01-05-2014

The prototype decadal prediction system

At the end of development stage 1 (DS1) in February 2013, all projects were given the opportunity to provide recommendations for the next (third) generation of the MiKlip decadal prediction system, the so-called prototype system. The recommendations included the suggestion to used full-field initialisation for the ocean component of the MPI-ESM, rather than the anomaly initialisation approach used for the first two generations of the system (Baseline-0 and Baseline-1). Furthermore, it was suggested to use the GECCO2 ocean synthesis (Köhl, 2014) for the initialisation, instead of the ORA-S4 (Balmeseda et al., 2012) that was used for Baseline-1.

Based on these recommendations, a test suite was started. These prototype testing experiments reveal an increase in skill when considering full-field initialization compared to anomaly initialization of the ocean component of MPI-ESM as used in the baseline versions. The skill was examined in parameters such as surface temperature and heat content, but also first year storm track diagnostics applied by project partners from Module E. Based on the test experiments it was not possible to distinguish a difference in skill when using GECCO2 or ORAS4 for the initial fields.

Therefore, full field initialization of the ocean component with both ORAS4 and GECCO2 is being used for the prototype system. Although small improvements from a higher model resolution with MPI-ESM-MR are shown for baseline-1 (Pohlmann et al. 2013), the prototype system will employ the MPI-ESM-LR model in order to produce a large ensemble size with more than 30 ensemble members, which will increase the robustness of the deterministic and probabilistic evaluation of the system considerably.

Another recommendation concerned the way the ensemble is generated, suggesting to use a more sophisticated way of generating the ensemble than the lagged initialisation method that was used for the first two generations of the systems. However, none of the tested alternative methods showed an improvement in spread characteristics compared to the simple method. Thus the lagged initialisation method is still being used for the prototype system.

References:

  • Balmaseda, M. A., K. Mogensen, and A. T. Weaver,Evaluation of the ECMWF ocean reanalysis system ORAS4, Q. J. R. Meteorol. Soc., published online, doi:10.1002/qj.2063, 2012.
  • Köhl A., Evaluation of the GECCO2 Ocean Synthesis: Transports of Volume, Heat and Freshwater in the Atlantic, Q. J. R. Meteor. Soc., doi:10.1002/qj.2347, 2014.
  • Pohlmann, H., W. A. Müller, K. Kulkarni, M. Kameswarrao, D. Matei, F. S. E. Vamborg, C. Kadow, S. Illing, J. Marotzke, Improved forecast skill in the tropics in the new MiKlip decadal climate predictions, Geophys. Res. Lett., 40, 5798-5802, doi:10.1002/2013GL058, 2013.
Back to List