Understanding the Uncertainty in the 21st CenturyDynamic Sea Level Projections: The Role ofthe AMOCChanglin Chen1,2 , Wei Liu2, and Guihua Wang1

1Department of Atmospheric and Oceanic Sciences & Institute of Atmospheric Sciences, Fudan University, Shanghai,China, 2Department of Earth Sciences, University of California, Riverside, CA, USA

Abstract Climate models show that the largest uncertainties in the 21st century dynamic sea level (DSL)projections are in the high latitudes of the North Atlantic and Southern Oceans. We conduct an intermodelsingular value decomposition analysis and find that the DSL uncertainties in these two oceans are bothintrinsically connected to the uncertainty in the change of the Atlantic meridional overturning circulation(AMOC). We further conduct a freshwater hosing experiment to show that the AMOC decline not onlyaccounts for the dipole pattern in the DSL change in the North Atlantic but also remotely induces a polewardshift in the Southern Hemisphere westerlies that helps build a belted pattern of DSL change in the SouthernOcean.Our results suggest that reducing the intermodel spread in the change of theAMOCcan greatly improvethe consistency of DSL projection among models not only in individual basins but over the global ocean.

Plain Language Summary State-of-art climate models project distinct patterns of regional sealevel change. Most of these patterns arise from a dynamic change in the ocean, which has been namedthe dynamic sea level (DSL) change. However, climate models project different regional DSL changes in the21st century, especially those in the North Atlantic and Southern Oceans. Any connection betweenintermodel uncertainties in the DSL projections in different basins has largely been ignored. We argue herethat the change in the Atlantic meridional overturning circulation (AMOC) is a key factor that controls theintermodel uncertainty of regional DSL projections on a global scale. A weakening AMOC alters the DSLpattern in the North Atlantic and concurrently remotely affects the DSL over the Southern Ocean. Inparticular, we find that a decline in the AMOC could cause a poleward shift of the Southern Hemispherewesterly winds and thereby lead to a zonal belt-like change of the DSL in the Southern Ocean. These resultsimply that improving the intermodel consistency of the projected changes in the AMOCwill greatly improvethe reliability of future sea level projections on a global scale.

1. Introduction

Sea level has been observed to have undergone a rapid rise during the past century (Church & White, 2011;Dangendorf et al., 2017; Ray & Douglas, 2011) and is predicted to rise even more rapidly throughout the cur-rent century (Church et al., 2013). However, this rise is not globally uniform (Bouttes & Gregory, 2014;Church et al., 2013; Pardaens et al., 2011; Slangen et al., 2014; Yin, 2012). Regional sea level change is mainlydriven by alterations in ocean dynamics (ocean density and circulation) and land-ice discharge (Bamber &Riva, 2010; Bouttes & Gregory, 2014; Kopp et al., 2010; Perrette et al., 2013). Particularly, the former contri-bution, also named the dynamic sea level (DSL) change, is considered to be dominant even though land-icechange is predicted to become increasingly important later in the century (Bamber & Riva, 2010; Churchet al., 2013; Horton et al., 2018; Kopp et al., 2014; Perrette et al., 2013).

State-of-art climate models participating in the Coupled Model Intercomparison Project Phase 5 (CMIP5)project distinct patterns of regional DSL change (Perrette et al., 2013; Yin, 2012). However, these models dis-agree on some of the details in the DSL change patterns, especially for those in the North Atlantic andSouthern Oceans (Bouttes et al., 2012; Little et al., 2015; Yin, 2012). The intermodel spread in the DSL projec-tions can be substantial when compared to the global mean sea level rise (Church et al., 2013) and hencebecome a serious impediment to sea level projection. To reduce the uncertainty in DSL and improve the relia-bility of sea level projection, it is important to understand the physical mechanisms governing the intermodelspread in regional DSL.

CHEN ET AL. 210

RESEARCH LETTER10.1029/2018GL080676

Key Points:• The intermodel uncertainties in

regional dynamic sea level (DSL)projections are large over both theNorth Atlantic and Southern Oceans

• The Atlantic meridional overturningcirculation (AMOC) is the key tounderstanding these uncertaintieson a global scale

• An AMOC decline can remotelycause a poleward shift of SouthernHemisphere westerlies and thuscontrol the DSL over the SouthernOcean

Supporting Information:• Supporting Information S1

Correspondence to:C. Chen and W. Liu,chencl@fudan.edu.cn;wei.liu@ucr.edu

Citation:Chen, C., Liu, W., & Wang, G. (2019).Understanding the uncertainty in the21st century dynamic sea levelprojections: The role of the AMOC.Geophysical Research Letters, 46,210–217. https://doi.org/10.1029/2018GL080676

Received 27 SEP 2018Accepted 4 DEC 2018Accepted article online 10 DEC 2018Published online 3 JAN 2019

©2018. American Geophysical Union.All Rights Reserved.

In previous studies, several mechanisms were proposed to explain the intermodel uncertainty in theregional DSL projections for individual basins. For example, over the Atlantic, the magnitude of regionalDSL rise was found to be significantly correlated with the degree of weakening of the Atlantic meridio-nal overturning circulation (AMOC) projected by each model (Pardaens et al., 2011). In the SouthernOcean, it was suggested that the poleward-intensification of the westerly wind stress accounts for mostof the intermodel disagreement in the projected change of regional DSL (Bouttes et al., 2012). However,the mechanisms driving the intermodel uncertainty in regional DSL projections were examined basin-by-basin and interpreted separately ignoring any intrinsic connections in the DSL responses betweenbasins. The intermodel uncertainty of regionally varying DSL change had rarely been examined on aglobal scale.

In this study, we explore the key factor controlling the uncertainty in the projection of regional DSLon a global scale. We perform a singular value decomposition (SVD) analysis of CMIP5 model projec-tions and find that intermodel uncertainties in the regional DSL projections are commonly linked tothe uncertainty in the AMOC response. The mechanism by which the AMOC affects the regionalDSL pattern is further elaborated in a freshwater hosing experiment. Here it is worth noting thatthe aim of the freshwater hosing experiment is different from previous studies. In previous hosingexperiments, a weakening of the AMOC was suggested to be responsible for the pattern of the DSLchange in the North Atlantic (Bouttes & Gregory, 2014; Hu et al., 2011; Levermann et al., 2005;Yin et al., 2009, 2010) and partially responsible for the DSL change in the Southern Ocean (Yinet al., 2010). However, the dynamical processes through which the AMOC affects DSL changes—espe-cially those over the Southern Ocean—were not explored in detail. Consequently, understanding thephysical processes that couple the interbasin response serve as the focus in our analysis of freshwaterhosing experiment.

2. Data, Methods, and Model2.1. CMIP5 Models and Simulations

We use historical and representative concentration pathways 4.5 (RCP4.5) simulations (Taylor et al.,2012) from 30 available CMIP5 models (Table S1 in the supporting information). Only one memberrun (r1i1p1) is selected from each model to ensure equal model weighting in the intermodel analysis.DSL is the sea level deviation from the global mean (its global average is zero). Some models showDSL drifts in their preindustrial control runs in which greenhouse gas concentrations are held constant(Sen Gupta et al., 2013). Therefore, a linear drift correction is applied to all climate models at each gridpoint for the simulations. For each model, variables are interpolated onto a common 1° × 1° latitude-longitude grid prior to analysis. For each variable, the projected change is calculated by subtracting the25-year mean of 1976–2000 of the historical simulation from the 2076–2100 mean of the RCP4.5 runs.For reference, we also make a parallel analysis under the RCP8.5 scenario (Figure S1 in the supportinginformation) to show that our results are not scenario-dependent.

2.2. Methods

We calculate the AMOC streamfunction (ψ) from the meridional velocity (v) of the model outputs as

ψ y; zð Þ ¼ ∫0z ∫xexwv x; y; z0ð Þdxdz0

where x, y, and z are the zonal, meridional, and vertical coordinates and xe and xw are the eastern and wes-tern boundaries of the Atlantic Ocean. Also, we perform an intermodel SVD analysis, which is similar to atraditional SVD analysis (Wallace et al., 1992) but with the space-model field instead of the traditional space-time field. Intermodel SVD is usually applied to two data fields together to identify pairs of coupled spatialpatterns, which explain as much as possible the covariance between the two variables (Long et al., 2016;Long & Xie, 2015; Shiogama et al., 2011; Wang et al., 2014). The detailed description of the intermodelSVD method can be found in Abe et al. (2011) and Shiogama et al. (2011). In this study, we apply the inter-model SVD to the changes in the DSL and AMOC to extract the statistically correlated modes. Note that themultimodel ensemble means of the DSL and AMOC are removed before the SVD analysis is performed.

10.1029/2018GL080676Geophysical Research Letters

CHEN ET AL. 211

2.3. CESM and the Freshwater Hosing Experiment

We use the Community Earth SystemModel version 1.2.2 (CESM; Hurrell et al., 2013), which is maintainedby the National Center for Atmospheric Research (NCAR). The f19gx1v6 configuration used here has afinite-volume dynamical core with a nominal 2° horizontal atmosphere and land grid (1.9° latitude × 2.5°longitude) with 26 vertical layers in the atmosphere and a nominal 1° horizontal ocean and ice grid with60 vertical layers in the ocean. The freshwater hosing experiment is conducted based upon the preindustrialCESM control run by adding freshwater to the North Atlantic between 50 to 70°N at a rate of 0.5 Sv andmaintaining that input throughout the 150-year integration. Because the added freshwater inhibits deepconvection and deep water formation in the North Atlantic, the AMOC rapidly weakens (within 50 years)from 18 to 1–2 Sv and remains in this weakened state throughout the simulation (Figure S2). The climaticresponse to the AMOCweakening is then calculated as the difference between the average of the last 50 yearsof the freshwater hosing experiment and the average of the 50 years of the control run.

3. Results3.1. Effect of the Change in the AMOC on the Intermodel DSL Uncertainty

We first examine the ensemble mean and intermodel spread in the DSL change among CMIP5 models at theend of the 21st century. The ensemble-mean DSL changes exhibit distinct spatial patterns in high latitudes: ameridional dipole pattern in the North Atlantic Ocean and a belted pattern in the Southern Ocean(Figure 1a, also see Perrette et al., 2013; Saenko et al., 2005; Yin, 2012). The dipole pattern is characterizedby larger sea level rise north of the Gulf Stream and smaller sea level rise to the south, while the belted pat-tern is characterized by smaller sea level rise south of 50°S and larger sea level rise to the north. In addition,the intermodel spread of the DSL change is large (Figure 1b) and comparable in magnitude to the ensemblemean (Figure 1a), indicative of large intermodel uncertainty in DSL projections. This DSL uncertainty is lar-gest in the high latitudes of the North Atlantic and Southern Oceans where the ensemble-mean DSL changesare also large.

Figure 1. Multimodel (a) ensemble mean and (b) spread (one standard deviation) of the Coupled Model IntercomparisonProject Phase 5 (CMIP5) dynamic sea level change under the representative concentration pathway (RCP) 4.5 scenario.The dynamic sea level change is shown relative to the global mean sea level rise.

10.1029/2018GL080676Geophysical Research Letters

CHEN ET AL. 212

We have been seeking a key factor that controls the intermodel uncertainty of regional DSL projections on aglobal scale, and it appears that the AMOC plays this role. To verify this, we perform an intermodel SVD ana-lysis of the changes in the DSL and AMOC streamfunction under the RCP4.5 scenario (Figure 2). The resultsshow that the DSL pattern of the first SVDmode possesses a dipole pattern in the North Atlantic Ocean and abelted pattern in the Southern Ocean (Figure 2a), which resembles the pattern of the ensemble-mean DSLalterations (Figure 1a). This DSL mode corresponds to a weakening AMOC mode, as indicated by the nega-tive values of the AMOC streamfunction in the upper 2,000 m (Figure 2b). As such, this first SVDmode sug-gests that the DSL uncertainties over the North Atlantic and Southern Oceans are both connected throughthe variations in the AMOC. Statistically, this first coupled mode of the intermodel variance accounts for68.7% of total covariance, meaning that the majority of the coupled variability between DSL and theAMOC is captured by this mode. The correlation coefficient (0.84) between the expansion coefficients ofDSL and the AMOC is statistically significant at the 99% confidence level. Therefore, the SVD results suggesta strong and significant relationship between the changes of DSL and the AMOC. In addition, we perform alinear regression of the DSL alterations upon changes in the AMOC strength for CMIP5 models (Figure S3).We find that the regressionmap also shows a robust dipole pattern in the North Atlantic and a belted patternin the Southern Ocean, which indicates that a larger (smaller) reduction in the AMOC will correspond to amore (less) pronounced change in the DSL pattern.

The results from the SVD and regression analyses show that the impact of the change in the AMOC onDSL change is not limited to the North Atlantic but extends to the Southern Ocean. To demonstrate theexistence of this remote AMOC effect, we compose the changes in the zonal mean DSL, surface winds,and sea level pressures based on the degree of weakening of the AMOC in CMIP5 models (Figure S4).Compared to models with smaller AMOC weakening, models with larger AMOC weakening project astronger intensification and larger poleward shifts of the westerly winds (Figure S4b) and sea level pres-sures (Figure S4c) over the Southern Ocean, which match the change in DSL (Figure S4a). This resultsuggests that change in the strength of the AMOC plays an important role in altering Southern

Figure 2. Spatial patterns of changes in the (a) dynamic sea level (DSL) and (b) Atlantic meridional overturning circula-tion (AMOC) for the first inter-model singular value decomposition mode (accounting for 68.7% of total covariance) underthe representative concentration pathway 4.5 scenario. (c) Scatter plot and least squares regression line of the firstmode expansion coefficients of DSL versus AMOC for the 30 Coupled Model Intercomparison Project Phase 5 (CMIP5)models (see Table S1 for model list). Both expansion coefficients have been normalized by their own standard deviations.

10.1029/2018GL080676Geophysical Research Letters

CHEN ET AL. 213

Ocean winds and therefore impacts the Southern Ocean DSL. However, the role of the AMOC cannot bedemonstrated solely from the statistical analysis of CMIP5 model outputs alone. To examine the physicalprocesses by which the variations in the strength of the AMOC affect regional DSL on a global scale, wecarry out a freshwater hosing experiment in which anthropogenic forcing is excluded to isolate theAMOC change as the only factor driving changes of the climate system.

3.2. DSL Response in the Freshwater Hosing Experiment

Figure 3a shows the DSL response to the weakening of the AMOC in the freshwater hosing experiment. Thechange in the AMOC has a global impact on the change in DSL, with regional changes most pronounced inthe North Atlantic and Southern Oceans. In the North Atlantic, the decline of the AMOC induces an anom-alous heat flux and ocean circulation, which creates a strong north-south sea level gradient around 40°N(Bouttes & Gregory, 2014; Pardaens et al., 2011; Yin et al., 2009, 2010). This dipole pattern well reproducesthe uncertainty pattern in the CMIP5 DSL in the North Atlantic from the previous SVD analysis (Figure 2a).

In the Southern Ocean, we find that the weakening of the AMOC causes an anomalously low (high) sea levelaround 65°S (50°S; Figures 3a and 3b). This belted pattern in the Southern Ocean DSL also captures themainstructure of the first DSL mode in the SVD analysis (Figure 2a) and is also consistent with the results fromYin et al. (2010). Dynamically, the weakening of the AMOC reduces northward ocean heat transport, whichleads to strong tropospheric cooling in the Northern Hemisphere and a mild tropospheric warming at mid-dle and high latitudes of the Southern Hemisphere. The strongest warming in the Southern Hemisphere islocated around 50°S (Figure 4a), which increases the poleward temperature gradient south of 50°S butdecreases it to the north. Corresponding to these changes in the temperature gradient, the winds on theequatorward flank of the upper level subtropical westerlies weaken while those on its poleward flank inten-sify (Figure 4b), thus indicative of poleward shifts of the Southern Hemisphere westerly jet (Chavaillaz et al.,2013; Lu et al., 2008; Wilcox et al., 2012) and the Ferrell cell (Figure 4c). Similar poleward shifts are also evi-dent in the surface winds (Figures S5 and S6) and sea level pressures (Figure S5).

Consistent with this poleward shift of the westerly winds, the zonal mean wind stress curl anomaly over theSouthern Ocean possesses a dipole structure, with positive values north of 60°S and negative values to the

Figure 3. Dynamic sea level (DSL) and wind stress responses in the freshwater hosing experiment. Spatial patterns of (a) DSL and (c) surface wind stress. The zonalmean (b) DSL and (d) surface wind stress (red) and wind stress curl (black).

10.1029/2018GL080676Geophysical Research Letters

CHEN ET AL. 214

south (Figure 3d). This change in the surface wind forcing displaces the Deacon cell and in turn the residualMOC in the Southern Ocean (Figure S7), leading to anomalous positive (negative) heat storage north (south)of 50°S (Bouttes et al., 2012; Fyfe et al., 2007; Liu et al., 2018; Morrison et al., 2016; Yin et al., 2010). As aresult of the vertically integrated effect of this anomalous heat storage, the sea level change forms a beltedpattern over the Southern Ocean (Figures 3a and 3b).

4. Summary

We investigate the intermodel uncertainty in the regional DSL projection on a global scale based on 30CMIP5 models and focus on the role of the AMOC. We find that the largest uncertainties occur in the highlatitudes of the North Atlantic and Southern Oceans, which lower the confidence of future DSL projection.An intermodel SVD analysis shows that the DSL uncertainties in the North Atlantic and the Southern

Figure 4. Meridional structure (color shading) of the zonal mean (a) temperature and (b) zonal winds, and (c) meridionalstreamfunction in the freshwater hosing experiment. The overlapping black contour lines in each panel denote the cli-matology of the control run.

10.1029/2018GL080676Geophysical Research Letters

CHEN ET AL. 215

Oceans are intrinsically connected, both being governed by the magnitude of the change in the AMOC.Further, a CESM freshwater hosing experiment shows that a weakening of the AMOC can cause an anom-alous DSL dipole in the North Atlantic and remotely induce a belted DSL change over the Southern Ocean.Specifically, the weakening of the AMOC drives an increase (decrease) in the poleward tropospheric tem-perature gradient south (north) of 50°S. These changes in tropospheric temperature gradient cause a pole-ward shift of the westerly winds in the Southern Hemisphere, which in turn, generate a belted changepattern in DSL in the Southern Ocean. Our results suggest that the AMOC is likely the key to understandingthe uncertainty of regional DSL projection on a global scale. Consequently, reducing the uncertainty in theprojection of the strength of the AMOC will greatly improve the fidelity of DSL projections.

ReferencesAbe, M., Shiogama, H., Nozawa, T., & Emori, S. (2011). Estimation of future surface temperature changes constrained using the future-

present correlated modes in inter-model variability of CMIP3 multimodel simulations. Journal of Geophysical Research, 116, D18104.https://doi.org/10.1029/2010JD015111

Bamber, J., & Riva, R. (2010). The sea level fingerprint of recent ice mass fluxes. The Cryosphere, 4(4), 621–627. https://doi.org/10.5194/tc-4-621-2010

Bouttes, N., & Gregory, J. (2014). Attribution of the spatial pattern of CO2-forced sea level change to ocean surface flux changes.Environmental Research Letters, 9(3), 034004. https://doi.org/10.1088/1748-9326/9/3/034004

Bouttes, N., Gregory, J. M., Kuhlbrodt, T., & Suzuki, T. (2012). The effect of windstress change on future sea level change in the SouthernOcean. Geophysical Research Letters, 39, L23602. https://doi.org/10.1029/2012GL054207

Chavaillaz, Y., Codron, F., & Kageyama, M. (2013). Southern westerlies in LGM and future (RCP4.5) climates. Climate of the Past, 9(2),517–524. https://doi.org/10.5194/cp-9-517-2013

Church, J., Clark, P. U., Cazenave, A., Gregory, J. M., Jevrejeva, S., Levermann, A., Merrifield, M. A., et al. (2013). Sea level change. InT. Stocker, et al. (Eds.), Climate change 2013: The physical science basis. Contribution of Working Group I to the Fifth AssessmentReport of the Intergovernmental Panel on Climate Change (pp. 1137–1216). Cambridge, UK and New York: Cambridge University Press.

Church, J. A., & White, N. J. (2011). Sea-level rise from the late 19th to the early 21st century. Surveys in Geophysics, 32(4–5), 585–602.https://doi.org/10.1007/s10712-011-9119-1

Dangendorf, S., Marcos, M., Wöppelmann, G., Conrad, C. P., Frederikse, T., & Riva, R. (2017). Reassessment of 20th century global meansea level rise. Proceedings of the National Academy of Sciences of the United States of America, 114(23), 5946–5951. https://doi.org/10.1073/pnas.1616007114

Fyfe, J. C., Saenko, O. A., Zickfeld, K., Eby, M., & Weaver, A. J. (2007). The role of poleward-intensifying winds on Southern Oceanwarming. Journal of Climate, 20(21), 5391–5400. https://doi.org/10.1175/2007JCLI1764.1

Horton, B. P., Kopp, R. E., Garner, A. J., Hay, C. C., Khan, N. S., Roy, K., & Shaw, T. A. (2018). Mapping sea-level change in time, space, andprobability. Annual Review of Environment and Resources, 43, 13.1–13.41. https://doi.org/10.1146/annurev-environ-102017-025826

Hu, A., Meehl, G. A., Han, W., & Yin, J. (2011). Effect of the potential melting of the Greenland Ice Sheet on the meridional overturningcirculation and global climate in the future. Deep-Sea Research Part II: Topical Studies in Oceanography, 58(17-18), 1914–1926. https://doi.org/10.1016/j.dsr2.2010.10.069

Hurrell, J. W., Holland, M. M., Gent, P. R., Ghan, S., Kay, J. E., Kushner, P. J., Lamarque, J. F., et al. (2013). The Community Earth SystemModel: A framework for collaborative research. Bulletin of the American Meteorological Society, 94(9), 1339–1360. https://doi.org/10.1175/BAMS-D-12-00121.1

Kopp, R. E., Horton, R. M., Little, C. M., Mitrovica, J. X., Oppenheimer, M., Rasmussen, D. J., Strauss, B. H., et al. (2014). Probabilistic 21stand 22nd century sea-level projections at a global network of tide-gauge sites. Earth’s Future, 2(8), 383–406. https://doi.org/10.1002/2014EF000239

Kopp, R. E., Mitrovica, J. X., Griffies, S. M., Yin, J., Hay, C. C., & Stouffer, R. J. (2010). The impact of Greenland melt on local sea levels: Apartially coupled analysis of dynamic and static equilibrium effects in idealized water-hosing experiments. Climatic Change, 103(3-4),619–625. https://doi.org/10.1007/s10584-010-9935-1

Levermann, A., Griesel, A., Hofmann, M., Montoya, M., & Rahmstorf, S. (2005). Dynamic sea level changes following changes in thethermohaline circulation. Climate Dynamics, 24, 347–354.

Little, C. M., Horton, R. M., Kopp, R. E., Oppenheimer, M., & Yip, S. (2015). Uncertainty in twenty-first-century CMIP5 sea level projec-tions. Journal of Climate, 28(2), 838–852. https://doi.org/10.1175/jcli-d-14-00453.1

Liu, W., Lu, J., Xie, S.-P., & Fedorov, A. (2018). Southern Ocean heat uptake, re-distribution, and storage in a warming climate: The role ofmeridional overturning circulation. Journal of Climate, 31(12), 4727–4743. https://doi.org/10.1175/JCLI-D-17-0761.1

Long, S.-M., & Xie, S.-P. (2015). Intermodel variations in projected precipitation change over the North Atlantic: Sea surface temperatureeffect. Geophysical Research Letters, 42, 4158–4165. https://doi.org/10.1002/2015GL063852

Long, S.-M., Xie, S.-P., & Liu, W. (2016). Uncertainty in tropical rainfall projections: Atmospheric circulation effect and the ocean coupling.Journal of Climate, 29(7), 2671–2687. https://doi.org/10.1175/JCLI-D-15-0601.1

Lu, J., Chen, G., & Frierson, D. M. W. (2008). Response of the zonal mean atmospheric circulation to El Nino versus global warming.Journal of Climate, 21(22), 5835–5851. https://doi.org/10.1175/2008JCLI2200.1

Morrison, A. K., Griffies, S. M., Winton, M., Anderson, W. G., & Sarmiento, J. L. (2016). Mechanisms of Southern Ocean heat uptake andtransport in a global eddying climate model. Journal of Climate, 29(6), 2059–2075. https://doi.org/10.1175/JCLI-D-15-0579.1

Pardaens, A. K., Gregory, J. M., & Lowe, J. A. (2011). A model study of factors influencing projected changes in regional sea level over thetwenty-first century. Climate Dynamics, 36(9-10), 2015–2033. https://doi.org/10.1007/s00382-009-0738-x

Perrette, M., Landerer, F., Riva, R., Frieler, K., & Meinshausen, M. (2013). A scaling approach to project regional sea level rise and itsuncertainties. Earth System Dynamics, 4(1), 11–29. https://doi.org/10.5194/ esd-4-11-2013

Ray, R. D., & Douglas, B. C. (2011). Experiments in reconstructing twentieth-century sea levels. Progress in Oceanography, 91(4), 496–515.https://doi.org/10.1016/ j.pocean.2011.07.021

Saenko, O. A., Fyfe, J. C., & England, M. H. (2005). On the response of the oceanic wind-driven circulation to atmospheric CO2 increase.Climate Dynamics, 25(4), 415–426. https://doi.org/10.1007/s00382-005-0032-5

10.1029/2018GL080676Geophysical Research Letters

CHEN ET AL. 216

AcknowledgmentsThis work is supported by the NationalNatural Science Foundation of China(41576022 and 91428206) and theProgram of ShanghaiAcademic/Technology ResearchLeader (17XD1400600). W. L. is sup-ported by the National ScienceFoundation (OPP-1741841), the startupfund from the Department of EarthSciences, University of California,Riverside and the Regents’ FacultyFellowship. All CMIP5 data are avail-able at http://pcmdi9.llnl.gov/. Theauthors would like to thank the tworeviewers for their constructive andvaluable comments.

Sen Gupta, A., Jourdain, N. C., Brown, J. N., & Monselesan, D. (2013). Climate drift in the CMIP5 models. Journal of Climate, 26(21),8597–8615. https://doi.org/10.1175/JCLI-D-12-00521.1

Shiogama, H., Emori, S., Hanasaki, N., Abe, M., Masutomi, Y., Takahashi, K., & Nozawa, T. (2011). Observational constrains indicate riskof drying in the Amazon basin. Nature Communications, 2(1), 253. https://doi.org/10.1038/ncomms1252

Slangen, A. B. A., Carson, M., Katsman, C. A., van de Wal, R. S. W., Köhl, A., Vermeersen, L. L. A., & Stammer, D. (2014). Projectingtwenty-first century regional sea-level changes. Climatic Change, 124(1-2), 317–332. https://doi.org/10.1007/s10584-014-1080-9

Taylor, K. E., Stouffer, R. J., & Meehl, G. A. (2012). An overview of CMIP5 and the experiment design. Bulletin of the AmericanMeteorological Society, 93(4), 485–498. https://doi.org/10.1175/BAMS-D-11-00094.1

Wallace, J. M., Smith, C., & Bretherton, C. S. (1992). Singular value decomposition of wintertime sea surface temperature and 500-mbheight anomalies. Journal of Climate, 5(6), 561–576. https://doi.org/10.1175/1520-0442(1992)005<0561:SVDOWS>2.0.CO;2

Wang, C., Zhang, L., Lee, S. -K., Wu, L., & Mechoso, C. R. (2014). A global perspective on CMIP5 climate model biases. Nature ClimateChange, 4(3), 201–205. https://doi.org/10.1038/nclimate2118

Wilcox, L. J., Charlton-Perez, A. J., & Gray, L. J. (2012). Trends in austral jet position in ensembles of high- and low-top CMIP5 models.Journal of Geophysical Research, 117, D13115. https://doi.org/10.1029/2012JD017597

Yin, J. (2012). Century to multi-century sea level rise projections from CMIP5 models. Geophysical Research Letters, 39, L17709. https://doi.org/10.1029/ 2012GL052947

Yin, J., Griffies, S. M., & Stouffer, R. J. (2010). Spatial variability of sea level rise in twenty-first century projections. Journal of Climate,23(17), 4585–4607. https://doi.org/10.1175/2010JCLI3533.1

Yin, J., Schlesinger, M. E., & Stouffer, R. J. (2009). Model projections of rapid sea-level rise on the northeast coast of the United States.Nature Geoscience, 2(4), 262–266. https://doi.org/10.1038/ngeo462

10.1029/2018GL080676Geophysical Research Letters

CHEN ET AL. 217