In response to the high demand for more skillful climate forecasts at the seasonal timescale, innovative climate prediction systems are developed with improved physics and increased spatial resolution. Alongside the model development process, seasonal predictions need to be evaluated on past years to provide robust information on the forecast performance. This work presents the quality assessment of the Météo-France coupled climate prediction system, taking advantage of an experiment performed with 90 ensemble members over a 37-year re-forecast period from 1979 to 2015. We focus on the boreal winter season initialised in November. Beyond typical skill measures we evaluate the model capability in reproducing ENSO and NAO teleconnections on precipitation and near surface temperature respectively. Such an assessment is carried out first through a composite analysis, and shows that the model succeeds in reproducing the main patterns for near surface temperature and precipitation. A covariance method leads to consistent results. Finally we find that the teleconnection representation of the model is not affected by shortening the verification period and reducing the ensemble size and therefore can be used to evaluate operational seasonal forecast systems.

为响应在季节性时标上对更熟练的气候预测的巨大需求，开发了具有改进的物理特性和更高的空间分辨率的创新性气候预测系统。除了模型开发过程外，还需要对过去几年的季节性预测进行评估，以提供有关预测绩效的可靠信息。这项工作利用了1979年至2015年为期37年的重新预测期中，对90位整体成员进行的实验，对法国-法国耦合气候预测系统进行了质量评估。我们关注11月启动的冬季北方季节。除了典型的技术指标外，我们还评估了分别在降水和近地表温度下再现ENSO和NAO遥相关的模型能力。首先通过综合分析进行评估，结果表明该模型成功地再现了近地表温度和降水的主要模式。协方差方法可得出一致的结果。最后，我们发现模型的远程连接表示不受缩短验证周期和减小集合大小的影响，因此可以用于评估运营季节预测系统。