Blocking index and Arctic Oscillation in decadal experiments with EC-Earth

Rossby Centre scientists have been analyzing the ability of our modelling system ability to reproduce Arctic Oscillation decadal variability. EC-Earth model version 2.3 (Hazeleger at al, 2013) was used in CMIP5 configuration and forcing setup for an extended decadal hindcast experiment. This consists of an ensemble of 5 members each with 46 decadal simulations, starting yearly on 1st November for the period 1960-2005. The coupled model was initialised using anomaly method for ocean and ice and the 5 members are obtained perturbing both: ocean and ice initial state.

Initialised decadal forecast skill during the first years is usually associated with higher skill coming from the internally generated component. Its main natural modes of variability are high teleconnection areas affecting the regional climate over various areas in most sensitive parameters (as temperature, precipitation) mainly through circulation changes and physics-dynamics interaction. Winter storms resulting from intense extra-tropical cyclones particularly causing severe weather anomalies over North_America and Eurasia are shown to be related to AO phase that controls tropospheric blocking variability over the Pacific basin and the Atlantic/European sector.

A preliminary analysis of our modeling system ability in reproducing AO decadal variability in unforced and forced climate is ongoing using the ensemble of simulations produced.

We compute decadal model skills for AO index, BLI index (Blocking index) and 2m temperature (T2m) - against ERA data and model climatology (uninitialised hindcast -Ctrl). We compute AO index as the PC time series of the main loading mode of 1000hPa pressure anomalies over 50N-90N, 10 years round. The BLI uses Tibaldi and Molteni (1990) modified to account for model variability in defining the latitudinal belts; here only persistent blocks are considered.

RC News Mihaela Fig 1
Figure 1: Time series of decadal normalised BLI for: model (black), Ctrl (yellow) and ERA (green): a 5 year running mean was applied for model data. The x axis shows starting dates for decades.

The time correlation between model and ERA BLI decadal skill over 46 decades is significant at 0.95 level (r=0.31). Ctrl and ERA appear uncorrelated (Figure 1) that implies that ensemble initialization may supply skill for longer-period oscillation of decadal BLI (AO).

Moreover, the modelled AO index is positively (statistically significant) correlated with the observed AO index on decadal scale, in those cases when the AO phase is initially accurate, while worse correlations correspond to wrong phasing at initial time, as shown in Figure 2a. That indicates AO predictability in our system -or feature that should be improved at initial time to get a good decadal AO skill.

Then, besides known regional AO teleconnected-areas that would gain from that, Figure 2b. shows an interesting feature connecting AO prediction to global skill in 2m temperature: AO decadal skill is leading by 1 year the decadal skill of temperature at 2m (t2m) with a correlation coefficient of r=+0.507 over 36 decades (unlagged r=0.253). This might have been emphasised due to the use of a continuous forced chain for producing initial conditions - but indicates AO has a predictive potential driving global t2m decadal predictability.

RC News Mihaela Fig 2a
Figure 2a: Time series of model AO index correlation to ERA AO index computed for: years1-6 (black). Green squares show the accuracy of initial AO phase sign compared to ERA (positive value shows same sign,i.e. correct phase sign). Red circles show the correlation of time evolution of the AO phase at initial time (first season) between ERA and model (both positive indicate accurately initialised phase)
RC News Mihaela Fig 2b
Figure 2b: Time series of model 2m temperature decadal skill (black) against AO index model decadal skill (green). Each point represents the decadal skill i.e. the value of time correlation with ERA, for each decade with starting date indicated on the x axis.


Mitchell et al. (2012) The Influence of Stratospheric Vortex Displacements and Splits on Surface Climate. J. Climate, 2012

W. Hazeleger at al, 2013: “Multiyear climate predictions using two initialization strategies”,
GRL, DOI:10.1002/grl.50355