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

=== 4.2.2 Observed Changes in Sea Level (Past and Present) ===

<div id="section-4-2-2observed-changes-in-sea-level-past-and-present-block-1"></div>

Sea level changes in the distant geologic past provide information on the size of the ice sheets in climate states different from today. Past intervals with temperatures comparable to or warmer than today are of particular interest, and since AR5 (Masson-Delmotte et al., 2013 <sup>[[#fn:r46|46]]</sup> ) they have been increasingly used to test and calibrate process-based ice sheet models used in future projections (DeConto and Pollard, 2016 <sup>[[#fn:r47|47]]</sup> ; Edwards et al., 2019 <sup>[[#fn:r48|48]]</sup> ; Golledge et al., 2019 <sup>[[#fn:r49|49]]</sup> ) . These intervals include the mPWP around 3.3–3.0 Ma, when atmospheric CO <sub>2</sub> concentrations were similar to today (~300–450 ppmv; Badger et al., 2013 <sup>[[#fn:r50|50]]</sup> ; Martínez-Botí et al., 2015 <sup>[[#fn:r51|51]]</sup> ; Stap et al., 2016 <sup>[[#fn:r52|52]]</sup> ) and global mean temperature was 2ºC–4ºC warmer than pre-industrial (Dutton et al., 2015a <sup>[[#fn:r53|53]]</sup> ; Haywood et al., 2016 <sup>[[#fn:r54|54]]</sup> ) and the LIG around 129–116 ka, when global mean temperature was 0.5ºC–1.0ºC warmer (Capron et al., 2014 <sup>[[#fn:r55|55]]</sup> ; Dutton et al., 2015a <sup>[[#fn:r56|56]]</sup> ; Fischer et al., 2018 <sup>[[#fn:r57|57]]</sup> ) and sea surface temperatures were similar to today (Hoffman et al., 2017 <sup>[[#fn:r58|58]]</sup> ) . Updated reconstructions of GMSL (Dutton et al., 2015a <sup>[[#fn:r59|59]]</sup> ) based on ancient shoreline elevations corrected to account for geodynamic processes (4.2.1.5), and geochemical records extracted from marine sediment cores, indicate sea levels were &gt;5 m higher than today during these past warm periods ( ''medium confidence'' ).

Most estimates of peak GMSL during the mPWP range between 6 and 30 m higher than today (Miller et al., 2012 <sup>[[#fn:r60|60]]</sup> ; Rovere et al., 2014 <sup>[[#fn:r61|61]]</sup> ; Dutton et al., 2015a <sup>[[#fn:r62|62]]</sup> ) but with deep uncertainty (Cross-Chapter Box 5 in Chapter 1) and few constraints on the high end of the range. The large uncertainty is contributed by uncertain GIA corrections applied to palaeo shoreline indicators (Raymo et al., 2011 <sup>[[#fn:r63|63]]</sup> ; Rovere et al., 2014 <sup>[[#fn:r64|64]]</sup> ) , dynamic topography, the vertical land surface motion associated with Earth’s mantle flow (Rowley et al., 2013 <sup>[[#fn:r65|65]]</sup> ) , and possible biases in geochemical records of ice volume derived from marine sediments (Raymo et al., 2018 <sup>[[#fn:r66|66]]</sup> ) . Estimates of GMSL &gt;10 m higher than today require a meltwater contribution from the East Antarctic Ice Sheet in addition to the GIS and West Antarctic Ice Sheets (WAIS; Miller et al., 2012 <sup>[[#fn:r67|67]]</sup> ; Dutton et al., 2015a <sup>[[#fn:r68|68]]</sup> ) . Pliocene modelling studies appearing since AR5 (Masson-Delmotte et al., 2013 <sup>[[#fn:r69|69]]</sup> ) demonstrate the potential for substantial retreat of East Antarctic ice into deep submarine basins (Austermann and Mitrovica, 2015 <sup>[[#fn:r70|70]]</sup> ; Pollard et al., 2015 <sup>[[#fn:r71|71]]</sup> ; Aitken et al., 2016 <sup>[[#fn:r72|72]]</sup> ; DeConto and Pollard, 2016 <sup>[[#fn:r73|73]]</sup> ; Gasson et al., 2016 <sup>[[#fn:r74|74]]</sup> ; Golledge et al., 2019 <sup>[[#fn:r75|75]]</sup> ) , as does emerging geological evidence from marine sediment cores recovered from the East Antarctic margin (Cook et al., 2013 <sup>[[#fn:r76|76]]</sup> ; Patterson et al., 2014 <sup>[[#fn:r77|77]]</sup> ; Bertram et al., 2018 <sup>[[#fn:r78|78]]</sup> ) . However, the range of maximum Pliocene GMSL contributions from Antarctic modelling (Austermann and Mitrovica, 2015 <sup>[[#fn:r79|79]]</sup> ; Pollard et al., 2015 <sup>[[#fn:r80|80]]</sup> ; Yamane et al., 2015 <sup>[[#fn:r81|81]]</sup> ; DeConto and Pollard, 2016 <sup>[[#fn:r82|82]]</sup> ; Gasson et al., 2016 <sup>[[#fn:r83|83]]</sup> ) remains large (5.4–17.8 m), providing little additional constraint on the geological estimates. Land surface exposure measurements on sediment sourced from East Antarctica (Shakun et al., 2018 <sup>[[#fn:r84|84]]</sup> ) suggests Pliocene ice loss was limited to marine-based ice, where the bedrock is below sea level and possibly prone to marine ice sheet instabilities (Cross-Chapter box 8 in Chapter 3). The total potential contribution to GMSL rise from marine-based ice in Antarctica is ~22.5 m (Fretwell et al., 2013 <sup>[[#fn:r85|85]]</sup> ) . Combined with the complete loss of the GIS, this could conceivably produce ~30 m of GMSL rise. However, this would require maximum retreat of GIS and AIS to be synchronous, which is not probable due to orbitally paced, inter-hemispheric asymmetries in Greenland and Antarctic climate (de Boer et al., 2017) . As such, 25 m is found to be a reasonable upper bound on GMSL during the mPWP, but with ''low confidence'' .

An updated estimate of maximum GMSL during the more geologically recent LIG ranges between 6–9 m higher than today (Dutton et al., 2015a <sup>[[#fn:r86|86]]</sup> ) . This is close to the values reported by a probabilistic analysis of globally distributed sea level indicators (Kopp et al., 2009 <sup>[[#fn:r87|87]]</sup> ) , but slightly higher than AR5’s central estimate of 6 m. Like the mid-Pliocene, the LIG estimates also suffer from uncertainties in GIA corrections and dynamic topography. Düsterhus et al. (2016) <sup>[[#fn:r88|88]]</sup> applied data assimilation techniques including GIA corrections to the same LIG dataset used by Kopp et al. (2009) <sup>[[#fn:r89|89]]</sup> and found good agreement (7.5 ± 1.1 m likely range) with Kopp et al. (2009)  and Dutton et al. (2015a) , but the upper range remains poorly constrained. Their estimates of peak LIG sea level are sensitive to the assumed ice history before and after the LIG, consistent with the results of other studies (Lambeck et al., 2012 <sup>[[#fn:r90|90]]</sup> ; Dendy et al., 2017 <sup>[[#fn:r91|91]]</sup> ) . Austermann et al. (2017) compared a compilation of LIG shoreline indicators with dynamic topography simulations. They found that vertical surface motions driven by mantle convection can produce several metres of uncertainty in LIG sea level estimates, but their mean and most probable estimates of 6.7 m and 6.4 m are broadly in line with other estimates.

Greenland and Antarctic climate change on these time scales is influenced by inter-hemispheric differences in polar amplification (Stap et al., 2018 <sup>[[#fn:r93|93]]</sup> ) , changes in Earth’s orbit, and long-term climate system processes. This complicates relationships between global mean temperature and ice sheet response. On Greenland, the magnitude of LIG summer warming and changes in ice sheet volume continue to be contested. Extreme summer warming of 6ºC or more, reconstructed from ice cores (Dahl-Jensen et al., 2013 <sup>[[#fn:r94|94]]</sup> ; Landais et al., 2016 <sup>[[#fn:r95|95]]</sup> ; Yau et al., 2016 <sup>[[#fn:r96|96]]</sup> ) and lake archives (McFarlin et al., 2018 <sup>[[#fn:r97|97]]</sup> ) is in apparent conflict with a persistent, spatially extensive GIS reconstructed from ice cores and radar imaging (Dahl-Jensen et al., 2013 <sup>[[#fn:r98|98]]</sup> ) . Maximum retreat of the GIS during the LIG varies widely among modelling studies, ranging from ~1 m to ~6 m (Helsen et al., 2013 <sup>[[#fn:r99|99]]</sup> ; Quiquet et al., 2013 <sup>[[#fn:r100|100]]</sup> ; Dutton et al., 2015a <sup>[[#fn:r101|101]]</sup> ; Goelzer et al., 2016 <sup>[[#fn:r102|102]]</sup> ; Yau et al., 2016 <sup>[[#fn:r103|103]]</sup> ) ; however, the models consistently indicate a small Greenland contribution to GMSL early in the interglacial, implying Antarctica was the dominant contributor to the early interglacial highstand of 6 ± 1.5 m, beginning around 129 ka (Dutton et al., 2015b <sup>[[#fn:r104|104]]</sup> ) . An early LIG loss of Antarctic ice is consistent with recent ice sheet modelling (DeConto and Pollard, 2016 <sup>[[#fn:r105|105]]</sup> ; Goelzer et al., 2016 <sup>[[#fn:r106|106]]</sup> ) . Due to its bedrock configuration and susceptibility to marine ice sheet instabilities (Cross-Chapter Box 8 in Chapter 3), the WAIS would have been especially vulnerable to subsurface ocean warming during the LIG (Sutter et al., 2016 <sup>[[#fn:r107|107]]</sup> ) . However, most evidence of WAIS retreat during the LIG remains indirect (Steig et al., 2015 <sup>[[#fn:r108|108]]</sup> ) and firm geological evidence has yet to be uncovered. A recent analysis of East Antarctic sediments provides evidence of some ice retreat in Wilkes subglacial basin during the LIG (Wilson and Forsyth, 2018 <sup>[[#fn:r109|109]]</sup> ) , but the volume of ice loss is not quantified.

GMSL during the LIG was at times higher than today ( ''virtually certain'' ), with a ''likely range'' between 6–9 m, and not expected to be more than 10 m ( ''medium confidence'' ). Due to ongoing uncertainties in the evolution of atmospheric and oceanic warming over and around the ice sheets, and low confidence in the relative contributions of Antarctic versus Greenland meltwater to GMSL change, the LIG is not used here to directly assess the sensitivity of the ice sheets under current or future climate conditions. There is ''low confidence'' in the utility of changes in either mPWP or LIG sea level changes to quantitatively inform near-term future rates of GMSL rise.

An expanded summary of recent advances and ongoing difficulties in reconstructing these time periods in terms of climate, sea level, and implications for the future evolution of ice sheets and sea level is provided in SM4.1.

<div id="section-4-2-2observed-changes-in-sea-level-past-and-present-block-2"></div>

<span id="figure-4.4"></span>
<!-- START IMG -->
<!-- IMG TITLE -->
'''Figure 4.4'''

<span id="figure-4.4-a-schematic-illustration-of-the-climate-and-non-climate-driven-processes-that-can-influence-global-regional-green-colours-relative-and-extreme-sea-level-esl-events-red-colours-along-coasts.-major-ice-processes-are-shown-in-purple-and-general-terms-in-black.-sle-stands-for-sea-level-equivalent-and-reflects-the-increase-in-gmsl"></span>
<!-- IMG CAPTION -->
'''Figure 4.4 | A schematic illustration of the climate and non-climate driven processes that can influence global, regional (green colours), relative and extreme sea level (ESL) events (red colours) along coasts. Major ice processes are shown in purple and general terms in black. SLE stands for Sea Level Equivalent and reflects the increase in GMSL […]'''
<!-- IMG FILE -->
[[File:293bed953e086b90d833ac1cad5362b3 IPCC-SROCC-CH_4_4-3000x1491.jpg]]

Figure 4.4 | A schematic illustration of the climate and non-climate driven processes that can influence global, regional (green colours), relative and extreme sea level (ESL) events (red colours) along coasts. Major ice processes are shown in purple and general terms in black. SLE stands for Sea Level Equivalent and reflects the increase in GMSL if the mentioned ice mass is melted completely and added to the ocean.

<!-- END IMG -->
<div id="section-4-2-2-1global-mean-sea-level-changes-during-the-instrumental-period"></div>

<span id="global-mean-sea-level-changes-during-the-instrumental-period"></span>
==== 4.2.2.1 Global Mean Sea Level Changes During the Instrumental Period ====

<div id="section-4-2-2-1global-mean-sea-level-changes-during-the-instrumental-period-block-1"></div>

Observational estimates of the sea level variations over past millennia rely essentially on proxy-based regional relative sea level reconstructions corrected for GIA. Since AR5, the increasing availability of regional proxy-based reconstructions enables the estimation of GMSL change over the last ∼ 3 kyr. The first statistical integration of the available reconstructions shows that the GMSL experienced variations of ±9 [±7 to ±11] cm (5–95% uncertainty range; Kopp et al., 2016 <sup>[[#fn:r110|110]]</sup> ) over the 2400 years preceding the 20th century ( ''medium confidence'' ). This is more tightly bound than the AR5 assessment which indicated a variability in GMSL that was &lt;±25 cm over the same period. This progress since AR5 confirms that it is ''virtually certain'' that the mean rate of GMSL has increased during the last two centuries from relatively low rates of change during the late Holocene (order tenths of mm yr <sup>–1</sup> ) to modern rates (order mm yr <sup>–1</sup> ; Woodruff et al., 2013 <sup>[[#fn:r111|111]]</sup> ) .

Over the last two centuries, sea level observations have mostly relied on tide gauge measurements. These records, beginning around 1700 in some locations (Holgate et al., 2012 <sup>[[#fn:r112|112]]</sup> ; PSMSL, 2019 <sup>[[#fn:r113|113]]</sup> ) , provide insight into historic sea level trends. Since 1992, the emergence of precise satellite altimetry has advanced our knowledge on GMSL and regional sea level changes considerably through a combination of near global ocean coverage and high spatial resolution. It has also enabled more detailed monitoring of land ice loss. Since 2002, high precision gravity measurements provided by the Gravity Recovery and Climate Experiment (GRACE) and GRACE Follow-On missions show the loss of land ice in Greenland and Antarctica, and confirm independent assessments of ice sheet mass changes based on satellite altimetry (Shepherd et al., 2012 <sup>[[#fn:r114|114]]</sup> ; The Imbie team, 2018) and InSAR measurements combined with ice sheet SMB estimates (Noël et al., 2018 <sup>[[#fn:r115|115]]</sup> ; Rignot et al., 2019 <sup>[[#fn:r116|116]]</sup> ) . Since 2006, when the array of Argo profiling floats reached near-global coverage, it has been possible to get an accurate estimate of the ocean thermal expansion (down to 2000 m depth) and test the closure of the sea level budget. T he combined analysis of the different observing systems that are available has improved significantly the understanding of the magnitude and relative contributions of the different processes causing sea level change. In particular, important progress has been achieved since AR5 on estimating and understanding the increasing contribution of the ice sheets to SLR.

<div id="section-4-2-2-1global-mean-sea-level-changes-during-the-instrumental-period-block-2"></div>

<span id="tide-gauge-records"></span>
===== 4.2.2.1.1 Tide gauge records =====

The number of tide gauges has increased over time from only a few in northern Europe in the 18th century to more than 2000 today along the world’s coastlines. Because of their location and limited number, tide gauges sample the ocean sparsely and non-uniformly with a bias towards the Northern Hemisphere. Most tide gauge records are short and have significant gaps. In addition, tide gauges are anchored on land and are affected by the vertical motion of Earth’s crust caused by both natural processes (e.g., GIA, tectonics and sediment compaction; Wöppelmann and Marcos, 2016 <sup>[[#fn:r117|117]]</sup> ; Pfeffer et al., 2017 <sup>[[#fn:r118|118]]</sup> ) and anthropogenic activities (e.g., groundwater depletion, dam building or settling of landfill in urban areas; Raucoules et al., 2013 <sup>[[#fn:r119|119]]</sup> ; Pfeffer et al., 2017 <sup>[[#fn:r120|120]]</sup> ) . When estimating the GMSL due to the ocean thermal expansion and land ice melt, tide gauges must be corrected for this VLM, where VLM = GIA + anthropogenic subsidence + (tectonics, natural subsidence). This is possible with stations of the Global Positioning System (GPS) network when they are co-located with tide gauges (Santamaría-Gómez et al., 2017 <sup>[[#fn:r121|121]]</sup> ; Kleinherenbrink et al., 2018 <sup>[[#fn:r122|122]]</sup> ) . However, this approach provides information on the VLM over the past two to three decades and has limited value over longer time scales for places where the VLM has varied significantly through the last century (Riva et al., 2017 <sup>[[#fn:r123|123]]</sup> ) .

AR5 assessed the different strategies to estimate the 20th century GMSL changes. These strategies only accounted for the inhomogeneous space and time coverage of tide gauge data and for the VLM induced by GIA (Figure 4.5). Since AR5 two new approaches have been developed. The first one uses a Kalman smoother which combines tide gauge records with the spatial patterns associated with ocean dynamic change, change in land ice and GIA. It enables accounting for the inhomogeneous distribution of tide gauges and the VLM associated with both GIA and current land ice loss (Hay et al., 2015 <sup>[[#fn:r124|124]]</sup> ; Figure 4.5). The second approach uses ad hoc corrections to tide gauge records with an additional spatial pattern associated with changes in terrestrial water storage to account for the inhomogeneous distribution in tide gauges. It also accounts for the total VLM (Dangendorf et al., 2017 <sup>[[#fn:r125|125]]</sup> ; Figure 4.5). Both methods yield significantly lower GMSL changes over the period 1950–1970 than previous estimates, leading to long-term trends since 1900 that are smaller than previous estimates by 0.4 mm yr <sup>–1</sup> (Figure 4.5). Different arguments including biases in the tide gauge datasets (Hamlington and Thompson, 2015 <sup>[[#fn:r126|126]]</sup> ) , biases in the averaging technique and absence of VLM correction (Dangendorf et al., 2017 <sup>[[#fn:r127|127]]</sup> ) , or in the spatial patterns associated with the sea level contributions (Hamlington et al., 2018 <sup>[[#fn:r128|128]]</sup> ) have been proposed to explain these smaller GMSL rates. There is no agreement yet on which is the primary reason for the differences and it is not clear whether all the reasons invoked can actually explain all the differences across reconstructions. As there is no clear evidence to discard any reconstruction, this assessment considers the ensemble of AR5 sea level reconstructions augmented by the two recent reconstructions from Hay et al. (2015) <sup>[[#fn:r129|129]]</sup> and Dangendorf et al. (2017) <sup>[[#fn:r130|130]]</sup> to evaluate the GMSL changes over the 20th century. On this basis, it is estimated that it is ''very likely'' that the long-term trend in GMSL estimated from tide gauge records is 1.5 (1.1–1.9) mm yr <sup>–1</sup> between 1902 and 2010 for a total SLR of 0.16 (0.12–0.21) m (see also Table 4.1). This estimate is consistent with the AR5 assessment (but with an increased uncertainty range) and confirms that it is ''virtually certain'' that GMSL rates over the 20th century are several times as large as GMSL rates during the late Holocene (see 4.2.2.1). Over the 20th century the GMSL record also shows an acceleration ( ''high confidence'' ) as now four out of five reconstructions extending back to at least 1902 show a robust acceleration (Jevrejeva et al., 2008 <sup>[[#fn:r131|131]]</sup> ; Church and White, 2011 <sup>[[#fn:r132|132]]</sup> ; Ray and Douglas, 2011 <sup>[[#fn:r133|133]]</sup> ; Haigh et al., 2014b <sup>[[#fn:r134|134]]</sup> ; Hay et al., 2015 <sup>[[#fn:r135|135]]</sup> ; Watson, 2016 <sup>[[#fn:r136|136]]</sup> ; Dangendorf et al., 2017 <sup>[[#fn:r137|137]]</sup> ) . The estimates of the acceleration ranges between -0.002–0.019 mm yr <sup>–1</sup> over 1902–2010 are consistent with AR5.

<div id="section-4-2-2-1global-mean-sea-level-changes-during-the-instrumental-period-block-3"></div>

<span id="satellite-altimetry"></span>
===== 4.2.2.1.2 Satellite altimetry =====

High precision satellite altimetry started in October 1992 with the launch of the TOPEX/Poseidon and Jason series of spacecraft. Since then, 11 satellite altimeters have been launched providing nearly global sea level measurements (up to ±82° latitude) over more than 25 years. Six groups (AVISO/CNES, SL_cci/ESA, University of Colorado, CSIRO, NASA/GSFC, NOAA; Nerem et al., 2010; <sup>[[#fn:r138|138]]</sup> Henry et al., 2014 <sup>[[#fn:r139|139]]</sup> ; Leuliette, 2015 <sup>[[#fn:r140|140]]</sup> ; Watson et al., 2015 <sup>[[#fn:r141|141]]</sup> ; Beckley et al., 2017 <sup>[[#fn:r142|142]]</sup> ; Legeais et al., 2018 <sup>[[#fn:r143|143]]</sup> ) provide altimetry-based GMSL time series. Since AR5, several studies using two independent approaches based on tide gauge records (Watson et al., 2015 <sup>[[#fn:r144|144]]</sup> ) and the sea level budget closure (Chen et al., 2017 <sup>[[#fn:r145|145]]</sup> ; Dieng et al., 2017 <sup>[[#fn:r146|146]]</sup> ) identified a drift of 1.5 (0.4–3.4) mm yr <sup>–1</sup> in TOPEX A from January 1993 to February 1999. Accounting for this drift leads to a revised GMSL rate from satellite altimetry of 3.16 (2.79–3.53) for 1993–2015 (WCRP Global Sea Level Budget Group, 2018 <sup>[[#fn:r147|147]]</sup> ; see Table 4.1) compared to 3.3 mm yr <sup>–1</sup> (2.7–3.9) for 1993–2010 in AR5. Compared to AR5, the revised satellite altimetry GMSL estimates now show with ''high confidence'' an acceleration of 0.084 (0.059–0.090) mm yr <sup>–1</sup> over 1993–2015 (5–95% uncertainty range; Watson et al., 2015 <sup>[[#fn:r148|148]]</sup> ; Nerem et al., 2018 <sup>[[#fn:r149|149]]</sup> ) . This acceleration is due to an increase in Greenland mass loss since the 2000s (Chen et al., 2017 <sup>[[#fn:r150|150]]</sup> ; Dieng et al., 2017 <sup>[[#fn:r151|151]]</sup> ) and a slight increase in all other contributions probably partly due to the recovery from the Pinatubo volcanic eruption in 1991 (Fasullo et al., 2016 <sup>[[#fn:r152|152]]</sup> ) and partly due to increased GHG concentrations e.g., (Slangen et al., 2016 <sup>[[#fn:r153|153]]</sup> ; ''high confidence'' ). The current sea level rise is 3.6 ± 0.3 mm yr <sup>–1</sup> over 2006–2015 (90% confidence level). This is the highest rate measured by satellite altimetry (Ablain et al., 2019 <sup>[[#fn:r154|154]]</sup> ; ''medium confidence'' ). Before the satellite altimetry era, the highest rate of sea level rise recorded was reached during the period 1935–1944. It amounted 2.5 ± 0.7 mm yr <sup>–1</sup> (estimate at the 90% confidence level from sea level reconstructions; Church and White, 2011 <sup>[[#fn:r155|155]]</sup> ; Ray and Douglas, 2011 <sup>[[#fn:r156|156]]</sup> ; Jevrejeva et al., 2008 <sup>[[#fn:r157|157]]</sup> ; Hay et al., 2015 <sup>[[#fn:r158|158]]</sup> ; Dangendorf et al., 2017 <sup>[[#fn:r159|159]]</sup> ). This is expected to be smaller than the current rate of sea level rise, making the current sea level rise the highest on instrumental record ( ''medium confidence'' ).

<div id="section-4-2-2-2contributions-to-global-mean-sea-level-change-during-the-instrumental-period"></div>

<span id="contributions-to-global-mean-sea-level-change-during-the-instrumental-period"></span>
==== 4.2.2.2 Contributions to Global Mean Sea Level Change During the Instrumental Period ====

<div id="section-4-2-2-2contributions-to-global-mean-sea-level-change-during-the-instrumental-period-block-1"></div>

The different contributions to the GMSL rise are independently observed over various time scales. They are compared with simulated estimates from climate model experiments of CMIP5 (Taylor et al., 2012 <sup>[[#fn:r160|160]]</sup> ) when available (see Table 4.1). The observations are compared with experiments beginning in the mid-19th century, forced with past time-dependent anthropogenic changes in atmospheric composition, natural forcings due to volcanic aerosols and variations in solar irradiance (Taylor et al., 2012 <sup>[[#fn:r161|161]]</sup> ). The objective is first, to assess understanding of the causes of observed sea level changes and second, to evaluate the ability of coupled climate models to simulate these causes. It enables the evaluation of the confidence level there is in current coupled climate models that form the basis of future sea level projections.

<div id="section-4-2-2-2contributions-to-global-mean-sea-level-change-during-the-instrumental-period-block-2"></div>

<span id="thermal-expansion-contribution"></span>
===== 4.2.2.2.1 Thermal expansion contribution =====

The ocean thermal expansion is caused by excess heat being absorbed by the ocean, as the climate warms. Thermal expansion is estimated from ''in situ'' ocean observations and ocean heat content reanalyses that rely on assimilation of data into numerical models (Storto et al., 2017 <sup>[[#fn:r162|162]]</sup> ; Sections 1.8.1.1 and 1.8.1.4; WCRP Global Sea Level Budget Group, 2018 <sup>[[#fn:r163|163]]</sup> ) . Full-depth, high-quality and unbiased ocean temperature profile data with adequate metadata and spatio-temporal coverage are required to estimate thermal expansion and to understand drivers of variability and long-term change (Pfeffer et al., 2018 <sup>[[#fn:r164|164]]</sup> ; Section 5.2.2.2.2) .

Historically, however, observational gaps exist and some ocean regions remain under-sampled to date (Sections 1.8.1.1 and 5.2.2.2.2; Figure 1.3; Appendix 1.A, Figure 1.1). Other factors also introduce uncertainty in estimates of thermal expansion like changes in instrumentation, systematic instrumental errors, changes in the quality control of the data and the mapping method used to produce regular grids (Section 5.2.2.2.2; Palmer et al., 2010 <sup>[[#fn:r165|165]]</sup> ) . In the upper 700 m, the largest sources of uncertainty for estimates of global mean thermal expansion from 1970 to 2004 are the choice of mapping methods (Boyer et al., 2016 <sup>[[#fn:r166|166]]</sup> ) , followed by the choice of bias correction for the bathythermographic observations (Cheng et al., 2016 <sup>[[#fn:r167|167]]</sup> ; Section 5.2.2.2.2). From 2006 onwards, the uncertainty is considerably reduced (Roemmich et al., 2015 <sup>[[#fn:r168|168]]</sup> ; von Schuckmann et al., 2016 <sup>[[#fn:r169|169]]</sup> ; Wijffels et al., 2016 <sup>[[#fn:r170|170]]</sup> ) , because the Argo array reached its targeted near-global ( up to ±60° latitude) coverage for the upper 2000 m in November 2007 (Riser et al., 2016 <sup>[[#fn:r171|171]]</sup> ; Section 5.2.2.2.2) .

Since AR5, in a community effort, the (WCRP Global Sea Level Budget Group, 2018 <sup>[[#fn:r172|172]]</sup> ) revisited the global mean thermal expansion estimates based on observations only. On the basis of a full-depth 13-member ensemble of global mean thermal expansion time series developed with the latest data and corrections available, they estimated that the global thermal expansion was 1.40 (1.08 – 1.72) mm yr <sup>–1</sup> for 2006–2015, 1.36 (0.96 – 1.76) mm yr <sup>–1</sup> for 1993–2015 (see Table 4.1). While the relative contribution of the upper 300 m did not change (~70%) between 2006–2015 and 1993–2015, the 700–2000 m contribution increased around 10% over the Argo decade (2006–2015), when observations for that depth interval soared ( Figure 1.3; Appendix 1.A, Figure 1.1) . This suggests that observed changes for 700–2000 m may have been underestimated for 1993 – 2005. Before 1993, estimates are based on a smaller ensemble of 4 datasets in which no thermal expansion is assumed below 2000 m because of lack of data (see Section 5.2.2.2.2 for more details). This ensemble shows a thermal expansion linear rate of 0.89 (0.84 – 0.94) mm yr <sup>–1</sup> for 1970–2015 (see Table 4.1).

Coupled climate models simulate the historical thermal expansion (see Table 4.1). However, for models that omit the volcanic forcing in their control experiment, the imposition of the historical volcanic forcing during the 20th century results in a spurious time mean negative forcing and a spurious persistent ocean cooling related to the control climate (Gregory, 2010 <sup>[[#fn:r173|173]]</sup> ; Gregory et al., 2013 <sup>[[#fn:r174|174]]</sup> ) . Since AR5, the magnitude of this effect has been estimated from historical simulations forced by only natural radiative forcing. Then it has been used to correct the historical simulations forced with the full 20th century forcing (Slangen et al., 2016 <sup>[[#fn:r187|187]]</sup> ; Slangen et al., 2017b <sup>[[#fn:r188|188]]</sup> ) . The resulting ensemble mean of simulated thermal expansion provides a good fit to the observations within the uncertainty ranges of both models and observations (Slangen et al., 2017b <sup>[[#fn:r189|189]]</sup> ; Cheng et al., 2019 <sup>[[#fn:r190|190]]</sup> ; Table 4.1) . The spread, which is essentially due to uncertainty in radiative forcing and uncertainty in the modelled climate sensitivity and ocean heat uptake efficiency (Melet and Meyssignac, 2015 <sup>[[#fn:r191|191]]</sup> ) , is still larger than the observational uncertainties (Gleckler et al., 2016 <sup>[[#fn:r192|192]]</sup> ; Cheng et al., 2017 <sup>[[#fn:r193|193]]</sup> ; Table 4.1) . Compared to AR5, the availability of improved observed and modelled estimates of thermal expansion and the good agreement between both confirm the ''high confidence'' level in the simulated thermal expansion using climate models and the ''high confidence'' level in their ability to project future thermal expansion.

<div id="section-4-2-2-2contributions-to-global-mean-sea-level-change-during-the-instrumental-period-block-3"></div>

<span id="table-4.1"></span>
<!-- START IMG -->
<!-- TABLE IMG -->
<!-- IMG TITLE -->
'''Table 4.1'''
<!-- IMG CAPTION -->
Global mean sea level (GMSL) budget over different periods from observations and from climate model base contributions. All values are in mm yr–1. Values in brackets in 4.2 are uncertainties ranging from 5–95%. The climate model historical simulations end in 2005; projections for Representative Concentration Pathway (RCP)8.5 are used for 2006–2015. The modelled thermal expansion, glacier and ice sheet surface mass balance (SMB) contributions are computed from the Coupled Model Intercomparison Project Phase 5 (CMIP5) models as in Slangen et al. (2017b). For the model contributions, uncertainties are estimated from the spread of the ensemble of model simulations following Slangen et al. (2017b), see the footnotes for the details on the uncertainty propagation. GIS is Greenland Ice Sheet.
<!-- IMG FILE -->
[[File:bede59e395134bf8a28829a10c2702b3 table4.1.png]]

Notes:

# (a)  The number is built from WCRP Global Sea Level Budget Group (2018) estimate of the 0–700 m depth thermal expansion, assuming no trend below 2000 m depth before 1992 and the mean value from Purkey and Johnson (2010), and Desbruyères et al. (2017) afterwards.
# (b)  The number is calculated as the mean between the estimate from a reconstruction of glacier mass balance based on glacier length (update of Leclercq et al. (2011)) and the estimate from a mass balance model forced with atmospheric observations (Marzeion et al., 2015). The uncertainty is assumed to be a gaussian with a standard deviation of half the difference between the two estimates.
# (c)  The number is calculated as the sum of the Greenland Ice Sheet (GIS) contribution from Kjeldsen et al. (2015) and the peripheral glaciers’ contribution. The peripheral glaciers’ contribution and the associated uncertainty are computed from a mass balance model forced with atmospheric observations (Marzeion et al., 2015). The total uncertainty is computed assuming that both uncertainties from the GIS contribution and from the peripheral glaciers’ contribution are independent.
# (d)  Numbers from Bamber et al. (2018). See Section 3.3.1 for more details.
# (e)  These numbers are the weighted average of the numbers from Bamber et al. (2018) and from The Imbie team (2018). The weights in the average are based on the uncertainty associated to each estimate. See Section 3.3.1 for more details.
# (f)  Only direct anthropogenic contribution, from Wada et al. (2016).
# (g)  Land water storage estimated from Gravity Recovery and Climate Experiment (GRACE) excluding glaciers, from WCRP Global Sea Level Budget Group (2018).
# (h)  Direct estimate of ocean mass from GRACE from WCRP Global Sea Level Budget Group (2018).
# (i)  Sum of the thermal expansion and the contributions from glaciers, GIS, Antarctica Ice Sheet (AIS) and land water storage. Uncertainties in the different contributions are assumed as independent.
# (j)  Sea level reconstructions that end before 2015 have been extended to 2015 with the satellite altimetry record from Legeais et al. (2018). The uncertainty is derived from the uncertainty of individual sea level reconstructions over the longest period available that start in 1970. The uncertainty from different sea level reconstructions are assumed as independent.
# (k)  The mean estimate is from the satellite altimetry estimate in WCRP Global Sea Level Budget Group (2018) corrected for GIA and for the elastic response of the ocean crust to present day mass redistribution (Frederikse et al., 2017; Lickley et al., 2018). The uncertainty is computed using the updated error budget of Ablain et al. (2015).
# (l)  Land water storage is estimated from Wada et al. (2016) and ice discharge is deduced from Shepherd et al. (2012). The ice discharge contribution is assumed to be zero before 1992. The uncertainties in the different contributions from coupled climate models are assumed independent.
# (m)  The uncertainties in the observed GMSL and the coupled climate models’ estimate of GMSL are assumed independent for the computation of the uncertainties in the residuals.
# (n)  Numbers taken from Appendix 2.A.
# (o)  Numbers taken from Zemp et al. (2019), see Sections 2.2.3 and 3.3.2 for more details.
# (p)  The Number is calculated as the mean of the estimates of Zemp et al. (2019) and Bamber et al. (2018). The uncertainties of the two estimates are assumed to be independent of each other to obtain the uncertainty estimate of the mean.

<!-- END IMG -->
<div id="section-4-2-2-2contributions-to-global-mean-sea-level-change-during-the-instrumental-period-block-4"></div>

<span id="ocean-mass-observations-from-grace-and-grace-follow-on"></span>
===== 4.2.2.2.2 Ocean mass observations from GRACE and GRACE Follow-On =====

The ocean mass changes correspond to the sum of land ice and terrestrial water storage changes. Since 2002, the GRACE and GRACE follow-on missions provide direct estimates of the ocean mass changes and thus they provide an independent estimate of the sum of land ice and terrestrial water storage contributions to sea level. Since AR5, GRACE-based estimates of the ocean mass rates are increasingly consistent (WCRP Global Sea Level Budget Group, 2018 <sup>[[#fn:r194|194]]</sup> ) because of the extended length of GRACE missions’ observations (over 15 years), the improved understanding of data and methods for addressing GRACE limitations (e.g., noise filtering, leakage correction and low-degree spherical harmonics estimates), and the improved knowledge of geophysical corrections applied to GRACE data (e.g., GIA). The most recent estimates (Dieng et al., 2015b <sup>[[#fn:r195|195]]</sup> ; Reager et al., 2016 <sup>[[#fn:r196|196]]</sup> ; Rietbroek et al., 2016 <sup>[[#fn:r197|197]]</sup> ; Chambers et al., 2017 <sup>[[#fn:r198|198]]</sup> ; Blazquez et al., 2018 <sup>[[#fn:r199|199]]</sup> ; Uebbing et al., 2019 <sup>[[#fn:r200|200]]</sup> ) report a global ocean mass increase of 1.7 (1.4 – 2.0) mm yr <sup>–1</sup> over 2003–2015 (see also Table 4.1). The uncertainty arises essentially from differences in the inversion method to compute the ocean mass (Chen et al., 2013 <sup>[[#fn:r201|201]]</sup> ; Jensen et al., 2013 <sup>[[#fn:r202|202]]</sup> ; Johnson and Chambers, 2013 <sup>[[#fn:r204|204]]</sup> ; Rietbroek et al., 2016 <sup>[[#fn:r205|205]]</sup> ) , uncertainties in the geocentre motion and uncertainty in the GIA correction (Blazquez et al., 2018 <sup>[[#fn:r205|205]]</sup> ; Uebbing et al., 2019 <sup>[[#fn:r206|206]]</sup> ) . The consistency between estimates of the global mean ocean mass on a monthly time scale has also increased since AR5.

<div id="section-4-2-2-2contributions-to-global-mean-sea-level-change-during-the-instrumental-period-block-5"></div>

<span id="glaciers-1"></span>
===== 4.2.2.2.3 Glaciers =====

To assess the mass contribution of glaciers to sea level change, global estimates are required. Recent updates and temporal extensions of estimates obtained by different methods continue to provide ''very high confidence'' in continuing glacier mass loss on the global scale during the past decade (Bamber et al., 2018 <sup>[[#fn:r207|207]]</sup> ; Wouters et al., 2019; Zemp et al., 2019 <sup>[[#fn:r208|208]]</sup> ), see Section 2.2.3 and Appendix 2.A for a detailed discussion also on regional scales). Updates of the reconstructions of Cogley (2009) <sup>[[#fn:r210|210]]</sup> , Leclercq et al. (2011) <sup>[[#fn:r211|211]]</sup> and Marzeion et al. (2012) <sup>[[#fn:r212|212]]</sup> , presented and compared in Marzeion et al. (2015) <sup>[[#fn:r212|212]]</sup> , show increased agreement on rates of mass loss during the entire 20th century (Marzeion et al., 2015 <sup>[[#fn:r213|213]]</sup> ), compared to earlier estimates reported by AR5. The contribution of glaciers that may be missing in inventories or have already melted during the 20th century is hard to constrain (Parkes and Marzeion, 2018 <sup>[[#fn:r215|215]]</sup> ), and there is ''low confidence'' in their estimated contribution. These glaciers are thus neglected in the assessment of the sea level budget (Table 4.1).

While the agreement between the observational estimates of glacier mass changes and the modelled estimates from glacier models forced with climate model simulations has increased since AR5 (Slangen et al., 2017b <sup>[[#fn:r216|216]]</sup> ), there is only ''medium'' ''confidence'' in the use of glacier models to reconstruct sea level change because of the limited number of well-observed glaciers available to evaluate models on long time scales, and because of the small number of model-based global glacier reconstructions.

<div id="section-4-2-2-2contributions-to-global-mean-sea-level-change-during-the-instrumental-period-block-6"></div>

<span id="greenland-and-antarctic-ice-sheets"></span>
===== 4.2.2.2.4 Greenland and Antarctic ice sheets =====

Frequent observations of ice sheet mass changes have only been available since the advent of space observations (see Section 3.3.1). In the pre-satellite era, mass balance was geodetically reconstructed only for the GIS (Kjeldsen et al., 2015 <sup>[[#fn:r217|217]]</sup> ) . These geodetic reconstructions empirically constrain the contribution of the GIS to SLR between 1900 and 1983 to 17.2 (10.7 – 23.2; Kjeldsen et al., 2015 <sup>[[#fn:r218|218]]</sup> ) . During the satellite era, three approaches have been developed to estimate ice sheet mass balance: 1) Mass loss is estimated by direct measurements of ice sheet height changes with satellite laser or radar altimetry in combination with climatological/glaciological models for firn density and compaction, 2) the input–output method combines measurements of ice flow velocities estimated from satellite (synthetic aperture radar or optical imagery) across key outlets with estimates of net surface balance derived from ice thickness data, 3) space gravimetry data yields direct estimate of the mass changes by inversion of the anomalies in the gravity field (see Section 3.3.1 for more details). AR5 concluded that the three space-based methods give consistent results. They agree in showing that the rate of SLR due to the GIS and AIS’ contributions has increased since the early 1990s. Since AR5, up-to-date observations confirm this statement with increased confidence for both ice sheets (Rignot et al., 2019 <sup>[[#fn:r219|219]]</sup> ; see Section 3.3.1) . The assessment of the literature since AR5 made in Section 3.3.1 shows that the contribution from Greenland to SLR over 2012–2016 (0.68 (0.64 – 0.72) mm yr <sup>–1</sup> ) was similar to the contribution over 2002–2011 (0.73 (0.67 – 0.79) mm yr <sup>–1</sup> ) and ''extremely likely'' greater than over 1992–2001 (0.02 (0.21 – 0.25) mm yr <sup>–1</sup> ). The contribution from Antarctica over 2012–2016 (0.55 (0.48 – 0.62) mm yr <sup>–1</sup> ) was ''extremely likely'' greater than over the 2002–2011 period (0.23 (0.16 – 0.30) mm yr <sup>–1</sup> ) and ''likely'' greater than over the period 1992–2001 (0.14 (0.12 – 0.16); see Section 3.3.1 for more details).

Here, the approach of Section 3.3.1 is followed, using the two multi-method assessments from Bamber et al. (2018) <sup>[[#fn:r220|220]]</sup> and the IMBIE team (2018) to evaluate the contribution of ice sheet mass loss to SLR over 1993–2015 and 2006–2015 (see Table 4.1). These two studies agree with results from the WCRP Global Sea Level Budget Group (2018) . For the estimation of the AIS contribution, Bamber et al. (2018) <sup>[[#fn:r221|221]]</sup> and the The IMBIE team (2018) use similar but not identical data sources and processing. Both studies find consistent results within uncertainties over both periods. In Table 4.1, the results of these two studies were averaged, and weighted the average on the basis of their uncertainties, because there is no apparent reason to discount either study. For the estimation of the GIS contribution only the Bamber et al. (2018) <sup>[[#fn:r222|222]]</sup> estimate is used, as there is no other multi-method assessment available.

<div id="section-4-2-2-2contributions-to-global-mean-sea-level-change-during-the-instrumental-period-block-7"></div>

<span id="contributions-from-water-storage-on-land"></span>
===== 4.2.2.2.5 Contributions from water storage on land =====

Water is stored on land not only in the form of ice but snow, surface water, soil moisture and groundwater. Temporal changes in land water storage, defined as all forms of water stored on land excluding land ice, contribute to observed changes in ocean mass and thus sea level on annual to centennial time scales (Döll et al., 2016 <sup>[[#fn:r225|225]]</sup> ; Reager et al., 2016 <sup>[[#fn:r226|226]]</sup> ; Hamlington et al., 2017 <sup>[[#fn:r227|227]]</sup> ; Wada et al., 2017 <sup>[[#fn:r228|228]]</sup> ) . They are caused by both climate variability and direct human interventions, at the multi-decadal to centennial time scales. Over the past century, the main cause for land water storage changes are the groundwater depletion and impoundment of water behind dams in reservoirs (Döll et al., 2016 <sup>[[#fn:r229|229]]</sup> ; Wada et al., 2016 <sup>[[#fn:r230|230]]</sup> ) . While the rate of groundwater depletion and thus its contribution to SLR increased during the 20th century and up to today (Wada et al., 2016 <sup>[[#fn:r231|231]]</sup> ) , its effect on sea level was more than balanced by the increase in land water storage due to dam construction between 1950 and 2000 (Wada et al., 2016 <sup>[[#fn:r232|232]]</sup> ) . Since about 2000, based on hydrological models, the combined effect of both processes is a positive contribution to SLR (Wada et al., 2016 <sup>[[#fn:r233|233]]</sup> ) . Decreased water storage in lakes, wetlands and soils due to human activities are less important for ocean mass changes (Wada et al., 2016 <sup>[[#fn:r234|234]]</sup> ) . Overall, the integrated effects of the direct human intervention on land hydrology have reduced land water storage during the last decade, increasing the rate of SLR by 0.15–0.24 mm yr <sup>–</sup> <sup>1</sup> (Wada et al., 2016 <sup>[[#fn:r235|235]]</sup> ; Wada et al., 2017 <sup>[[#fn:r236|236]]</sup> ; Scanlon et al., 2018 <sup>[[#fn:r237|237]]</sup> ; WCRP Global Sea Level Budget Group, 2018 <sup>[[#fn:r238|238]]</sup> ) . Over periods of a few decades, land water storage was affected significantly by climate variability (Dieng et al., 2015a <sup>[[#fn:r239|239]]</sup> ; Reager et al., 2016 <sup>[[#fn:r240|240]]</sup> ; Dieng et al., 2017 <sup>[[#fn:r241|241]]</sup> ) . Net land water storage change driven by both climate and direct human interventions can be determined based on GRACE observations and global hydrological modelling. They indicate different estimates of the rate of SLR. Over the period 2002–2014 GRACE-based estimates of the net land water storage (i.e., not including glaciers) show a negative contribution to sea level (e.g., Scanlon et al., 2018) resulting in the negative value after 2006 in Table 1while hydrological models determined a slightly positive one. The reasons for this difference between estimates are not elucidated. There is scientific consensus that uncertainties of both net land water storage contribution to sea level and its individual contributions remain high (WCRP Global Sea Level Budget Group, 2018 <sup>[[#fn:r242|242]]</sup> ) . The differences in estimates and the lack of multiple consistent studies give ''low confidence'' in the net land water storage contribution to current SLR.

<div id="section-4-2-2-2contributions-to-global-mean-sea-level-change-during-the-instrumental-period-block-8"></div>

<span id="budget-of-global-mean-sea-level-change"></span>
===== 4.2.2.2.6 Budget of global mean sea level change =====

Drawing on previous sections, the budget of GMSL rise (Table 4.1, Figure 4.5) is assessed with observations over 4 periods: 1901–1990 (which corresponds to the period in the 20th century that is prior to the increase in ice sheet contributions to GMSL rise), 1970–2015 (when ocean observations are sufficiently accurate to estimate the global ocean thermal expansion and when glacier mass balance reconstructions start), 1993–2015 (when precise satellite altimetry is available) and 2006–2015 (when GRACE data is available in addition to satellite altimetry and when the Argo network reaches a near-global coverage). The budget of GMSL rise is also assessed with sea level contributions simulated by climate models over the same periods (Table 4.1, Figure 4.5). The periods 1993–2015 and 2006–2015 are only 23 and 10 years long respectively, short enough so that they can be affected by internal climate variability. Therefore, it is not expected that observations over these periods will be precisely reproduced by climate model historical experiments. For the contribution from land water storage, the estimated effect of direct human intervention was used, neglecting climate-related variations until 2002 (Ngo ‐ Duc et al., 2005) . From 2002 to 2015, total land water storage estimated with GRACE was used. In general, historical simulations of climate models end in 2005. Historical simulations were extended here to 2015 using the RCP8.5 scenario. This choice of RCP scenario is not critical for the simulated sea level, as the different scenarios only start to diverge significantly after the year 2030 (Church et al., 2013 <sup>[[#fn:r243|243]]</sup> ) .

For 1993–2015 and 2006–2015, the observed GMSL rise is consistent within uncertainties with the sum of the estimated observed contributions (Table 4.1). Over the period 1993–2015 the two largest terms are the ocean thermal expansion (accounting for 43% of the observed GMSL rise) and the glacier mass loss (accounting for a further 20%). Compared to AR5, the extended observations corrected for the TOPEX-A drift (see Section 4.2.2.1.2) allow us now to identify an acceleration in the observed SLR over 1993–2015 and to attribute this acceleration mainly to Greenland ice loss along with an acceleration in Antarctic ice loss (Velicogna et al., 2014 <sup>[[#fn:r244|244]]</sup> ; Harig and Simons, 2015 <sup>[[#fn:r245|245]]</sup> ; Chen et al., 2017 <sup>[[#fn:r246|246]]</sup> ; Dieng et al., 2017 <sup>[[#fn:r247|247]]</sup> ; Yi et al., 2017 <sup>[[#fn:r248|248]]</sup> ; see also Sections 4.2.2.2.2, 4.2.2.3.4, 3.3.1) . Since 2006, land ice, collectively from glaciers and the ice sheets has become the most important contributor to GMSL rise over the thermal expansion with mountain glaciers contributing 20% and ice sheets 33% (see Table 4.1) . Over the periods 1993–2015, the sum of the observed sea level contributions is consistent with the total observed sea level within uncertainties at monthly-scales (not shown, e.g., Dieng et al., 2017) . This is also true for the period 2006–2015, when uncertainties are significantly smaller. This agreement at monthly time scales represents a significant advance since the AR5 in physical understanding of the causes of past GMSL change. It provides an improved basis for the evaluation of models. Given these elements there is ''high confidence'' that the current observing system is capable of resolving decadal to multidecadal changes in GMSL and its components (with an uncertainty of &lt;0.7 mm yr <sup>–1</sup> at decadal and longer time scales, see Table 4.1 and for example, WCRP Global Sea Level Budget Group, 2018) . However, despite this advance since AR5 there are still no comprehensive observations of ocean thermal expansion below 2000 m, in regions covered by sea ice and in marginal seas. The understanding of glacier mass loss can be improved at regional scale and the understanding of the land water storage contribution is still limited. Thus, for smaller changes in sea level of the order of a few tenths of a mm yr <sup>–1</sup> at decadal time scales and shorter time scales there is ''medium confidence'' in the capability of the current observing system to resolve them (e.g., WCRP Global Sea Level Budget Group, 2018) .

Before 1992, observations are not sufficient to confidently estimate the ice sheet mass balance and before 1970, the space and time sampling of ocean observations are not sufficient to estimate the global ocean thermal expansion. For these reasons, it is difficult to assess the closure of the GMSL rise budget over 1901–1990 and 1970–2015 (Church et al., 2013 <sup>[[#fn:r249|249]]</sup> ; Gregory et al., 2013 <sup>[[#fn:r250|250]]</sup> ; Jevrejeva et al., 2017 <sup>[[#fn:r251|251]]</sup> ; Meyssignac et al., 2017c <sup>[[#fn:r252|252]]</sup> ; Slangen et al., 2017b <sup>[[#fn:r253|253]]</sup> ; Parkes and Marzeion, 2018 <sup>[[#fn:r254|254]]</sup> ) . For the period 1970–2015, the thermal expansion of the ocean represents 43% of the observed GMSL rise while the glaciers’ contribution represents 22% (see Table 4.1). This result indicates a slightly smaller contribution from glaciers than reported by AR5. If the GIS contribution and the Antarctic SMB is added, then the sum of the contributors to sea level is in agreement with the low end observed SLR estimates over 1970–2015 (Frederikse et al., 2018 <sup>[[#fn:r255|255]]</sup> ) . This result suggests that the contribution of Antarctica ice sheet dynamics to SLR has been small, if any, before the 1990s.

Since AR5, extended simulations along with recent findings in observations and improved model estimates allow for a new more robust, consistent and comprehensive comparison between sea level estimates based on observations and climate model simulations (e.g., Meyssignac et al., 2017c; Slangen et al., 2017b <sup>[[#fn:r256|256]]</sup> ; Parkes and Marzeion, 2018 <sup>[[#fn:r257|257]]</sup> ) . Compared to AR5, the simulated thermal expansion from climate models has improved with a new correction for the volcanic activity (see Section 4.2.2.2.1). The glacier contribution from glacier models forced with inputs from climate models is updated with a new glacier inventory and improvements to the glacier mass balance model (Marzeion et al., 2015 <sup>[[#fn:r258|258]]</sup> ) . The simulated Greenland SMB is estimated with a new regional SMB-component downscaling technique, which accounts for the regional variations in components of the Greenland SMB (Noël et al., 2015 <sup>[[#fn:r259|259]]</sup> ; Meyssignac et al., 2017a) <sup>[[#fn:r260|260]]</sup> . In addition, an updated groundwater extraction contribution from Döll et al. (2014) <sup>[[#fn:r261|261]]</sup> is now used for the land water storage contribution.

For the periods 1970–2015, 1993–2015 and 2006–2015 the simulated contributions from thermal expansion, glaciers mass loss and Greenland SMB explain respectively 84%, 81% and 77% of the observed GMSL (see Table 4.1). For all these periods the residual is consistent within uncertainty with the sum of the contribution from land water storage and ice discharge from Greenland and Antarctica. For each period the consistency is improved compared to AR5 (see Table 4.1) although the uncertainty on the residual is slightly larger because of a larger uncertainty in simulated Glaciers and Greenland SMB contributions.

For the period 1901–1990 the simulated contributions from thermal expansion, glaciers mass loss and Greenland SMB explain only 60% of the observed GMSL and the residual is too large to be explained by the sum of the contribution from land water storage and ice discharge from Greenland and Antarctica. The gap can be explained by a bias in the simulated Greenland SMB and glacier ice loss around Greenland in the early 20th century (Slangen et al., 2017b <sup>[[#fn:r262|262]]</sup> ) . When the glacier model and the Greenland SMB downscaling technique are forced with observed climate from atmospheric reanalyses, rather than the simulated climate from coupled climate models, simulated SLR becomes consistent with the observed SLR (see the dashed blue line on Figure 4.5). This is because atmospheric reanalyses show an increase in air temperatures in and around Greenland over the period 1900–1940, which lead to increased melt in Greenland (Bjørk et al., 2012 <sup>[[#fn:r267|267]]</sup> ; Fettweis et al., 2017 <sup>[[#fn:r268|268]]</sup> ) and surrounding glaciers in the first half of the 20th century. This increase in air temperature over 1900–1940 is not reproduced by climate models (Slangen et al., 2017b <sup>[[#fn:r269|269]]</sup> ) . It may be because this increase in air temperature was due to internal climate variability on temporal and spatial scales that cannot be precisely reproduced by climate models. It may also be due to a bias in atmospheric circulation in climate models (Fettweis et al., 2017 <sup>[[#fn:r270|270]]</sup> ) , or an issue with the spatial pattern of the historical aerosol forcing.

In summary, the agreement between climate model simulations and observations of the global thermal expansion, glacier mass loss and Greenland SMB has improved compared to AR5 for periods starting after 1970. However, for periods prior to 1970, significant discrepancies between climate models and observations arise from the inability of climate models to reproduce some observed regional changes in glacier and GIS SMB around the southern tip of Greenland. It is not clear whether this bias in climate models is due to the internal variability of the climate system or deficiencies in climate models. For this reason, there is still ''medium confidence'' in the ability of climate models to simulate past and future changes in glaciers mass loss and Greenland SMB.

<div id="section-4-2-2-2contributions-to-global-mean-sea-level-change-during-the-instrumental-period-block-9"></div>

<span id="figure-4.5"></span>
<!-- START IMG -->
<!-- IMG TITLE -->
'''Figure 4.5'''

<span id="figure-4.5-comparison-of-simulated-by-coupled-climate-models-as-in-section-4.4.2.6-and-observed-global-mean-sea-level-change-gmsl-since-1901-a-and-since-1993-b.-the-average-estimate-of-12-coupled-model-intercomparison-project-phase-5-cmip5-climate-model-simulations-is-shown-in-blue-with-the-595-uncertainty-range-shaded-in"></span>
<!-- IMG CAPTION -->
'''Figure 4.5 | Comparison of simulated (by coupled climate models as in Section 4.4.2.6) and observed global mean sea level change (GMSL) since 1901 (a) and since 1993 (b). The average estimate of 12 Coupled Model Intercomparison Project Phase 5 (CMIP5) climate model simulations is shown in blue with the 5–95% uncertainty range shaded in […]'''
<!-- IMG FILE -->
[[File:a361928b5f4ba7db98666b4e46217e56 IPCC-SROCC-CH_4_5-3000x2599.jpg]]

Figure 4.5 | Comparison of simulated (by coupled climate models as in Section 4.4.2.6) and observed global mean sea level change (GMSL) since 1901 (a) and since 1993 (b). The average estimate of 12 Coupled Model Intercomparison Project Phase 5 (CMIP5) climate model simulations is shown in blue with the 5–95% uncertainty range shaded in blue and calculated according to the procedures in Church et al. (2013) <sup>[[#fn:r263|263]]</sup> . The average of the 12 model estimates corrected for the bias in glaciers mass loss and Greenland surface mass balance (SMB) over 1900–1940 (see Section 4.2.2.2.6) is shown in dashed blue. The estimates from tide gauge reconstructions is shown in other colours in panel a), with the 5–95% uncertainty range shaded in grey. The satellite altimetry observations from Legeais et al. (2018) <sup>[[#fn:r264|264]]</sup> is shown in black in panel b). GMSL from altimetry corrected for the TOPEX-A drift (Watson et al., 2015 <sup>[[#fn:r265|265]]</sup> ) in orange as well as the tide gauge reconstruction. The 5–95% uncertainty range is shaded in orange (Ablain et al., 2015 <sup>[[#fn:r266|266]]</sup> ). All curves in (a) represent anomalies in sea level with respect to the period 1986–2005 (i.e., with zero time-mean over the period 1986–2005) in order to be consistent with sea level projections in Section 4.2.3. Vertical lines indicate the occurrence of major volcanic eruptions, which cause temporary drops in GMSL. Updated from Slangen et al. (2017b).

<!-- END IMG -->
<div id="section-4-2-2-3regional-sea-level-changes-during-the-instrumental-period"></div>

<span id="regional-sea-level-changes-during-the-instrumental-period"></span>
==== 4.2.2.3 Regional Sea Level Changes During the Instrumental Period ====

<div id="section-4-2-2-3regional-sea-level-changes-during-the-instrumental-period-block-1"></div>

Sea level does not rise uniformly. Observations from tide gauges and satellite altimetry (Figure 4.6) indicate that sea level shows substantial regional variability at decadal to multi-decadal time scales (e.g., Carson et al., 2017; Hamlington et al., 2018 <sup>[[#fn:r271|271]]</sup> ). These regional changes are essentially due to changing winds, air-sea heat and freshwater fluxes, atmospheric pressure loading and the addition of melting ice into the ocean, which alters the ocean circulation (Stammer et al., 2013 <sup>[[#fn:r272|272]]</sup> ; Forget and Ponte, 2015 <sup>[[#fn:r273|273]]</sup> ; Meyssignac et al., 2017b <sup>[[#fn:r274|274]]</sup> ). The addition of water into the ocean also change the geoid, alter the rotation of the Earth and deform the ocean floor which in turn change sea level (e.g., Tamisiea, 2011; Stammer et al., 2013 <sup>[[#fn:r275|275]]</sup> ).

Sea level is rising in all ocean basins ( ''virtually certain'' ; Legeais et al. 2018 <sup>[[#fn:r276|276]]</sup> ). Part of this regional sea level rise is due to global sea level rise of which a majority is attributable to anthropogenic greenhouse gas emissions ( ''high confidence'' ; Slangen et al. 2016 <sup>[[#fn:r277|277]]</sup> ). The remaining part of the regional sea-level rise in ocean basins is a combination of the response to anthropogenic GHG emissions and internal variability (e.g., Stammer et al. 2013; ''medium confidence'' ).

In the open ocean, the spatial variability and trends in sea level observed during the recent altimetry era or reconstructed over the previous decades are dominated by the thermal expansion of the ocean. In shallow shelf seas and at high latitudes (&gt;60°N and &lt;55°S), the effect of dynamic mass redistribution becomes important. At local scale, salinity changes can also generate sizeable changes in the ocean density similar to thermal expansion and lead to significant variability in sea level (Forget and Ponte, 2015 <sup>[[#fn:r278|278]]</sup> ; Meyssignac et al., 2017b <sup>[[#fn:r279|279]]</sup> ). On global average, the heat and freshwater fluxes from the atmosphere into the ocean are responsible for the total heat that enters the ocean and for the associated GMSL rise. At regional scale and local scale, both the ocean transport divergences caused by wind stress anomalies and the spatial variability in atmospheric heat fluxes are responsible for the spatial variability in thermal expansion and thus for most of the regional sea level departures around the GMSL rise (e.g., Stammer et al., 2013; Forget and Ponte, 2015 <sup>[[#fn:r280|280]]</sup> ).

Over the Pacific, the surface wind anomalies responsible for the sea level spatio-temporal variability are associated with the ENSO, Pacific Decadal Oscillation (PDO) and North Pacific Gyre Oscillation modes (Hamlington et al., 2013 <sup>[[#fn:r281|281]]</sup> ; Moon et al., 2013 <sup>[[#fn:r282|282]]</sup> ; Palanisamy et al., 2015 <sup>[[#fn:r283|283]]</sup> ; Han et al., 2017 <sup>[[#fn:r284|284]]</sup> ). In the Indian Ocean they are associated with the ENSO and Indian Ocean Dipole (IOD) modes (Nidheesh et al., 2013 <sup>[[#fn:r285|285]]</sup> ; Han et al., 2014 <sup>[[#fn:r286|286]]</sup> ; Thompson et al., 2016 <sup>[[#fn:r287|287]]</sup> ; Han et al., 2017 <sup>[[#fn:r288|288]]</sup> ). In particular, the PDO is responsible for most of the intensified SLR that has been observed in the western tropical Pacific Ocean since the 1990s (Moon et al., 2013 <sup>[[#fn:r289|289]]</sup> ; Han et al., 2014 <sup>[[#fn:r290|290]]</sup> ; Thompson and Mitchum, 2014 <sup>[[#fn:r291|291]]</sup> ). Several studies suggested that in addition to the PDO signal, warming of the tropical Indian and Atlantic Oceans enhanced surface easterly trade winds and thus also contributes to the intensified SLR in the western tropical Pacific (England et al., 2014 <sup>[[#fn:r292|292]]</sup> ; Hamlington et al., 2014 <sup>[[#fn:r293|293]]</sup> ; McGregor et al., 2014 <sup>[[#fn:r294|294]]</sup> ).

Over the Atlantic, the regional sea level variability at interannual to multi-decadal time scales, is generated by surface wind anomalies and heat fluxes associated with the North Atlantic Oscillation (NAO; Han et al., 2017 <sup>[[#fn:r295|295]]</sup> ) and also by ocean heat transport due to changes in the Atlantic Meridional Overturning Circulation (AMOC; McCarthy et al., 2015 <sup>[[#fn:r296|296]]</sup> ). Both mechanisms are not independent as heat fluxes and wind stress anomalies associated with NAO can induce changes in the AMOC (Schloesser et al., 2014 <sup>[[#fn:r297|297]]</sup> ; Yeager and Danabasoglu, 2014 <sup>[[#fn:r298|298]]</sup> ). In the Southern Ocean, the sea level variability is dominated by the SAM influence in particular in the Indian and Pacific sectors. The Southern Annular Mode (SAM) influence becomes weaker equator-wards in these sectors while the influence of PDO, ENSO and IOD increases (Frankcombe et al., 2015 <sup>[[#fn:r299|299]]</sup> ). In the southern ocean, the zonal asymmetry in westerly winds associated to the SAM, generates convergent and divergent transport in the Antarctic Circumpolar Current which may have contributed to the regional asymmetry of decadal sea level variations during most of the twentieth century (Thompson and Mitchum, 2014 <sup>[[#fn:r300|300]]</sup> ).

As for GMSL, net regional sea level changes can be estimated from a combination of the various contributions to sea level change. The contributions from dynamic sea level, atmospheric loading, glacier mass changes and ice sheet SMB can be derived from CMIP5 climate model outputs either directly or through downscaling techniques (Perrette et al., 2013 <sup>[[#fn:r301|301]]</sup> ; Kopp et al., 2014 <sup>[[#fn:r302|302]]</sup> ; Slangen et al., 2014a <sup>[[#fn:r303|303]]</sup> ; Bilbao et al., 2015 <sup>[[#fn:r304|304]]</sup> ; Carson et al., 2016 <sup>[[#fn:r305|305]]</sup> ; Meyssignac et al., 2017a <sup>[[#fn:r306|306]]</sup> ). The contributions from groundwater depletion, reservoir storage and dynamic ice sheet mass changes are not simulated by coupled climate models over the 20th century and have to be estimated from observations. The sum of all contributions, including the GIA contribution, provides a modelled estimate of the 20th century net regional sea level changes that can be compared with observations from satellite altimetry and tide-gauge records (see Figure 4.6).

In terms of interannual to multi-decadal variability, there is a general agreement between the simulated regional sea level and tide gauge records, over the period 1900–2015 (see inset figures in Figure 4.6). The relatively large, short-term oscillations in observed sea level (black lines in insets in Figure 4.6), which are due to the natural internal climate variability, are included in general within the modelled internal variability of the climate system represented by the blue shaded area (5–95% uncertainty). But, as for GMSL, climate models tend to systematically underestimate the observed sea level trends from tide gauge records, particularly in the first half of the 20th century. This underestimation is explained by a bias identified in modelled Greenland SMB, and glacier ice loss around Greenland in the early 20th century (see Section 4.2.2.2.6; Slangen et al., 2017b <sup>[[#fn:r307|307]]</sup> ). The correction of this bias improves the agreement between the spatial variability in sea level trends from observations and from climate models (see Figure 4.6). Climate models indicate that the spatial variability in sea level trends observed by tide-gauge records over the 20th century is dominated by the GIA contribution and the thermal expansion contribution over 1900–2015. Locally all contributions to sea level changes are important as any contribution can cause significant local deviations. Around India for example, groundwater depletion is responsible for the low 20th century SLR (because the removal of groundwater mass generated a local decrease in geoid that made local SLR slower; Meyssignac et al., 2017c <sup>[[#fn:r308|308]]</sup> )

These results show the ability of models to reproduce the major 20th century regional sea level changes due to GIA, thermal expansion, glacier mass loss and ice sheet SMB. This is tangible progress since AR5. But some doubts remain regarding the ability of climate models to reproduce local variations such as the glaciers and the Greenland SMB contributions to sea level in the region around the southern tip of Greenland (Slangen et al., 2017b <sup>[[#fn:r309|309]]</sup> ) or such as the thermal expansion in some eddy active regions (Sérazin et al., 2016 <sup>[[#fn:r310|310]]</sup> ). Because of these doubts there is still ''medium confidence'' in climate models to project future regional sea level changes associated with thermal expansion, glacier mass loss and ice sheet SMB. Coupled climate models have not simulated the other contributions to 20th century sea level, including the growing ice sheet dynamical contribution and land water storage changes.

<div id="section-4-2-2-3regional-sea-level-changes-during-the-instrumental-period-block-2"></div>

<span id="figure-4.6"></span>
<!-- START IMG -->
<!-- IMG TITLE -->
'''Figure 4.6'''

<span id="figure-4.6-20th-century-simulated-regional-sea-level-changes-by-coupled-climate-models-and-comparison-with-a-selection-of-local-tide-gauge-time-series.-in-the-upper-left-corner-map-of-changes-in-simulated-relative-sea-level-rsl-for-the-period-19011920-to-19962015-estimated-from-climate-model-outputs.-insets-observed-rsl-changes-black"></span>
<!-- IMG CAPTION -->
'''Figure 4.6 | 20th century simulated regional sea level changes by coupled climate models and comparison with a selection of local tide gauge time series. In the upper left corner: map of changes in simulated relative sea level (RSL) for the period 1901–1920 to 1996–2015 estimated from climate model outputs. Insets: Observed RSL changes (black […]'''
<!-- IMG FILE -->
[[File:ce92a975877f501626ca4d1003740770 IPCC-SROCC-CH_4_6-3000x1993.jpg]]

Figure 4.6 | 20th century simulated regional sea level changes by coupled climate models and comparison with a selection of local tide gauge time series. In the upper left corner: map of changes in simulated relative sea level (RSL) for the period 1901–1920 to 1996–2015 estimated from climate model outputs. Insets: Observed RSL changes (black lines) from selected tide gauge stations for the period 1900–2015. For comparison, the estimate of the simulated RSL change at the tide gauge station is also shown (blue plain line for the model estimates and blue dashed line for the model estimates corrected for the bias in glaciers mass loss and Greenland surface mass balance (SMB) over 1900–1940, see Section 4.2.2.2.6). The relatively large, short-term oscillations in observed local sea level (black lines) are due to the natural internal climate variability. For Mediterranean tide gauges, that is, Venice and Alexandria, the local simulated sea level has been computed with the simulated sea level in the Atlantic ocean at the entrance of the strait of Gibraltar following (Adloff et al., 2018). Tide gauge records have been corrected for vertical land motion (VLM) not associated with GIA where available, that is, for New York, Balboa and Lusi. Updated from Meyssignac et al. (2017b) to mimic RSL as good as possible.

<!-- END IMG -->
<div id="section-4-2-2-4local-coastal-sea-level"></div>

<span id="local-coastal-sea-level"></span>
==== 4.2.2.4 Local Coastal Sea Level ====

<div id="section-4-2-2-4local-coastal-sea-level-block-1"></div>

Since the local coastal sea level (scale ~10 km) is affected by global, regional (scale ~100 km) and coastal scale features and processes like anthropogenic subsidence, it may differ substantially from the regional sea level. At the coast, the sea level change is additionally affected by wave run up, tidal level, wind forcing, sea level pressure (SLP), the dominant modes of climate variability, seasonal climatic periodicities, mesoscale eddies, changes in river flow, as well as anthropogenic subsidence (see also Box 4.1). These local contributions, combined with sea level events generated by storm surges and tides result in anomalous conditions (ESL) which last for a short time in contrast to the gradual increase over time from for instance ice mass loss. Flood risk due to ESL is exacerbated due to its interaction with RSL and hence physical vulnerability assessments combine uncertainties around ESL and RSL, both in terms of contemporary assessments and future projections (Little et al., 2015b <sup>[[#fn:r312|312]]</sup> ; Vousdoukas, 2016 <sup>[[#fn:r313|313]]</sup> ; Vousdoukas et al., 2016 <sup>[[#fn:r314|314]]</sup> ; Wahl et al., 2017 <sup>[[#fn:r315|315]]</sup> ). Changes in mean sea level have been dealt with in previous sections (e.g., Section 4.2.2.2.6). Here the focus is on some of the components of ESL that have been assessed in combination with changes in RSL. Church et al. (2013) concluded that change in sea level extremes is ''very'' ''likely'' to be caused by a RSL increase, and that storminess and surges will contribute towards these extremes; however, it was noted that there was ''low confidence'' in region-specific projections as there was only a limited number of studies with a poor geographical coverage available.

Recent advances in statistical and dynamical modelling of wave effects at the coast, storm surges and inundation risk have reduced the uncertainties around the inundation risks at the coast (Vousdoukas et al., 2016 <sup>[[#fn:r316|316]]</sup> ; Vitousek et al., 2017 <sup>[[#fn:r317|317]]</sup> ; Melet et al., 2018 <sup>[[#fn:r318|318]]</sup> ; Vousdoukas et al., 2018c <sup>[[#fn:r319|319]]</sup> ) and assessments of the resulting highly resolved coastal sea levels are now emerging (Cid et al., 2017 <sup>[[#fn:r320|320]]</sup> ; Muis et al., 2017 <sup>[[#fn:r321|321]]</sup> ; Wahl et al., 2017 <sup>[[#fn:r322|322]]</sup> ). This progress was facilitated due to the availability of, for example, the Global Extreme Sea Level Analysis (GESLA-2; Woodworth et al., 2016 <sup>[[#fn:r323|323]]</sup> ) high-frequency (hourly) datasets, advances in the Coordinated Ocean Wave Climate Project (COWCLIP; Hemer et al., 2013 <sup>[[#fn:r324|324]]</sup> ), coastal altimetry datasets (Cipollini et al., 2017 <sup>[[#fn:r325|325]]</sup> ), and the Global Tide and Surge Reanalysis (GTSR; Muis et al., 2016 <sup>[[#fn:r326|326]]</sup> ), while new analyses of datasets that have been available since before the publication of AR5 have continued (e.g., PSML; Holgate et al., 2012 <sup>[[#fn:r327|327]]</sup> ).

Although ESL is experienced episodically by definition, Marcos et al. (2015) <sup>[[#fn:r328|328]]</sup> examined the long-term behaviour of storm surge models and detected decadal and multidecadal variations in storm surge that are not related to changes in RSL. They found that, although 82% of their observed time series showed synchronous patterns at regional scales, the pattern tended to be non-linear, implying that it would be difficult to infer future behaviour unless the physical basis for the responses was understood. An analysis of the relative contributions of SLR and ESL due to storminess showed that in the US Pacific northwest since the early 1980s, increases in wave height and period have had a larger effect on coastal flooding and erosion than RSL (Ruggiero, 2012 <sup>[[#fn:r329|329]]</sup> ) since the early 1980s. This is also true in other regions distributed over the entire globe (Melet et al., 2016 <sup>[[#fn:r330|330]]</sup> ; Melet et al., 2018 <sup>[[#fn:r331|331]]</sup> ). Changes since 1990 in the sea level harmonics and seasonal phases and amplitudes of the wave period and significant wave height were found for the Gulf of Mexico coast and along the US east coast (Wahl et al., 2014 <sup>[[#fn:r332|332]]</sup> ; Wahl and Plant, 2015 <sup>[[#fn:r333|333]]</sup> ). These authors found that high waters have increased twice as much as one would expect from long-term SLR alone, because of additional changes in the seasonal cycle, yielding a 30% increase in risk of flooding. Such effects are ''likely'' to be highly dependent on the local conditions. For example, using WAVEWATCH III, TOPEX/Poseidon altimetry tide model data and atmospheric forcing physically downscaled using Delft3D-WAVE and Delft3D-FLOW in what they call the Coastal Storm Modeling System (CoSMoS), Vitousek et al. (2017) were able to detect local inundation hazards (at a scale of hundreds of metres) across regions along the Californian coast. Similarly, Castrucci and Tahvildari (2018) <sup>[[#fn:r334|334]]</sup> simulated the impact of SLR along the Mid-Atlantic region in the USA. A study for the Maldives shows that the contribution of wave setup is essential to estimate flood risks (Wadey et al., 2017 <sup>[[#fn:r335|335]]</sup> ).

In deltas, the local sea level can be dominated by anthropogenic subsidence more than by the processes outlined above. It is often a primary driver of elevated local SLR and increased flood hazards in those regions. This is particularly true for deltaic systems, where fertile soils, low-relief topography, freshwater access, and strategic ports have encouraged the development of many of the world’s most densely populated coastlines and urban centres. For example, globally, one in fourteen humans resides in mid-to-low latitude deltas (Day et al., 2016 <sup>[[#fn:r336|336]]</sup> ). Although in these areas RSL is dominated by anthropogenic subsidence, climate effects need to be included for estimating risks associated with RSL (Syvitski et al., 2009 <sup>[[#fn:r337|337]]</sup> ).

Deltas are formed by the accumulation of unconsolidated river born sediments and porous organic material, both of which are particularly prone to compaction. It is the compaction which causes a drop in land elevation that increases the rate of local SLR above what would be observed along a static coastline or one where only climatological forced processes control the RSL. Under stable deltaic conditions, the accumulation of fluvially-sourced surficial sediment and organic matter offsets this natural subsidence (Syvitski and Saito, 2007 <sup>[[#fn:r338|338]]</sup> ); however, in many cases this natural process of delta construction has been disturbed by reductions in fluvial sediment supply via upstream dams and fluvial channelisation (Vörösmarty et al., 2003 <sup>[[#fn:r339|339]]</sup> ; Syvitski and Saito, 2007 <sup>[[#fn:r340|340]]</sup> ; Syvitski et al., 2009 <sup>[[#fn:r341|341]]</sup> ; Luo et al., 2017 <sup>[[#fn:r342|342]]</sup> ). Further, the extraction of fluids and gas that fill the pore space of deltaic sediments and provide support for overlying material has significantly increased the rate of compaction and resultant anthropogenic subsidence along many populated deltas (Higgins, 2016 <sup>[[#fn:r343|343]]</sup> ). In addition, Nicholls (2011) pointed to anthropogenic subsidence by the weight of buildings in megacities in South-East Asia.

Average natural and anthropogenic subsidence rates of 6–9 mm yr <sup>–1</sup> are reported for the highly populated areas of Ganges-Brahmaputra-Meghna delta in the urban centres of Kolkata and Dhaka (Brown and Nicholls, 2015 <sup>[[#fn:r344|344]]</sup> ). A fraction of these subsidence rates might be caused by long-term processes of increased sediment loading during the Holocene resulting from changes in the monsoon system (Karpytchev et al., 2018 <sup>[[#fn:r345|345]]</sup> ). Subsidence rates are expected to decrease in the Ganges-Brahmaputra-Meghna delta in the near future due to planned dam projects and an estimated 21% drop in resulting sediment supply (Tessler et al., 2018 <sup>[[#fn:r346|346]]</sup> ). Observations of enhanced natural and anthropogenic subsidence on the Ganges-Brahmaputra-Meghna are common to most heavily populated deltaic systems. Coastal mega-cities that have been particularly prone to human-enhanced subsidence include Bangkok, Ho Chi Minh city (Vachaud et al., 2018 <sup>[[#fn:r347|347]]</sup> ), Jakarta, Manila, New Orleans, West Netherlands and Shanghai (Yin et al., 2013 <sup>[[#fn:r348|348]]</sup> ; Cheng et al., 2018 <sup>[[#fn:r349|349]]</sup> ). On a global scale, observed rates of modern deltaic anthropogenic subsidence range from 6–100 mm yr <sup>–1</sup> (Bucx et al., 2015 <sup>[[#fn:r350|350]]</sup> ; Higgins, 2016 <sup>[[#fn:r351|351]]</sup> ). Rates of recent deltaic subsidence over the last few decades have been at least twice the 3 mm yr <sup>–1</sup> rate of GMSL rise observed over this same interval (Higgins, 2016 <sup>[[#fn:r352|352]]</sup> ; Tessler et al., 2018 <sup>[[#fn:r353|353]]</sup> ). Numerical models that have reproduced these observed rates of anthropogenic deltaic subsidence by considering human-induced compaction and reduced sediment supply, support anthropogenic causes for elevated rates of subsidence (Tessler et al., 2018 <sup>[[#fn:r354|354]]</sup> ).

In summary, ESL interacts with RSL rise including anthropogenic subsidence in many vulnerable areas (see Box 4.1). Therefore, it is concluded with ''high confidence'' that the inclusion of local processes (wave effects, storm surges, tides, erosion, sedimentation and compaction) is essential to estimate local, relative and changes in ESL events. Although the effect of anthropogenic subsidence may be very large locally, it is not accounted for in the projection sections of this chapter as no global data sets are available which are consistent with RCP scenarios, and because the scale at which these processes take place is often smaller than the spatial scale used in climate models.

<div id="section-4-2-2-5attribution-of-sea-level-change-to-anthropogenic-forcing"></div>

<span id="attribution-of-sea-level-change-to-anthropogenic-forcing"></span>
==== 4.2.2.5 4.2.2.5 Attribution of Sea Level Change to Anthropogenic Forcing ====

<div id="section-4-2-2-5attribution-of-sea-level-change-to-anthropogenic-forcing-block-1"></div>

Bindoff et al. (2013) concluded that it is ''very likely'' that there has been a substantial contribution to ocean heat content from anthropogenic forcing (i.e., anthropogenic greenhouse gases, anthropogenic aerosols and land use change) since the 1970s, that it is ''likely'' that loss of land ice is partly caused by anthropogenic forcing, and that as a result, it is ''very likely'' that there is an anthropogenic contribution to the observed trend in GMSL rise since 1970. However, these conclusions were based on the understanding of the responsible physical processes, since formal attribution studies dedicated to quantifying the effect of individual external forcings were not available for GMSLR. Since AR5, such formal studies have attributed changes in individual components of sea level change (i.e., thermosteric sea level change and glacier mass loss), and in the total GMSL, to anthropogenic forcing.

<div id="section-4-2-2-5attribution-of-sea-level-change-to-anthropogenic-forcing-block-2"></div>

<span id="attribution-of-individual-components-of-sea-level-change-to-anthropogenic-forcing"></span>
===== 4.2.2.5.1 Attribution of individual components of sea level change to anthropogenic forcing =====

Marcos and Amores (2014) found that during the period 1970–2005, 87% (95% confidence interval: 72–100%) of the observed thermosteric SLR in the upper 700 m of the ocean was anthropogenic. Slangen et al. (2014b)  included the full ocean depth in their analysis. They concluded that a combination of anthropogenic and natural forcing is necessary to explain the temporal evolution of observed global mean thermosteric sea level change during the period 1957–2005. Anthropogenic forcing was responsible for the amplitude of observed thermosteric sea level change, while natural forcing caused the forced variability of observations. Observations could best be reproduced by scaling the patterns from ‘natural-only’ forcing experiments by using a factor of 0.70 ± 0.30 (2 standard deviations of the CMIP5 ensemble subset used), indicating a potential overestimation of forced variability in the CMIP5 ensemble. Patterns from the ‘anthropogenic-only’ forcing experiments needed to be scaled by a factor of 1.08 ± 0.13 (2 standard deviations of the CMIP5 ensemble subset used), indicating a realistic response of the CMIP5 ensemble to anthropogenic forcing.

For the glacier contribution to GMSL, Marzeion et al. (2014) concluded that while natural climate forcing and long-term adjustment of the glaciers to the end of the preceding Little Ice Age lead to continuous glacier mass loss throughout the simulation period of 1851–2010, the observed rates of glacier mass loss since 1990 can only be explained by including anthropogenic forcing. During the period 1851–2010, only 25 ± 35% of global glacier mass loss can be attributed to anthropogenic forcing, but 69 ± 24% during the period 1991–2010 (see Section 2.2.3 for a more detailed discussion of attribution of glacier mass change on regional scales).

There is ''medium confidence'' in evidence linking GIS mass loss to anthropogenic climate change, and ''low confidence'' in the evidence that AIS mass balance can be attributed to anthropogenic forcing (see Section 3.3.1.6 for a detailed discussion). The effects of groundwater depletion and reservoir impoundment on sea level change are anthropogenic by definition (e.g., Wada et al., 2012) .

<div id="section-4-2-2-5attribution-of-sea-level-change-to-anthropogenic-forcing-block-3"></div>

<span id="attribution-of-global-mean-sea-level-change-to-anthropogenic-forcing"></span>
===== 4.2.2.5.2 Attribution of global mean sea level change to anthropogenic forcing =====

By estimating a probabilistic upper range of long-term persistent natural sea level variability, Dangendorf et al. (2015) <sup>[[#fn:r356|356]]</sup> detected a fraction of observed sea level change that is unexplained by natural variability and concluded by inference that it is ''virtually certain'' that at least 45% of the observed increase in GMSL since 1900 is attributable to anthropogenic forcing. Similarly, Becker et al. (2014) provided statistical evidence that the observed sea level trend, both in the global mean and at selected tide gauge locations, is not consistent with unforced, internal variability. They inferred that more than half of the observed GMSL trend during the 20th century is attributable to anthropogenic forcing.

Slangen et al. (2016) <sup>[[#fn:r357|357]]</sup> reconstructed GMSL from 1900 to 2005 based on CMIP5 model simulations separating individual components of radiative climate forcing and combining the contributions of thermosteric sea level change with glacier and ice sheet mass loss. They found that the naturally caused sea level change, including the long-term adjustment of sea level to climate change preceding 1900, caused 67 ± 23% of observed change from 1900 to 1950, but only 9 ± 18% between 1970 and 2005. Anthropogenic forcing was found to have caused 15 ± 55% of observed sea level change during 1900–1950, but 69 ± 31% during 1970–2005. The sum of all contributions explains only 74 ± 22% of observed GMSL change during the period 1900–2005 considering the mean of the reconstructions of Church and White (2011) <sup>[[#fn:r358|358]]</sup> , Ray and Douglas (2011) , Jevrejeva et al. (2014b) and Hay et al. (2015) . However, the budget could be closed taking into contribution of glaciers that are missing from the global glacier inventory or have already melted (Parkes and Marzeion, 2018 <sup>[[#fn:r361|361]]</sup> ) which were not considered in Slangen et al. (2016) <sup>[[#fn:r362|362]]</sup> .

Based on these multiple lines of evidence, there is ''high confidence'' that anthropogenic forcing ''very likely'' is the dominant cause of observed GMSL rise since 1970.

<span id="projections-of-sea-level-change"></span>