CMIP5 - Ensemble decadal simulations with the EC-EARTH model

An ensemble of decadal simulations with actualised forcing covering the 1960-2015 decades has been performed with the Earth System Model EC-EARTH. The data will be provided into the CMIP5 archive, and their analysis will investigate several actual aspects of decadal predictability.

Decadal simulations has been accomplished for the CMIP5 databasis with the coupled Earth System Model EC-EARTH, for an ensemble of three members. Each member contains 10 decadal simulations based on the 1st of November for the years: 1960, 1965, 1970, 1975, 1980, 1985, 1990, 1995, 2000 and 2005. The members differ in the initial state of the ocean (we used the perturbed NEMOVAR – ocean model analysis) and in the control climatology reported to which we computed the initial anomaly state.

The coupled system was initialised for all members using an anomaly method in which the initial state is projected onto the coupled model climate. The forcing is based on observed values up to 2010.

Analysis and processing of results

The processing of the results will be based on commonly established metrics computed by partners that contribute to the project decadal simulations databasis. These involve atmospheric, ocean and sea-ice metrics.

Main issues like the benefit from initialising decadal simulations versus forcing accuracy (figure 1), assessement of decadal predictability sources for the coupled system and the model ability to represent those will be addressed during this analysis.

Concluding remarks

The fact that the initialisation process shows an impact during the first years, even overestimating the later effect of the forcing (in agreement with results of Smith, 2007 ) emphasises that we need that, and that we can gain predictability over decadal scales through this. In the next step, the decadal metrics evaluation will quantify that benefit.

EC-EARTH decadal simulations
Figure 1. The difference in T2m averaged over years 1 to 5 between an initialised and non-initialised simulation started the same day (left). The difference in T2m averaged over years 1 to 10 between a simulation forced by background and observed values (right). We note that the magnitude of main reponse patterns (NH) have comparable order of magnitude, indicating the importance of initialisation. Enlarge Image