2016 for both the Eurasian and North American sectors of the Arctic (Brown et al., 2017 <sup>[[#fn:r1367|1367]]</sup> ). While the snow water equivalent anomaly time series show reasonable consistency between products when averaged at the continental scale, considerable inter-dataset variability in the spatial patterns of change (Liston and Hiemstra, 2011 <sup>[[#fn:r1368|1368]]</sup> ; Park et al., 2012 <sup>[[#fn:r1369|1369]]</sup> ; Brown et al., 2017 <sup>[[#fn:r1370|1370]]</sup> ) mean there is only ''medium confidence'' in these trends.

<div id="section-3-4-1-1seasonal-snow-cover-block-4"></div>

<span id="drivers"></span>
===== 3.4.1.1.3 Drivers =====

Despite uncertainties due to sparse observations (Cowtan and Way, 2014 <sup>[[#fn:r1371|1371]]</sup> ), surface temperature has increased across Arctic land areas in recent decades (Hawkins and Sutton, 2012 <sup>[[#fn:r1372|1372]]</sup> ; Fyfe et al., 2013 <sup>[[#fn:r1373|1373]]</sup> ), driving reductions in Arctic snow extent and duration ( ''high confidence'' ) ''.'' Changes in Arctic snow extent can be directly related to extratropical temperature increases (Brutel-Vuilmet et al., 2013 <sup>[[#fn:r1374|1374]]</sup> ; Thackeray et al., 2016 <sup>[[#fn:r1375|1375]]</sup> ; Mudryk et al., 2017 <sup>[[#fn:r1376|1376]]</sup> ). Based on multiple historical datasets, there is a consistent temperature sensitivity for Arctic snow extent, with approximately 800,000 km 2 of snow cover lost per degrees Celsius warming in spring (Brown and Derksen, 2013 <sup>[[#fn:r1377|1377]]</sup> ; Brown et al., 2017), and 700,000–800,000 km 2 lost in autumn (Derksen and Brown, 2012 <sup>[[#fn:r1379|1379]]</sup> ; Brown and Derksen, 2013 <sup>[[#fn:r1380|1380]]</sup> ) ( ''high confidence'' ).

There is ''high confidence'' that darkening of snow through the deposition of black carbon and other light absorbing particles enhances snow melt (Bullard et al., 2016 <sup>[[#fn:r1381|1381]]</sup> ; Skiles et al., 2018 <sup>[[#fn:r1382|1382]]</sup> ; Boy et al., 2019 <sup>[[#fn:r1383|1383]]</sup> ). The global direct radiative forcing for black carbon in seasonal snow and over sea ice is estimated to be 0.04 W m –2 , but the effective forcing can be up to threefold greater at regional scales due to the enhanced albedo feedback triggered by the initial darkening (Bond et al., 2013). Lawrence et al. (2011) <sup>[[#fn:r1393|1393]]</sup> estimate the present-day radiative effect of black carbon and dust in land-based snow to be 0.083 W m –2 , only marginally greater than the simulated 1850 effect (0.075 W m –2 ) due to offsetting effects from increased black carbon emissions and reductions in dust darkening ( ''medium confidence'' ). Kylling et al. (2018) <sup>[[#fn:r1394|1394]]</sup> estimate a surface radiative effect of 0.292 W m –2 caused by dust deposition (largely transported from Asia) to Arctic snow, approximately half of the black carbon central scenario estimate of Flanner et al. (2007) <sup>[[#fn:r1395|1395]]</sup> . The forcing from brown carbon deposited in snow (associated with both combustion and secondary organic carbon) is estimated to be 0.09−0.25 W m –2 , with the range due to assumptions of particle absorptivity (Lin et al., 2014 <sup>[[#fn:r1396|1396]]</sup> ) ( ''low confidence'' ).

Precipitation remains a sparse and highly uncertain measurement over Arctic land areas: ''in situ'' datasets remain uncertain (Yang, 2014 <sup>[[#fn:r1397|1397]]</sup> ) and are largely regional (Kononova, 2012 <sup>[[#fn:r1398|1398]]</sup> ; Vincent et al., 2015 <sup>[[#fn:r1399|1399]]</sup> ). Atmospheric reanalyses show increases in Arctic precipitation in recent decades (Lique et al., 2016 <sup>[[#fn:r1400|1400]]</sup> ; Vihma et al., 2016 <sup>[[#fn:r1401|1401]]</sup> ), but there remains ''low confidence'' in reanalysis-based closure of the Arctic freshwater budget due to a wide spread between available reanalysis derived precipitation estimates (Lindsay et al., 2014 <sup>[[#fn:r1484|1484]]</sup> ). Despite improved process understanding, estimates of sublimation loss during blowing snow events remain a key uncertainty in the mass budget of the Arctic snowpack (Sturm and Stuefer, 2013 <sup>[[#fn:r1485|1485]]</sup> ).

<div id="section-3-4-1-2permafrost"></div>

<span id="permafrost"></span>
==== 3.4.1.2 Permafrost ====

<div id="section-3-4-1-2permafrost-block-1"></div>

<span id="temperature-1"></span>
===== 3.4.1.2.1 Temperature =====

Record high temperatures at ~10–20 m depth in the permafrost (near or below the depths affected by intra-annual fluctuation in temperature) have been documented at many long-term monitoring sites in the Northern Hemisphere circumpolar permafrost region (AMAP, 2017d <sup>[[#fn:r1386|1386]]</sup> ) (Figure 3.10) ( ''very high confidence'' ). At some locations, the temperature is 2°C–3°C higher than 30 years ago. During the decade between 2007 and 2016, the rate of increase in permafrost temperatures was 0.39°C ± 0.15°C for colder continuous zone permafrost monitoring sites, 0.20°C ± 0.10°C for warmer discontinuous zone permafrost, giving a global average of 0.29 ± 0.12°C across all polar and mountain permafrost (Biskaborn et al., 2019 <sup>[[#fn:r1387|1387]]</sup> ). Relatively smaller increases in permafrost temperature in warmer sites indicate that permafrost is thawing with heat absorbed by the ice-to-water phase change, and as a result, the active layer may be increasing in thickness. In contrast to temperature, there is only ''medium confidence'' that active layer thickness across the region has increased. This confidence level is because decadal trends vary across regions and sites (Shiklomanov et al., 2012 <sup>[[#fn:r1388|1388]]</sup> ) and because mechanical probing of the active layer can underestimate the degradation of permafrost in some cases because the surface subsides when ground ice melts and drains (Mekonnen et al., 2016 <sup>[[#fn:r1389|1389]]</sup> ; AMAP, 2017d <sup>[[#fn:r1390|1390]]</sup> ; Streletskiy et al., 2017 <sup>[[#fn:r1391|1391]]</sup> ). Permafrost in the Southern Hemisphere polar region occurs in ice-free exposed areas (Bockheim et al., 2013 <sup>[[#fn:r1392|1392]]</sup> ), 0.18% of the total land area of Antarctica (Burton-Johnson et al., 2016). This area is three orders of magnitude smaller than the 13–18 x 10 6 km 2 area underlain by permafrost in the Northern Hemisphere terrestrial permafrost region (Gruber, 2012). Antarctic permafrost temperatures are generally colder (Noetzli et al., 2017 <sup>[[#fn:r1403|1403]]</sup> ) and increased 0.37°C ± 0.10°C between 2007 and 2016 (Biskaborn et al., 2019 <sup>[[#fn:r1404|1404]]</sup> ).

<div id="section-3-4-1-2permafrost-block-2"></div>

<span id="figure-3.10"></span>
<!-- START IMG -->
<!-- IMG TITLE -->
'''Figure 3.10'''

<span id="schematic-of-important-land-surface-components-influenced-by-the-arctic-terrestrial-cryosphere-permafrost-1-ground-ice-2-river-discharge-3-abrupt-thaw-4-surface-water-5-fire-6-tundra-7-shrubs-8-boreal-forest-9-lake-ice-10-seasonal-snow-11.-time-series-of-snow-cover-extent-anomalies-in-june-relative-to-19812010-climatology-from"></span>
<!-- IMG CAPTION -->
'''Schematic of important land surface components influenced by the Arctic terrestrial cryosphere: permafrost (1); ground ice (2); river discharge (3); abrupt thaw (4); surface water (5); fire (6); tundra (7); shrubs (8); boreal forest (9); lake ice (10); seasonal snow (11). Time series of snow cover extent anomalies in June (relative to 1981–2010 climatology) from […]'''
<!-- IMG FILE -->
[[File:1f09753851439dd6e693fe4f8198d24b IPCC-SROCC-CH_3_10.jpg]]

Schematic of important land surface components influenced by the Arctic terrestrial cryosphere: permafrost (1); ground ice (2); river discharge (3); abrupt thaw (4); surface water (5); fire (6); tundra (7); shrubs (8); boreal forest (9); lake ice (10); seasonal snow (11). Time series of snow cover extent anomalies in June (relative to 1981–2010 climatology) from 5 products based on the approach of Mudryk et al. (2017) (a); permafrost temperature change normalised to a baseline period (Romanovsky et al., 2017), Region A: Continuous to discontinuous permafrost in Scandanavia, Svalbard, and Russia/Siberia, Region B: Cold continuous permafrost in northern Alaska, Northwest Territories, and NE Siberia, Region C: Cold continuous permafrost in Eastern and High Arctic Canada, Region D: Discontinuous permafrost in Interior Alaska and Northwest Canada (b), and runoff from northern flowing watersheds normalised to a baseline period (1981–2010) (Holmes et al., 2018), multi-station average (± 1 standard deviation) (c). Coupled Model Intercomparison Project Phase 5 (CMIP5) multi-model average (± 1 standard deviation) projections for different Representative Concentration Pathway (RCP) scenarios for June snow cover extent change (based on Thackeray et al., 2016) (d), area change of near-surface permafrost (e), and runoff change to the Arctic Ocean (based on McGuire et al., 2018) (f).

<!-- END IMG -->
<div id="section-3-4-1-2permafrost-block-3"></div>

<span id="ground-ice"></span>
===== 3.4.1.2.2 Ground ice =====

Permafrost thaw and loss of ground ice causes the land surface to subside and collapse into the volume previously occupied by ice, resulting in disturbance to overlying ecosystems and human infrastructure (Kanevskiy et al., 2013 <sup>[[#fn:r1405|1405]]</sup> ; Raynolds et al., 2014 <sup>[[#fn:r1406|1406]]</sup> ). Excess ice in permafrost is typical, varying for example from 40% of total volume in some sands up to 80–90% of total volume in fine-grained soil/sediments (Kanevskiy et al., 2013 <sup>[[#fn:r1407|1407]]</sup> ). Ice rich permafrost areas where impacts of thaw could be greatest include the Yedoma deposits in Siberia, Alaska, and the Yukon in Canada, with ice divided between massive wedges interspersed with frozen soil/sediment containing pore ice and smaller ice features (Schirrmeister et al., 2011 <sup>[[#fn:r1408|1408]]</sup> ; Strauss et al., 2017 <sup>[[#fn:r1409|1409]]</sup> ). Other areas including, for example, Northwestern Canada, the Canadian Archipelago, the Yamal and Gydan peninsulas of West Siberia, and smaller portions of Eastern Siberia and Alaska contain buried glacial ice bodies of significant thickness and extent (Lantuit and Pollard, 2008; Leibman et al., 2011; Kokelj et al., 2017; Coulombe et al., 2019). The location and volume of ground ice integrated across the northern permafrost region (5.63–36.55 x 10 3 km 3 , equivalent to 2–10 cm sea level rise) is known with ''medium confidence'' and with no recent updates at the circumpolar scale (Zhang et al., 2008 <sup>[[#fn:r1410|1410]]</sup> ).

<div id="section-3-4-1-2permafrost-block-4"></div>

<span id="carbon"></span>
===== 3.4.1.2.3 Carbon =====

The permafrost region represents a large, climate sensitive reservoir of organic carbon with the potential for some of this pool to be rapidly decayed and transferred to the atmosphere as CO 2 and methane as permafrost thaws in a warming climate, thus accelerating the pace of climate change (Schuur et al., 2015 <sup>[[#fn:r1414|1414]]</sup> ). The current best mean estimate of total (surface plus deep) organic soil carbon (terrestrial) in the northern circumpolar permafrost region (17.8 x 10 6 km 2 area) is 1460 to 1600 petagrams ( ''medium confidence'' ) (Pg; 1 Pg = 1 billion metric tonnes) (Schuur et al., 2018 <sup>[[#fn:r1415|1415]]</sup> ). All permafrost region soils estimated to 3 m in depth (surface) contain 1035 ± 150 Pg C (Tarnocai et al., 2009 <sup>[[#fn:r1416|1416]]</sup> ; Hugelius et al., 2014 <sup>[[#fn:r1417|1417]]</sup> ) ( ''high confidence'' ). Of the carbon in the surface, 800–1000 Pg C is perennially frozen, with the remainder contained in seasonally-thawed soils. The northern circumpolar permafrost region occupies only 15% of the total global soil area, but the 1035 Pg C adds another 50% to the rest of the 3 m soil carbon inventory (2050 Pg C for all global biomes excluding tundra and boreal; Jobbágy and Jackson, 2000 <sup>[[#fn:r1418|1418]]</sup> ; Schuur et al., 2015 <sup>[[#fn:r1419|1419]]</sup> ).

Substantial permafrost carbon exists below 3 m depth ( ''medium confidence'' ). Deep carbon (&gt;3 m) has been best quantified for the Yedoma region of Siberia and Alaska, characterised by wind- and water-moved permafrost sediments tens of meters thick. The Yedoma region covers a 1.4 x 10 6 km 2 area that remained ice-free during the last Ice Age (Strauss et al., 2013 <sup>[[#fn:r1420|1420]]</sup> ) and accounts for 327–466 Pg C in deep sediment accumulations below 3 m (Strauss et al., 2017).

The current inventory has also highlighted additional carbon pools that are likely to be present but are so poorly quantified ( ''low confidence'' ) that they cannot yet be added into the number reported above. There are deep terrestrial soil/sediment deposits outside of the Yedoma region that may contain about 400 Pg C (Schuur et al., 2015 <sup>[[#fn:r1421|1421]]</sup> ). An additional pool is organic carbon remaining in permafrost but that is now submerged on shallow Arctic sea shelves that were formerly exposed as terrestrial ecosystems during the Last Glacial Maximum ~20,000 years ago (Walter et al., 2007 <sup>[[#fn:r1423|1423]]</sup> ). This permafrost is degrading slowly due to seawater intrusion, and it is not clear what amounts of permafrost and organic carbon still remain in the sediment versus what has already been converted to greenhouse gases. A recent synthesis of permafrost extent for the Beaufort Sea shelf showed that most remaining subsea permafrost in that region exists near shore with much reduced area ( ''high confidence'' ) as compared to original subsea permafrost maps that outlined the entire 3 x 10 6 km 2 shelf area (&lt;120 m below sea level depth) that was formerly exposed as land (Ruppel et al., 2016 <sup>[[#fn:r1424|1424]]</sup> ). These observations are supported by similar studies in the Siberian Arctic Seas (Portnov et al., 2013 <sup>[[#fn:r1425|1425]]</sup> ), and by modelling that suggests that subsea permafrost would be thawed many meters below the seabed under current submerged conditions (Anisimov et al., 2012 <sup>[[#fn:r1426|1426]]</sup> ; AMAP, 2017d <sup>[[#fn:r1427|1427]]</sup> ; Angelopoulos et al., 2019 <sup>[[#fn:r1428|1428]]</sup> ).

<div id="section-3-4-1-2permafrost-block-5"></div>

<span id="drivers-1"></span>
===== 3.4.1.2.4 Drivers =====

Changes in temperature and precipitation act as gradual ‘press’ (i.e., continuous) disturbances that directly affect permafrost by modifying the ground thermal regime, as discussed in Section 3.4.1.2.1. Climate change can also modify the occurrence and magnitude of abrupt physical disturbances such as fire, and soil subsidence and erosion resulting from ice rich permafrost thaw (thermokarst). These ‘pulse’ (i.e., discrete) disturbances (Smith et al., 2009 <sup>[[#fn:r1429|1429]]</sup> ) often are part of the ongoing disturbance and successional cycle in Arctic and boreal ecosystems (Grosse et al., 2011 <sup>[[#fn:r1430|1430]]</sup> ), but changing rates of occurrence alter the landscape distribution of successional ecosystem states, with permafrost characteristics defined by the ecosystem and climate state (Kanevskiy et al., 2013 <sup>[[#fn:r1431|1431]]</sup> ).

Pulse disturbances often rapidly remove the insulating soil organic layer, leading to permafrost degradation (Gibson et al., 2018 <sup>[[#fn:r1423|1423]]</sup> ). Of all pulse disturbance types, wildfire affects the most high-latitude land area annually at the continental scale. In some well-studied regions, there is ''high confidence'' that area burned, fire frequency and extreme fire years are higher now than the first half of the last century, or even the last 10,000 years (Kasischke and Turetsky, 2006 <sup>[[#fn:r1433|1433]]</sup> ; Flannigan et al., 2009 <sup>[[#fn:r1434|1434]]</sup> ; Kelly et al., 2013 <sup>[[#fn:r1435|1435]]</sup> ; Hanes et al., 2019 <sup>[[#fn:r1436|1436]]</sup> ) ''.'' Recent climate warming has been linked to increased wildfire activity in the boreal forest regions in Alaska and western Canada where this has been studied (Gillett, 2004 <sup>[[#fn:r1437|1437]]</sup> ; Veraverbeke et al., 2017 <sup>[[#fn:r1438|1438]]</sup> ). Based on satellite imagery, an estimated 80,000 km 2 of boreal area was burned globally per year from 1997 to 2011 (van der Werf et al., 2010 <sup>[[#fn:r1439|1439]]</sup> ; Giglio et al., 2013 <sup>[[#fn:r1440|1440]]</sup> ). Extreme fire years in northwest Canada during 2014 and Alaska during 2015 doubled the long-term (1997–2011) average area burned annually in this region (Canadian Forest Service, 2017), surpassing Eurasia to contribute 60% of the global boreal area burned (van der Werf et al., 2010 <sup>[[#fn:r1441|1441]]</sup> ; Randerson et al., 2012 <sup>[[#fn:r1442|1442]]</sup> ; Giglio et al., 2013 <sup>[[#fn:r1443|1443]]</sup> ). These extreme North American fire years were balanced by lower-than-average area burned in Eurasian forests, resulting in a 5% overall increase in global boreal area burned. The annual area burned in Arctic tundra is generally small compared to the forested boreal biome. In Alaska—the only region where estimates of burned area exist for both boreal forest and tundra vegetation types—tundra burning averaged approximately 270 km 2 yr -1 during the last half century (French et al., 2015 <sup>[[#fn:r1445|1445]]</sup> ), accounting for 7% of the average annual area burned throughout the state (Pastick et al., 2017 <sup>[[#fn:r1446|1446]]</sup> ). There is ''high confidence'' that changes in the fire regime are degrading permafrost faster than had occurred over the historic successional cycle (Turetsky et al., 2011 <sup>[[#fn:r1447|1447]]</sup> ; Rupp et al., 2016 <sup>[[#fn:r1448|1448]]</sup> ; Pastick et al., 2017 <sup>[[#fn:r1449|1449]]</sup> ), and that the effect of this driver of permafrost change is under-represented in the permafrost temperature observation network.

Abrupt permafrost thaw occurs when changing environmental and ecological conditions interact with geomorphological processes. Melting ground ice causes the ground surface to subside. Pooling or flowing water causes localised permafrost thaw and sometimes mass erosion. Together, these localised feedbacks can thaw through meters of permafrost within a short time, much more rapidly than would be caused by increasing air temperature alone. This process is a pulse disturbance to permafrost that can occur in response to climate, such as an extreme precipitation event (Balser et al., 2014 <sup>[[#fn:r1450|1450]]</sup> ; Kokelj et al., 2015 <sup>[[#fn:r1451|1451]]</sup> ), or coupled with other disturbances such as wildfire that affects the ground thermal regime (Jones et al., 2015a <sup>[[#fn:r1452|1452]]</sup> ). There is ''medium confidence'' in the importance of abrupt thaw for driving change in permafrost at the circumpolar scale because it occurs at point locations rather than continuously across the landscape, but the risk for widespread change from this mechanism remains high because of the rapidity of change in these locations (Kokelj et al., 2017 <sup>[[#fn:r1453|1453]]</sup> ; Nitze et al., 2018 <sup>[[#fn:r1454|1454]]</sup> ). New research at the global scale has revealed that 3.6 x 10 6 km 2 , about 20% of the northern permafrost region, appears to be vulnerable to abrupt thaw (Olefeldt et al., 2016 <sup>[[#fn:r1455|1455]]</sup> ).

<div id="section-3-4-1-3freshwater-systems"></div>

<span id="freshwater-systems"></span>
==== 3.4.1.3 Freshwater Systems ====

<div id="section-3-4-1-3freshwater-systems-block-1"></div>

There is increasing awareness of the influence of a changing climate on freshwater systems across the Arctic, and associated impacts on hydrological, biogeophysical and ecological processes (Prowse et al., 2015 <sup>[[#fn:r1456|1456]]</sup> ; Walvoord and Kurylyk, 2016 <sup>[[#fn:r1457|1457]]</sup> ), and northern populations (Takakura, 2018 <sup>[[#fn:r1458|1458]]</sup> ) (Section 3.4.3.3.1). Assessing these impacts requires consideration of complex interconnected processes, many of which are incompletely observed. The increasing imprint of human development, such as flow regulation on major northerly flowing rivers adds complexity to the determination of climate-driven changes.

<div id="section-3-4-1-3freshwater-systems-block-2"></div>

<span id="freshwater-ice"></span>
===== 3.4.1.3.1 Freshwater ice =====

Long-term ''in situ'' river ice records indicate that the duration of ice cover in Russian Arctic rivers decreased by 7–20 days between 1955 and 2012 (Shiklomanov and Lammers, 2014 <sup>[[#fn:r1459|1459]]</sup> ) ( ''high confidence'' ). This is consistent with historical reductions in Arctic river ice cover derived from models (Park et al., 2015) and regional analysis of satellite data (Cooley and Pavelsky, 2016 <sup>[[#fn:r1461|1461]]</sup> ).

Analysis of satellite imagery between 2000 and 2013 identified a significant trend of earlier spring ice break-up across all regions of the Arctic (Šmejkalová et al., 2016 <sup>[[#fn:r1462|1462]]</sup> ); independent satellite data showed approximately 80% of Arctic lakes experienced declines in ice cover duration during 2002–2015, due to both a later freeze-up and earlier break-up (Du et al., 2017 <sup>[[#fn:r1463|1463]]</sup> ) ( ''high confidence'' ). There are indications that lake ice across Alaska has thinned in recent decades (Alexeev et al., 2016 <sup>[[#fn:r1464|1464]]</sup> ), but ice thickness trends are not available at the pan-Arctic scale. Analysis of satellite data over northern Alaska show that approximately one-third of bedfast lakes (the entire water volume freezes by the end of winter) experienced a regime change to floating ice over the 1992–2011 period (Surdu et al., 2014 <sup>[[#fn:r1465|1465]]</sup> ; Arp et al., 2015 <sup>[[#fn:r1466|1466]]</sup> ). This can result in degradation of underlying permafrost (Arp et al., 2016 <sup>[[#fn:r1467|1467]]</sup> ; Bartsch et al., 2017 <sup>[[#fn:r1468|1468]]</sup> ). Lakes of the central and eastern Canadian High Arctic are transitioning from a perennial to seasonal ice regime (Surdu et al., 2016 <sup>[[#fn:r1469|1469]]</sup> ).

<div id="section-3-4-1-3freshwater-systems-block-3"></div>

<span id="runoff-and-surface-water"></span>
===== 3.4.1.3.2 Runoff and surface water =====

A general trend of increasing discharge has been observed for large Siberian (Troy et al., 2012 <sup>[[#fn:r1470|1470]]</sup> ; Walvoord and Kurylyk, 2016 <sup>[[#fn:r1471|1471]]</sup> ) and Canadian (Ge et al., 2013 <sup>[[#fn:r1472|1472]]</sup> ; Déry et al., 2016 <sup>[[#fn:r1473|1473]]</sup> ) rivers that drain to the Arctic Ocean ( ''medium confidence'' ). Between 1976 and 2017, trends are 3.3 ± 1.6% for Eurasian rivers and 2.0 ± 1.8% for North American rivers (Holmes et al., 2018 <sup>[[#fn:r1474|1474]]</sup> ) (Figure 3.10). Extreme regional runoff events have also been identified (Stuefer et al., 2017 <sup>[[#fn:r1475|1475]]</sup> ). An observed increase in baseflow in the North American (Walvoord and Striegl, 2007 <sup>[[#fn:r1476|1476]]</sup> ; St. Jacques and Sauchyn, 2009) and Eurasian Arctic (Smith et al., 2007 <sup>[[#fn:r1477|1477]]</sup> ; Duan et al., 2017 <sup>[[#fn:r1478|1478]]</sup> ) over the last several decades is attributable to permafrost thaw and concomitant enhancement in groundwater discharge. The timing of spring season peak flow is generally earlier (Ge et al., 2013 <sup>[[#fn:r1479|1479]]</sup> ; Holmes et al., 2015 <sup>[[#fn:r1480|1480]]</sup> ). There is consistent evidence of decreasing summer season discharge for the Yenisei, Lena, and Ob watersheds in Siberia (Ye et al., 2003 <sup>[[#fn:r1481|1481]]</sup> ; Yang et al., 2004a <sup>[[#fn:r1482|1482]]</sup> ; Yang et al., 2004b <sup>[[#fn:r1483|1483]]</sup> ) and the majority of northern Canadian rivers (Déry et al., 2016 <sup>[[#fn:r1484|1484]]</sup> ). Long-term records indicate water temperature increases (Webb et al., 2008 <sup>[[#fn:r1485|1485]]</sup> ; Yang and Peterson, 2017 <sup>[[#fn:r1486|1486]]</sup> ); attribution to rising air temperatures is complicated by the influence of reservoir regulation over Siberian regions (Liu et al., 2005 <sup>[[#fn:r1487|1487]]</sup> ; Lammers et al., 2007 <sup>[[#fn:r1488|1488]]</sup> ). Increases in discharge and water temperature in the spring season represent notable freshwater and heat fluxes to the Arctic Ocean (Yang et al., 2014 <sup>[[#fn:r1489|1489]]</sup> ).

A large proportion of low-lying Arctic land areas are covered by lakes because permafrost limits surface water drainage and supports ponding even across areas with high moisture deficits (Grosse et al., 2013 <sup>[[#fn:r1490|1490]]</sup> ). While thaw in continuous permafrost is linked to intensified thermokarst activity and subsequent ponding (resulting in lake/wetland expansion), observations of change in surface water coverage across the Arctic are regionally variable (Nitze et al., 2017 <sup>[[#fn:r1491|1491]]</sup> ; Ulrich et al., 2017 <sup>[[#fn:r1492|1492]]</sup> ; Pastick et al., 2019 <sup>[[#fn:r1493|1493]]</sup> ). In landscapes with degrading ice-wedge polygons, subsidence can reduce inundation, increase runoff, and decrease surface water (Liljedahl et al., 2016 <sup>[[#fn:r1494|1494]]</sup> ; Perreault et al., 2017 <sup>[[#fn:r1495|1495]]</sup> ). In discontinuous permafrost, thaw opens up pathways of subsurface flow, improving the connection among inland water systems which supports the drainage of lakes and overall reduction in surface water cover (Jepsen et al., 2013 <sup>[[#fn:r1496|1496]]</sup> ). Enhanced subsurface connectivity from thaw in discontinuous permafrost serves tempers short-term lake fluctuations (Rey et al., 2019 <sup>[[#fn:r1497|1497]]</sup> ).

<div id="section-3-4-1-3freshwater-systems-block-4"></div>

<span id="drivers-2"></span>
===== 3.4.1.3.3 Drivers =====

There is ''high confidence'' that environmental drivers of Arctic surface water change are diverse and depend on local and regional factors such as permafrost properties and geomorphology (Nitze et al., 2018 <sup>[[#fn:r1498|1498]]</sup> ). Thermokarst lake expansion has been observed in the continuous permafrost of northern Siberia (Smith et al., 2005 <sup>[[#fn:r1499|1499]]</sup> ; Polishchuk et al., 2015 <sup>[[#fn:r1500|1500]]</sup> ) and Alaska (Jones et al., 2011 <sup>[[#fn:r1501|1501]]</sup> ); surface water area reduction has been observed in discontinuous permafrost of central and southern Siberia (Smith et al., 2005 <sup>[[#fn:r1502|1502]]</sup> ; Sharonov et al., 2012 <sup>[[#fn:r1503|1503]]</sup> ), western Canada (Labrecque et al., 2009 <sup>[[#fn:r1504|1504]]</sup> ; Carroll et al., 2011 <sup>[[#fn:r1505|1505]]</sup> ; Lantz and Turner, 2015 <sup>[[#fn:r1506|1506]]</sup> ) and interior Alaska (Chen et al., 2012 <sup>[[#fn:r1507|1507]]</sup> ; Rover et al., 2012 <sup>[[#fn:r1508|1508]]</sup> ). Increased evaporation from warmer/longer summers, decreased recharge due to reductions in snow melt volume, and dynamic processes such as ice-jam flooding (Chen et al., 2012 <sup>[[#fn:r1509|1509]]</sup> ; Bouchard et al., 2013 <sup>[[#fn:r1510|1510]]</sup> ; Jepsen et al., 2015 <sup>[[#fn:r1511|1511]]</sup> ) are important considerations for understanding observed surface water area change across the Arctic.

Satellite and model-derived estimates of evapotranspiration show increases across the Arctic (Rawlins et al., 2010 <sup>[[#fn:r1512|1512]]</sup> ; Liu et al., 2014 <sup>[[#fn:r1513|1513]]</sup> ; Liu et al., 2015b <sup>[[#fn:r1514|1514]]</sup> ; Fujiwara et al., 2016 <sup>[[#fn:r1515|1515]]</sup> ; Suzuki et al., 2018 <sup>[[#fn:r1516|1516]]</sup> ) ( ''medium confidence'' ). Increases in the seasonal active layer thickness impact temporary water storage and thus runoff regimes in drainage basins. Formation of taliks underneath lakes and rivers may result in reconnection of surface with sub-permafrost ground water aquifers with varying hydrological consequences depending on local geological and hydraulic settings (Wellman et al., 2013 <sup>[[#fn:r1517|1517]]</sup> ).

<span id="projections-1"></span>
=== 3.4.2 Projections ===

<div id="section-3-4-2-1seasonal-snow"></div>

<span id="seasonal-snow"></span>
==== 3.4.2.1 Seasonal Snow ====

<div id="section-3-4-2-1seasonal-snow-block-1"></div>

Historical simulations from CMIP5 models tend to underestimate observed reductions in spring snow cover extent due to uncertainty in the parameterisation of snow processes (Essery, 2013 <sup>[[#fn:r1518|1518]]</sup> ; Thackeray et al., 2014 <sup>[[#fn:r1519|1519]]</sup> ), challenges in simulating snow-albedo feedback (Qu and Hall, 2014 <sup>[[#fn:r1520|1520]]</sup> ; Fletcher et al., 2015 <sup>[[#fn:r1521|1521]]</sup> ; Li et al., 2016b <sup>[[#fn:r1522|1522]]</sup> ), unrealistic temperature sensitivity (Brutel-Vuilmet et al., 2013 <sup>[[#fn:r1523|1523]]</sup> ; Mudryk et al., 2017 <sup>[[#fn:r1524|1524]]</sup> ), and biases in climatological spring snow cover (Thackeray et al., 2016 <sup>[[#fn:r1525|1525]]</sup> ). The role of precipitation biases is not well understood (Thackeray et al., 2016 <sup>[[#fn:r1526|1526]]</sup> ).

Reductions in Arctic snow cover duration are projected by the CMIP5 multi-model ensemble due to later snow onset in the autumn and earlier snow melt in spring (Brown et al., 2017 <sup>[[#fn:r1527|1527]]</sup> ) driven by increased surface temperature over essentially all Arctic land areas (Hartmann et al., 2013). There is ''high confidence'' that projected snow cover declines are proportional to the amount of future warming in each model realisation (Thackeray et al., 2016 <sup>[[#fn:r1528|1528]]</sup> ; Mudryk et al., 2017 <sup>[[#fn:r1529|1529]]</sup> ). Projections to mid-century are primarily dependent on natural variability and model dependent uncertainties rather than the choice of forcing scenario (Hodson et al., 2013 <sup>[[#fn:r1530|1530]]</sup> ). By end of century, however, differences between scenarios emerge. Under RCP2.6 and RCP4.5, Arctic snow cover duration stabilises at 5–10% reduction (compared to a 1986–2005 reference period); under RCP8.5, snow cover duration declines reach –15 to –25% (Brown et al., 2017 <sup>[[#fn:r1531|1531]]</sup> ) (Figure 3.10) ( ''high confidence'' ).

Positive Arctic snow water equivalent changes emerge across the eastern Eurasian Arctic by mid-century for both RCP4.5 and RCP8.5 (Brown et al., 2017 <sup>[[#fn:r1532|1532]]</sup> ) ( ''medium confidence'' ). Projected snow water equivalent increases across the North American Arctic are only modest, emerge later in the century, and only under RCP8.5 (Brown et al., 2017 <sup>[[#fn:r1533|1533]]</sup> ). These projected increases are due to enhanced snowfall (Krasting et al., 2013 <sup>[[#fn:r1534|1534]]</sup> ) from a more moisture-rich Arctic atmosphere coupled with winter season temperatures that remain sufficiently low for precipitation to fall as snow. There is ''low confidence'' in changes to snow properties such as density and stratigraphy (relevant for understanding the impacts of changes to Arctic snow on ecosystems) which are not resolved directly by climate model simulations, but require detailed snow physics models.

<div id="section-3-4-2-2permafrost"></div>

<span id="permafrost-1"></span>
==== 3.4.2.2 Permafrost ====

<div id="section-3-4-2-2permafrost-block-1"></div>

Circumpolar- or global-scale models represent permafrost degradation in response to warming scenarios as increases in thaw depth only. The CMIP5 models project with ''high confidence'' that thaw depth will increase and areal extent of near-surface permafrost will decrease substantially (Koven et al., 2013 <sup>[[#fn:r1535|1535]]</sup> ; Slater and Lawrence, 2013 <sup>[[#fn:r1536|1536]]</sup> ) (Figure 3.10). However, there is only ''medium confidence'' in the magnitude of these changes due to at least a five-fold range of estimated present day near-surface permafrost area (&lt;5 – &gt;25 x 10 6 km 2 ) by these models. This was caused by a wide range of model sensitivity in permafrost area to air temperature change, resulting in a large range of projected near-surface permafrost loss by 2100: 2–66% for RCP2.6 (24 ± 16%; ''likely'' range), 15–87% under RCP4.5 and 30–99% (69 ± 20%; ''likely'' range) under RCP8.5. A more recent analysis of near-surface permafrost trends from a subset of models that self-identified as structurally representing the permafrost region had a significantly smaller range of estimated present day near-surface permafrost area (13.1–19.3 x 10 6 km 2 ; mean ± SD, 14.1 ± 3.5 x 10 6 km 2 ) (McGuire et al., 2018 <sup>[[#fn:r1537|1537]]</sup> ). This subset of models also showed large reductions of near-surface permafrost area, averaging a 90% loss (12.7 ± 5.1×10 6 km 2 ) of permafrost area by 2300 for RCP8.5 and 29% loss (4.1 ± 0.6×10 6 km 2 ) for RCP4.5, with much of that long-term loss already occurring by 2100.

Pulse disturbances are not included in the permafrost projections described above, and there is ''high confidence'' that fire and abrupt thaw will accelerate change in permafrost relative to climate effects alone, if the rates of these disturbances increase. The observed trend of increasing fire is projected to continue for the rest of the century across most of the tundra and boreal region for many climate scenarios, with the boreal region projected to have the greatest increase in total area burned (Balshi et al., 2009 <sup>[[#fn:r1538|1538]]</sup> ; Kloster et al., 2012 <sup>[[#fn:r1539|1539]]</sup> ; Wotton et al., 2017 <sup>[[#fn:r1540|1540]]</sup> ). Due to vegetation-climate interactions, there is only ''medium confidence'' in projections of future area burned. As fire activity increases, flammable vegetation, such as the black spruce forest that dominates boreal Alaska, is projected to decline as it is replaced by low-flammability deciduous forest (Johnstone et al., 2011 <sup>[[#fn:r1541|1541]]</sup> ; Pastick et al., 2017 <sup>[[#fn:r1542|1542]]</sup> ). In other regions such as western Canada, by contrast, black spruce could be replaced by the even more flammable jack pine, creating regional-scale feedbacks that increase the spread of fire on the landscape (Héon et al., 2014 <sup>[[#fn:r1543|1543]]</sup> ). A regional process-model study of Alaska projected annual median area burned during the 21st century to be 1.3-1.7 times higher compared with the historical average  (Pastick et al., 2017 <sup>[[#fn:r1544|1544]]</sup> ). Fire also appears to be expanding as a novel disturbance into tundra and forest-tundra boundary regions previously protected by a cool, moist climate (Jones et al., 2009 <sup>[[#fn:r1545|1545]]</sup> ; Hu et al., 2010 <sup>[[#fn:r1546|1546]]</sup> ; Hu et al., 2015 <sup>[[#fn:r1547|1547]]</sup> ) ( ''medium confidence'' ). Annual tundra area burned in Alaska is projected to double under RCP6.0 from a historic rate of 270 km 2 yr -1 to 500–610 km 2 yr -1 over the 21st century (Hu et al., 2015 <sup>[[#fn:r1548|1548]]</sup> ). A statistical approach projected a fourfold increase in the 30-year probability of fire occurrence in the forest-tundra boundary by 2100 (Young et al., 2017 <sup>[[#fn:r1549|1549]]</sup> ). In contrast to fire, there has not yet been a comprehensive circumpolar projection of how abrupt thaw rates may change in the future, but one component of abrupt thaw, change in abrupt thaw lake area, has been projected to increase to increase by 53% under RCP8.5 (Walter Anthony et al., 2018 <sup>[[#fn:r1550|1550]]</sup> ) above the 1.4 x 10 6 km 2 of small lakes and ponds that currently exist in the permafrost region (Muster et al., 2017 <sup>[[#fn:r1551|1551]]</sup> ). As a result, there is ''low confidence'' in the ability to assess the magnitude by which abrupt thaw across the entire landscape will affect regional permafrost, even though this mechanism for rapid change appears critically important for projecting future change (Kokelj et al., 2017 <sup>[[#fn:r1552|1552]]</sup> ).

<div id="section-3-4-2-3freshwater-systems"></div>

<span id="freshwater-systems-1"></span>
==== 3.4.2.3 Freshwater Systems ====

<div id="section-3-4-2-3freshwater-systems-block-1"></div>

Climate model simulations project a warmer and wetter Arctic (Krasting et al., 2013 <sup>[[#fn:r1553|1553]]</sup> ), with increased specific humidity due to enhanced evaporation (Laîné et al., 2014 <sup>[[#fn:r1554|1554]]</sup> ), and moisture flux convergence increases into the Arctic (Skific and Francis, 2013 <sup>[[#fn:r1555|1555]]</sup> ). Increased cold-season precipitation is projected across the Arctic by CMIP5 models (Lique et al., 2016 <sup>[[#fn:r1556|1556]]</sup> ) due to increased moisture flux convergence from outside the Arctic (Zhang et al., 2012 <sup>[[#fn:r1557|1557]]</sup> ) and enhanced moisture availability from reduced sea ice cover (Bintanja and Selten, 2014 <sup>[[#fn:r1558|1558]]</sup> ) ( ''high confidence'' ). Increases in precipitation extremes are also projected over northern watersheds (Kharin et al., 2013 <sup>[[#fn:r1559|1559]]</sup> ; Sillmann et al., 2013 <sup>[[#fn:r1560|1560]]</sup> ), while rain on snow events are expected to increase (Hansen et al., 2014 <sup>[[#fn:r1561|1561]]</sup> ). A net increased ratio of precipitation minus evaporation is projected, resulting in increased freshwater flux from the land surface to the Arctic Ocean, projected to be 30% above current values by 2100 under RCP4.5 (Haine et al., 2015 <sup>[[#fn:r1562|1562]]</sup> ) (Figure 3.10). This is consistent with CMIP5 model projections of increased discharge from Arctic watersheds (van Vliet et al., 2013 <sup>[[#fn:r1563|1563]]</sup> ; Gelfan et al., 2016 <sup>[[#fn:r1564|1564]]</sup> ; MacDonald et al., 2018 <sup>[[#fn:r1565|1565]]</sup> ). The water temperature of this increased discharge is projected to be approximately 1°C warmer than current conditions, increasing the heat flux to Arctic Ocean (van Vliet et al., 2013 <sup>[[#fn:r1566|1566]]</sup> ).

Lake ice phenology is sensitive to projected changes in surface temperature (Sharma et al., 2019 <sup>[[#fn:r1567|1567]]</sup> ). Lake ice models project an earlier spring break-up of between 10–25 days by mid-century (compared with 1961–1990), and up to a 15-day delay in the freeze-up for lakes in the North American Arctic, with more extreme reductions for coastal regions (Brown and Duguay, 2011 <sup>[[#fn:r1568|1568]]</sup> ; Dibike et al., 2011 <sup>[[#fn:r1569|1569]]</sup> ; Prowse et al., 2011 <sup>[[#fn:r1570|1570]]</sup> ) ( ''medium confidence'' ). Mean maximum ice thickness is projected to decrease by 10–50 cm over the same period (Brown and Duguay, 2011 <sup>[[#fn:r1571|1571]]</sup> ). High-latitude warming is projected to drive earlier river ice break-up in spring due to both decreasing ice strength, and earlier onset of peak discharge (Cooley and Pavelsky, 2016 <sup>[[#fn:r1572|1572]]</sup> ). Complex interplay between hydrology and hydraulics in controlling spring flooding and ice jam events complicate projections of these events (Prowse et al., 2010 <sup>[[#fn:r1573|1573]]</sup> ; Prowse et al., 2011 <sup>[[#fn:r1574|1574]]</sup> ).

<span id="consequences-and-impacts-1"></span>
=== 3.4.3 Consequences and Impacts ===

<div id="section-3-4-3-1global-climate-feedbacks"></div>

<span id="global-climate-feedbacks"></span>
==== 3.4.3.1 Global Climate Feedbacks ====

<div id="section-3-4-3-1global-climate-feedbacks-block-1"></div>

<span id="carbon-cycle"></span>
===== 3.4.3.1.1 Carbon cycle =====

Climate warming is expected to change the storage of carbon in vegetation and soils in northern regions, and net carbon transferred to the atmosphere as CO 2 and methane acts as a feedback to accelerate global climate change. There is ''high confidence'' that the northern region acted as a net carbon sink as carbon accumulated in terrestrial ecosystems over the Holocene (Loisel et al., 2014 <sup>[[#fn:r1575|1575]]</sup> ; Lindgren et al., 2018 <sup>[[#fn:r1576|1576]]</sup> ). There is ''medium evidence'' with ''low agreement'' whether changing climate in the modern period has shifted these ecosystems into net carbon sources. Syntheses of ecosystem CO 2 fluxes have alternately showed tundra ecosystems as carbon sinks or neutral averaged across the circumpolar region for the 1990s and 2000s (McGuire et al., 2012 <sup>[[#fn:r1577|1577]]</sup> ), or carbon sources over the same time period (Belshe et al., 2013 <sup>[[#fn:r1578|1578]]</sup> ). Both syntheses agree that the summer growing season is a period of net carbon uptake into terrestrial ecosystems ( ''high confidence'' ), and this uptake appears to be increasing as a function of vegetation density/biomass (Ueyama et al., 2013 <sup>[[#fn:r1579|1579]]</sup> ). The discrepancy between these syntheses may be a result of CO 2 release rates during the non-summer season that are now thought to be higher than previously estimated ( ''high confidence'' ) (Webb et al., 2016 <sup>[[#fn:r1580|1580]]</sup> ) or the separation of upland and wetland ecosystem types, which was done in one synthesis but not the other. Moisture status is a primary control over ecosystem carbon sink/source strength with wetlands more often than not still acting as annual net carbon sinks even while methane is emitted (Lund et al., 2010 <sup>[[#fn:r1581|1581]]</sup> ). Recent aircraft measurements of atmospheric CO 2 concentrations over Alaska showed that tundra regions of Alaska were a consistent net CO 2 source to the atmosphere, whereas boreal forest regions were either neutral or net CO 2 sinks for the period 2012–2014 (Commane et al., 2017 <sup>[[#fn:r1582|1582]]</sup> ). That study region as a whole was estimated to be a net carbon source of 25 ± 14 Tg CO 2 -C yr -1 averaged over the land area of both biomes for the entire study period. For comparison to projected global emissions, this would be equivalent to a net source of 0.3 Pg CO 2 -C yr -1 assuming the Alaska study region (1.6 x 10 6 km 2 ) could be scaled to the entire northern circumpolar permafrost region soil area (17.8 x 10 6 km 2 ).

The permafrost soil carbon pool is climate sensitive and an order of magnitude larger than carbon stored in plant biomass (Schuur et al., 2018 <sup>[[#fn:r1583|1583]]</sup> ) ( ''very high confidence'' ). Initial estimates were converging on a range of cumulative emissions from soils to the atmosphere by 2100, but recent studies have actually widened that range somewhat (Figure 3.11) ( ''medium confidence'' ). Expert assessment and laboratory soil incubation studies suggest that substantial quantities of C (tens to hundreds Pg C) could potentially be transferred from the permafrost carbon pool into the atmosphere under RCP8.5 (Schuur et al., 2013 <sup>[[#fn:r1584|1584]]</sup> ; Schädel et al., 2014 <sup>[[#fn:4|4]]</sup> ) . Global dynamical models supported these findings, showing potential carbon release from the permafrost zone ranging from 37–174 Pg C by 2100 under high emission climate warming trajectories, with an average across models of 92 ± 17 Pg C (mean ± SE) (Zhuang et al., 2006 <sup>[[#fn:r1585|1585]]</sup> ; Koven et al., 2011 <sup>[[#fn:r1586|1586]]</sup> ; Schaefer et al., 2011 <sup>[[#fn:r1587|1587]]</sup> ; MacDougall et al., 2012 <sup>[[#fn:r1588|1588]]</sup> ; Burke et al., 2013 <sup>[[#fn:r1589|1589]]</sup> ; Schaphoff et al., 2013 <sup>[[#fn:r1590|1590]]</sup> ; Schneider von Deimling et al., 2015 <sup>[[#fn:r1591|1591]]</sup> ). This range is generally consistent with several newer data-driven modelling approaches that estimated that soil carbon releases by 2100 (for RCP8.5) will be 57 Pg C (Koven et al., 2015 <sup>[[#fn:r1592|1592]]</sup> ) and 87 Pg C (Schneider von Deimling et al., 2015 <sup>[[#fn:r1593|1593]]</sup> ), as well as an updated estimate of 102 Pg C from one of the previous models (MacDougall and Knutti, 2016 <sup>[[#fn:r1594|1594]]</sup> ). However, the latest model runs performed with either structural enhancements to better represent permafrost carbon dynamics (Burke et al., 2017a <sup>[[#fn:r1595|1595]]</sup> ), or common environmental input data (McGuire et al., 2016 <sup>[[#fn:r1596|1596]]</sup> ) show similar soil carbon losses, but also indicate the potential for stimulated plant growth (nutrients, temperature/growing season length, CO 2 fertilisation) to offset some (Kleinen and Brovkin, 2018 <sup>[[#fn:r1597|1597]]</sup> ) or all of these losses, at least during this century, by sequestering new carbon into plant biomass and increasing carbon inputs into the surface soil (McGuire et al., 2018 <sup>[[#fn:r1598|1598]]</sup> ). These future carbon emission levels would be a significant fraction of those projected from fossil fuels with implications for allowable carbon budgets that are consistent with limiting global warming, but will also depend on how vegetation responds ( ''high confidence'' ). Furthermore, there is ''high confidence'' that climate scenarios that involve mitigation (e.g., RCP4.5) will help to dampen the response of carbon emissions from the Arctic and boreal regions.

Northern ecosystems contribute significantly to the global methane budget, but there is ''low confidence'' about the degree to which additional methane from northern lakes, ponds, wetland ecosystems, and the shallow Arctic Ocean shelves is currently contributing to increasing atmospheric concentrations. Analyses of atmospheric concentrations in Alaska concluded that local ecosystems surrounding the observation site have not changed in the exchange of methane from the 1980s until the present, which suggests that either the local wetland ecosystems are responding similarly to other northern wetland ecosystems, or that increasing atmospheric methane concentrations in northern observation sites is derived from methane coming from mid-latitudes (Sweeney et al., 2016 <sup>[[#fn:r1600|1600]]</sup> ). However, this contrasts with indirect integrated estimates of methane emissions from observations of expanding permafrost thaw lakes that suggest a release of an additional 1.6–5 Tg CH 4 yr –1 over the last 60 years (Walter Anthony et al., 2014 <sup>[[#fn:r1601|1601]]</sup> ). At the same time, there is ''high confidence'' that methane fluxes at the ecosystem to regional scale have been under-observed, in part due to the low  solubility of methane in water leading to ebullution (bubbling) flux to the atmosphere that is heterogeneous in time and space. Some new quantifications include: cold-season methane emissions that can be &gt;50% of the annual budget of terrestrial ecosystems (Zona et al., 2016 <sup>[[#fn:r1602|1602]]</sup> ); geological methane seeps that may be climate sensitive if permafrost currently serves as a cap preventing atmospheric release (Walter Anthony et al., 2012 <sup>[[#fn:r1603|1603]]</sup> ; Ruppel and Kessler, 2016 <sup>[[#fn:r1604|1604]]</sup> ; Kohnert et al., 2017 <sup>[[#fn:r1605|1605]]</sup> ); estimates of shallow Arctic Ocean shelf methane emissions where the range of estimates based on methane concentrations in air and water has widened with more observations and now ranges from 3 Tg CH 4 yr –1 (Thornton et al., 2016 <sup>[[#fn:r1606|1606]]</sup> ) to 17 Tg CH 4 yr –1 (Shakhova et al., 2013 <sup>[[#fn:r1607|1607]]</sup> ). Observations such as these underlie the fact that source estimates for methane made from atmospheric observations are typically lower than methane source estimates made from upscaling of ground observations (e.g., Berchet et al., 2016), and this problem has not improved, even at the global scale, over several decades of research (Saunois et al., 2016 <sup>[[#fn:r1608|1608]]</sup> ; Crill and Thornton, 2017 <sup>[[#fn:r1609|1609]]</sup> ).

In many of the dynamical model projections previously discussed, methane release is not explicitly represented because fluxes are small even though higher global warming potential of methane makes these emissions relatively more important than on a mass basis alone. Global models that do include methane show that emissions may already (from 2000 to 2012) be increasing at a rate of 1.2 Tg CH 4 yr –1 in the northern region as a direct response to temperature (Riley et al., 2011 <sup>[[#fn:r1610|1610]]</sup> ; Gao et al., 2013 <sup>[[#fn:r1611|1611]]</sup> ; Poulter et al., 2017 <sup>[[#fn:r1612|1612]]</sup> ). A model intercomparison study forecast northern methane emissions to increase from 18 Tg CH 4 yr –1 to 42 Tg CH 4 yr –1 under RCP8.5 by 2100 largely as a result of an increase in wetland extent (Zhang et al., 2017 <sup>[[#fn:r1613|1613]]</sup> ). However, projected methane emissions are sensitive to changes in surface hydrology (Lawrence et al., 2015 <sup>[[#fn:r1614|1614]]</sup> ) and a suite of models that were thought to perform well in high-latitude ecosystems showed a general soil drying trend even as the overall water cycle intensified (McGuire et al., 2018 <sup>[[#fn:r1615|1615]]</sup> ). Furthermore, most models described above do not include many of the abrupt thaw processes that can result in lake expansion, wetland formation, and massive erosion and exposure to decomposition of previously frozen carbon-rich permafrost, leading to ''medium confidence'' in future model projections of methane. Recent studies that addressed some of these landscape controls over future emissions projected increases in methane above the current levels on the order 10–60 Tg CH 4 yr -1 under RCP8.5 by 2100 (Schuur et al., 2013 <sup>[[#fn:r1616|1616]]</sup> ; Koven et al., 2015 <sup>[[#fn:r1617|1617]]</sup> ; Lawrence et al., 2015 <sup>[[#fn:r1618|1618]]</sup> ; Schneider von Deimling et al., 2015 <sup>[[#fn:r1619|1619]]</sup> ; Walter Anthony et al., 2018 <sup>[[#fn:r1620|1620]]</sup> ). These additional methane fluxes are projected to cause 40–70% of total permafrost-affected radiative forcing in this century even though methane emissions are much less than CO 2 by mass (Schneider von Deimling et al., 2015 <sup>[[#fn:r1621|1621]]</sup> ; Walter Anthony et al., 2018 <sup>[[#fn:r1622|1622]]</sup> ). As with total carbon emissions, there is ''high confidence'' that mitigation of anthropogenic methane sources could help to dampen the impact of increased methane emissions from the Arctic and boreal regions (Christensen et al., 2019 <sup>[[#fn:r1623|1623]]</sup> ).

<div id="section-3-4-3-1global-climate-feedbacks-block-2"></div>

<span id="energy-budget"></span>
===== 3.4.3.1.2 Energy budget =====

Warming induced reductions in the duration and extent of Arctic spring snow cover (Section 3.4.1.1) lower albedo because snow-free land reflects much less solar radiation than snow. The corresponding increase in net radiation absorption at the surface provides a positive feedback to global temperatures (Flanner et al., 2011 <sup>[[#fn:r1624|1624]]</sup> ; Qu and Hall, 2014 <sup>[[#fn:r1625|1625]]</sup> ; Thackeray and Fletcher, 2016 <sup>[[#fn:r1626|1626]]</sup> ) ( ''high confidence'' ). Estimates of increases in global net solar energy flux due to snow cover loss range from 0.10–0.22 W m –2 (± 50%; ''medium confidence'' ) depending on dataset and time period (Flanner et al., 2011 <sup>[[#fn:r1627|1627]]</sup> ; Chen et al., 2015 <sup>[[#fn:r1628|1628]]</sup> ; Singh et al., 2015 <sup>[[#fn:r1629|1629]]</sup> ; Chen et al., 2016b <sup>[[#fn:r1630|1630]]</sup> ). Sources of uncertainty include the range in observed spring snow cover extent trends (Hori et al., 2017 <sup>[[#fn:r1631|1631]]</sup> ) and the influence of clouds on shortwave feedbacks (Sedlar, 2018 <sup>[[#fn:r1632|1632]]</sup> ; Sledd and L’Ecuyer, 2019 <sup>[[#fn:r1633|1633]]</sup> ). Terrestrial snow changes also affect the longwave energy budget via altered surface emissivity (Huang et al., 2018 <sup>[[#fn:r1634|1634]]</sup> ). Climate model simulations show that changes in snow cover dominate land surface related positive feedbacks to atmospheric heating (Euskirchen et al., 2016 <sup>[[#fn:r1635|1635]]</sup> ), but regional variations in surface albedo are also influenced by vegetation (Loranty et al., 2014 <sup>[[#fn:r1636|1636]]</sup> ). There is evidence for positive sensitivity of surface temperatures to increased northern hemisphere boreal and tundra leaf area index, which contributes a positive feedback to warming (Forzieri et al., 2017 <sup>[[#fn:r1637|1637]]</sup> ).

<div id="section-3-4-3-2ecosystems-and-their-services"></div>

<span id="ecosystems-and-their-services"></span>
==== 3.4.3.2 Ecosystems and their Services ====

<div id="section-3-4-3-2ecosystems-and-their-services-block-1"></div>

<span id="vegetation"></span>
===== 3.4.3.2.1 Vegetation =====

Changes in tundra vegetation can have important ecosystem effects, in particular on hydrology, carbon and nutrient cycling and surface energy balance, which together impact permafrost (e.g., Myers-Smith and Hik, 2013; Frost and Epstein, 2014 <sup>[[#fn:r1638|1638]]</sup> ; Nauta et al., 2014 <sup>[[#fn:r1639|1639]]</sup> ). Aside from physical impacts, changing vegetation influences the diversity and abundance of herbivores (e.g., Fauchald et al., 2017b; Horstkotte et al., 2017 <sup>[[#fn:r1640|1640]]</sup> ) in the Arctic. The overall trend for tundra vegetation across the 36–year satellite period (1982–2017) shows increasing above ground biomass (greening) throughout a majority of the circumpolar Arctic ( ''high confidence'' ) (Xu et al., 2013a <sup>[[#fn:r1641|1641]]</sup> ; Ju and Masek, 2016 <sup>[[#fn:r1642|1642]]</sup> ; Bhatt et al., 2017 <sup>[[#fn:r1643|1643]]</sup> ). Increasing greenness has been in some cases linked with shifts in plant species dominance away from graminoids (grasses and sedges) towards shrubs ( ''high confidence'' ) (Myers-Smith et al., 2015 <sup>[[#fn:r1644|1644]]</sup> ). Within the overall trend of increases (greening), some tundra areas show declines (browning) (Bhatt et al., 2017).

The spatial variation in greening and browning trends in tundra are also not consistent over time (decadal scale) and can vary across landform/ecosystem types (Lara et al., 2018 <sup>[[#fn:r1645|1645]]</sup> ), suggesting interactions between the changing environment and the biological components of the system that control these trends. There is ''high confidence'' that increases in summer, spring and winter temperatures lead to tundra greening, as well as increases in growing season length (e.g., Vickers et al., 2016; Myers-Smith and Hik, 2018 <sup>[[#fn:r1646|1646]]</sup> ) that are in part linked to reductions in Arctic Ocean sea ice cover (Bhatt et al., 2017 <sup>[[#fn:r1647|1647]]</sup> ; Macias-Fauria et al., 2017 <sup>[[#fn:r1648|1648]]</sup> ). Other factors that stimulate tundra greening include increases in snow water equivalent and soil moisture (Westergaard-Nielsen et al., 2017 <sup>[[#fn:r1649|1649]]</sup> ), increases in active layer thickness (via nutrient availability or changes in moisture), changes in herbivore activity, and to a lesser degree, human use of the land (e.g., Salmon et al., 2016; Horstkotte et al., 2017; Martin et al., 2017; Yu et al., 2017). Research on tundra browning is more limited but suggests causal mechanisms that include changes in winter climate—specifically reductions in snow cover due to winter warming events that expose tundra to subsequent freezing and desiccation—, insect and pathogen outbreaks, increased herbivore grazing and ground ice melting and subsidence that increases surface water (Phoenix and Bjerke, 2016; Bjerke et al., 2017) ( ''medium confidence'' ).

Projections of tundra vegetation distribution across the Arctic by 2050 in response to changing environmental conditions suggest that the areal extent of most tundra types will decrease by at least 50% (Pearson et al., 2013) ( ''medium confidence'' ). Woody shrubs and trees are projected to expand to cover 24–52% of the current tundra region by 2050, or 12–33% if tree dispersal is restricted. Adding to this, the expansion of fire into tundra that has not experienced large-scale disturbance for centuries causes large reductions in soil carbon stocks (Mack et al., 2011), shifts in vegetation composition and productivity (Bret-Harte et al., 2013), and can lead to widespread permafrost degradation (Jones et al., 2015a) at faster rates than would occur by changing environmental conditions alone. In tundra regions, graminoid (grasses and sedges) tundra is projected to be replaced by more flammable shrub tundra in future climate scenarios, and tree migration into tundra could further increase fuel loading (Pastick et al., 2017) ( ''medium confidence'' ).

Similar to tundra, boreal forest vegetation shows trends of both greening and browning over multiple years in different regions across the satellite record (Beck and Goetz, 2011; Ju and Masek, 2016) ( ''high confidence'' ). Here, patterns of changing vegetation are a result of direct responses to changes in climate (temperature, precipitation and seasonality) and other driving factors for vegetation (nutrients, disturbance) similar to what has been reported in tundra. While boreal forest may expand at the northern edge (Pearson et al., 2013), climate projections suggest that it could diminish at the southern edge and be replaced by lower biomass woodland/shrublands (Koven, 2013; Gauthier et al., 2015) ( ''medium confidence'' ). Furthermore, changes in fire disturbance are leading to shifts in landscape distribution of early and late successional ecosystem types, which is also a major factor in satellite trends. Fires that burn deeply into the organic soil layer can alter permafrost stability, hydrology and vegetation. Loss of the soil organic layer exposes mineral soil seedbeds (Johnstone et al., 2009), leading to recruitment of deciduous tree and shrub species that do not establish on organic soil (Johnstone et al., 2010). This recruitment has been shown to shift post-fire vegetation to alternate successional trajectories (Johnstone et al., 2010). Model projections suggest that Alaskan boreal forest soon may cross a point where recent increases in fire activity have made deciduous stands as abundant as spruce stands on the landscape (Mann et al., 2012) ( ''medium confidence'' ). This projected trend of increasing deciduous forest at the expense of evergreen forest is mirrored in Russian and Chinese boreal forests as well (Shakhova et al., 2013; Shuman et al., 2015; Wu et al., 2017) ( ''medium confidence'' ).

<div id="section-3-4-3-2ecosystems-and-their-services-block-2"></div>

<span id="figure-3.11"></span>
<!-- START IMG -->
<!-- IMG TITLE -->
'''Figure 3.11'''

<span id="figure-3.11-estimates-of-cumulative-net-soil-carbon-pool-change-for-the-northern-circumpolar-permafrost-region-by-2100-following-medium-and-high-emission-scenarios-e.g.-representative-concentration-pathway-rcp4.5-and-rcp8.5-or-equivalent.-cumulative-carbon-amounts-are-shown-in-gigatons-c-1-gt-c1-billion-metric-tonnes-with-source-negative-values-indicating-net-carbon"></span>
<!-- IMG CAPTION -->
'''Figure 3.11 | Estimates of cumulative net soil carbon pool change for the northern circumpolar permafrost region by 2100 following medium and high emission scenarios (e.g., Representative Concentration Pathway (RCP)4.5 and RCP8.5 or equivalent). Cumulative carbon amounts are shown in Gigatons C (1 Gt C=1 billion metric tonnes), with source (negative values) indicating net carbon […]'''
<!-- IMG FILE -->
[[File:3a8fa4a00a17dd1acc4f3014b3ee2577 IPCC-SROCC-CH_3_11.jpg]]

Figure 3.11 | Estimates of cumulative net soil carbon pool change for the northern circumpolar permafrost region by 2100 following medium and high emission scenarios (e.g., Representative Concentration Pathway (RCP)4.5 and RCP8.5 or equivalent). Cumulative carbon amounts are shown in Gigatons C (1 Gt C=1 billion metric tonnes), with source (negative values) indicating net carbon movement from soil to the atmosphere and sink (positive values) indicating the reverse. Some data-constrained models differentiated CO2 and CH4; bars show total carbon by weight, paired bars with * indicate CO2-equivalent, which takes into account the global warming potential of CH4. Ensemble mean bars refer to the model average for the Permafrost Carbon Model Intercomparison Project [5 models]. Bars that do not start at zero are in part informed by expert assessment and are shown as ranges; all other bars represent model mean estimates. Data are from 1 (Schuur et al., 2013 <sup>[[#fn:r1650|1650]]</sup> ); 2 (Schaefer et al., 2014 <sup>[[#fn:r1651|1651]]</sup> ) [8 models]; 3 (Schuur et al., 2015 <sup>[[#fn:r1652|1652]]</sup> ); 4 (Koven et al., 2015 <sup>[[#fn:r1653|1653]]</sup> ; Schneider von Deimling et al., 2015 <sup>[[#fn:r1654|1654]]</sup> ; Walter Anthony et al., 2018 <sup>[[#fn:r1655|1655]]</sup> ); 5 (MacDougall and Knutti, 2016 <sup>[[#fn:r1656|1656]]</sup> ; Burke et al., 2017a <sup>[[#fn:r1657|1657]]</sup> ; Kleinen and Brovkin, 2018 <sup>[[#fn:r1658|1658]]</sup> ); 6 (McGuire et al., 2018 <sup>[[#fn:r1659|1659]]</sup> ).

<!-- END IMG -->
<div id="section-3-4-3-2ecosystems-and-their-services-block-3"></div>

<span id="wildlife"></span>
===== 3.4.3.2.2 Wildlife =====

Reindeer and caribou ( ''Rangifer tarandus'' ), through their numbers and ecological role as a large-bodied herbivores, are a key driver of Arctic ecology. The seasonal migrations that characterise ''Rangifer'' link the coastal tundra to the continental boreal forests for some herds, while others live year-round on the tundra. Population estimates and trends exist for most herds, and indicate that pan-Arctic migratory tundra ''Rangifer'' have declined from about 5 million in the 1990s to about 2 million in 2017 (Gunn, 2016 <sup>[[#fn:r1682|1682]]</sup> ; Fauchald et al., 2017a <sup>[[#fn:r1683|1683]]</sup> ) ( ''high confidence'' ). Numbers have recently increased for two Alaska herds and the Porcupine caribou herd straddling Yukon and Alaska is at a historic high.

There is ''low confidence'' in understanding the complex drivers of observed ''Rangifer'' changes. Hunting and predation (the latter exacerbated by modification of the landscape for exploration and resource extraction; Dabros et al., 2018 <sup>[[#fn:r1684|1684]]</sup> ) increase in importance as populations decline. Climate strongly influences productivity: extremes in heat, drought, winter icing and snow depth reduce ''Rangifer'' survival (Mallory and Boyce, 2017 <sup>[[#fn:r1685|1685]]</sup> ). Changes in the timing of sea ice formation have direct effects on risks during ''Rangifer'' migration via inter-island movement and connection to the mainland (Poole et al., 2010 <sup>[[#fn:r1686|1686]]</sup> ). Summer warming is changing the composition of tundra plant communities, modifying the relationship between climate, forage and ''Rangifer'' (Albon et al., 2017 <sup>[[#fn:r1687|1687]]</sup> ), which also impacts other Arctic species such as musk ox ( ''Ovibos moschatus)'' (Schmidt et al., 2015 <sup>[[#fn:r1688|1688]]</sup> ). As polar trophic systems are highly connected (Schmidt et al., 2017 <sup>[[#fn:r1689|1689]]</sup> ), changes will propagate through the ecosystem with effects on other herbivores such as geese and voles, as well as predators such as wolves (Hansen et al., 2013 <sup>[[#fn:r1690|1690]]</sup> ; Klaczek et al., 2016 <sup>[[#fn:r1691|1691]]</sup> ).

In northern Fennoscandia, there are approximately 600,000 semi-domesticated reindeer. Lichen rangelands are key to sustaining reindeer carrying capacity, with variable response to climate change: enhanced summer precipitation increases lichen biomass, while an increase in winter precipitation lowers it (Kumpula et al., 2014 <sup>[[#fn:r1692|1692]]</sup> ). Fire disturbance reduces the amount of pasture available for domestic reindeer and increases predation on herding lands (Lavrillier and Gabyshev, 2017 <sup>[[#fn:r1693|1693]]</sup> ). Later ice formation on waterbodies can impact herding activities (Turunen et al., 2016 <sup>[[#fn:r1694|1694]]</sup> ). Ice formation from rain-on-snow events is associated with population changes including cases of catastrophic mass starvation (Bartsch et al., 2010 <sup>[[#fn:r1695|1695]]</sup> ; Forbes et al., 2016 <sup>[[#fn:r1696|1696]]</sup> ), but there is no evidence of trends in rain-on-snow events (Cohen et al., 2015 <sup>[[#fn:r1697|1697]]</sup> ; Dolant et al., 2017 <sup>[[#fn:r1698|1698]]</sup> ).

Management of keystone species requires an understanding of pathogens and disease in the context of climate warming, but evidence of changing patterns across northern ecosystems (spanning terrestrial, aquatic, and marine environments) is hindered by an incomplete picture of pathogen diversity and distribution (Hoberg, 2013 <sup>[[#fn:r1699|1699]]</sup> ; Jenkins et al., 2013 <sup>[[#fn:r1700|1700]]</sup> ; Cook et al., 2017 <sup>[[#fn:r1701|1701]]</sup> ). Among ungulates, it is ''virtually certain'' that the emergence of disease attributed to nematode pathogens has accelerated since 2000 in the Canadian Arctic islands and Fennoscandia (Kutz et al., 2013 <sup>[[#fn:r1702|1702]]</sup> ; Hoberg and Brooks, 2015 <sup>[[#fn:r1703|1703]]</sup> ; Laaksonen et al., 2017 <sup>[[#fn:r1704|1704]]</sup> ; Kafle et al., 2018 <sup>[[#fn:r1705|1705]]</sup> ). Discovery of the pathogenic bacterium ''Erysipelothrix rhusiopathiae'' has been linked to massive and widespread mortality among muskoxen from the Canadian Arctic Archipelago; loss of &gt;50% of the population since 2010 may be attributable to disease interacting with extreme temperature events, although unequivocal links to climate have not been established (Kutz et al., 2015 <sup>[[#fn:r1706|1706]]</sup> ; Forde et al., 2016a <sup>[[#fn:r1707|1707]]</sup> ; Forde et al., 2016b <sup>[[#fn:r1708|1708]]</sup> ). Anthrax is projected to expand northward in response to warming, and resulted in substantial mortality events for reindeer on the Yamal Peninsula of Russia in 2016 with mobilisation of bacteria possibly from a frozen reindeer carcass or melting permafrost (Walsh et al., 2018 <sup>[[#fn:r1709|1709]]</sup> ). In concert with climate forcing, pathogens are ''very likely'' responsible for increasing mortality in Arctic ungulates (muskox, caribou/reindeer) and alteration of transmission patterns in marine food chains, broadly threatening sustainability of subsistence and commercial hunting and fishing and safety of traditional foods for northern cultures at high latitudes (Jenkins et al., 2013 <sup>[[#fn:r1710|1710]]</sup> ; Kutz et al., 2014 <sup>[[#fn:r1711|1711]]</sup> ; Hoberg et al., 2017 <sup>[[#fn:r1712|1712]]</sup> ).

<div id="section-3-4-3-2ecosystems-and-their-services-block-4"></div>

<span id="freshwater"></span>
===== 3.4.3.2.3 Freshwater =====

Climate-driven changes in seasonal ice and permafrost conditions influence water quality ( ''high confidence'' ). Shortened duration of freshwater ice cover (more light absorption, increased nutrient input) is expected to result in higher primary productivity (Hodgson and Smol, 2008 <sup>[[#fn:r1713|1713]]</sup> ; Vincent et al., 2011 <sup>[[#fn:r1714|1714]]</sup> ; Griffiths et al., 2017b <sup>[[#fn:r1715|1715]]</sup> ) and may also encourage greater methane emissions from Arctic lakes (Greene et al., 2014 <sup>[[#fn:r1716|1716]]</sup> ; Tan and Zhuang, 2015 <sup>[[#fn:r1717|1717]]</sup> ). Thaw slumps, active layer detachments and peat plateau collapse affect surface water connectivity (Connon et al., 2014 <sup>[[#fn:r1718|1718]]</sup> ) and enhance sediment, particulate and solute fluxes in river and stream networks (Kokelj et al., 2013 <sup>[[#fn:r1719|1719]]</sup> ). The transfer of enhanced nutrients from land to water (driven by active layer thickening and thermokarst processes; Abbott et al., 2015 <sup>[[#fn:r1720|1720]]</sup> ; Vonk et al., 2015 <sup>[[#fn:r1721|1721]]</sup> ) has been linked to heightened autotrophic productivity in freshwater ecosystems (Wrona et al., 2016 <sup>[[#fn:r1722|1722]]</sup> ). Still, there is ''low confidence'' in the influence of permafrost changes on dissolved organic carbon, because of competing mechanisms that influence carbon export. Permafrost thaw could contribute to the mobilisation of previously frozen organic carbon (Abbott et al., 2014 <sup>[[#fn:r1723|1723]]</sup> ; Wickland et al., 2018 <sup>[[#fn:r1724|1724]]</sup> ; Walvoord et al., 2019 <sup>[[#fn:r1725|1725]]</sup> ) thereby enhancing both particulate and dissolved organic carbon export to aquatic systems. Increased delivery of this dissolved carbon from enhanced river discharge to the Arctic Ocean (Section 3.4.3.1.2) can exacerbate regionally extreme aragonite undersaturation of shelf waters (Semiletov et al., 2016 <sup>[[#fn:r1726|1726]]</sup> ) driven by ocean uptake of anthropogenic CO 2 (Section 3.2.1.2.4). Conversely, reduced dissolved organic carbon export could accompany permafrost thaw as (1) water infiltrates deeper and has longer residence times for decomposition (Striegl et al., 2005 <sup>[[#fn:r1727|1727]]</sup> ) and (2) the proportion of groundwater (typically lower in dissolved organic carbon and higher in DIC than runoff) to total streamflow increases (Walvoord and Striegl, 2007 <sup>[[#fn:r1728|1728]]</sup> ). Increased thermokarst also has the potential to impact freshwater cycling of inorganic carbon (Zolkos et al., 2018 <sup>[[#fn:r1729|1729]]</sup> ).

Enhanced subsurface water fluxes resulting from permafrost degradation has consequences for inorganic natural and anthropogenic constituents. Emerging evidence suggests large natural stores of mercury (Schuster et al., 2018 <sup>[[#fn:r1730|1730]]</sup> ; St Pierre et al., 2018 <sup>[[#fn:r1731|1731]]</sup> ) and other trace elements in permafrost (Colombo et al., 2018 <sup>[[#fn:r1732|1732]]</sup> ) may be released upon thaw, thereby having effects (largely unknown at this point) on aquatic ecosystems. In parallel, increased development activity in the Arctic is ''likely'' to lead to enhanced local sources of anthropogenic chemicals of emerging Arctic concern, including siloxanes, parabens, flame retardants, and per- and polyfluoroalkyl substances (AMAP, 2017c <sup>[[#fn:r1733|1733]]</sup> ). For legacy pollutants, there is ''high confidence'' that black carbon and persistent organic pollutants (e.g., hexachlorocyclohexanes, polycyclic aromatic hydrocarbons, and polychlorinated biphenyls) can be transferred downstream and affect water quality (Hodson, 2014 <sup>[[#fn:r1734|1734]]</sup> ). Lakes can become sinks of these contaminants, while floodplains can be contaminated (Sharma et al., 2015).

There is ''high confidence'' that habitat loss or change due to climate change impact Arctic fishes. Thinning ice on lakes and streams changes the overwintering habitat for aquatic fauna by impacting winter water volumes and dissolved oxygen levels (Leppi et al., 2016 <sup>[[#fn:r1735|1735]]</sup> ). Surface water loss, reduced surface water connectivity among aquatic habitats, and changes to the timing and magnitude of seasonal flows (Section 3.4.1.2) result in a direct loss of spawning, feeding, or rearing habitats (Poesch et al., 2016 <sup>[[#fn:r1736|1736]]</sup> ). Changes to permafrost landscapes have reduced freshwater habitat available for fish and other aquatic biota, including aquatic invertebrates upon which the fish depend for food (Chin et al., 2016 <sup>[[#fn:r1737|1737]]</sup> ). Gullying deepens channels (Rowland et al., 2011 <sup>[[#fn:r1738|1738]]</sup> ; Liljedahl et al., 2016 <sup>[[#fn:r1739|1739]]</sup> ) that otherwise may connect lake habitats occupied by fishes. This can lead to the loss of surface water connectivity, limit fish access to key habitats, and lower fish diversity (Haynes et al., 2014 <sup>[[#fn:r1740|1740]]</sup> ; Laske et al., 2016 <sup>[[#fn:r1741|1741]]</sup> ). Small connecting stream channels, which are vulnerable to drying, provide necessary migratory pathways for fishes, allowing them to access spawning and summer rearing grounds (Heim et al., 2016 <sup>[[#fn:r1742|1742]]</sup> ; McFarland et al., 2017 <sup>[[#fn:r1743|1743]]</sup> ).

Changes to the timing, duration and magnitude of high surface flow events in early and late summer threaten Arctic fish dispersal and migration activities (Heim et al., 2016 <sup>[[#fn:r1744|1744]]</sup> ) ( ''high confidence'' ). Timing of important life history events such as spawning can become mismatched with changing stream flows (Lique et al., 2016 <sup>[[#fn:r1745|1745]]</sup> ). There is regional evidence that migration timing has shifted earlier and winter egg incubation temperature has increased for pink salmon ( ''Oncorhynchus gorbuscha'' ), directly related to warming (Taylor, 2007 <sup>[[#fn:r1746|1746]]</sup> ). While long-term, pan-Arctic data on run timing of fishes are limited, phenological shifts could create mismatches with food availability or habitat suitability in both marine and freshwater environments for anadromous species, and in freshwater environments for freshwater resident species. Changes to the Arctic growing season (Xu et al., 2013a <sup>[[#fn:r1747|1747]]</sup> ) increase the risk of drying of surface water habitats and pose a potential mismatch in seasonal availability of food in rearing habitats.

Freshwater systems across the Arctic are relatively shallow, and thus are expected to warm ( ''high confidence'' ). This may make some surface waters inhospitably warm for cold water species such as Arctic Grayling ( ''Thymallus arcticus'' ) and whitefishes ( ''Coregonus spp'' .), or may increase the risk of ''Saprolegnia'' ''fungus'' that appears to have recently spread rapidly, infecting whitefishes at much higher rates in Arctic Alaska than noted in the past (Sformo et al., 2017 <sup>[[#fn:r1748|1748]]</sup> ). High infection rates may be driven by stress or nutrient enrichment from thawing permafrost, which increases pathogen virulence with fish (Wedekind et al., 2010 <sup>[[#fn:r1749|1749]]</sup> ). Warmer water and longer growing seasons will also affect food abundance because invertebrate life histories and production are temperature and degree-day dependent (Régnière et al., 2012 <sup>[[#fn:r1750|1750]]</sup> ). Increased nutrient export from permafrost loss (Frey et al., 2007 <sup>[[#fn:r1751|1751]]</sup> ), facilitated by warmer temperatures, will ''likely'' increase food resources for consumers, but the impact on lower trophic levels within food webs is not clearly understood.

<div id="section-3-4-3-2ecosystems-and-their-services-block-5" class="box"></div>

<span id="box-3.4-impacts-and-risks-for-polar-biodiversity-from-range-shifts-and-species-invasions-related-to-climate-change"></span>