Seasonal Climate Prediction - final report for the CERFACS-EDF project

ABSTRACT

Abstract As a necessary first step in using a coupled model for seasonal prediction of European weather, the northern hemisphere circulation of the CERFACS coupled simulation is analysed and compared with both the ECMWF analyses and with the results of an atmosphere-only simulation, with an emphasis on the wintertime circulation. There is a large response to coupling an ocean model especially in the North Atlantic with a more realistic Icelandic low and weaker jet than in the AMIP simulation. This has a beneficial impact on the modelled European wintertime climate which becomes cooler with weaker mean westerlies. The transient behaviour also improves with stormtracks centered correctly over the oceans and increased low frequency variability.
The atmospheric response is due to a dipole SST anomaly in the Atlantic which resembles that found in previous observational and modelling studies due to a weakened thermohaline circulation. The atmospheric heat flux acts as a negative feedback on the SST anomaly and is dominated by the latent and sensible heating. A sub-surface analysis of the North Atlantic behaviour in the coupled GCM is made. The temperature anomalies are in general in the upper 500m of the ocean and are correlated with similar errors in salinity. The mixed-layer is reduced in the central North Atlantic and the thermal structure shows some pathologies such as abrupt jumps which could be due to insufficient vertical mixing. In the western part of the basin, the Gulf stream does not separate at Cape Hatteras but continues northward and flows over the Grand Banks causing a warm anomaly. This is perhaps partly due to a poor representation of the Labrador current in the model.
The atmosphere blocking has been studied and is improved in the coupled model run compared to the atmosphere only run. The blocking frequency is increased and the position improved in the Atlantic sector although blocking is still severely underestimated compared to in the ECMWF analyses.
Both Topex/Poseidon observed and model sea level height data have been extracted, calculated and interpolated into Vairmer format thus enabling detailed comparison. Some preliminary altimetric results are presented. Seasonal fluctuations of sea level heights reconstructed from both forced and coupled runs of the OPA model are then analysed. The separation is made between steric height and dynamical contribution to those fluctuations. Steric height is found dominated by the thermal variations of the upper 200 m. Salinity plays a minor role, especially in the coupled experiment and it is not clear whether the forced experiment salinity changes can be trusted. Sea level corrected from steric variations show weak variations, but spatially coherent. Interannual variability in the North Atlantic ocean is investigated, as a function of time and space. In the interior of the gyre, the coupled model show more variability than the forced one.
Interannual variations of sea level heights and sea surface temperature are characterized through a Principal Component (PC) analysis. In general, more variance is explained by the PCs of sea-level than those of temperature. In the forced case, the first PC is related to ENSO variability, whereas the coupled case exhibits a very significant drift. However, the second coupled PC shows very good agreement with ENSO variability, but is also very weak, so that teleconnections can not clearly be evidenced. Results on seasonal teleconnections have been obtained. 200hPa geopotential height and total ozone column are analysed through the same method. Conclusions reached are that the interannual variability of the coupled model is probably too weak to produce teleconnections.
Analyses of five coupled experiments beginning three month after each other during the year 1992 have been carried out. Though the initial drift is a major drawback of the present coupled experiments for climate prediction, it is found that the model may reproduce some known features of the actual climate predictibility, such as the spring prediction barrier, known in ENSO forecast activities. A brief review of data assimilation into oceanic models is given in the perspective of sea level data assimilation into coupled models. Emphasis is put on the specificities of the oceanographic observations, and particularly on the use of altimeter data to constrain the deep oceanic flow.
Back to the previous page