UCL Department of Geography


UCL Geography banner



    UCL Home / Geography / PMIP Var / Individual Models

    Summary Figures

    This pages contains a series of figures that collect the response of several modes of variability across the runs and ensembles. They are not meant to be authorative, just helpful. I am happy to take suggestions of additional figures and will try my best to actually make them.

    Decadal Variability

    The decadal variability throughout the simulation is calculated across the globe for in surface (air) temperature, precipitation, SST and sea level pressure. The PMIP3 ensemble can be used to look to se if one would expect more or less "noise" in a proxy record at a site. Here I'm showing the ensemble mean change in standard deviation between the mid-Holocene and preindustrial for the decadally-smoothed surface temperature. The stippling shows that there is a robust reduction for regions in the mid-latitudes - it means at two-thirds or more the ensemble members have the same sign change.


    The change in both ENSO and the Annual Cycle across the different experiments. This is based on IPCC AR5 Fig 5.10.. It compares the percentage change in standard deviation of the Nino3.4 SST anomalies and the annual climatology (from which the anomalies are computed). The 5%,25%,50%,75%,95% percentiles are shown. [Note: currently the 1pctCO2 run takes the average throughout the whole simulation, and the changes are computed w.r.t piControl].

    North Atlantic Oscillation

    The existence of large ice sheets over North America at the Last Glacial Mximum altered the location of the stormtracks downstream of them. This in turn lead to changes in the variability, as exmplified by the North Atlantic Oscillation. The figure above shows the PMIP3 ensemble mean change in the winter precipition regressed onto the positive phase of the NAO. This looks the reverse of the normal pattern over the Atlantic - indicating that during the LGM the NAO played less of a role in rainfal variations. The dots on the model give a measure of confidence as they indicate when two-thirds of the ensemble show the same sign.