Editing IPCC:AR6/SR15/Chapter-3 (section)

== Box 3.2: Droughts in the Mediterranean Basin and the Middle East ==

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

Human society has developed in tandem with the natural environment of the Mediterranean basin over several millennia, laying the groundwork for diverse and culturally rich communities. Even if advances in technology may offer some protection from climatic hazards, the consequences of climatic change for inhabitants of this region continue to depend on the long-term interplay between an array of societal and environmental factors (Holmgren et al., 2016) <sup>[[#fn:r170|170]]</sup> . As a result, the Mediterranean is an example of a region with high vulnerability where various adaptation responses have emerged. Previous IPCC assessments and recent publications project regional changes in climate under increased temperatures, including consistent climate model projections of increased precipitation deficit amplified by strong regional warming (Section 3.3.3; Seneviratne et al., 2012; Christensen et al., 2013; Collins et al., 2013; Greve and Seneviratne, 2015 <sup>[[#fn:r171|171]]</sup> ).

The long history of resilience to climatic change is especially apparent in the eastern Mediterranean region, which has experienced a strong negative trend in precipitation since 1960 (Mathbout et al., 2017) <sup>[[#fn:r172|172]]</sup> and an intense and prolonged drought episode between 2007 and 2010 (Kelley et al., 2015) <sup>[[#fn:r173|173]]</sup> . This drought was the longest and most intense in the last 900 years (Cook et al., 2016) <sup>[[#fn:r174|174]]</sup> . Some authors (e.g., Trigo et al., 2010; Kelley et al., 2015) <sup>[[#fn:r175|175]]</sup> assert that very low precipitation levels have driven a steep decline in agricultural productivity in the Euphrates and Tigris catchment basins, and displaced hundreds of thousands of people, mainly in Syria. Impacts on the water resources (Yazdanpanah et al., 2016) <sup>[[#fn:r176|176]]</sup> and crop performance in Iran have also been reported (Saeidi et al., 2017) <sup>[[#fn:r177|177]]</sup> . Many historical periods of turmoil have coincided with severe droughts, for example the drought which occurred at the end of the Bronze Age approximately 3200 years ago (Kaniewski et al., 2015) <sup>[[#fn:r178|178]]</sup> . In this instance, a number of flourishing eastern Mediterranean civilizations collapsed, and rural settlements re-emerged with agro-pastoral activities and limited long-distance trade. This illustrates how some vulnerable regions are forced to pursue drastic adaptive responses, including migration and societal structure changes.

The potential evolution of drought conditions under 1.5°C or 2°C of global warming (Section 3.3.4) can be analysed by comparing the 2008 drought (high temperature, low precipitation) with the 1960 drought (low temperature, low precipitation) (Kelley et al., 2015) <sup>[[#fn:r179|179]]</sup> . Though the precipitation deficits were comparable, the 2008 drought was amplified by increased evapotranspiration induced by much higher temperatures (a mean increase of 1°C compared with the 1931–2008 period in Syria) and a large population increase (from 5 million in 1960 to 22 million in 2008). Koutroulis et al. (2016) <sup>[[#fn:r180|180]]</sup> reported that only 6% out of the total 18% decrease in water availability projected for Crete under 2°C of global warming at the end of the 21st century would be due to decreased precipitation, with the remaining 12% due to an increase in evapotranspiration. This study and others like it confirm an important risk of extreme drought conditions for the Middle East under 1.5°C of global warming (Jacob et al., 2018) <sup>[[#fn:r181|181]]</sup> , with risks being even higher in continental locations than on islands; these projections are consistent with current observed changes (Section 3.3.4; Greve et al., 2014) <sup>[[#fn:r182|182]]</sup> . Risks of drying in the Mediterranean region could be substantially reduced if global warming is limited to 1.5°C compared to 2°C or higher levels of warming (Section 3.4.3; Guiot and Cramer, 2016) <sup>[[#fn:r183|183]]</sup> . Higher warming levels may induce high levels of vulnerability exacerbated by large changes in demography.

<span id="runoff-and-fluvial-flooding"></span>
=== 3.3.5 Runoff and Fluvial Flooding ===

<div id="section-3-3-5-1"></div>

<span id="observed-and-attributed-changes-in-runoff-and-river-flooding"></span>
==== 3.3.5.1 Observed and attributed changes in runoff and river flooding ====

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

There has been progress since AR5 in identifying historical changes in streamflow and continental runoff. Using the available streamflow data, Dai (2016) <sup>[[#fn:r184|184]]</sup> showed that long‐term (1948–2012) flow trends are statistically significant only for 27.5% of the world’s 200 major rivers, with negative trends outnumbering the positive ones. Although streamflow trends are mostly not statistically significant, they are consistent with observed regional precipitation changes. From 1950 to 2012, precipitation and runoff have increased over southeastern South America, central and northern Australia, the central and northeastern United States, central and northern Europe, and most of Russia, and they have decreased over most of Africa, East and South Asia, eastern coastal Australia, the southeastern and northwestern United States, western and eastern Canada, the Mediterranean region and some regions of Brazil (Dai, 2016) <sup>[[#fn:r185|185]]</sup> .

A large part of the observed regional trends in streamflow and runoff might have resulted from internal multi-decadal and multi-year climate variations, especially the Pacific decadal variability (PDV), the Atlantic Multi-Decadal Oscillation (AMO) and the El Niño–Southern Oscillation (ENSO), although the effect of anthropogenic greenhouse gases and aerosols could also be important (Hidalgo et al., 2009; Gu and Adler, 2013, 2015; Chiew et al., 2014; Luo et al., 2016; Gudmundsson et al., 2017) <sup>[[#fn:r186|186]]</sup> . Additionally, other human activities can influence the hydrological cycle, such as land-use/land-cover change, modifications in river morphology and water table depth, construction and operation of hydropower plants, dikes and weirs, wetland drainage, and agricultural practices such as water withdrawal for irrigation. All of these activities can also have a large impact on runoff at the river basin scale, although there is less agreement over their influence on global mean runoff (Gerten et al., 2008; Sterling et al., 2012; Hall et al., 2014; Betts et al., 2015; Arheimer et al., 2017) <sup>[[#fn:r187|187]]</sup> . Some studies suggest that increases in global runoff resulting from changes in land cover or land use (predominantly deforestation) are counterbalanced by decreases resulting from irrigation (Gerten et al., 2008; Sterling et al., 2012) <sup>[[#fn:r188|188]]</sup> . Likewise, forest and grassland fires can modify the hydrological response at the watershed scale when the burned area is significant (Versini et al., 2013; Springer et al., 2015; Wine and Cadol, 2016) <sup>[[#fn:r189|189]]</sup> .

Few studies have explored observed changes in extreme streamflow and river flooding since the IPCC AR5. Mallakpour and Villarini (2015) <sup>[[#fn:r190|190]]</sup> analysed changes of flood magnitude and frequency in the central United States by considering stream gauge daily records with at least 50 years of data ending no earlier than 2011. They showed that flood frequency has increased, whereas there was ''limited evidence'' of a decrease in flood magnitude in this region. Stevens et al. (2016) <sup>[[#fn:r191|191]]</sup> found a rise in the number of reported floods in the United Kingdom during the period 1884–2013, with flood events appearing more frequently towards the end of the 20th century. A peak was identified in 2012, when annual rainfall was the second highest in over 100 years. Do et al. (2017) <sup>[[#fn:r192|192]]</sup> computed the trends in annual maximum daily streamflow data across the globe over the 1966–2005 period. They found decreasing trends for a large number of stations in western North America and Australia, and increasing trends in parts of Europe, eastern North America, parts of South America, and southern Africa.

In summary, streamflow trends since 1950 are not statistically significant in most of the world’s largest rivers ( ''high confidence'' ) '','' while flood frequency and extreme streamflow have increased in some regions ( ''high confidence'' ) ''.''

<div id="section-3-3-5-2"></div>

<span id="projected-changes-in-runoff-and-river-flooding-at-1.5c-versus-2c-of-global-warming"></span>
==== 3.3.5.2 Projected changes in runoff and river flooding at 1.5°C versus 2°C of global warming ====

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

Global-scale assessments of projected changes in freshwater systems generally suggest that areas with either positive or negative changes in mean annual streamflow are smaller for 1.5°C than for 2°C of global warming (Betts et al., 2018; Döll et al., 2018) <sup>[[#fn:r193|193]]</sup> . Döll et al. (2018) <sup>[[#fn:r194|194]]</sup> found that only 11% of the global land area (excluding Greenland and Antarctica) shows a statistically significantly larger hazard at 2°C than at 1.5°C. Significant decreases are found for 13% of the global land area for both global warming levels, while significant increases are projected to occur for 21% of the global land area at 1.5°C, and rise to between 26% (Döll et al., 2018) <sup>[[#fn:r195|195]]</sup> and approximately 50% (Betts et al., 2018) <sup>[[#fn:r196|196]]</sup> at 2°C.

At the regional scale, projected runoff changes generally follow the spatial extent of projected changes in precipitation (see Section 3.3.3). Emerging literature includes runoff projections for different warming levels. For 2°C of global warming, an increase in runoff is projected for much of the high northern latitudes, Southeast Asia, East Africa, northeastern Europe, India, and parts of, Austria, China, Hungary, Norway, Sweden, the northwest Balkans and Sahel (Schleussner et al., 2016b; Donnelly et al., 2017; Döll et al., 2018; Zhai et al., 2018) <sup>[[#fn:r197|197]]</sup> . Additionally, decreases are projected in the Mediterranean region, southern Australia, Central America, and central and southern South America (Schleussner et al., 2016b; Donnelly et al., 2017; Döll et al., 2018) <sup>[[#fn:r198|198]]</sup> . Differences between 1.5°C and 2°C would be most prominent in the Mediterranean, where the median reduction in annual runoff is expected to be about 9% ( ''likely'' range 4.5–15.5%) at 1.5°C, while at 2°C of warming runoff could decrease by 17% ( ''likely'' range 8–25%) (Schleussner et al., 2016b) <sup>[[#fn:r199|199]]</sup> . Consistent with these projections, Döll et al. (2018) <sup>[[#fn:r200|200]]</sup> found that statistically insignificant changes in the mean annual streamflow around the Mediterranean region became significant when the global warming scenario was changed from 1.5°C to 2°C, with decreases of 10–30% between these two warming levels. Donnelly et al. (2017) <sup>[[#fn:r201|201]]</sup> found an intense decrease in runoff along both the Iberian and Balkan coasts with an increase in warming level.

Basin-scale projections of river runoff at different warming levels are available for many regions. Betts et al. (2018) <sup>[[#fn:r202|202]]</sup> assessed runoff changes in 21 of the world’s major river basins at 1.5°C and 2°C of global warming (Figure 3.15). They found a general tendency towards increased runoff, except in the Amazon, Orange, Danube and Guadiana basins where the range of projections indicate decreased mean flows (Figure 3.13). In the case of the Amazon, mean flows are projected to decline by up to 25% at 2°C global warming Betts et al., 2018, Gosling et al., (2017) <sup>[[#fn:r204|204]]</sup> analysed the impact of global warming of 1°C, 2°C and 3°C above pre-industrial levels on river runoff at the catchment scale, focusing on eight major rivers in different continents: Upper Amazon, Darling, Ganges, Lena, Upper Mississippi, Upper Niger, Rhine and Tagus. Their results show that the sign and magnitude of change with global warming for the Upper Amazon, Darling, Ganges, Upper Niger and Upper Mississippi is unclear, while the Rhine and Tagus may experience decreases in projected runoff and the Lena may experience increases. Donnelly et al. (2017) <sup>[[#fn:r205|205]]</sup> analysed the mean flow response to different warming levels for six major European rivers: Glomma, Wisla, Lule, Ebro, Rhine and Danube. Consistent with the increases in mean runoff projected for large parts of northern Europe, the Glomma, Wisla and Lule rivers could experience increased discharges with global warming while discharges from the Ebro could decrease, in part due to a decrease in runoff in southern Europe. In the case of the Rhine and Danube rivers, Donnelly et al. (2017) <sup>[[#fn:r206|206]]</sup> did not find clear results. Mean annual runoff of the Yiluo River catchment in northern China is projected to decrease by 22% at 1.5°C and by 21% at 2°C, while the mean annual runoff for the Beijiang River catchment in southern China is projected to increase by less than 1% at 1.5°C and 3% at 2°C in comparison to the studied baseline period (L. Liu et al., 2017) <sup>[[#fn:r207|207]]</sup> . Chen et al. (2017) <sup>[[#fn:r208|208]]</sup> assessed the future changes in water resources in the Upper Yangtze River basin for the same warming levels and found a slight decrease in the annual discharge at 1.5°C but a slight increase at 2°C. Montroull et al. (2018) <sup>[[#fn:r209|209]]</sup> studied the hydrological impacts of the main rivers (Paraguay, Paraná, Iguazú and Uruguay) in La Plata basin in South America under 1.5°C and 2°C of global warming and for two emissions scenarios. The Uruguay basin shows increases in streamflow for all scenarios/warming targets except for the combination of RCP8.5/1.5°C of warming. The increase is approximately 15% above the 1981–2000 reference period for 2°C of global warming and the RCP4.5 scenario. For the other three rivers the sign of the change in mean streamflow depends strongly on the RCP and GCM used.

Marx et al. (2018) <sup>[[#fn:r210|210]]</sup> analysed how hydrological low flows in Europe are affected under different global warming levels (1.5°C, 2°C and 3°C). The Alpine region showed the strongest low flow increase, from 22% at 1.5°C to 30% at 2°C, because of the relatively large snow melt contribution, while in the Mediterranean low flows are expected to decrease because of the decreases in annual precipitation projected for that region. Döll et al. (2018) <sup>[[#fn:r211|211]]</sup> found that extreme low flows in the tropical Amazon, Congo and Indonesian basins could decrease by 10% at 1.5°C, whereas they could increase by 30% in the southwestern part of Russia under the same warming level. At 2°C, projected increases in extreme low flows are exacerbated in the higher northern latitudes and in eastern Africa, India and Southeast Asia, while projected decreases intensify in the Amazon basin, western United States, central Canada, and southern and western Europe, although not in the Congo basin or Indonesia, where models show less agreement.

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

<span id="figure-3.15"></span>
<!-- START IMG -->
<!-- IMG TITLE -->
'''Figure 3.15'''

<span id="runoff-changes-in-twenty-one-of-the-worlds-major-river-basins-at-1.5c-blue-and-2c-orange-of-global-warming-simulated-by-the-joint-uk-land-environment-simulator-jules-ecosystemhydrology-model-under-the-ensemble-of-six-climate-projections."></span>
<!-- IMG CAPTION -->
'''Runoff changes in twenty-one of the world’s major river basins at 1.5°C (blue) and 2°C (orange) of global warming, simulated by the Joint UK Land Environment Simulator (JULES) ecosystem–hydrology model under the ensemble of six climate projections.'''
<!-- IMG FILE -->
[[File:e7341fe1f642366233511a12267322a3 Figure_3.15-1024x576.jpg]]

Boxes show the 25th and 75th percentile changes, whiskers show the range, circles show the four projections that do not define the ends of the range, and crosses show the ensemble means. Numbers in square brackets show the ensemble-mean flow in the baseline (millimetres of rain equivalent) (Source: Betts et al., 2018) <sup>[[#fn:r212|212]]</sup> .

<!-- END IMG -->
<div id="section-3-3-5-2-block-3"></div>

Recent analyses of projections in river flooding and extreme runoff and flows are available for different global warming levels. At the global scale, Alfieri et al. (2017) <sup>[[#fn:r213|213]]</sup> assessed the frequency and magnitude of river floods and their impacts under 1.5°C, 2°C and 4°C global warming scenarios. They found that flood events with an occurrence interval longer than the return period of present-day flood protections are projected to increase in all continents under all considered warming levels, leading to a widespread increment in the flood hazard. Döll et al. (2018) <sup>[[#fn:r214|214]]</sup> found that high flows are projected to increase significantly on 11% and 21% of the global land area at 1.5°C and 2°C, respectively. Significantly increased high flows are expected to occur in South and Southeast Asia and Central Africa at 1.5°C, with this effect intensifying and including parts of South America at 2°C.

Regarding the continental scale, Donnelly et al. (2017) <sup>[[#fn:r215|215]]</sup> and Thober et al. (2018) <sup>[[#fn:r216|216]]</sup> explored climate change impacts on European high flows and/or floods under 1.5°C, 2°C and 3°C of global warming. Thober et al. (2018) <sup>[[#fn:r217|217]]</sup> identified the Mediterranean region as a hotspot of change, with significant decreases in high flows of −11% and –13% at 1.5°C and 2°C, respectively, mainly resulting from reduced precipitation (Box 3.2). In northern regions, high flows are projected to rise by 1% and 5% at 1.5°C and 2°C, respectively, owing to increasing precipitation, although floods could decrease by 6% in both scenarios because of less snowmelt. Donnelly et al. (2017) <sup>[[#fn:r218|218]]</sup> found that high runoff levels could rise in intensity, robustness and spatial extent over large parts of continental Europe with an increasing warming level. At 2°C, flood magnitudes are expected to increase significantly in Europe south of 60°N, except for some regions (Bulgaria, Poland and southern Spain); in contrast, they are projected to decrease at higher latitudes (e.g., in most of Finland, northwestern Russia and northern Sweden), with the exception of southern Sweden and some coastal areas in Norway where flood magnitudes may increase (Roudier et al., 2016) <sup>[[#fn:r219|219]]</sup> . At the basin scale, Mohammed et al. (2017) <sup>[[#fn:r220|220]]</sup> found that floods are projected to be more frequent and flood magnitudes greater at 2°C than at 1.5°C in the Brahmaputra River in Bangladesh. In coastal regions, increases in heavy precipitation associated with tropical cyclones (Section 3.3.6) combined with increased sea levels (Section 3.3.9) may lead to increased flooding (Section 3.4.5).

In summary, there is ''medium confidence'' that global warming of 2°C above the pre-industrial period would lead to an expansion of the area with significant increases in runoff, as well as the area affected by flood hazard, compared to conditions at 1.5°C of global warming. A global warming of 1.5°C would also lead to an expansion of the global land area with significant increases in runoff ( ''medium confidence'' ) and to an increase in flood hazard in some regions ( ''medium confidence'' ) compared to present-day conditions.

<span id="tropical-cyclones-and-extratropical-storms"></span>
=== 3.3.6 Tropical Cyclones and Extratropical Storms ===

<div id="section-3-3-6-block-1"></div>

Most recent studies on observed trends in the attributes of tropical cyclones have focused on the satellite era starting in 1979 (Rienecker et al., 2011) <sup>[[#fn:r221|221]]</sup> , but the study of observed trends is complicated by the heterogeneity of constantly advancing remote sensing techniques and instrumentation during this period (e.g., Landsea, 2006; Walsh et al., 2016) <sup>[[#fn:r222|222]]</sup> . Numerous studies leading up to and after AR5 have reported a decreasing trend in the global number of tropical cyclones and/or the globally accumulated cyclonic energy (Emanuel, 2005; Elsner et al., 2008; Knutson et al., 2010; Holland and Bruyère, 2014; Klotzbach and Landsea, 2015; Walsh et al., 2016) <sup>[[#fn:r223|223]]</sup> . A theoretical physical basis for such a decrease to occur under global warming was recently provided by Kang and Elsner (2015) <sup>[[#fn:r224|224]]</sup> . However, using a relatively short (20 year) and relatively homogeneous remotely sensed record, Klotzbach (2006) <sup>[[#fn:r225|225]]</sup> reported no significant trends in global cyclonic activity, consistent with more recent findings of Holland and Bruyère (2014) <sup>[[#fn:r226|226]]</sup> . Such contradictions, in combination with the fact that the almost four-decade-long period of remotely sensed observations remains relatively short to distinguish anthropogenically induced trends from decadal and multi-decadal variability, implies that there is only ''low confidence'' regarding changes in global tropical cyclone numbers under global warming over the last four decades.

Studies in the detection of trends in the occurrence of very intense tropical cyclones (category 4 and 5 hurricanes on the Saffir-Simpson scale) over recent decades have yielded contradicting results. Most studies have reported increases in these systems (Emanuel, 2005; Webster et al., 2005; Klotzbach, 2006; Elsner et al., 2008; Knutson et al., 2010; Holland and Bruyère, 2014; Walsh et al., 2016) <sup>[[#fn:r227|227]]</sup> , in particular for the North Atlantic, North Indian and South Indian Ocean basins (e.g., Singh et al., 2000; Singh, 2010; Kossin et al., 2013; Holland and Bruyère, 2014; Walsh et al., 2016) <sup>[[#fn:r228|228]]</sup> . In the North Indian Ocean over the Arabian Sea, an increase in the frequency of extremely severe cyclonic storms has been reported and attributed to anthropogenic warming (Murakami et al., 2017) <sup>[[#fn:r229|229]]</sup> . However, to the east over the Bay of Bengal, tropical cyclones and severe tropical cyclones have exhibited decreasing trends over the period 1961–2010, although the ratio between severe tropical cyclones and all tropical cyclones is increasing (Mohapatra et al., 2017) <sup>[[#fn:r230|230]]</sup> . Moreover, studies that have used more homogeneous records, but were consequently limited to rather short periods of 20 to 25 years, have reported no statistically significant trends or decreases in the global number of these systems (Kamahori et al., 2006; Klotzbach and Landsea, 2015) <sup>[[#fn:r231|231]]</sup> . Likewise, CMIP5 model simulations of the historical period have not produced anthropogenically induced trends in very intense tropical cyclones (Bender et al., 2010; Knutson et al., 2010, 2013; Camargo, 2013; Christensen et al., 2013) <sup>[[#fn:r232|232]]</sup> , consistent with the findings of Klotzbach and Landsea (2015) <sup>[[#fn:r233|233]]</sup> . There is consequently ''low confidence'' in the conclusion that the number of very intense cyclones is increasing globally.

General circulation model (GCM) projections of the changing attributes of tropical cyclones under high levels of greenhouse gas forcing (3°C to 4°C of global warming) consistently indicate decreases in the global number of tropical cyclones (Knutson et al., 2010, 2015; Sugi and Yoshimura, 2012; Christensen et al., 2013; Yoshida et al., 2017) <sup>[[#fn:r234|234]]</sup> . A smaller number of studies based on statistical downscaling methodologies contradict these findings, however, and indicate increases in the global number of tropical cyclones under climate change (Emanuel, 2017) <sup>[[#fn:r235|235]]</sup> . Most studies also indicate increases in the global number of very intense tropical cyclones under high levels of global warming (Knutson et al., 2015; Sugi et al., 2017) <sup>[[#fn:r236|236]]</sup> , consistent with dynamic theory (Kang and Elsner, 2015) <sup>[[#fn:r237|237]]</sup> , although a few studies contradict this finding (e.g., Yoshida et al., 2017) <sup>[[#fn:r238|238]]</sup> . Hence, it is assessed that under 3°C to 4°C of warming that the global number of tropical cyclones would decrease whilst the number of very intense cyclones would increase ( ''medium confidence'' ).

To date, only two studies have directly explored the changing tropical cyclone attributes under 1.5°C versus 2°C of global warming. Using a high resolution global atmospheric model, Wehner et al. (2018a) <sup>[[#fn:r239|239]]</sup> concluded that the differences in tropical cyclone statistics under 1.5°C versus 2°C stabilization scenarios, as defined by the HAPPI protocols (Mitchell et al., 2017) <sup>[[#fn:r240|240]]</sup> are small. Consistent with the majority of studies performed for higher degrees of global warming, the total number of tropical cyclones is projected to decrease under global warming, whilst the most intense (categories 4 and 5) cyclones are projected to occur more frequently. These very intense storms are projected to be associated with higher peak wind speeds and lower central pressures under 2°C versus 1.5°C of global warming. The accumulated cyclonic energy is projected to decrease globally from 1.5°C to 2°C, in association with a decrease in the global number of tropical cyclones under progressively higher levels of global warming. It is also noted that heavy rainfall associated with tropical cyclones was assessed in the IPCC SREX as ''likely'' to increase under increasing global warming (Seneviratne et al., 2012) <sup>[[#fn:r241|241]]</sup> . Two recent articles suggest that there is ''high confidence'' that the current level of global warming (i.e., about 1°C, see Section 3.3.1) increased the heavy precipitation associated with the 2017 Hurricane Harvey by about 15% or more (Risser and Wehner, 2017; van Oldenborgh et al., 2017) <sup>[[#fn:r242|242]]</sup> . Hence, it can be inferred, under the assumption of linear dynamics, that further increases in heavy precipitation would occur under 1.5°C, 2°C and higher levels of global warming ( ''medium confidence'' ). Using a high resolution regional climate model explored the effects of different degrees of global warming on tropical cyclones over the southwest Indian Ocean, using transient simulations that downscaled a number of RCP8.5 GCM projections. Decreases in tropical cyclone frequencies are projected under both 1.5°C and 2°C of global warming. The decreases in cyclone frequencies under 2°C of global warming are somewhat larger than under 1.5°C, but no further decreases are projected under 3°C. This suggests that 2°C of warming, at least in these downscaling simulations, represents a type of stabilization level in terms of tropical cyclone formation over the southwest Indian Ocean and landfall over southern Africa (Muthige et al., 2018) <sup>[[#fn:r244|244]]</sup> . There is thus ''limited evidence'' that the global number of tropical cyclones will be lower under 2°C compared to 1.5°C of global warming, but with an increase in the number of very intense cyclones ( ''low confidence'' ).

The global response of the mid-latitude atmospheric circulation to 1.5°C and 2°C of warming was investigated using the HAPPI ensemble with a focus on the winter season (Li et al., 2018) <sup>[[#fn:r245|245]]</sup> . Under 1.5°C of global warming a weakening of storm activity over North America, an equatorward shift of the North Pacific jet exit and an equatorward intensification of the South Pacific jet are projected. Under an additional 0.5°C of warming a poleward shift of the North Atlantic jet exit and an intensification on the flanks of the Southern Hemisphere storm track are projected to become more pronounced. The weakening of the Mediterranean storm track that is projected under low mitigation emerges in the 2°C warmer world (Li et al., 2018) <sup>[[#fn:r246|246]]</sup> . AR5 assessed that under high greenhouse gas forcing (3°C or 4°C of global warming) there is ''low confidence'' in projections of poleward shifts of the Northern Hemisphere storm tracks, while there is ''high confidence'' that there would be a small poleward shift of the Southern Hemisphere storm tracks (Stocker et al., 2013) <sup>[[#fn:r247|247]]</sup> . In the context of this report, the assessment is that there is ''limited evidence'' and ''low confidence'' in whether any projected signal for higher levels of warming would be clearly manifested under 2°C of global warming.

<span id="ocean-circulation-and-temperature"></span>
=== 3.3.7 Ocean Circulation and Temperature ===

<div id="section-3-3-7-block-1"></div>

It is ''virtually certain'' that the temperature of the upper layers of the ocean (0–700 m in depth) has been increasing, and that the global mean for sea surface temperature (SST) has been changing at a rate just behind that of GMST. The surfaces of three ocean basins has warmed over the period 1950–2016 (by 0.11°C, 0.07°C and 0.05°C per decade for the Indian, Atlantic and Pacific Oceans, respectively; Hoegh-Guldberg et al., 2014 <sup>[[#fn:r248|248]]</sup> ), with the greatest changes occurring at the highest latitudes. Isotherms (i.e., lines of equal temperature) of sea surface temperature (SST) are shifting to higher latitudes at rates of up to 40 km per year (Burrows et al., 2014; García Molinos et al., 2015) <sup>[[#fn:r249|249]]</sup> . Long-term patterns of variability make detecting signals due to climate change complex, although the recent acceleration of changes to the temperature of the surface layers of the ocean has made the climate signal more distinct (Hoegh-Guldberg et al., 2014) <sup>[[#fn:r250|250]]</sup> . There is also evidence of significant increases in the frequency of marine heatwaves in the observational record (Oliver et al., 2018) <sup>[[#fn:r251|251]]</sup> , consistent with changes in mean ocean temperatures ( ''high confidence'' ). Increasing climate extremes in the ocean are associated with the general rise in global average surface temperature, as well as more intense patterns of climate variability (e.g., climate change intensification of ENSO) (Section 3.5.2.5). Increased heat in the upper layers of the ocean is also driving more intense storms and greater rates of inundation in some regions, which, together with sea level rise, are already driving significant impacts to sensitive coastal and low-lying areas (Section 3.3.6).

Increasing land–sea temperature gradients have the potential to strengthen upwelling systems associated with the eastern boundary currents (Benguela, Canary, Humboldt and Californian Currents; Bakun, 1990) <sup>[[#fn:r252|252]]</sup> . Observed trends support the conclusion that a general strengthening of longshore winds has occurred (Sydeman et al., 2014) <sup>[[#fn:r253|253]]</sup> , but the implications of trends detected in upwelling currents themselves are unclear (Lluch-Cota et al., 2014) <sup>[[#fn:r254|254]]</sup> . Projections of the scale of changes between 1°C and 1.5°C of global warming and between 1.5°C and 2°C are only informed by the changes during the past increase in GMST of 0.5°C ( ''low confidence'' ). However, evidence from GCM projections of future climate change indicates that a general strengthening of the Benguela, Canary and Humboldt upwelling systems under enhanced anthropogenic forcing (D. Wang et al., 2015) <sup>[[#fn:r255|255]]</sup> is projected to occur ( ''medium confidence'' ). This strengthening is projected to be stronger at higher latitudes. In fact, evidence from regional climate modelling is supportive of an increase in long-shore winds at higher latitudes, whereas long-shore winds may decrease at lower latitudes as a consequence of the poleward displacement of the subtropical highs under climate change (Christensen et al., 2007; Engelbrecht et al., 2009) <sup>[[#fn:r256|256]]</sup> .

''It is more likely than not'' that the Atlantic Meridional Overturning Circulation (AMOC) has been weakening in recent decades, given the detection of the cooling of surface waters in the North Atlantic and evidence that the Gulf Stream has slowed since the late 1950s (Rahmstorf et al., 2015b; Srokosz and Bryden, 2015; Caesar et al., 2018) <sup>[[#fn:r257|257]]</sup> . There is only ''limited evidence'' linking the current anomalously weak state of AMOC to anthropogenic warming (Caesar et al., 2018) <sup>[[#fn:r258|258]]</sup> . It is ''very likely'' that the AMOC will weaken over the 21st century. The best estimates and ranges for the reduction based on CMIP5 simulations are 11% (1– 24%) in RCP2.6 and 34% (12– 54%) in RCP8.5 (AR5). There is no evidence indicating significantly different amplitudes of AMOC weakening for 1.5°C versus 2°C of global warming.

<span id="sea-ice"></span>
=== 3.3.8 Sea Ice ===

<div id="section-3-3-8-block-1"></div>

Summer sea ice in the Arctic has been retreating rapidly in recent decades. During the period 1997 to 2014, for example, the monthly mean sea ice extent during September (summer) decreased on average by 130,000 km² per year (Serreze and Stroeve, 2015) <sup>[[#fn:r259|259]]</sup> . This is about four times as fast as the September sea ice loss during the period 1979 to 1996. Sea ice thickness has also decreased substantially, with an estimated decrease in ice thickness of more than 50% in the central Arctic (Lindsay and Schweiger, 2015) <sup>[[#fn:r260|260]]</sup> . Sea ice coverage and thickness also decrease in CMIP5 simulations of the recent past, and are projected to decrease in the future (Collins et al., 2013) <sup>[[#fn:r261|261]]</sup> . However, the modelled sea ice loss in most CMIP5 models is much smaller than observed losses. Compared to observations, the simulations are less sensitive to both global mean temperature rise (Rosenblum and Eisenman, 2017) <sup>[[#fn:r262|262]]</sup> and anthropogenic CO <sub>2</sub> emissions (Notz and Stroeve, 2016) <sup>[[#fn:r263|263]]</sup> . This mismatch between the observed and modelled sensitivity of Arctic sea ice implies that the multi-model-mean responses of future sea ice evolution probably underestimates the sea ice loss for a given amount of global warming. To address this issue, studies estimating the future evolution of Arctic sea ice tend to bias correct the model simulations based on the observed evolution of Arctic sea ice in response to global warming. Based on such bias correction, pre-AR5 and post-AR5 studies generally agree that for 1.5°C of global warming relative to pre-industrial levels, the Arctic Ocean will maintain a sea ice cover throughout summer in most years (Collins et al., 2013; Notz and Stroeve, 2016; Screen and Williamson, 2017; Jahn, 2018; Niederdrenk and Notz, 2018; Sigmond et al., 2018) <sup>[[#fn:r264|264]]</sup> . For 2°C of global warming, chances of a sea ice-free Arctic during summer are substantially higher (Screen and Williamson, 2017; Jahn, 2018; Niederdrenk and Notz, 2018; Screen et al., 2018; Sigmond et al., 2018) <sup>[[#fn:r265|265]]</sup> . Model simulations suggest that there will be at least one sea ice-free Arctic <sup>[[#fn:8|8]]</sup> summer after approximately 10 years of stabilized warming at 2°C, as compared to one sea ice-free summer after 100 years of stabilized warming at 1.5°C above pre-industrial temperatures (Jahn, 2018; Screen et al., 2018; Sigmond et al., 2018) <sup>[[#fn:r266|266]]</sup> . For a specific given year under stabilized warming of 2°C, studies based on large ensembles of simulations with a single model estimate the likelihood of ice-free conditions as 35% without a bias correction of the underlying model (Sanderson et al., 2017; Jahn, 2018) <sup>[[#fn:r267|267]]</sup> ; as between 10% and &gt;99% depending on the observational record used to correct the sensitivity of sea ice decline to global warming in the underlying model (Niederdrenk and Notz, 2018) <sup>[[#fn:r268|268]]</sup> ; and as 19% based on a procedure to correct for biases in the climatological sea ice coverage in the underlying model (Sigmond et al., 2018) <sup>[[#fn:r269|269]]</sup> . The uncertainty of the first year of the occurrence of an ice-free Arctic Ocean arising from internal variability is estimated to be about 20 years (Notz, 2015; Jahn et al., 2016) <sup>[[#fn:r270|270]]</sup> .

The more recent estimates of the warming necessary to produce an ice-free Arctic Ocean during summer are lower than the ones given in AR5 (about 2.6°C–3.1°C of global warming relative to pre-industrial levels or 1.6°C–2.1°C relative to present–day conditions), which were similar to the estimate of 3°C of global warming relative to pre-industrial levels (or 2°C relative to present-day conditions) by Mahlstein and Knutti (2012) <sup>[[#fn:r271|271]]</sup> based on bias-corrected CMIP3 models. Rosenblum and Eisenman (2016) <sup>[[#fn:r272|272]]</sup> explained why the sensitivity estimated by Mahlstein and Knutti (2012) <sup>[[#fn:r273|273]]</sup> might be too low, estimating instead that September sea ice in the Arctic would disappear at 2°C of global warming relative to pre-industrial levels (or about 1°C relative to present-day conditions), in line with the other recent estimates. Notz and Stroeve (2016) <sup>[[#fn:r274|274]]</sup> used the observed correlation between September sea ice extent and cumulative CO <sub>2</sub> emissions to estimate that the Arctic Ocean would become nearly free of sea ice during September with a further 1000 Gt of emissions, which also implies a sea ice loss at about 2°C of global warming. Some of the uncertainty in these numbers stems from the possible impact of aerosols (Gagne et al., 2017) <sup>[[#fn:r275|275]]</sup> and of volcanic forcing (Rosenblum and Eisenman, 2016) <sup>[[#fn:r276|276]]</sup> . During winter, little Arctic sea ice is projected to be lost for either 1.5°C or 2°C of global warming (Niederdrenk and Notz, 2018) <sup>[[#fn:r277|277]]</sup> .

A substantial number of pre-AR5 studies found that there is no indication of hysteresis behaviour of Arctic sea ice under decreasing temperatures following a possible overshoot of a long-term temperature target (Holland et al., 2006; Schröder and Connolley, 2007; Armour et al., 2011; Sedláček et al., 2011; Tietsche et al., 2011; Boucher et al., 2012; Ridley et al., 2012) <sup>[[#fn:r278|278]]</sup> . In particular, the relationship between Arctic sea ice coverage and GMST was found to be indistinguishable between a warming scenario and a cooling scenario. These results have been confirmed by post-AR5 studies (Li et al., 2013; Jahn, 2018) <sup>[[#fn:r279|279]]</sup> , which implies ''high confidence'' that an intermediate temperature overshoot has no long-term consequences for Arctic sea ice coverage.

In the Antarctic, sea ice shows regionally contrasting trends, such as a strong decrease in sea ice coverage near the Antarctic peninsula but increased sea ice coverage in the Amundsen Sea (Hobbs et al., 2016) <sup>[[#fn:r280|280]]</sup> . Averaged over these contrasting regional trends, there has been a slow long-term increase in overall sea ice coverage in the Southern Ocean, although with comparably low ice coverage from September 2016 onwards. Collins et al. (2013) <sup>[[#fn:r281|281]]</sup> assessed ''low confidence'' in Antarctic sea ice projections because of the wide range of model projections and an inability of almost all models to reproduce observations such as the seasonal cycle, interannual variability and the long-term slow increase. No existing studies have robustly assessed the possible future evolution of Antarctic sea ice under low-warming scenarios.

In summary, the probability of a sea-ice-free Arctic Ocean during summer is substantially higher at 2°C compared to 1.5°C of global warming relative to pre-industrial levels, and there is ''medium confidence'' that there will be at least one sea ice-free Arctic summer after about 10 years of stabilized warming at 2°C, while about 100 years are required at 1.5°C. There is ''high confidence'' that an intermediate temperature overshoot has no long-term consequences for Arctic sea ice coverage with regrowth on decadal time scales.

<span id="sea-level"></span>
=== 3.3.9 Sea Level ===

<div id="section-3-3-9-block-1"></div>

Sea level varies over a wide range of temporal and spatial scales, which can be divided into three broad categories. These are global mean sea level (GMSL), regional variation about this mean, and the occurrence of sea-level extremes associated with storm surges and tides. GMSL has been rising since the late 19th century from the low rates of change that characterized the previous two millennia (Church et al., 2013) <sup>[[#fn:r282|282]]</sup> . Slowing in the reported rate over the last two decades (Cazenave et al., 2014) <sup>[[#fn:r283|283]]</sup> may be attributable to instrumental drift in the observing satellite system (Watson et al., 2015) <sup>[[#fn:r284|284]]</sup> and increased volcanic activity (Fasullo et al., 2016) <sup>[[#fn:r285|285]]</sup> . Accounting for the former results in rates (1993 to mid-2014) between 2.6 and 2.9 mm yr <sup>–1</sup> (Watson et al., 2015) <sup>[[#fn:r286|286]]</sup> . The relative contributions from thermal expansion, glacier and ice-sheet mass loss, and freshwater storage on land are relatively well understood (Church et al., 2013; Watson et al., 2015) <sup>[[#fn:r287|287]]</sup> and their attribution is dominated by anthropogenic forcing since 1970 (15 ± 55% before 1950, 69 ± 31% after 1970) (Slangen et al., 2016) <sup>[[#fn:r288|288]]</sup> .

There has been a significant advance in the literature since AR5, which has included the development of semi-empirical models (SEMs) into a broader emulation-based approach (Kopp et al., 2014; Mengel et al., 2016; Nauels et al., 2017) <sup>[[#fn:r289|289]]</sup> that is partially based on the results from more detailed, process-based modelling Church et al. (2013) <sup>[[#fn:r290|290]]</sup> assigned ''low confidence'' to SEMs because these models assume that the relation between climate forcing and GMSL is the same in the past (calibration) and future (projection). Probable future changes in the relative contributions of thermal expansion, glaciers and (in particular) ice sheets invalidate this assumption. However, recent emulation-based studies overcame this shortcoming by considering individual GMSL contributors separately, and they are therefore employed in this assessment. In this subsection, the process-based literature of individual contributors to GMSL is considered for scenarios close to 1.5°C and 2°C of global warming before emulation-based approaches are assessed.

A limited number of processes-based studies are relevant to GMSL in 1.5°C and 2°C worlds. Marzeion et al. (2018) <sup>[[#fn:r291|291]]</sup> used a global glacier model with temperature-scaled scenarios based on RCP2.6 to investigate the difference between 1.5°C and 2°C of global warming and found little difference between scenarios in the glacier contribution to GMSL for the year 2100 (54–97 mm relative to present-day levels for 1.5°C and 63–112 mm for 2°C, using a 90% confidence interval). This arises because glacier melt during the remainder of the century is dominated by the response to warming from pre-industrial to present-day levels, which is in turn a reflection of the slow response times of glaciers. Fürst et al. (2015) <sup>[[#fn:r292|292]]</sup> made projections of the Greenland ice sheet’s contribution to GMSL using an ice-flow model forced by the regional climate model Modèle Atmosphérique Régional (MAR; considered by Church et al. (2013) <sup>[[#fn:r293|293]]</sup> to be the ‘most realistic’ such model). They projected an RCP2.6 range of 24–60 mm (1 standard deviation) by the end of the century (relative to the year 2000 and consistent with the assessment of Church et al. (2013) <sup>[[#fn:r294|294]]</sup> ; however, their projections do not allow the difference between 1.5°C and 2°C worlds to be evaluated.

The Antarctic ice sheet can contribute both positively, through increases in outflow (solid ice lost directly to the ocean), and negatively, through increases in snowfall (owing to the increased moisture-bearing capacity of a warmer atmosphere), to future GMSL rise. Frieler et al. (2015) <sup>[[#fn:r295|295]]</sup> suggested a range of 3.5–8.7% °C <sup>–1</sup> for this effect, which is consistent with AR5. Observations from the Amundsen Sea sector of Antarctica suggest an increase in outflow (Mouginot et al., 2014) <sup>[[#fn:r296|296]]</sup> over recent decades associated with grounding line retreat (Rignot et al., 2014) <sup>[[#fn:r297|297]]</sup> and the influx of relatively warm Circumpolar Deepwater (Jacobs et al., 2011) <sup>[[#fn:r298|298]]</sup> . Literature on the attribution of these changes to anthropogenic forcing is still in its infancy (Goddard et al., 2017; Turner et al., 2017a) <sup>[[#fn:r299|299]]</sup> . RCP2.6-based projections of Antarctic outflow (Levermann et al., 2014; Golledge et al., 2015; DeConto and Pollard, 2016 <sup>[[#fn:r300|300]]</sup> , who include snowfall changes) are consistent with the AR5 assessment of Church et al. (2013) <sup>[[#fn:r301|301]]</sup> for end-of-century GMSL for RCP2.6, and do not support substantial additional GMSL rise by Marine Ice Sheet Instability or associated instabilities (see Section 3.6). While agreement is relatively good, concerns about the numerical fidelity of these models still exist, and this may affect the quality of their projections (Drouet et al., 2013; Durand and Pattyn, 2015) <sup>[[#fn:r302|302]]</sup> . An assessment of Antarctic contributions beyond the end of the century, in particular related to the Marine Ice Sheet Instability, can be found in Section 3.6.

While some literature on process-based projections of GMSL for the period up to 2100 is available, it is insufficient for distinguishing between emissions scenarios associated with 1.5°C and 2°C warmer worlds. This literature is, however, consistent with the assessment by Church et al. (2013) <sup>[[#fn:r303|303]]</sup> of a ''likely'' range of 0.28–0.61 m in 2100 (relative to 1986–2005), suggesting that the AR5 assessment is still appropriate. Recent emulation-based studies show convergence towards this AR5 assessment (Table 3.1) and offer the advantage of allowing a comparison between 1.5°C and 2°C warmer worlds. Table 3.1 features a compilation of recent emulation-based and SEM studies.

<div id="section-3-3-9-block-2"></div>

<span id="table-3.1"></span>
<!-- START TABLE -->
'''Table 3.1:'''

<span id="compilation-of-recent-projections-for-sea-level-at-2100-in-cm-for-representative-concentration-pathway-rcp2.6-and-1.5c-and-2c-scenarios.-upper-and-lower-limits-are-shown-for-the-17-84-and-5-95-confidence-intervals-quoted-in-the-original-papers."></span>
'''Compilation of recent projections for sea level at 2100 (in cm) for Representative Concentration Pathway (RCP)2.6, and 1.5°C and 2°C scenarios. Upper and lower limits are shown for the 17-84% and 5-95% confidence intervals quoted in the original papers.'''

<!-- TABLE -->
{| class="wikitable"
|-
! rowspan="2"| Study
! rowspan="2"| Baseline
! colspan="2"| RCP2.6
! colspan="2"| 1.5°C
! colspan="2"| 2°C
|-
! 67%
! 90%
! 67%
! 90%
! 67%
! 90%
|-
| AR5
| 1986–2005
| 28–61
|
|-
| Kopp et al. (2014)  <sup>[[#fn:r304|304]]</sup>
| 2000
| 37–65
| 29–82
|
|-
| Jevrejeva et al. (2016) <sup>[[#fn:r305|305]]</sup>
| 1986–2005
|
| 29–58
|
|-
| Kopp et al. (2016) <sup>[[#fn:r306|306]]</sup>
| 2000
| 28–51
| 24–61
|
|-
| Mengel et al. (2016) <sup>[[#fn:r307|307]]</sup>
| 1986–2005
| 28–56
|
|-
| Nauels et al. (2017) <sup>[[#fn:r308|308]]</sup>
| 1986–2005
| 35–56
|
|-
| Goodwin et al. (2017) <sup>[[#fn:r309|309]]</sup>
| 1986–2005
|
| 31–59<br />
45–70<br />
45–72
|
|-
| Schaeffer et al. (2012) <sup>[[#fn:r310|310]]</sup>
| 2000
|
| 52–96
|
| 54–99
|
| 56–105
|-
| Schleussner et al. (2016b) <sup>[[#fn:r311|311]]</sup>
| 2000
|
| 26–53
|
| 36–65
|
|-
| Bittermann et al. (2017) <sup>[[#fn:r312|312]]</sup>
| 2000
|
| 29–46
|
| 39–61
|-
| Jackson et al. (2018) <sup>[[#fn:r313|313]]</sup>
| 1986–2005
|
| 30–58<br />
40–77
| 20–67<br />
28–93
| 35–64<br />
47–93
| 24–74<br />
32–117
|-
| Sanderson et al. (2017) <sup>[[#fn:r314|314]]</sup>
|
| 50–80
|
| 60–90
|-
| Nicholls et al. (2018) <sup>[[#fn:r315|315]]</sup>
| 1986–2005
|
| 24–54
|
| 31–65
|-
| Rasmussen et al. (2018) <sup>[[#fn:r316|316]]</sup>
| 2000
|
| 35–64
| 28–82
| 39–76
| 28–96
|-
| Goodwin et al. (2018) <sup>[[#fn:r317|317]]</sup>
| 1986–2005
|
| 26–62
|
| 30–69
|}
<!-- END TABLE -->

<div id="section-3-3-9-block-3"></div>

There is little consensus between the reported ranges of GMSL rise (Table 3.1). Projections vary in the range 0.26–0.77 m and 0.35–0.93 m for 1.5°C and 2°C respectively for the 17–84% confidence interval (0.20–0.99 m and 0.24–1.17 m for the 5–95% confidence interval). There is, however, ''medium agreement'' that GMSL in 2100 would be 0.04–0.16 m higher in a 2°C warmer world compared to a 1.5°C warmer world based on the 17–84% confidence interval (0.00–0.24 m based on 5–95% confidence interval) with a value of around 0.1m. There is ''medium confidence'' in this assessment because of issues associated with projections of the Antarctic contribution to GMSL that are employed in emulation-based studies (see above) and the issues previously identified with SEMs (Church et al., 2013) <sup>[[#fn:r318|318]]</sup> .

Translating projections of GMSL to the scale of coastlines and islands requires two further steps. The first step accounts for regional changes associated with changing water and ice loads (such as Earth’s gravitational field and rotation, and vertical land movement), as well as spatial differences in ocean heat uptake and circulation. The second step maps regional sea level to changes in the return periods of particular flood events to account for effects not included in global climate models, such as tides, storm surges, and wave setup and runup. Kopp et al. (2014) <sup>[[#fn:r319|319]]</sup> presented a framework to do this and gave an example application for nine sites located in the US, Japan, northern Europe and Chile. Of these sites, seven (all except those in northern Europe) were found to experience at least a quadrupling in the number of years in the 21st century with 1-in-100-year floods under RCP2.6 compared to under no future sea level rise. Rasmussen et al. (2018) <sup>[[#fn:r320|320]]</sup> used this approach to investigate the difference between 1.5°C and 2°C warmer worlds up to 2200. They found that the reduction in the frequency of 1-in-100-year floods in a 1.5°C compared to a 2°C warmer world would be greatest in the eastern USA and Europe, with ESL event frequency amplification being reduced by about a half and with smaller reductions for small island developing states (SIDS). This last result contrasts with the finding of Vitousek et al. (2017) <sup>[[#fn:r321|321]]</sup> that regions with low variability in extreme water levels (such as SIDS in the tropics) are particularly sensitive to GMSL rise, such that a doubling of frequency may be expected for even small (0.1–0.2 m) rises. Schleussner et al. (2011) <sup>[[#fn:r322|322]]</sup> emulated the AMOC based on a subset of CMIP-class climate models. When forced using global temperatures appropriate for the CP3-PD scenario (1°C of warming in 2100 relative to 2000 or about 2°C of warming relative to pre-industrial) the emulation suggests an 11% median reduction in AMOC strength at 2100 (relative to 2000) with an associated 0.04 m dynamic sea level rise along the New York City coastline.

In summary, there is ''medium confidence'' that GMSL rise will be about 0.1 m (within a 0.00–0.20 m range based on 17–84% confidence-interval projections) less by the end of the 21st century in a 1.5°C compared to a 2°C warmer world. Projections for 1.5°C and 2°C global warming cover the ranges 0.2–0.8 m and 0.3–1.00 m relative to 1986–2005, respectively ( ''medium confidence'' ). Sea level rise beyond 2100 is discussed in Section 3.6; however, recent literature strongly supports the assessment by Church et al. (2013) <sup>[[#fn:r323|323]]</sup> that sea level rise will continue well beyond 2100 ( ''high confidence'' ).

<div id="section-3-3-9-block-4" class="box"></div>

<span id="box-3.3-lessons-from-past-warm-climate-episodes"></span>