Home
This Title All WIREs
WIREs RSS Feed
How to cite this WIREs title:
WIREs Clim Change
Impact Factor: 4.571

Modelling interdecadal climate variability and the role of the ocean

Full article on Wiley Online Library:   HTML PDF

Can't access this content? Tell your librarian.

To estimate the anthropogenic contribution to climate signals in the recent past and future decades implies a certain degree of confidence in both understanding and simulating natural internal variability at interdecadal time scales. If we are to embark on the challenge of decadal prediction, we must be able to mechanistically attribute events to known processes and phenomena, and reproduce their features and statistics within our models. To date, models have succeeded in reproducing only partially spatial patterns, statistics and climatic impacts of interdecadal modes of variability. Reasons for the partial success and agreement among models are to be attributed to the short observational record, the different and complex flavours of coupling between the many subcomponents of the climate system, and the present inability to resolve all climate processes. At an even more fundamental level, this difficulty is aggravated by the limited understanding of the physical mechanisms involved. Here, we review the proposed mechanisms giving rise to interdecadal climate variability, we discuss the hypotheses explaining the main interdecadal modes of variability, and present an overview on the ability and level of agreement in their simulation by the latest generation of coupled climate models. To achieve any progress, the modeling community should focus on both improving the representation and parameterization of key ocean physical processes and obtaining a firmer grasp on the physical mechanisms generating the variability. Both goals can benefit from process studies, intercomparisons with perturbation experiments to study model's sensitivities, and the use of a hierarchy of climate models. WIREs Clim Change 2017, 8:e441. doi: 10.1002/wcc.441

(a) Seasonal mean global mean surface temperature (GMST, in °C) anomalies, relative to the 20th‐century mean, for the 1910–2015 period. GMST data are a blended product derived from Global Historical Climatology Network‐Monthly (GHCN‐M) data set and International Comprehensive Ocean–atmosphere Data Set (ICOADS) available at http://www.ncdc.noaa.gov/cag. (b) Seasonal mean PDO index, derived as the leading principal component of de‐meaned monthly SST anomalies in the North Pacific Ocean, poleward of 20°N, and for the period 1910–2015. Data sources for this index are the United Kingdom Met Office (UKMO) Historical SST data set for 1900–81, Reynold's Optimally Interpolated (OI) SST V1 for January 1982 to December 2001 and OI SST Version 2 (V2) beginning January 2002 (data available from: http://jisao.washington.edu/pdo/PDO.latest). Decades shaded in light blue mark negative phases of the PDO and corresponding ‘hiatus’ periods. (c) The Atlantic Multidecadal Variability (AMV) index, calculated as the detrended low‐pass filtered SST area‐weighted average over the North Atlantic (0°–70°N) from the Kaplan SST data set for the 1910–2015 period (data available from: http://www.esrl.noaa.gov/psd/data/timeseries/AMO/). Decades shaded in light blue mark negative phases of the AMV. Black curves in (a), (b) and (c) represent decadal variations after applying a gaussian filter to each time series.
[ Normal View | Magnified View ]
Power spectra of atmosphere (thin lines) and ocean (thick lines) temperature for the stochastically forced, one‐dimensional, linear, energy balance Barsugli and Battisti's model in the coupled (solid lines) and uncoupled (dotted lines) case. In the coupled case, thermal variance is increased in both atmosphere and ocean due to the reduced damping of thermal anomalies resulting from decreased surface energy fluxes. The stochastic coupled model is highly sensitive to the strength of the atmospheric response to SST anomalies, or equivalently the strength of coupling. Decreasing the value of the atmospheric response parameter α results in a weaker enhancement of the variance in the system (gray lines).
[ Normal View | Magnified View ]
The AMV as simulated by a suite of CMIP5 historical runs for the period 1900–2005. The AMV is computed as the North Atlantic (0°–60°N, 80°W–0°E) SST anomalies minus global (60°S:60°N) SST anomalies. The AMV pattern is created by regressing global SST anomalies onto the index time series and smoothing with a 9‐point spatial filter. The observed AMV (bottom right panel) is computed from the Hadley Centre Global Sea Ice and Sea Surface Temperature (HadISST). Data can be obtained from the NCAR Climate Analysis Section's Climate Variability Diagnostics Package at https://www2.cesm.ucar.edu/working‐groups/cvcwg/cvdp.
[ Normal View | Magnified View ]
(a) Time‐mean equatorward pycnocline volume transport convergence across 9°N and 9°S. Grey areas represent interior transport (defined as the equatorward transport at 9°N and 9°S, zonally averaged from the eastern boundary to 145°E in the NH and 160°E in the SH). (b) Correlation between interior pycnocline volume transport convergence and tropical SST averaged over 9°N–9°S and 90°–180°W. (c) Interior pycnocline transport standard deviations and (d) SST standard deviations. All values are computed for the period 1950–2000. Observational values, hatched, are taken from Ref The CMIP5 simulations are taken from the historical climate simulation. The last three bars in each plot are not from coupled models but represent values computed from GFDL‐MOM global ocean‐sea ice simulations using the interannually varying CORE atmospheric forcing data sets from 1948 to 2007. The main difference between the three simulations is the horizontal grid spacing, ranging from 0.25° to 2° of nominal resolution.
[ Normal View | Magnified View ]
Time series of global mean surface temperature (GMST) anomalies as in Figure but for the period 1950–2010 and observed subtropical cell (STC) equatorward volume transport convergence anomalies for the period 1954–2000. During STC positive anomalies (before 1976–1977) GMST is in a hiatus period, whereas when STC transport convergence is anomalously weak GMST anomalies grow fast consistent with periods of accelerated warming. The recent hiatus period (starting in 2000) was supposedly accompanied by a strengthening of the STC transport convergence and related upwelling of cool waters. The STC rebound was observed during 1998–2003, but a longer observed time series is missing in the literature. In the period considered, shifts in STC transport convergence anomalies correspond to PDO phases or ‘regime shifts’ (denoted here by the grey vertical bars; see also Figure ).
[ Normal View | Magnified View ]
The PDO as simulated by a suite of CMIP5 historical runs for the period 1900–2005. The PDO is computed as the leading principal component (PC) of North Pacific (20°–70°N, 110°E–100°W) area‐weighted SST anomalies, with global mean SST removed at each timestep prior to the calculation. The PDO pattern is obtained by regressing global SST anomalies onto the normalized PC timeseries. The observed PDO (bottom right panel) is computed from the Hadley Centre Global Sea Ice and Sea Surface Temperature (HadISST). Data can be obtained from the NCAR Climate Analysis Section's Climate Variability Diagnostics Package at https://www2.cesm.ucar.edu/working‐groups/cvcwg/cvdp.
[ Normal View | Magnified View ]
Annual‐mean AMOC time series anomalies for a suite of GFDL (a–h) and NCAR (i and j) models. The AMOC index is defined as the maximum value of the stream function at 27°N (no significant difference was found when the AMOC maximum was taken at 45°N). At the top of each panel the mean and standard deviation are also given. All data are from a pre‐industrial simulation with greenhouse gases fixed at 1860 levels, and thus only natural internal variability is simulated. Note the different length of simulation in each model. Models basic characteristics are given in Table .
[ Normal View | Magnified View ]
Wavelet power spectra for the AMOC at 27°N in (a) CM2.1, (b) CM3 and (c) CM2.5. Thick black lines encircle statistical significant signals at the 95% level and the white line is the ‘cone of influence’ below which results are not reliable. Note the intermittent presence of statistically significant interdecadal variability (20–30 years) in CM2.1 and CM3 and its absence in CM2.5. Note also the common multicentennial variability in all three coupled models. For CM2.1 and CM3 only a segment of the spectrum is plotted.
[ Normal View | Magnified View ]
Power spectral densities for annual‐mean AMOC time series at 27°N (black solid lines in a–f) and the 95% confidence levels based on a first‐order Markov process (black dotted lines in a–f). (a) CM2.5, (b) CM2.5_FLOR, (c) ESM2M, (d) ESM2G, (e) CM3, (f) CM2.1. Blue lines in (a–f) are for the AMOC spectra computed at 30°S. In (g) CM2.1 has a different spectral peak than in (f) because only a 600‐year long segment is used for comparison with CM2M‐A and CM2M‐C runs which are 600‐year long and both derive from the CM2.1 model. CM2M‐A and CM2.1 use the same parameterization of mesoscale eddy‐induced transport whereas CM2M‐C uses a different scheme (see text for details). In (h) two NCAR coupled models with the same ocean but using an atmosphere which differs in horizontal resolution (T42 and T85) are compared.
[ Normal View | Magnified View ]

Browse by Topic

Climate Models and Modeling > Earth System Models

Access to this WIREs title is by subscription only.

Recommend to Your
Librarian Now!

The latest WIREs articles in your inbox

Sign Up for Article Alerts