The effect of model bias on Atlantic freshwater transport and implications for AMOC bi-stability

J.V. Mecking; S.S. Drijfhout; L.C. Jackson; M.B. Andrews

doi:10.1080/16000870.2017.1299910

The effect of model bias on Atlantic freshwater transport and implications for AMOC bi-stability

Tellus

Volume 69 (2017): Issue 1

By: J.V. Mecking, S.S. Drijfhout, L.C. Jackson and M.B. Andrews

Open Access

|Jan 2017

Figures & Tables

Table 1.

Table containing a list of all the markers used to identify the models, model name, the institute that conducted the experiments, the values of AMOC at 26.5°N, $M_{o v}$ at 34°S and $M_{o v}$ at 26.5°N. The values with a grey shading for $M_{o v}$ fall within the range of observational estimates. The stars behind the model names indicate the models for which freshwater fluxes were available in the CMIP5 model database.

Institute IDAMOC 26.5?N (Sv)Mov 34?S (Sv)Mov 26.5?S (Sv)

ACCESS1-0*CSIRO-BOM15.97-0.038-0.695

ACCESS1-3*CSIRO-BOM17.26-0.004-0.826

bcc-csm1-1BCC15.85+0.099-0.388

bcc-csm1-1-mBCC19.28+0.112-0.519

BNU-ESMGCESS24.03+0.513-0.741

CanESM2CCCMA13.89+0.153-0.589

CCSM4NCAR18.33+0.198-0.730

CESM1-BGCNSF-DOE-NCAR19.31+0.188-0.777

CESM1-CAM5NSF-DOE-NCAR19.39+0.240-0.677

CESM1-CAM5-1-FV2NSF-DOE-NCAR18.87+0.303-0.627

CESM1-FASTCHEMNSF-DOE-NCAR18.09+0.194-0.733

CESM1-WACCMNSF-DOE-NCAR20.42+0.301-0.713

CMCC-CESM*CMCC13.46-0.233-0.544

CMCC-CM*CMCC12.01-0.130-0.474

CMCC-CMS*CMCC12.53-0.224-0.506

CNRM-CM5*CNRM-CERFACS10.63+0.002-0.487

CNRM-CM5-2*CNRM-CERFACS13.76+0.039-0.564

CSIRO-Mk3-6-0*CSIRO-QCCCE19.54+0.342-0.651

EC-EARTH*EC-EARTH13.77+0.019-0.580

FGOALS-g2LASG-CESS24.52+0.721-0.706

FIO-ESMFIO14.04-0.160-0.645

GFDL-CM3*NOAA GFDL20.5+0.104-0.739

GFDL-ESM2G*NOAA GFDL18.16+0.176-0.733

GFDL-ESM2M*NOAA GFDL20.31+0.150-0.716

GISS-E2-H-CCNASA GISS9.49-0.103-0.275

GISS-E2-RNASA GISS16.54-0.001-0.426

GISS-E2-R-CCNASA GISS17.91-0.000-0.487

HadGEM2-AONIMR/KMA15.67+0.177-0.439

HadGEM2-CC*MOHC16.34+0.182-0.443

HadGEM2-ES*MOHC14.49+0.143-0.387

inmcm4INM15.4+0.230-0.216

IPSL-CM5A-LR*IPSL9.27-0.112-0.343

IPSL-CM5A-MR*IPSL11.93-0.062-0.483

MIROC-ESM*MIROC11.9-0.030-0.530

MIROC-ESM-CHEM*MIROC11.62-0.019-0.506

MIROC5*MIROC15.89-0.036-0.648

MPI-ESM-LR*MPI-M19.5+0.002-0.554

MPI-ESM-MR*MPI-M15.61-0.117-0.584

MPI-ESM-P*MPI-M18.09-0.014-0.515

MRI-CGCM3*MRI10.36-0.006-0.327

MRI-ESM1*MRI10.1-0.023-0.329

NorESM1-M*NCC30.73+0.328-0.480

NorESM1-ME*NCC31.15+0.359-0.432CMIP5 mean16.65+0.092-0.553

SODA18.58-0.093-0.611

GloSea5*MOHC15.87-0.086-0.620

HadGEM3-GC2*MOHC17.61+0.020-0.653

The range of the values of a) $M_{o v}$ at 34°S and b) $M_{o v}$ at 26.5°N for the historical mean (blue), salinity bias corrected (orange) and for 26.5°N velocity bias corrected (purple). For the CMIP5 models the range of model values (vertical line), multi-model mean (dot) and the range of the models falling in the middle half (shaded box). The shading indicates the observational estimates of $M_{o v}$ at 34°S as estimated by Garzoli et al. (2013) (note that other observational estimates fall within that range) and $M_{o v}$ at 26.5°N as computed by McDonagh et al. (2015).

Scatter plots of the strength of the AMOC at 26.5°N and a) $M_{o v}$ at 34°S, b) $M_{o v}$ at 26.5°N, c) salinity bias corrected $M_{o v}$ at 34°S and d) salinity bias corrected $M_{o v}$ at 26.5°N. For panels a) and c) shading indicates range of values of $M_{o v}$ at 34°S as estimated by Garzoli et al. (2013) with the vertical black line indicating 0 Sv and in panels b) and d) shading indicates the range of values of $M_{o v}$ at 26.5°N as estimated by McDonagh et al. (2015). In all the panels the horizontal black line indicates the estimated value of AMOC strength at 26.5°N by Smeed et al. (2014). In the title correlation values that are significant at the 99% level are indicated in red. The dashed lines on the figures show the lines of best fit using a least squares estimate, one line minimises the error in the vertical direction and the other in the horizontal direction. The lines of best fit are computed using the CMIP5 models only.

Top panels show the mutli-model mean of (a) $M_{o v}$ and (b) $M_{a z}$ of the CMIP5 models with the thin lines indicating the range of the model values of both the historical mean (blue) and the salinity bias corrected mean (orange) with the range of the model values indicated by the thinner lighter lines of the same colour. Estimates of $M_{o v}$ from observations (black): triangle based on McDonagh and King (2005); cross McDonagh et al. (2010); stars Bryden et al. (2011); circles Garzoli et al. (2013) with thevertical line representing the range in estimates; and diamond McDonagh et al. (2015). The bottom panels show (c) the multi-model mean salinity bias from the CMIP5 models for the zonal mean Atlantic salinity bias and (d) the zonal mean Atlantic salinity in the EN3 data.

The zonal values of $M_{o v}$ for a) SODA and b) GloSea5 for the historical mean (black), salinity bias corrected (grey), CMIP5 multi-model mean (blue) and salinity bias corrected multi-model mean (orange). With black markers indicating the same observational estimates as used in Fig. 3.

Correlation of the strength of the AMOC at 26.5°N with $M_{o v}$ (blue), the salinity bias corrected $M_{o v}$ (orange) and the difference between the $M_{o v}$ and salinity bias corrected $M_{o v}$ (the correction applied to $M_{o v}$ ) (yellow). Everything outside the shaded region is significant at the 99% level.

Integrated upward surface freshwater fluxes starting from the Bering Strait to 34°S for the CMIP5 multi-model mean (blue line), the range of all the CMIP5 models (light blue line) and the range of the models falling in the middle half (light blue shading). Black markers indicate the observational estimates and black lines show the error bars on these estimates, where available. Only the CMIP5 models which have surface freshwater fluxes available were used in this computation (See Table 1).

(a) Correlations between AMOC at 26.5°N, $M_{o v}$ at 34°S, integrated freshwater fluxes from Bering Strait to 34°S ( $P - E + R$ ), South Atlantic SSS bias (SSS 34°S-EQ), North Atlantic SSS bias (SSS 26.5-65°N) and salinity bias in the full depth Atlantic from 34°S-65°N (Atl. Salinity). (b–d) Spatial patterns of correlations between AMOC at 26.5°N and (b) Atlantic zonal mean salinity bias, (c) $P - E + R$ , and (d) SSS bias. Statistically significant correlations at the 99% level are circled in white (b,c,d) or written in white text (a). Only the CMIP5 models which have surface freshwater fluxes available were used to compute $P - E + R$ (See Table 1).

(a) The zonal values of $M_{o v}$ for HadGEM3-GC2 for the historical mean (black) and the salinity bias corrected values (grey), and for the CMIP5 multi-model mean (blue) and salinity bias corrected multi-model mean (orange). Black markers indicate the same observational estimates as used in Fig. 3c. The zonal mean Atlantic salinity bias in HadGEM3-GC2.

Same as Fig. 9 but showing the mean salinity bias in the AAIW formation region (130°W-45°W, 60°S-50°S, upper 1000 m) and (a) $M_{o v}$ at 34°S and (b) AMOC at 26.5°N.

References

Authors

Metrics

Articles in this issue

DOI: https://doi.org/10.1080/16000870.2017.1299910 | Journal eISSN: 3035-9554

Journal RSS Feed

Language: English

Page range: 1299910 - 1299910

Submitted on: Dec 22, 2016

Accepted on: Feb 16, 2017

Published on: Jan 1, 2017

Published by: Stockholm University Press

In partnership with: Paradigm Publishing Services

Keywords:

AMOC,

AMOC collapse,

abrupt climate change,

CGCM,

CMIP5

© 2017 J.V. Mecking, S.S. Drijfhout, L.C. Jackson, M.B. Andrews, published by Stockholm University Press
This work is licensed under the Creative Commons Attribution 4.0 License.

Volume 69 (2017): Issue 1

The effect of model bias on Atlantic freshwater transport and implications for AMOC bi-stability

Figures & Tables

Table 1.

Figure 1.

Figure 2.

Figure 3.

Figure 4.

Figure 5.

Figure 6.

Figure 7.

Figure 8.

Figure 9.

Paradigm

My account