Editing IPCC:AR6/WGII/Chapter-4 (section)

== 4.4 Projected Changes in the Hydrological Cycle Due to Climate Change ==

<div id="h1-5-siblings" class="h1-siblings"></div>

The terrestrial hydrological cycle is projected to intensify through a higher exchange of water between the land surface and the atmosphere. A rise of near-surface atmospheric water capacity is projected because of greater warming leading to changes in the atmospheric circulation patterns, the intensification of the convection processes, and the increased temperature of the underlying surface. Continuation of projected warming and other physical mechanisms will further accelerate the melting of snow cover and glaciers and thawing of permafrost ( ''high confidence'' ).

Methodologically, the projected changes in the hydrological cycle due to climate change are assessed directly from climate models or hydrological system models driven by the climate models’ projections (SM4.1). The latter is simulated by the CMIP-based multi-model experiments carried out under the scenarios of future climate forcing and socioeconomic changes (e.g., RCPs, SSPs scenarios) or the pre-assigned global warming levels over the 21st century. Since AR5, there has been an improvement of the physical basis of the climate change impact projections owing to the advances in modelling clouds, precipitation, surface fluxes, vegetation, snow, floodplains, groundwater and other processes relevant to the water cycle ( [[#Douville--2021|Douville et al., 2021]] ) (SM4.1).

The subsections highlight the projected responses of these hydrological systems/processes to multiple drivers, high variability and the uncertainty of the projections, depending on regions, seasons, temporal and spatial scales, and the influence of the non-climatic factors.

<div id="4.4.1" class="h2-container"></div>

<span id="projected-changes-in-precipitation-evapotranspiration-and-soil-moisture"></span>
=== 4.4.1 Projected Changes in Precipitation, Evapotranspiration and Soil Moisture ===

<div id="h2-19-siblings" class="h2-siblings"></div>

<div id="4.4.1.1" class="h3-container"></div>

<span id="projected-changes-in-precipitation"></span>
==== 4.4.1.1 Projected Changes in Precipitation ====

<div id="h3-4-siblings" class="h3-siblings"></div>

WGI ( [[#Douville--2021|Douville et al., 2021]] ) concludes with ''high confidence'' that without large-scale reduction in GHG emissions, global warming is projected to cause substantial changes in the water cycle at both global and regional scales. However, WGI also noted large uncertainties in many aspects of regional water cycle projections by climate models. Water cycle variability and extremes are projected to increase faster than average changes in most regions of the world and under all emission scenarios ( ''high confidence'' ). The concept of ‘wetter regions get wetter, drier regions get drier’ from AR5 ( [[#Collins--2013)|Collins et al., 2013)]] is assessed by AR6 WGI ( [[#Douville--2021|Douville et al., 2021]] ) as too simplistic. WGI ( [[#Seneviratne--2021|Seneviratne et al., 2021]] ) further concludes that heavy precipitation will generally become more frequent and more intense with additional global warming.

In the CMIP6 multi-model ensemble, as in previous generations of ensembles, the projected changes in annual mean precipitation vary substantially across the world. Importantly, in most land regions, the future changes are subject to high uncertainty even in the sign of the projected change ( ''high confidence'' ). Figure 4.10 illustrates this using the 5th, 50th and 95th percentile changes across the ensemble at individual grid points. For any given location, the range of projected changes generally increases with global warming ( ''high confidence'' ).

<div id="_idContainer051" class="Figure"></div>

[[File:25e3489875648caacd743003f03ca5fb IPCC_AR6_WGII_Figure_4_010.png]]

'''Figure 4.10 |''' '''Projected percentage changes in annual mean precipitation at global warming levels (GWLs) of 4°C (top), 2°C (middle) and 1''' '''.''' '''5°C (bottom) for the CMIP6 multi-model ensemble of GCMs driven by the SSP5-8.5 scenario.''' For any given GWL, similar ranges of changes are seen with other scenarios that reach that GWL, and the difference between scenarios is smaller than the ensemble uncertainty ( [[#Seneviratne--2021|Seneviratne et al., 2021]] ). The distribution of outcomes is shown at local scales with the 5th, 50th and 95th percentile precipitation changes in individual grid boxes. Note that these are uncertainties at the individual point and are not spatially coherent, that is, they do not represent plausible global patterns of change. Results for 1.5°C, 2°C and 4°C global warming are defined as 20-year means relative to 1850–1900 and use 40, 40 and 31 ensemble members, respectively, due to some members not reaching 4°C global warming.

For example, in parts of the Indian sub-continent, the projected changes in mean precipitation at 1.5°C global warming range from a 10–20% decrease to a 40–50% increase. The multi-model median change is close to zero. Most other regions show a smaller range of changes (except for very dry regions where a small absolute change in precipitation appears as a larger percentage change). Nevertheless, across most global land regions, both increases and decreases in precipitation are projected across the ensemble. At 1.5°C global warming, a complete consensus on increased precipitation is seen only in the central and eastern Sahel, south-central Asia, parts of Greenland and Antarctica, and the far northern regions of North America and Asia, with projected increases in the latter ranging up to 20–30%. No land regions see a complete consensus on decreased precipitation, but South America, southern Africa and the Mediterranean region show a stronger consensus towards reduced precipitation.

The geographical patterns of local agreement/disagreement in projected precipitation change remain broadly similar with increased global warming, but the range of uncertainty generally increases ( ''high confidence'' ). For example, in northeastern Amazonia, the driest projections increase from a 10% decrease at 1.5°C global warming to a 40% decrease at 4°C global warming. In comparison, the wettest projections remain at up to a 10% increase. In the far north of North America and Asia, the higher end of projected increases in precipitation extends to approximately 40–60%. A few regions are projected to see a shift in the consensus on the sign of the change. These include parts of the Indian sub-continent where at 4°C global warming, the projected changes shift to a consensus on increased precipitation ranging between a few percent to over 70%.

Notably, the multi-model median change in precipitation is relatively small in many regions—less than 10% over most of the global land surface at 1.5°C global warming. In contrast, in many locations, the 5 th to 95 th percentile range can include changes that are much larger changes than the median and also changes that are relatively large but opposite in sign. At 4°C global warming, the median projected changes are larger, ranging from a 20% decrease to a 40% increase (excluding very dry areas, where percentage changes can be much larger due to very small baseline values), but nevertheless often remain a poor indicator of the range of changes across the ensemble. Therefore, use of the median or mean projected changes for future adaptation decisions could substantially underestimate the risk of large changes in precipitation. It could mean that the risk of the opposite sign of changes is not accounted for. Indeed, for mean precipitation, different multi-model ensembles can show different levels of significance of the central estimate of change ( [[#Uhe--2021|Uhe et al., 2021]] : Figure 4.11a). Consequently, information on the range of possible outcomes can be valued by users for effectively informing risk assessments ( [[#Lowe--2018|Lowe et al., 2018]] ).

<div id="_idContainer053" class="Figure"></div>

[[File:a3d58facdeb025463fa8aa3a933020fa IPCC_AR6_WGII_Figure_4_011.png]]

'''Figure 4.11 |''' '''Agreement between different multi-model ensembles on significant changes in (a) annual mean precipitation and (b) annual maximum 1-d precipitation (Rx1day) at 2°C global warming (Uhe et a''' '''l.''' ''', 2021).''' Using central estimates from five ensembles of climate models (CMIP5, CMIP6, HAPPI, HELIX and UKCP18) using different models and different experimental designs for the ensembles, the maps show the number of ensembles for which the central estimate shows a significant drying or wetting change at 2°C global warming relative to pre-industrial levels. The different ensembles reach 2°C global warming at different times. The projected changes are aggregated over the new climatic regions defined for IPCC AR6 ( [[#Iturbide--2020|Iturbide et al., 2020]] ). Hatched regions show where different ensembles project significant changes in opposite directions, i.e., there is no agreement on either drying or wetting. Regions with thick outlines are where CMIP6 disagrees with three of the other four ensembles on the significance of the change, highlighting where over-relying on CMIP6 alone may not fully represent the level of confidence in the projections.

There is a stronger consensus on changes in heavy precipitation than mean precipitation within individual ensembles such as CMIP6 (Figure 4.12) and especially between the means of the different ensembles (Figure 4.11b). At 4°C global warming, the 50th percentile projection is for increased annual maximum 1-d precipitation over virtually all global land, with the median increase being over 20% for a majority of the land. The 95 th percentile increase is 20–40% over most mid-latitude areas and at least 40-–70% over the tropics and subtropics, exceeding 80% over western Amazonia, central Africa and most of the Indian sub-continent. The 5 th percentile also shows an increase over most global land; in other words, decreased heavy precipitation has less than a 5% probability in these regions (Figure 4.12a), although decreases remain possible but of low probability in some regions, particularly northern South America and northern and western Africa. At the 50th and 95th percentiles, similar global patterns of change are projected at 2°C and 1.5°C global warming, with smaller local magnitudes (Figure 4.12e,f,h,i). At the 5 th percentile, decreased Rx1day is seen over much larger land areas (Figure 4.12d,g), which may be a result of internal climate variability being relatively larger than the long-term trend at lower GWLs. In CMIP5, precipitation extremes are projected to be ''more likely'' to increase than to decrease on average over both the humid and arid regions of the world, but with larger uncertainty in arid areas ( [[#Donat--2019|Donat et al., 2019]] ).

<div id="_idContainer055" class="Figure"></div>

[[File:bcb69aa3116106fce9ffd5d63df9e442 IPCC_AR6_WGII_Figure_4_012.png]]

'''Figure 4.12 |''' '''Projected percentage changes in annual maximum daily precipitation (Rx1day) averaged over 20 years centred at the time of first passing (a–c) 4°C, (d–f) 2°C and (g–i) 1''' '''.''' '''5°C global warming levels (GWLs) relative to 1851–1900.''' Results are based on simulations from the CMIP6 multi-model ensemble under the SSP1-1.9, SSP1-2.6, SSP2-4.5, SSP3-7.0 and SSP5-8.5 scenarios. Uncertainties in the projections are quantified with the (a, d, g) 5th, (b, e, h) 50th and (c, f, i) 95th percentile local values from the ensemble at each GWL. Note that these are uncertainties at the individual point and are not spatially coherent, that is, they do not represent plausible global patterns of change. The 50th percentile maps (b, e, h) present the same data over land as Figure 11.16 of [[#Seneviratne--2021|Seneviratne et al. (2021)]] . The numbers on the left indicate the number of simulations included at each warming level, including multiple realisations from some models with varying initial conditions, depending on data availability. Results for the 1.5°C GWL include 37 unique models. Fewer models and realisations are available for the 2°C and 4°C GWLs, as fewer scenarios and/or models reach those warming levels. For individual models, the global patterns of changes are very similar across scenarios, and any differences between scenarios are smaller than the ensemble uncertainty for an individual scenario. The CMIP6 projections of changes in mean and extreme precipitation are discussed in more detail by WGI ( [[#Doblas-Reyes--2021|Doblas-Reyes et al., 2021]] ; [[#Seneviratne--2021|Seneviratne et al., 2021]] ).

In the 50th percentile projections at 4°C global warming, dry spells are projected to become up to 40 d longer in South America and southern Africa and up to 20 d shorter in large parts of Asia (Figure 4.13a,b,c). In most regions, the projected changes in dry spell lengths are highly uncertain. In southern Africa, the increase in dry spell length ranges from 10 d to over 40 d. In northeast Asia, dry spells are projected to become shorter by up to 20–30 d. In much of South America, dry spells could increase by over 40 d or decrease by over 10 d. Similar global patterns with smaller magnitudes of change are projected for 2°C and 1.5°C global warming in all three percentiles (Figure 4.13d,e,f,g,h,i).

<div id="_idContainer057" class="Figure"></div>

[[File:f4a906401685e34f14a28077646c2671 IPCC_AR6_WGII_Figure_4_013.png]]

'''Figure 4.13 |''' '''As Figure 4.''' '''12 for projected changes in annual consecutive dry days (CDD), the highest number of days yr''' –1 '''with precipitation &lt; 1 mm.''' The 50th percentile maps (b, e, h) present the same data as Figure 11.19a,b,c of [[#Seneviratne--2021|Seneviratne et al. (2021)]] .

Taken together, these projections of more intense precipitation and changes in the length of dry spells give a clear picture of increasingly volatile precipitation regimes, with many regions seeing both longer dry spells and heavier events when precipitation does occur ( ''high confidence'' ).

The critical knowledge gap for precipitation projections is the ability to make precise projections. With such large uncertainties in many regions, climate model projections can inform risk assessments, but cannot provide confident predictions of specific outcomes.

In summary, the annual mean precipitation range is projected to increase or decrease by up to 40% or more at 4°C global warming over many land areas. The ranges of projected precipitation changes are smaller at lower levels of global warming ( ''high confidence'' ). Either an increase or decrease is possible in most regions, but there is an agreement among models on the increase in the far north ( ''high confidence'' ). There is a stronger model consensus on heavy precipitation increasing with global warming over most land areas ( ''high confidence'' ). There are widely varying projections of change in dry spell length ( ''high confidence'' ), but in regions with increasing projected dry spells, the potential increase is larger at higher levels of global warming ( ''high confidence'' ).

<div id="4.4.1.2" class="h3-container"></div>

<span id="projected-changes-in-evapotranspiration"></span>
==== 4.4.1.2 Projected Changes in Evapotranspiration ====

<div id="h3-5-siblings" class="h3-siblings"></div>

AR5 ( [[#Collins--2013)|Collins et al., 2013)]] found that the CMIP5 model projections of ET increases or decreases followed the same pattern over land as precipitation projections, with additional impacts of reduced transpiration due to plant stomatal closure in response to rising CO 2 concentrations. AR6 WGI ( [[#Douville--2021|Douville et al., 2021]] ) assessed that it is ''very likely'' that ET will increase over land, with regional exceptions in drying areas.

In most CMIP5 and CMIP6 models, projected ET changes are driven not just by meteorological conditions and soil moisture but also by plant physiological responses to elevated CO 2 , which themselves influence meteorology and soil moisture through surface fluxes ( [[#Halladay--2017|Halladay and Good, 2017]] ; [[#Lemordant--2019|Lemordant and Gentine, 2019]] ). Elevated CO 2 causes stomatal closure which decreases ET, but also increases leaf area index (LAI) which in turn increases ET, but these do not necessarily compensate ( [[#Skinner--2017|Skinner et al., 2017]] ). Higher LAI increases transpiration, depleting soil moisture but increasing shading, thus reducing soil evaporation ( [[#Skinner--2017|Skinner et al., 2017]] ), but LAI may not increase in areas where it is already high ( [[#Lemordant--2018|Lemordant et al., 2018]] ). Projected ET decreases from physiological effects alone are widespread but greatest in tropical forests ( [[#Swann--2016|Swann et al., 2016]] ; [[#Kooperman--2018|Kooperman et al., 2018]] ).

Future changes in regional ET are therefore highly uncertain. The CMIP6 multi-model ensemble projects changes in ET varying both in magnitude and sign across the ensemble members (Figure 4.14). At 4°C global warming, the ensemble median projection shows increased ET of approximately 25% in mid/high latitudes but decreases of up to 10% across most of tropical South America, southern Africa and Australia. These CMIP6 ensemble projections resemble ET changes projected by the CMIP5 ensemble, except over central Africa and Southeast Asia ( [[#Berg--2019|Berg and Sheffield, 2019]] ). However, the ensemble ranges are wide and include both increases and decreases in projected ET in many locations, with mid-latitude ET increases being up to approximately 50% and ET decreases in southern Africa being up to approximately 30%. Projected changes are proportionally smaller at lower levels of global warming, while patterns of change remain similar.

<div id="_idContainer059" class="Figure"></div>

[[File:25173552ec9590928aaa2881f38ef4f1 IPCC_AR6_WGII_Figure_4_014.png]]

'''Figure 4.14 |''' '''Projected percentage changes in annual mean ET at global warming levels (GWLs) of 4°C (top), 2°C (middle) and 1''' '''.''' '''5°C (bottom) for the CMIP6 multi-model ensemble of GCMs driven by SSP5-8.5 concentrations.''' The distribution of outcomes is shown at local scales with the 5th, 50th and 95th percentile ET changes in individual grid boxes. Note that these are uncertainties at the individual point and are not spatially coherent, that is, they do not represent plausible global patterns of change. Results for 1.5°C, 2°C and 4°C global warming are defined as 20-year means relative to 1850–1900 and use 40, 40 and 31 ensemble members, respectively, due to some members not reaching 4°C global warming.

The relative importance of the physiological and radiative effects of CO 2 on future ET is a crucial knowledge gap, partly because many ESM land surface schemes still use representations of this process based on older experimental studies. Furthermore, large-scale experimental studies using free-air CO 2 enrichment (FACE) techniques to constrain the models have not yet been performed in certain critical ecosystems, such as tropical forests. Finally, uncertainties in equilibrium climate sensitivity (ECS) imply uncertainties in the CO 2 concentration accompanying any given level of warming (Betts and McNeall, 2018).

In summary, the sign of projected ET change depends on region, but there is ''medium confidence'' that ET will increase in the global mean and mid/high latitudes and decrease in northern South America and southern Africa. In addition, the impacts of rising CO 2 concentrations on plant stomata and leaf area play a role in model projections of ET change ( ''high confidence'' ), but there is ''low confidence'' in their overall contribution to global ET change.

<div id="4.4.1.3" class="h3-container"></div>

<span id="projected-changes-in-soil-moisture"></span>
==== 4.4.1.3 Projected Changes in Soil Moisture ====

<div id="h3-6-siblings" class="h3-siblings"></div>

AR5 ( [[#Collins--2013)|Collins et al., 2013)]] mainly focused on surface (upper 10 cm) soil moisture, summarising multi-model projections of 21st century annual mean soil moisture changes as broadly decreasing in the subtropics and Mediterranean region and increasing in east Africa and central Asia across the RCPs, with the changes tending to become stronger as global warming increases. AR6 WGI ( [[#Douville--2021|Douville et al., 2021]] ) draw broadly similar conclusions based on new ESMs, noting that compared to CMIP5, the CMIP6 models project more consistent drying in the Amazon basin, Siberia, westernmost North Africa and southwestern Australia. WGI ( [[#Douville--2021|Douville et al., 2021]] ) also note that soil moisture in the upper 10 cm shows more widespread drying than in the total soil column.

The CMIP6 multi-model ensemble of ESMs show varying levels of consensus on projected changes in surface soil moisture with global warming (Figure 4.15). As in CMIP5 ( [[#Cheng--2017|Cheng et al., 2017]] ), uncertainties are substantial, often associated with uncertainties in projected regional precipitation changes ( [[#4.4.1.1|Section 4.4.1.1]] ), and in most regions, both increases and decreases are projected across the ensemble. In the far north of North America and Asia, projected changes in soil moisture at 4°C global warming range from a 20–30% decrease to an increase of 30–40%. In northern mid-latitudes, projections range from a 10–20% decrease to an increase of 20–30%, except for eastern North America, where the projected changes (both increases and decreases) are less than 10%, and western Europe and the Mediterranean where there is a stronger consensus towards decreased soil moisture of up to 25%. South America, southern Africa and Asia also show a stronger consensus towards decreased soil moisture of up to 40% or more in some regions.

<div id="_idContainer061" class="Figure"></div>

[[File:c61ce16a1331a03cf2da0e5e9856ffea IPCC_AR6_WGII_Figure_4_015.png]]

'''Figure 4.15 |''' '''Projected percentage changes in annual mean total column soil moisture relative to 1981–2010 at global warming levels (GWLs) of 4°C (top), 2°C (middle) and 1''' '''.''' '''5°C (bottom) for the CMIP6 multi-model ensemble of GCMs driven by SSP5-8.5 concentrations.''' The distribution of outcomes is shown at local scales with the 5th, 50th and 95th percentile soil moisture changes in individual grid boxes. Note that these are uncertainties at individual points and are not spatially coherent, that is, they do not represent plausible global patterns of change. Results for 1.5°C, 2°C and 4°C global warming are defined as 20-year means relative to 1850–1900 and use 34, 34 and 26 ensemble members, respectively, due to some members not reaching 4°C global warming. Fewer models are shown here than in Figure 4.10 on precipitation and Figure 4.14 on ET because some do not provide soil moisture output.

Most CMIP6 models simulate direct CO 2 effects on plant transpiration, which has been shown to be a strong influence on projected future changes in soil moisture ( [[#Milly--2016|Milly and Dunne, 2016]] ). Approaches that neglect this process project greater decreases in soil moisture availability than the climate models ( [[#Roderick--2015|Roderick et al., 2015]] ; [[#Swann--2016|Swann et al., 2016]] ). Therefore, although several studies project increased global aridity and dryland expansion ( [[#Feng--2013|Feng and Fu, 2013]] ; [[#Sherwood--2014|Sherwood and Fu, 2014]] ; [[#Huang--2016a|Huang et al., 2016a]] ), these may overestimate future drying (Berg et al., 2017). Nevertheless, land surface models, including vegetation responses to CO 2, still project reduced soil moisture in many regions ( [[#Grillakis--2019|Grillakis, 2019]] ).

A critical knowledge gap concerns the relative importance of climate and CO 2 physiological effects on soil moisture, in relation to uncertainties in climate sensitivity. For a given level of global warming, the relative importance of climate effects and the direct effects of CO 2 on transpiration depend on the CO 2 concentration accompanying that level of warming (Betts and McNeall, 2018). Some CMIP6 models have very high climate sensitivities ( [[#Meehl--2020|Meehl et al., 2020]] ), which are assessed as being of low probability on the basis of other lines of evidence ( [[#Sherwood--2020|Sherwood et al., 2020]] ). This means that the CO 2 concentration accompanying specific global warming levels may be too low and lead to overly large projections of soil moisture decrease in those models.

In summary, projected soil moisture changes increase with levels of global warming ( ''high confidence'' ), although there remains substantial disagreement on specific regional changes. In the CMIP6 multi-model ensemble at 4°C global warming, decreased soil moisture of up to 40% is projected in Amazonia, southern Africa and western Europe in all models ( ''high confidence'' ). In all other regions, there is no consensus on the sign of projected soil moisture changes, and projected changes at 4°C global warming include decreases of up to 30% and increases of up to 40%. Projected changes are smaller at lower levels of global warming, with similar geographical patterns of change.

<div id="4.4.2" class="h2-container"></div>

<span id="projected-changes-in-the-cryosphere-snow-glaciers-and-permafrost"></span>
=== 4.4.2 Projected Changes in the Cryosphere (Snow, Glaciers and Permafrost) ===

<div id="h2-20-siblings" class="h2-siblings"></div>

AR5 noted that global glacier mass loss is ''very likely'' to increase further during the 21st century ( [[#Jiménez%20Cisneros--2014|Jiménez Cisneros et al., 2014]] ). According to the SROCC ( [[#Hock--2019b|Hock et al., 2019b]] ), it is ''very likely'' that glaciers will continue to lose mass throughout the 21st century: from 18% (by 2100, relative to 2015) for RCP2.6 to 36% for RCP8.5. AR5 ( [[#Collins--2013)|Collins et al., 2013)]] and SROCC ( [[#Meredith--2019|Meredith et al., 2019]] ) reported with ''high confidence'' that permafrost would continue to thaw in the 21st century, but the projections are uncertain. Constraining warming to 1.5°C would prevent the thawing of a permafrost area of 1.5 to 2.5 million km 2 compared to thawing under 2°C ( ''medium confidence'' ) ( [[#IPCC--2018b|IPCC, 2018b]] ). AR5 ( [[#Collins--2013)|Collins et al., 2013)]] and SROCC ( [[#Meredith--2019|Meredith et al., 2019]] ) concluded that Northern Hemisphere snow extent and mass would likely reduce by the end of the 21st century, both in plain and mountain regions. AR6 assessed with ''medium confidence'' that under RCP2.6 and RCP8.5 from 2015 to 2100, glaciers are expected to lose 18% and 36% of their early 21st-century mass, respectively (AR6 WGI, ( [[#Fox-Kemper--2021|Fox-Kemper et al., 2021]] )).

Global glacier mass loss since 2015 and 2100 was projected to be 18 ± 13% by 2100 with 0.9 – 2.3°C global warming and 36 ± 20% with 3.2 – 5.4°C global warming ( [[#Marzeion--2020|Marzeion et al., 2020]] ), which corresponds with previous findings ( [[#Radić--2014|Radić et al., 2014]] ; [[#Hock--2019a|Hock et al., 2019a]] ; [[#Shannon--2019|Shannon et al., 2019]] ). The regional glacier loss rate projections are unevenly distributed worldwide and considerably vary between scenarios ( [[#Huss--2018|Huss and Hock, 2018]] ; [[#Hock--2019a|Hock et al., 2019a]] ). In most regions, ‘peak water’ has already been reached or is expected to be reached before mid-century (with an earlier ‘peak water’ for RCP2.6 scenario compared with RCP8.5) ( [[#Huss--2018|Huss and Hock, 2018]] ; [[#Pritchard--2019|Pritchard, 2019]] ; [[#Marzeion--2020|Marzeion et al., 2020]] ; [[#Rounce--2020|Rounce et al., 2020]] ). The influence of the expected subsequent decrease in glacier runoff by the end of the 21st century will be more pronounced during droughts and dry seasons ( [[#Farinotti--2016|Farinotti et al., 2016]] ; [[#Huss--2016|Huss and Fischer, 2016]] ; [[#Hanzer--2018|Hanzer et al., 2018]] ; [[#Brunner--2019|Brunner et al., 2019]] ).

Such changes in runoff could potentially lead to water shortages for over 200 million people in the high mountains of Asia ( [[#Pritchard--2019|Pritchard, 2019]] ; [[#Shahgedanova--2020|Shahgedanova et al., 2020]] ). There is ''medium confidence'' that under a 4°C warming scenario, 40% of current irrigated demand in sub-basins relying primarily on snowmelt runoff would need to be supplemented from other water sources ( [[#Qin--2020|Qin et al., 2020]] ). Basins where such alternate sources are not available will face agricultural water scarcity ( [[#4.5.1|Section 4.5.1]] ). Globally, 1.5 billion people are projected to critically depend on runoff from the mountains by the mid-21 st century under the RCP6.0 scenario ( [[#Viviroli--2020|Viviroli et al., 2020]] ). Furthermore, there is ''medium confidence'' that projected changes in snow and glacier melt runoff will affect water inputs to hydropower, leading to a decline in hydroelectricity production in mountain basins, for example, in India ( [[#Ali--2018|Ali et al., 2018]] ), Switzerland ( [[#Schaefli--2019|Schaefli et al., 2019]] ) and the USA ( [[#Lee--2016|Lee et al., 2016]] ) ( [[#4.5.2|Section 4.5.2]] ) (IPCC AR6 WGI, 2021) (Sections 9.5.1.3 and 8.4.1.7.1).

Projections of snow cover metrics [IPCC AR6 WGI, 2021 ( [[IPCC:Wg2:Chapter:Chapter-9#9.5.3|Section 9.5.3.3]] )] suggest a further decrease in snow water equivalent (SWE) and snow cover extent (SCE), though the inter-model spread is considerable ( [[#Lute--2015|Lute et al., 2015]] ; [[#Thackeray--2016|Thackeray et al., 2016]] ; [[#Kong--2017|Kong and Wang, 2017]] ; [[#Henderson--2018|Henderson et al., 2018]] ) ( ''high confidence'' ). The projected CMIP6 SCE and SWE changes share the broad features of the CMIP5 projections: SCE is expected to decrease in the Northern Hemisphere by approximately 20%, relative to the 1995–2014 mean value, around 2060 and stabilise afterwards under the RCP2.6 scenario, while the RCP8.5 scenario leads to snow cover losses up to 60% by 2100 ( [[#Mudryk--2020|Mudryk et al., 2020]] ). Regionally, the SWE loss will probably lead to more frequent snow droughts; for example, the frequency of consecutive snow droughts is projected to increase to 80–100% of years at 4 ° C warming in western Canada ( [[#Shrestha--2021|Shrestha et al., 2021]] ) and 42% of years under the RCP8.5 scenario in the western USA ( [[#Marshall--2019|Marshall et al., 2019]] ) by 2100. Thus, by the mid- to late 21st century, for more than 2/3 of snow-dominated areas in the western USA, the ability to predict seasonal droughts and prepare robust water management plans will decline ( [[#Livneh--2020|Livneh and Badger, 2020]] ) ( [[#4.4.5|Section 4.4.5]] ).

There is a ''high agreement'' between the CMIP6 projections and the previous findings that permafrost will undergo increasing thaw and degradation during the 21st century worldwide ( [[#Fox-Kemper--2021|Fox-Kemper et al., 2021]] ) . The CMIP6 models project that the annual mean frozen volume in the top 2 m of the soil could decrease by 10–40% for every degree increase of global temperature ( [[#Burke--2020|Burke et al., 2020]] ; [[#Yokohata--2020b|Yokohata et al., 2020b]] ). The CMIP5-based equilibrium sensitivity of permafrost extent to stabilised global mean warming is established to be about 4.0 × 10 6 km 2 °C –1 ( [[#Chadburn--2017|Chadburn et al., 2017]] ). The southern boundary of the permafrost is projected the move to the north: 1°–3.5° northward (relative to 1986–2005) at the level of 1.5°C temperature rise ( [[#Kong--2017|Kong and Wang, 2017]] ).

The observational knowledge gaps ( [[#4.2.2|Section 4.2.2]] ) impede efforts to calibrate and evaluate models that simulate the past and future evolution of the cryosphere and its social impacts.

In summary, in most basins fed by glaciers, runoff is projected to increase initially in the 21st century and then decline ( ''medium confidence'' ). Projections suggest a further decrease in seasonal snow cover extent and mass in mid to high latitudes and high mountains ( ''high confidence'' ), though the projection spread is considerable. Permafrost will continue to thaw throughout the 21st century ( ''high confidence'' ). There is ''medium confidence'' that future changes in cryospheric components will negatively affect irrigated agriculture and hydropower production in regions dependent on snowmelt runoff.

<div id="4.4.3" class="h2-container"></div>

<span id="projected-changes-in-streamflow"></span>
=== 4.4.3 Projected Changes in Streamflow ===

<div id="h2-21-siblings" class="h2-siblings"></div>

AR5 ( [[#Jiménez%20Cisneros--2014|Jiménez Cisneros et al., 2014]] ) concluded that increases in the mean annual runoff are projected in high latitudes and the wet tropics and decreases in dry tropical regions, but with very considerable uncertainty. Both the patterns of change and uncertainties were found to be primarily driven by projected changes in precipitation. SR1.5 ( [[#Hoegh-Guldberg--2018|Hoegh-Guldberg et al., 2018]] ) concluded with ''medium confidence'' that areas with either positive or negative changes in mean annual runoff/streamflow are projected to be smaller for 1.5°C than for 2°C of global warming. AR6 WGI ( [[#Douville--2021|Douville et al., 2021]] ) conclude with ''medium confidence'' that global runoff will increase with global warming but with significant regional and seasonal variations. WGI further concluded with ''high confidence'' that runoff will increase in the high northern latitudes and decrease in the Mediterranean and southern Africa. However, there was ''medium confidence'' that runoff will increase in central and eastern African regions and decrease in Central America and parts of southern South America. The magnitude of the change is projected to increase with emissions. There is ''medium confidence'' that the seasonality of runoff and streamflow will increase with global warming in the subtropics. In snow-dominated regions, there is ''high confidence'' that peak flows associated with spring snowmelt will occur earlier in the year and ''medium confidence'' that snowmelt-induced runoff will decrease with reduced snow, except in glacier-fed basins where runoff may increase in the near term.

Changes in runoff and streamflow are projected over most of the ice-free land surface with all recent climate and hydrological model ensembles (Figure 4.16). Changes in streamflow could increase the number of people facing water scarcity or insecurity ( ''high confidence'' ) ( [[#Schewe--2014|Schewe et al., 2014]] ; [[#Gosling--2016|Gosling and Arnell, 2016]] ; [[#McMillan--2016|McMillan et al., 2016]] ). Projections of future runoff at basin scales show considerable uncertainty in many regions, including differences in signs in many regions (Figure 4.16). This uncertainty is driven by uncertainties in regional precipitation patterns and hydrological models ( [[#Koirala--2014|Koirala et al., 2014]] ; [[#Asadieh--2016|Asadieh et al., 2016]] ), including vegetation responses to CO 2 and its effects on ET ( [[#Betts--2015|Betts et al., 2015]] ). This uncertainty in future water availability contributes to the policy challenges for adaptation, for example, for managing risks of water scarcity ( [[#Greve--2018|Greve et al., 2018]] ; Box 4.1). In many regions, some models project large changes in runoff/streamflow but with low consistency between models on the sign of the change (Figure 4.16). In streamflow projections driven by 11 CMIP5 models with the RCP8.5 scenario, strong model consistency (agreement by at least 10 models) is only seen over 21% of global land ( [[#Koirala--2014|Koirala et al., 2014]] ). Consensus on the sign of projected change is smaller with the RCP4.5 scenario.

<div id="_idContainer063" class="Figure"></div>

[[File:4d5b2328d8a479f459559917873daec2 IPCC_AR6_WGII_Figure_4_016.png]]

'''Figure 4.16 |''' '''Projected changes in the annual mean runoff in selected river basins at global warming levels (GWLs) of 1''' '''.''' '''5°C, 2°C and 4°C in a combined ensemble.''' For each named basin, the sinaplot dots show individual model outcomes for percentage increased flows (blue) and decreased flows (red) at each GWL. Black circles show the ensemble median, and black bars show the 95% confidence range in the median. See inset with the Rio Grande sinaplot for additional guidance on interpretation. In the map, the colours in the basins show the percentage model agreement on the sign of the projected change in streamflow at the 4°C GWL. The combined ensemble is comprised of four multi-model ensembles: the CMIP5 multi-model ensemble of GCMs driven with RCP8.5; the CMIP6 multi-model ensemble of GCMs driven with SSP5-8.5; varying combinations of hydrological models with five GCMs in the Inter-Sectoral Impacts Model Intercomparison Project (ISIMIP); and the JULES land ecosystems and hydrology model driven by GCMs from the HELIX study ( [[#Betts--2018|Betts et al., 2018]] ; [[#Koutroulis--2019|Koutroulis et al., 2019]] ). In CMIP5 and CMIP6, the projected runoff changes are directly from the GCM land surface schemes without bias correction. In ISIMIP and HELIX, bias-corrected climate model outputs were used to drive the hydrology models. A comparison of the projected changes at the 4°C GWL for the four individual ensembles is shown in Figure Cross-Chapter Box CLIMATE.1 in Chapter 1.

Considering a wider set of projections, the consensus on increased flows becomes stronger at higher GWLs in (for example) the Yukon, Mackenzie, Kemijoki, Amur, Hwang Ho, Yangtze, Mekong, Ganges-Brahmaputra, Nile, Zaire and Parana basins (Figure 4.16). The consensus on decreased flows becomes stronger for higher GWLs in (for example) the Colorado, Tagus, Helmand, Tigris-Euphrates and Amazon. However, in both cases, some models have projected changes of the opposite sign to the consensus. Moreover, the distribution of projected outcomes becomes notably broader at higher GWLs in (for example) the Mississippi, Yangtze and Amazon. Therefore, even with a strong global climate change signal, uncertainties in changes in mean runoff/streamflow can remain large or even increase. Nevertheless, since projected changes typically increase with global warming, limiting warming to 1.5°C or 2°C substantially reduces the potential for either large increases or decreases in mean streamflow compared to 3°C or 4°C ( [[#Warszawski--2014|Warszawski et al., 2014]] ; [[#Falkner--2016|Falkner, 2016]] ; [[#Gosling--2017|Gosling et al., 2017]] ; Figure 4.16) ( ''high confidence'' ).

In CMIP5, strong model consistency on changes in high and low streamflows is seen with similar global patterns to the mean flows, but over smaller areas ( [[#Koirala--2014|Koirala et al., 2014]] ). By the end of the 21st century, with RCP8.5, increases in mean, high and low flows are projected for the Lena, and mean and low flows for the MacKenzie ( [[#Gelfan--2017|Gelfan et al., 2017]] ; [[#Pechlivanidis--2017|Pechlivanidis et al., 2017]] ; [[#Döll--2018|Döll et al., 2018]] ) ''.'' Increased mean and high flows are projected in the Ganges, high flow in the Rhine and Mississippi, while decreasing mean and low flows are projected in the Rhine ( [[#Krysanova--2017|Krysanova et al., 2017]] ; [[#Pechlivanidis--2017|Pechlivanidis et al., 2017]] ; [[#Vetter--2017|Vetter et al., 2017]] ). Decreases in mean, high and low flows are projected for the Tagus (Krysanova et al. 2017; Vetter et al. 2017). Low flows are projected to decrease in the Mediterranean region and increase in the Alps and northern Europe ( [[#Marx--2018|Marx et al., 2018]] ). A general shift in the runoff distribution towards more extreme low runoff is projected in Mexico, western USA, western Europe, southeastern China and the West Siberian Plain, and more extreme high runoff is projected in Alaska, northern Canada and large parts of Asia ( [[#Zhai--2020|Zhai et al., 2020]] ).

While projected changes in high and low flows are similar to those in mean flows in many regions, this is not the case everywhere. When a single hydrological model and a sample of climate models are selected to explore uncertainties systematically, approximately 56% of the global population is projected to be affected by increased extreme high flows at 1.5°C warming, rising to 61% at 2°C warming ( [[#Zhai--2020|Zhai et al., 2020]] ). Those affected by extreme low flows decrease is projected to remain close to 45% at both 1.5°C and 2°C warming. However, these results are based on the median of the ensemble projections, so they are subject to high uncertainty. At 1.5°C global warming, 15% of the population is projected to be affected concurrently by decreased extreme low flows and increased extreme high flows, increasing to 20% at 2°C warming. In 25 combinations of five CMIP5 climate models and five global hydrological models under the RCP8.5 scenario reaching approximately 4°C GWL at the end of the century, 10% of the global land area is projected to face simultaneously increasing high extreme streamflow and decreasing low extreme streamflow. These regions include the British Isles and the shores of the North Sea, large parts of the Tibetan Plateau, South Asia and western Oceania, and smaller regions of Africa and North and South America, affecting over 2.1 billion people with 2015 population distributions ( [[#Asadieh--2017|Asadieh and Krakauer, 2017]] ). With 11 CMIP5 models driving a single hydrological model, simultaneous increases in high flows and decreases in low flows are projected over 7% of global land ( [[#Koirala--2014|Koirala et al., 2014]] ).

By the end of the 21st century, global changes in streamflow extremes are projected to be approximately twice as large with RCP8.5 (over 4°C GWL) than with RCP2.6 (approximately 2°C GWL) ( [[#Asadieh--2017|Asadieh and Krakauer, 2017]] ).

Glacier retreat and associated runoff changes represent a major global sustainability concern ( [[#4.4.2|Section 4.4.2]] ). By 2100, using an ensemble of 14 CMIP5 climate models driven by the RCP4.5 scenario, one third of the 56 large-scale glacierised catchments are projected to experience a mean annual runoff decline by over 10%, with the most significant reductions in central Asia and the Andes ( [[#Huss--2018|Huss and Hock, 2018]] ). Thus, communities dependent on glacier runoff are particularly vulnerable ( [[#Jiménez%20Cisneros--2014|Jiménez Cisneros et al., 2014]] ).

Societal impacts of change in runoff spread throughout several socioeconomic sectors, such as agriculture, health and energy production, affecting overall water security ( [[#Wang--2021a|Wang et al., 2021a]] ). Decreases in runoff may lead to water scarcity and result in increased multi-sectoral effects in sub-Saharan Africa ( [[#Serdeczny--2017|Serdeczny et al., 2017]] ), western Africa, the Middle East, Mexico, Northeastern Brazil, central Argentina, Mediterranean Africa and Europe ( [[#Gosling--2016|Gosling and Arnell, 2016]] ; [[#Greve--2018|Greve et al., 2018]] ), and southeastern Australia ( [[#Barnett--2015|Barnett et al., 2015]] ).

In summary, mean and extreme streamflow changes are projected over most of the ice-free land surface ( ''high confidence'' ). The magnitude of streamflow change is projected to increase with global warming in most regions ( ''high confidence'' ), but there is often high uncertainty on the sign of change. There is ''high confidence'' that mean streamflows will increase in the northern high latitudes and decrease in the Mediterranean and southern Africa. Annual mean runoff in one third of assessed glacierised catchments is projected to decline by at least 10% by 2100 under RCP4.5, with the most significant reductions in central Asia and the Andes ( ''medium confidence'' ). Elsewhere, projections include both increased and decreased flows. Substantial fractions of ensemble projections disagree with the multi-model mean ( ''high confidence'' ), with implications for long-term planning for water management. With 1.5 and 2°C global warming, approximately 15 and 20% of the current global population, respectively, would experience both an increase in high streamflows and a decrease in low streamflows ( ''medium confidence'' ). At 4°C global at the end of the century, 10% of the global land area is projected to simultaneously experience an increase in high extreme streamflow and decrease in low extreme streamflow.

<div id="4.4.4" class="h2-container"></div>

<span id="projected-changes-in-floods"></span>
=== 4.4.4 Projected Changes in Floods ===

<div id="h2-22-siblings" class="h2-siblings"></div>

SR1.5 ( [[#Hoegh-Guldberg--2018|Hoegh-Guldberg et al., 2018]] ) concluded with ''medium confidence'' that global warming of 2°C would lead to an expansion of the area affected by flood hazards, compared to conditions at 1.5°C global warming. Both AR5 ( [[#Jiménez%20Cisneros--2014|Jiménez Cisneros et al., 2014]] ) and SROCC ( [[#Hock--2019b|Hock et al., 2019b]] ) concluded that spring snowmelt floods would be earlier ( ''high confidence'' ), and hazards from floods involving meltwater will gradually diminish, particularly at low elevation ( ''medium confidence'' ). SROCC ( [[#Hock--2019b|Hock et al., 2019b]] ) and AR6 WGI [[IPCC:Wg2:Chapter:Chapter-9|Chapter 9]] stated that given ''limited evidence'' and the complexity of the process, the changes of glacier-related floods under climate change are not clear. AR6 WGI Chapters 8 and 11 summarised that there is ''medium confidence'' for a general increase in flooding due to warming, but there are significant regional and seasonal variations.

There is ''high confidence'' that the frequency and magnitude of river floods are projected to change at a global scale. For example, the frequency of river floods is projected to increase in many regions, including Asia, central Africa, western Europe, Central and South America and eastern North America, and decrease in northern North America, southern South America, the Mediterranean and eastern Europe in 2050 and beyond ( [[#Koirala--2014|Koirala et al., 2014]] ; [[#Arnell--2016|Arnell et al., 2016]] ) (Figure 4.17). There is ''low agreement'' in projections in changes to snowmelt flood magnitude. A negative trend in snowmelt flood magnitude, together with an increase in rain-fed winter floods, is projected with ''medium confidence'' , for example, in mid-latitude and low-altitude basins of Scandinavia ( [[#Arheimer--2015|Arheimer and Lindström, 2015]] ; [[#Vormoor--2016|Vormoor et al., 2016]] ) and throughout Europe as a whole ( [[#Kundzewicz--2017|Kundzewicz et al., 2017]] ), and northeastern North America ( [[#Arnell--2014|Arnell and Lloyd-Hughes, 2014]] ). With ''medium confidence'' , a positive trend is projected in high-latitude basins, for example, for large Arctic rivers such as Lena and Mackenzie ( [[#Eisner--2017|Eisner et al., 2017]] ; [[#Gelfan--2017|Gelfan et al., 2017]] ; [[#Pechlivanidis--2017|Pechlivanidis et al., 2017]] ) and high-altitude upstreams, such as the Ganges, Brahmaputra, Salween, Mekong and the upper Indus Basin ( [[#Lutz--2014|Lutz et al., 2014]] ) and alpine catchments ( [[#Hall--2014|Hall et al., 2014]] ). Moderate decreasing trends or insignificant changes are projected for snowmelt floods in the Fraser River Basin of British Columbia ( [[#Shrestha--2017|Shrestha et al., 2017]] ).

<div id="_idContainer065" class="Figure"></div>

[[File:22f37989981eaf6f07d34c57a2537fe2 IPCC_AR6_WGII_Figure_4_017.png]]

'''Figure 4.17 |''' '''Multi-model median return period (years) in the 2080s for the 20th-century 100-year river flood, based on a global river and inundation model, CaMa-Flood, driven by runoff output of nine CMIP6 Models in the SSP1-2''' '''.''' '''6 (a), SSP2-4.5 (b) and SSP5-8.5 (c) scenario respectively.''' All changes are estimated in 2071–2100 relative to 1970-–2000. A dot indicates regions with high model consistency (more than seven models out of nine show the same direction of change).

'''(d)''' Global or regional potential exposure (% to the total population affected by flooding) under different global warming levels with a constant population scenario and climate of CMIP5-HELIX (circle, [[#Alfieri--2017|Alfieri et al., 2017]] ) and CMIP6 (triangle, [[#Hirabayashi--2021b|Hirabayashi et al., 2021b]] ), and with the population scenario of SSP5 and climate of CMIP6 (bar chart, [[#Hirabayashi--2021b|Hirabayashi et al., 2021b]] ). Inundation is calculated when the magnitude of flood exceeds current flood protection ( [[#Scussolini--2016|Scussolini et al., 2016]] ). Note that number of GCMs used to calculate global warming level (GWL) 4.0 is less than that for other GWLs, as the global mean temperature change of some GCMs did not exceed 4°C.

There is ''high confidence'' that climate change and projected socioeconomic development would increase exposure in inundation areas (Figure 4.17), resulting in a large increase in direct flood damages as several times more in all warming levels (Table 4.6). [[#Alfieri--2017|Alfieri et al. (2017)]] estimated a 120 and 400% increase in population affected by river flooding for 2°C and 4°C warming, respectively, and a 170% increase in damage for 2°C warming without socioeconomic impact development ( [[#4.7.5|Section 4.7.5]] ). [[#Dottori--2018|Dottori et al. (2018)]] estimated the same but with a 134% increase in fatalities with population increase under the SSP3 scenario. The highest numbers of people affected by river flooding are projected for countries in southern, eastern and southeastern Asia, with tens of millions of people per year per country projected to be affected (Figure 4.17; [[#Alfieri--2017|Alfieri et al., 2017]] ; [[#Hirabayashi--2021b|Hirabayashi et al., 2021b]] ). [[#Kinoshita--2018|Kinoshita et al. (2018)]] showed that climate change contributes a 2.8–28.8% increase in global fatality for the period 2071–2100 compared to 1991–2005, but socioeconomic change (~131.3% increase) and associated vulnerability change (~72.1% reduction) have a greater impact of the projected flood-related fatality rate than climate change alone. [[#Winsemius--2016|Winsemius et al. (2016)]] discussed that projected flood damage could be reduced to 1/20th in absolute value with adequate adaptation strategies. Direct flood damages are projected to increase by 4–5 times at 4°C compared to 1.5°C, highly depending on scenarios and assumptions (Table 4.6; Box 4.7).

'''Table 4.6 |''' Projected economic impact by river flooding in billion USD in different emission scenarios or for different global warming levels (GWLs). The percentage of the total GDP of the region is given in brackets.

{| class="wikitable"
|-
! Description

! The economic impact in billion USD (% of GDP)

! Reference

|-
| No adaptation with current flood protection, no economic development (fixed at the level of 2010), USD at 2010 purchasing power parity (PPP), mean of 7 GCMs with the RCP8.5 scenario

| * Current (1976–2005): 75 (0.11%)
* GWL 1.5°C: 145 (0.22%)
* (Asia 92, Australasia 8, Europe 29, Africa 7, North America 3, Central and South America 5)
* GWL 2°C: 172 (0.26%)
* (Asia 114, Australasia 7, Europe 32, Africa 9, North America 4, Central and South America 7)
* GWL 3°C: 249 (0.37%)
* (Asia 176, Australasia 9, Europe 38, Africa 11, North America 4, Central and South America 11)
* GWL 4°C: 343 (0.51%)
* (Asia 241, Australasia 19, Europe 55, Africa 9, North America 6, Central and South America 14)

| [[#Alfieri--2017|Alfieri et al. (2017)]] , with regional aggregation and currency conversion

|-
| No adaptation with current flood protection, USD at 2010 PPP, mean of five CMIP5 GCMs and 10 hydrological models

| * Current (1976–2005): 142 (0.21%)
* GWL 1.5°C, SSP3: 370 (0.55%), SSP5: 485 (0.72%)
* GWL 2°C, SSP3: 597 (0.89%), SSP5: 888 (1.32%)
* GWL 3°C, SSP3: 1024 (1.52%), SSP5: 1616 (2.40%)

| [[#Dottori--2018|Dottori et al. (2018)]] with currency conversion

|-
| No adaptation and no flood protection, mean value in 2030 (2010–2030) and 2080 (2010–2080), USD at 2010 PPP, mean of five CMIP5 GCMs

| * Current (1960–1999): 1,032 (1.6%)
* RCP2.6, SSP1: 2030: 2366 (1.44%), 2080: 7429 (1.43%)
* RCP6.0, SSP3: 2030: 1987 (1.44%), 2080: 3353(1.14%)
* RCP8.5, SSP5: 2030: 2304 (1.37%), 2080: 3684(1.77%)

| [[#Winsemius--2016|Winsemius et al. (2016)]]

|-
| Partial adaptation (protected against 100-year floods in high-income countries, against 5-year floods for all others), mean value in 2030 (2010–2030) and 2080 (2010–2080), USD at 2010 PPP, mean of five CMIP5 GCMs

| * Current (1960–1999): 163 (0.25%)
* RCP2.6, SSP1: 2030: 558 (0.34%), 2080: 851 (0.48%)
* RCP6.0, SSP3: 2030: 418 (0.29%), 2080: 413(0.32%)
* RCP8.5, SSP5: 2030: 418 (0.33%), 2080: 441 (0.57%)

| [[#Winsemius--2016|Winsemius et al. (2016)]]

|-
| A model calibrated to fit reported damages, future vulnerability scenarios considering autonomous adaptation, USD at 2005 PPP, mean of 11 CMIP5 GCMs,

| * Current (1991–2005): 14 (0.044%)
* RCP2.6, SSP1: 2081–2100, 121 (0.037%)
* RCP6.0, SSP2: 2081–2100, 133 (0.042%)
* RCP8.5, SSP3: 2081–2100, 130 (0.063%)

| [[#Kinoshita--2018|Kinoshita et al. (2018)]]

|-
| No adaptation and current flood protection, USD at 2005 PPP, mean of five CMIP5 GCMs

| * Current (1961–2005): 102 (0.39%)
* RCP2.6, SSP1: 2020–2100, 2333 (0.99%)
* RCP4.5, SSP2: 2020–2100, 2221 (0.99%)
* RCP6.0, SSP3: 2020–2100, 1328 (0.80%)
* RCP8.5, SSP5: 2020–2100, 4007 (1.21%)

| [[#Tanoue--2021|Tanoue et al. (2021)]]

|-
| Optimised adaptation, USD at 2005 PPP, mean of five CMIP5 GCMs

| * Current (1961–2005): 102 (0.39%)
* RCP2.6, SSP1: 2020–2100, 1621 (0.69%)
* RCP4.5, SSP2: 2020–2100, 1567 (0.70%)
* RCP6.0, SSP3: 2020–2100, 872 (0.52%)
* RCP8.5, SSP5: 2020–2100, 2558 (0.77%)

| [[#Tanoue--2021|Tanoue et al. (2021)]]

|}

In all climate scenarios projected, earlier snowmelt leads to earlier spring floods ( ''high confidence'' ), for example, in northern and eastern Europe ( [[#Gobiet--2014|Gobiet et al., 2014]] ; [[#Hall--2014|Hall et al., 2014]] ; [[#Etter--2017|Etter et al., 2017]] ; [[#Lobanova--2018|Lobanova et al., 2018]] ), northern North America ( [[#Vano--2015|Vano et al., 2015]] ; [[#Musselman--2018|Musselman et al., 2018]] ; [[#Islam--2019b|Islam et al., 2019b]] ), large Arctic rivers ( [[#Gelfan--2017|Gelfan et al., 2017]] ; [[#Pechlivanidis--2017|Pechlivanidis et al., 2017]] ) and high-altitude Asian basins ( [[#Lutz--2014|Lutz et al., 2014]] ; [[#Winsemius--2016|Winsemius et al., 2016]] ). There is ''high confidence'' that snowmelt floods will occur 25–30 d earlier in the year by the end of the 21st century with RCP8.5, but there is only ''low agreement'' in the projected magnitude of snowmelt flood ( [[#Arheimer--2015|Arheimer and Lindström, 2015]] ; [[#Vormoor--2016|Vormoor et al., 2016]] ; [[#Islam--2019b|Islam et al., 2019b]] ).

Challenges to projecting flood risk are large because of the complexity of the projecting snowmelt, high-intensity rainfall and soil wetness in large river basins. Even though increases in the number and area of glacier lakes may cause increases in glacier-related floods ( [[#4.2.2|Section 4.2.2]] ), knowledge of the frequency or magnitude of glacier-related projected floods is limited. Some local studies indicate that the severity of ice-jam flooding is projected to decrease ( [[#Rokaya--2019|Rokaya et al., 2019]] ; [[#Das--2020|Das et al., 2020]] ), but a model study in Canada projected increases in damage of ice-jam floods ( [[#Turcotte--2020|Turcotte et al., 2020]] ). While most flood risk projections do not consider the impact of urban expansion, Güneralp et al. (2015) estimate that urban areas exposed to flooding will increase by a factor of 2.7 due to urban growth by 2030 ( [[#4.5.4|Section 4.5.4]] ). Given the significant differences in assumption in flood protection, exposure or vulnerability scenario among studies, uncertainties in the global estimation of flood losses and damages are large are large (Table 4.6, 4.7.5).

Floods and their societal impacts, especially the enhancement of hazards and increase in vulnerability, depend on complex political, economic and cultural processes ( [[#Carey--2017|Carey et al., 2017]] ; [[#Caretta--2021|Caretta et al., 2021]] ). Thus, assessments that analyse long-term flood impacts need to account for the interplay of water and society relations. Unfortunately, such studies remain scarce ( [[#Pande--2017|Pande and Sivapalan, 2017]] ; [[#Ferdous--2018|Ferdous et al., 2018]] ; [[#Caretta--2021|Caretta et al., 2021]] ). In particular, projected socioeconomic, cultural and political impacts on the vulnerable group are understudied, as is their resourcefulness through LK, adaptive capacity and community-led adaptation (Sections 4.6.9; 4.8.4; Cross-Chapter Box INDIG in Chapter 18).

In summary, there is ''high confidence'' that the magnitude and frequency of floods are projected to increase in many regions, including Asia, central Africa, western Europe, Central and South America and eastern North America, and decrease in northern North America, southern South America, the Mediterranean and eastern Europe. Projected increases in flooding pose increasing risks, with a 1.2–1.8 and 4–5 times increase in global GDP loss at 2°C and 4°C compared to 1.5°C warming, respectively ( ''medium confidence'' ). Without adaptation, projected increases in flooding are 1.4 to 2.5 and 2.5 to 3.9 times in global GDP loss at 2°C and 3°C compared to 1.5°C warming, respectively ( ''medium confidence'' ). However, regional differences in risks are large because of the strong influence of socioeconomic conditions and significant uncertainty in flood hazard projection. In small river basins and urban areas, there is ''medium confidence'' that projected increases in heavy rainfall would contribute to increases in rain-generated local flooding. However, the snowmelt floods are projected to decrease ( ''medium confidence'' ) and occur 25–30 d earlier in the year by the end of the 21st century with RCP8.5 ( ''high confidence'' ).

<div id="4.4.5" class="h2-container"></div>

<span id="projected-changes-in-droughts"></span>
=== 4.4.5 Projected Changes in Droughts ===

<div id="h2-23-siblings" class="h2-siblings"></div>

AR6 WGI ( [[#Douville--2021|Douville et al., 2021]] ) concluded that the total land area subject to increasing drought frequency and severity would expand ( ''high confidence'' ), and in the Mediterranean, southwestern South America and western North America, future aridification will far exceed the magnitude of change seen in the last millennium ( ''high confidence'' ). WGI ( [[#Seneviratne--2021|Seneviratne et al., 2021]] ) also find many consistencies among projections of climate change effects on different forms of drought (meteorological, agricultural/ecological and hydrological drought, 4.2.5), but also significant differences in some regions, particularly in the levels of confidence in projected changes.

Many studies focus on precipitation-based drought indices ( [[#Carrão--2018|Carrão et al., 2018]] ), but higher evaporative demands and changes in snow cover are additional drivers of hydrological, agricultural and ecological drought ( ''medium confidence'' ) in many regions of the world ( [[#Koirala--2014|Koirala et al., 2014]] ; [[#Prudhomme--2014|Prudhomme et al., 2014]] ; [[#Touma--2015|Touma et al., 2015]] ; [[#Wanders--2015|Wanders et al., 2015]] ; [[#Zhao--2015|Zhao and Dai, 2015]] ; [[#Naumann--2018|Naumann et al., 2018]] ; [[#Cook--2020a|Cook et al., 2020a]] ). Furthermore, these droughts (hydrological, agricultural and ecological) are often modulated by prevailing soil and hydro-morphological characteristics. Therefore, the choice of drought definition can affect the magnitude and even the sign of the projected drought change.

In a study with multiple climate models, global water models and scenarios, the choice of drought definition was the dominant source of uncertainty in the sign of projected change in drought frequency in over 17% of global land by 2070–2099, including several major wheat- and maize-growing areas where agricultural (soil moisture) drought is of high importance ( [[#Satoh--2021|Satoh et al., 2021]] ). [[#Cook--2020a|Cook et al. (2020a)]] noted that in the CMIP6 projections, soil moisture and runoff drying are more robust, spatially extensive and severe than precipitation, resulting in the frequency of agricultural drought increasing over wider areas than for meteorological drought. At 1.5°C global warming, the likelihood of extreme agricultural (soil moisture) drought is projected to at least double (100% increase) over large areas of northern South America, the Mediterranean, western China and high latitudes in North America and Eurasia (Figure 4.18, left column). The likelihood is projected to increase by 150–200% in these regions at 2°C global warming, with an expansion of the affected areas, and increase by over 200% at 4°C global warming. Agricultural drought likelihood also increases by 100–250% at 4°C global warming in southwestern North America, southwest Africa, southern Asia and Australia. The likelihood of extreme drought is projected to decrease in central North America, the Sahel, the Horn of Africa, the eastern Indian sub-continent and parts of western and eastern Asia. Using eight global hydrological models driven by a subset of four of the CMIP5 climate models, [[#Lange--2020|Lange et al. (2020)]] projected a 370% (30–790%) increase of the global population annually exposed to agricultural (soil moisture) droughts in response to 2°C global warming. Therefore, it is essential to consider the drought type when applying drought projections to impact and risk in decision-making, especially for informing adaptation. For example, if responses are explicitly tailored to agricultural (soil moisture) drought changes, projected changes in a meteorological (precipitation) drought metric may not provide accurate information.

Compared to CMIP5, the CMIP6 ensemble projects more consistent drying in the Amazon basin ( [[#Parsons--2020|Parsons, 2020]] ), more extensive declines in total soil moisture in Siberia ( [[#Cook--2020a|Cook et al., 2020a]] ) and stronger declines in westernmost North Africa and southwestern Australia. Projected declines in soil moisture in these geographies would cause a significant risk of agricultural drought. Also, importantly, projected changes in drought in many regions depend on the season and may not be evident in annual mean changes. For example, in northwestern Asia, hydrological (runoff) drought frequency is projected to decrease by 50–100% in autumn and winter but increase by up to 250% in spring and summer ( [[#Cook--2020a|Cook et al., 2020a]] ). In contrast, meteorological (precipitation) drought frequency is projected to increase by up to 350% throughout the year.

Drought projections are subject to uncertainties due to limits of predictability and understanding of the relevant biophysical processes. Uncertainties in regional climate changes are significant in many regions (see Figure 4.10, Figure 4.13, Figure 4.15), and in climate model ensembles, the range of regional outcomes generally increases with global warming. This widening of the range of outcomes can contribute to the increased likelihood of extreme droughts across the ensemble as a whole (Figure 4.18, right column). The response of transpiration to elevated CO 2 is also a significant uncertainty. The inclusion of CO 2 physiological effects leads to smaller projected increases in agricultural, ecological or hydrological drought ( [[#Milly--2016|Milly and Dunne, 2016]] ; [[#Yang--2020|Yang et al., 2020]] ). However, the level of uncertainties in representing the effects of CO 2 is still very high, precluding conclusive results in a global analysis ( [[#de%20Kauwe--2013|de Kauwe et al., 2013]] ; [[#Prudhomme--2014|Prudhomme et al., 2014]] ; [[#Yang--2016|Yang et al., 2016]] ). Most CMIP6 climate models include CO 2 physiological effects, but many hydrological models used for impacts studies do not.

<div id="_idContainer068" class="Figure"></div>

[[File:18d4a266e2c8ef8940e5dc8cdd2619dd IPCC_AR6_WGII_Figure_4_018.png]]

'''Figure 4.18 |''' '''Projected changes in the likelihood of an extreme single-year agricultural (soil moisture) drought event, with extreme drought defined as the driest 10% of years from 1995 to 2014, using total soil moisture projections pooled from the CMIP6 ensemble following Cook et al.''' '''(2020a).''' All ensemble members are treated as equally likely potential outcomes, and likelihoods are calculated using the whole ensemble. Left: Percentage change in the likelihood of extreme drought at GWLs of 4°C (top), 2°C (middle) and 1.5°C (bottom), with ‘extreme drought’ defined locally as the 10th percentile in individual grid boxes. Right: probability distribution functions of regional mean soil moisture anomalies for the climatic regions Mediterranean (MED), South American Monsoon (SAM) and West Southern Africa (WSAF) ( [[#Iturbide--2020|Iturbide et al., 2020]] ), at 1.5°C, 2°C and 4°C GWLs. The solid vertical line shows the baseline, that is, the 50th percentile in 1995–2014. The dashed vertical line shows the 10th percentile for 1995–2014, defining ‘extreme drought’ at the regional scale. Projections used the SSP5-8.5 scenario to maximise the number of ensemble members at higher GWLs, but global patterns of change are very similar for all scenarios ( [[#Cook--2020a|Cook et al., 2020a]] ), and for any given GWL, similar results can be expected with other scenarios ( [[#Seneviratne--2021|Seneviratne et al., 2021]] ).

Terrestrial water storage (TWS) is the sum of continental water stored in canopies, snow and ice, rivers, lakes and reservoirs, wetlands, soil and groundwater ( [[#Pokhrel--2021|Pokhrel et al., 2021]] ). TWS drought can therefore be considered to be a combination of agricultural, ecological and hydrological drought. The proportion of the global population exposed to TWS drought is projected to increase with ongoing climate change (Figure 4.19). By the late 21st century, under RCP6.0, the global land area in extreme-to-exceptional TWS drought is projected to increase from 3% to 7% ( [[#Pokhrel--2021|Pokhrel et al., 2021]] ), with increasing uncertainty over time. Combined with a medium population growth scenario (SSP2), this leads to the global population in this level of drought increasing from 3% to 8%, again with increasing uncertainty over time. Hydrological droughts can also be driven by direct human impact via water abstraction ( [[#Javadinejad--2019|Javadinejad et al., 2019]] ).

<div id="_idContainer070" class="Figure"></div>

[[File:128fed671b10e398b6e6dd808896fe12 IPCC_AR6_WGII_Figure_4_019.png]]

'''Figure 4.19 |''' '''Projected changes in the area under drought and population affected, defined with changes in the Terrestrial Water Storage–Drought Severity Index (TWS-DSI) projected with seven terrestrial hydrology models driven by four CMIP5 climate models using RCP6.''' 0.

'''(a)''' Fractional global land area under moderate-to-severe drought (top), defined as −0.8 ≤ TWS-DSI &lt; −1.6, and extreme-to-exceptional drought (bottom), defined as TWS-DSI &lt; −1.6.

'''(b)''' Fraction of global population exposed to moderate-to-severe (top) and extreme-to-exceptional (bottom) drought, using the SSP2 population projection. Dark lines show the ensemble means; shaded areas indicate uncertainty as ± 1 standard deviation. Reproduced from [[#Pokhrel--2021|Pokhrel et al. (2021)]] .

Critical knowledge gaps include uncertainties in regional drought due to regional climate change uncertainties, challenges in constraining plant physiological responses to atmospheric CO 2, and the uncertainties in modelling the role of different population projections in projecting regional drought risk.

In summary, the likelihood of drought is projected to increase in many regions over the 21st century ( ''high confidence'' ) even with strong climate change mitigation, and more severely in the absence of this. Different forms of drought broadly show similar patterns of projected change in many regions ( ''high confidence'' ), but the frequency of agricultural drought is projected to increase over wider areas than for meteorological drought ( ''medium confidence'' ). Clarity on the definition of drought is therefore important for informing decision-making. With the RCP6.0 and SSP2 scenarios, the global population exposed to extreme-to-exceptional terrestrial water storage drought is projected to increase from 3% to 8% over the 21st century.

<div id="4.4.6" class="h2-container"></div>

<span id="projected-changes-in-groundwater"></span>
=== 4.4.6 Projected Changes in Groundwater ===

<div id="h2-24-siblings" class="h2-siblings"></div>

AR5 concluded that the range of projected future changes in groundwater storage was large, from statistically significant declines to increases due to several uncertainties in existing models ( [[#Jiménez%20Cisneros--2014|Jiménez Cisneros et al., 2014]] ). AR6 ( [[#Douville--2021|Douville et al., 2021]] ) concluded with ''high confidence'' that projected increases in precipitation alone cannot ensure an increase in groundwater storage under a warming climate unless unsustainable trends in groundwater extraction are also reversed.

Projected impacts of climate change on groundwater systems are commonly simulated using models at local to global scales ( [[#Bierkens--2019|Bierkens and Wada, 2019]] ). The relations between climate change and groundwater are more complex than those embedded in current numerical models ( [[#Cuthbert--2019b|Cuthbert et al., 2019b]] ). For instance, groundwater systems register effects of drought with several years of lag effect, and aquifer response times to changes in hydraulic forcing also vary across aquifers ( [[#Cuthbert--2019a|Cuthbert et al., 2019a]] ). For instance, long groundwater response times can buffer drought impacts and lengthen recovery times to sustained drought events ( [[#Van%20Lanen--2013|Van Lanen et al., 2013]] ; [[#Opie--2020|Opie et al., 2020]] ).

Global total and non-renewable groundwater withdrawals are projected to increase from 952 km 3 year –1 (2010) to 1621 km 3 year –1 (2099) and from 304 km 3 year –1 (2010) to 597 km 3 year –1 (2099), respectively ( [[#Bierkens--2019|Bierkens and Wada, 2019]] ). At the same time, groundwater depletion is projected to increase from approximately 204 (± 30) km 3 year –1 in 2000 to 427 (± 56) km 3 year –1 by 2099 ( [[#Wada--2016|Wada, 2016]] ). Much of the projected depletion is a function of increased future abstraction of groundwater for irrigation and increased ET ( [[#Condon--2020|Condon et al., 2020]] ) in a warmer climate. For example, the projected doubling of average water use by 2050 in Tunisia is attributed partly (3.8–16.4%) to climate change and mainly to socioeconomic policies ( [[#Guermazi--2019|Guermazi et al., 2019]] ). Similarly, groundwater depletion in the Bengal Basin and North China Plain is more due to irrigation development than climate change per se ( [[#Leng--2015|Leng et al., 2015]] ; [[#Kirby--2016|Kirby et al., 2016]] ).

A recent synthesis of modelling studies conducted in various climates showed that out of 33 studies, 21 reported a decrease in the projected groundwater recharge or storage, eight reported an increase and the rest showed no substantial change ( [[#Amanambu--2020|Amanambu et al., 2020]] ). A global-scale multi-model ensemble study projected decreasing recharge in southern Chile, Brazil, central continental USA, the Mediterranean and East China, but consistent and increasing recharge for northern Europe and East Africa ( [[#Reinecke--2021|Reinecke et al., 2021]] ). In continental Spain, a modelling study ( [[#Pulido-Velazquez--2018|Pulido-Velazquez et al., 2018]] ) projected significant reductions in groundwater recharge in the central and southeast region but a small and localised increase in east and northeastern areas. In subarctic Alaska, increased contribution of glacier melts to streamflow and aquifer recharge under a warming climate is projected ( [[#Liljedahl--2017|Liljedahl et al., 2017]] ). In contrast, over the Iranian and Anatolia Plateaus, groundwater recharge is projected to reduce by ~77% in the spring season (March–May) due to a decrease in snowfall ( [[#Wu--2020|Wu et al., 2020]] ). Overall, several recent studies of climate change impacts on groundwater in different parts of the world have concluded that projected groundwater recharge could either increase or decrease, and results are often uncertain ( ''high confidence'' ) ( [[#Meixner--2016|Meixner et al., 2016]] ; [[#Zaveri--2016|Zaveri et al., 2016]] ; [[#Hartmann--2017|Hartmann et al., 2017]] ; [[#Mehran--2017|Mehran et al., 2017]] ; [[#Tillman--2017|Tillman et al., 2017]] ; [[#Kahsay--2018|Kahsay et al., 2018]] ; [[#Herbert--2019|Herbert and Döll, 2019]] ).

[[#Wu--2020|Wu et al. (2020)]] reported a projected increase in future groundwater storage in the semiarid regions of northwest India, North China Plain, the Guarani Aquifer in South America and Canning Basin in Australia due to significant increases in projected precipitation, but the models do not consider local hydrogeological characteristics. However, the projected irrigation expansion could negate this positive gain in groundwater storage ( [[#Sishodia--2018|Sishodia et al., 2018]] ; [[#Wu--2020|Wu et al., 2020]] ). In drylands (e.g., playas in the southwestern USA), where focused groundwater recharge processes dominate, greater recharge is projected to occur from the increased number of significant runoff-generating extreme precipitation events in the future ( [[#McKenna--2018|McKenna and Sala, 2018]] ). Overall, an emerging body of studies have projected amplification of episodic recharge in the tropics and semiarid regions due to extreme precipitation under global warming ( ''medium confidence'' ).

Climate change is also projected to impact groundwater-dependent ecosystems and groundwater quality negatively ( ''medium confidence'' ). Projected increase in precipitation intensity and storms can contaminate groundwater by mobilising contaminants such as chemical fertilisers, pesticides and antibiotics and leaching of human waste from pit latrines into groundwater ( [[#Amanambu--2020|Amanambu et al., 2020]] ; [[#Lall--2020|Lall et al., 2020]] ). By 2050, environmentally critical streamflow is projected to be affected in 42–79% of the world’s watersheds. The majority of these watersheds currently experience intensive groundwater use, and changes in critical streamflow are projected to negatively impact aquatic ecosystems ( [[#de%20Graaf--2019|de Graaf et al., 2019]] ). Using a global synthesis of 9404 data points from 32 countries across six continents, [[#McDonough--2020|McDonough et al. (2020)]] report increases in DOC concentrations in groundwater following projected changes in precipitation and temperature. For example, hotspots of high DOC concentration (increases of up to 45%) are associated mainly with increased temperatures in the wettest quarter of the year in the southeastern USA under RCP8.5 scenarios.

The projected rise in sea levels can lead to saline intrusion into aquifers in low-lying areas and small islands and threaten coastal ecosystems and livelihood resilience; for example, in already vulnerable countries like Bangladesh and vulnerable ecosystems like the mangrove forest of Sundarbans ( [[#Befus--2020|Befus et al., 2020]] ; [[#Dasgupta--2020|Dasgupta et al., 2020]] ; [[#Shamsudduha--2020|Shamsudduha et al., 2020]] ). However, hydrogeological properties, aquifer settings and impacts of over-abstraction are more important determinants of salinisation of coastal aquifers than slowly rising sea levels ( [[#Michael--2013|Michael et al., 2013]] ; [[#Taylor--2013a|Taylor et al., 2013a]] ). The projected contribution of global groundwater depletion to sea level rise is expected to increase from 0.57 (± 0.09) mm year –1 in 2000 to 0.82 (± 0.13) mm year –1 by 2050, driven by a growing trend in groundwater extraction ( [[#Wada--2016|Wada, 2016]] ). However, several uncertainties around model parametrisation remain ( [[#Wada--2017|Wada et al., 2017]] ).

There are several knowledge gaps in our understanding of the global-scale sensitivity of groundwater systems to climate change and resulting feedbacks ( [[#Maxwell--2016|Maxwell and Condon, 2016]] ; [[#Cuthbert--2019a|Cuthbert et al., 2019a]] ). There are process uncertainties in groundwater recharge simulation due to the potential impact of atmospheric CO 2 on vegetation and resulting changes in ET ( [[#Reinecke--2021|Reinecke et al., 2021]] ). There are uncertainties in impact models due to poor representation of recharge pathways (diffuse compared to focused) and inability to adequately capture feedbacks among climate, land use and groundwater systems ( [[#Meixner--2016|Meixner et al., 2016]] ). Finally, there are gaps in long-term observational data, especially in less-developed countries ( [[#Amanambu--2020|Amanambu et al., 2020]] ), making it challenging to evaluate the performance of impact models ( [[#Gleeson--2020|Gleeson et al., 2020]] ).

In summary, groundwater abstraction is projected to deplete the long-term, non-renewable storage as withdrawals are projected to increase significantly in all major aquifers worldwide ( ''medium evidence, high agreement'' ). In the tropics and semiarid regions, growing precipitation intensification under global warming may enhance the resilience of groundwater through increased episodic recharge ( ''medium confidence'' ). However, in the semiarid areas, over-abstraction continues to be a threat to groundwater storage and can nullify the benefits of increased future recharge.

<div id="4.4.7" class="h2-container"></div>

<span id="projected-changes-in-water-quality"></span>
=== 4.4.7 Projected Changes in Water Quality ===

<div id="h2-25-siblings" class="h2-siblings"></div>

AR5 concluded that climate change was projected to reduce water quality ( [[#Jiménez%20Cisneros--2014|Jiménez Cisneros et al., 2014]] ). SR1.5 assessed with ''low confidence'' differences in projected impacts under 1.5°C compared with 2°C of warming ( [[#Hoegh-Guldberg--2018|Hoegh-Guldberg et al., 2018]] ). In addition, SROCC reported water quality degradation due to the release of legacy contaminants in glaciers and permafrost ( ''medium confidence'' ) ( [[#Hock--2019b|Hock et al., 2019b]] ). The AR6 WGI Report does not explicitly mention water quality issues.

Water insecurity due to water quality degradation is projected to increase under climate change due to warming, enhanced floods and sea level rise ( [[#Arnell--2014|Arnell and Lloyd-Hughes, 2014]] ; [[#Dyer--2014|Dyer et al., 2014]] ; [[#Whitehead--2015|Whitehead et al., 2015]] ) ( ''medium confidence'' ). Drought-driven diminishing river and lake levels ( [[#Jeppesen--2015|Jeppesen et al., 2015]] ) and continued water abstraction for irrigation ( [[#Aragüés--2015|Aragüés et al., 2015]] ) may contribute to the salinisation of soil and water. In addition, warming is projected to disrupt the historical sequestration of contaminants in permafrost in the Arctic and mountain regions ( [[#Bond--2018|Bond and Carr, 2018]] ).

Quantitative projections on climate-induced water quality degradation are sparse. Aminomethylphosphonic acid and glyphosate are projected to exceed drinking water quality standards in dry years in a high-emissions scenario in the Meuse River in Europe by 2050 ( [[#Sjerps--2017|Sjerps et al., 2017]] ). From 2020 to 2050, based on scenarios RCP2.6, RCP4.5 and RCP8.5, the incidences of total nitrogen pollution are projected as 97.3, 97.1 and 94.6%, respectively, in drought–flood abrupt alternation months compared to 69.3, 69.7 and 67.5% in normal months in the Luanhe River basin in China ( [[#Bi--2019|Bi et al., 2019]] ). From 2012 to 2050, freshwater river area is expected to decrease from 40.8% to 17.1–19.7% under different sea level rise scenarios in the southwest coastal zone of Bangladesh ( [[#Dasgupta--2013|Dasgupta et al., 2013]] ). Under the warming scenario of +4.8°C increase by the end of the century, the average nutrient abundance is projected to triple in a shallow lake in the northwest of England ( [[#Richardson--2019|Richardson et al., 2019]] ).

While there is some understanding of the potential effect of glacier and permafrost degradation on water quality, projections are lacking. Research is limited mainly in Europe and North America, and quantifying the future water quality changes is still incipient.

In summary, climate change is projected to increase water pollution incidences, salinisation and eutrophication due to increasing drought and flood events, sea level rise and water temperature rise, respectively, in some local rivers and lakes, but there is a dearth of exact quantification at a global scale ( ''medium confidence'' ).

<div id="4.4.8" class="h2-container"></div>

<span id="projected-changes-in-soil-erosion-and-sediment-load"></span>
=== 4.4.8 Projected Changes in Soil Erosion and Sediment Load ===

<div id="h2-26-siblings" class="h2-siblings"></div>

AR5 stated that soil erosion and sediment load are projected to change ( ''low confidence'' ) due to warming and increased rainfall intensity ( [[#Jiménez%20Cisneros--2014|Jiménez Cisneros et al., 2014]] ). SRCCL concluded that future climate change will increase, with ''medium confidence'' , the potential for water-driven soil erosion in many dryland areas, causing soil organic carbon decline ( [[#Mirzabaev--2019|Mirzabaev et al., 2019]] ). SR1.5 ( [[#Hoegh-Guldberg--2018|Hoegh-Guldberg et al., 2018]] ) concluded that because of the complex interactions among climate change, land cover, soil management, etc., the differences between mean annual sediment load under 1.5°C and 2°C of warming are unclear.

Globally, climate change is estimated to be responsible for 30–66% increase of soil erosion by 2070, while socioeconomic developments impacting land use may lead to ± 10% change of soil erosion ( [[#Borrelli--2020|Borrelli et al., 2020]] ). At a regional scale, different effects of the climate change impact on soil losses are found owing to the ensemble experiments with climate models coupled with regional/local models of soil erosion and sediment yield. In the 21st century, the soil erosion rates are projected to increase for the European countries (Czech Republic ( [[#Svoboda--2016|Svoboda et al., 2016]] ), Belgium ( [[#Mullan--2019|Mullan et al., 2019]] ), Spain ( [[#Eekhout--2018|Eekhout et al., 2018]] ; [[#Eekhout--2019a|Eekhout and de Vente, 2019a]] ; [[#Eekhout--2019b|Eekhout and De Vente, 2019b]] ), Germany ( [[#Gericke--2019|Gericke et al., 2019]] )) by 10–80% depending on the emission scenario and time period of the projection, as well as for the USA ( [[#Garbrecht--2015|Garbrecht and Zhang, 2015]] ) and Australia ( [[#Yang--2015|Yang et al., 2015]] b; [[#Zhu--2020|Zhu et al., 2020]] ). Only a few studies demonstrated decreasing trend in soil erosion, for example, up to 9% with RCP8.5 scenario in Greece ( [[#Vantas--2020|Vantas et al., 2020]] ). Sediment yield is projected to both increase (5–16% with the SRES A1, B1, B2 scenarios in Vietnam and Laos ( [[#Giang--2017|Giang et al., 2017]] ), 11% with the RCP8.5 scenario and 8% with the SRES A2 scenario in the USA ( [[#Yasarer--2017|Yasarer et al., 2017]] and [[#Wagena--2018|Wagena et al., 2018]] , respectively), 19–37% with the RCP4.5, RCP8.5 scenarios in Burkina Faso ( [[#Op%20de%20Hipt--2018|Op de Hipt et al., 2018]] )) and decrease (30% with the SRES A1B scenario in the southwest USA ( [[#Francipane--2015|Francipane et al., 2015]] ), 8–11% with the SRES A1B scenario in Spain ( [[#Rodríguez-Blanco--2016|Rodríguez-Blanco et al., 2016]] ), 11–52% with the RCP4.5, RCP8.5 scenarios in Ethiopia ( [[#Gadissa--2018|Gadissa et al., 2018]] ), 13–62% with the RCP2.6, RCP8.5 scenarios in Canada ( [[#Loiselle--2020|Loiselle et al., 2020]] )) over the different regions of the world in the 21st century.

Post-fire sedimentation is projected to increase for nearly nine tenths of watersheds by &gt;10% and for more than one third of watersheds by &gt;100% by the 2041 to 2050 decade in the western USA with the SRES A1B scenario ( [[#Sankey--2017|Sankey et al., 2017]] ).

In summary, soil losses mainly depend on the combined effects of climate and land use changes. Herewith, recent studies demonstrate increasing impact of the projected climate change (increase of precipitation, thawing permafrost) on soil erosion ( ''medium confidence'' ).

<div id="4.5" class="h1-container"></div>

<span id="projected-sectoral-water-related-risks"></span>