Regional downscaling of decadal predictions

Wednesday, 17 December 2014: 8:15 AM
Hendrik Feldmann, Karlsruhe Institute of Technology, Institute for Meteorology and Climate Research, Karlsruhe, Germany
During the last years the research field of decadal predictions gained increased attention. Its intention is to exploit the predictability derived from slowly varying components of the climate system on inter-annual to decadal time-scales. Such predictions are mostly performed using ensembles of global earth system models.

The prediction systems are able to achieve a relatively high predictive skill over some oceanic regions, like the North Atlantic sector. But potential users of decadal predictions are often interested in forecasts over land areas and require a higher resolution, too. Therefore, the German research program MiKlip develops a decadal ensemble predictions system with regional downscaling as an additional option. Dynamical downscaling and a statistical-dynamical downscaling approach are applied within the MiKlip regionalization module. The global prediction system consists of the MPI-ESM model. Different RCMs are used for the downscaling, e.g. CCLM and REMO. The focus regions are Europe and Western Africa.

Hindcast experiments for the period 1960 – 2013 were performed to assess the general skill of the prediction system. Of special interest is the value added by the regional downscaling. For mean quantities, like annual mean temperature and precipitation, the predictive skill is comparable between the global and the downscaled systems. For extremes on the other hand there seems to be an improvement by the RCM ensemble. The skill strongly varies on sub-continental regions and with the season. The lead time up to which a positive predictive skill can be achieved depends on the parameter and season, too.

A further goal is to assess the potential for valuable information, which can be derived from predicting long-term variations of the European climate. The leading mode of decadal variability in the European/Atlantic sector is the Atlantic Multidecadal Variation (AMV). The potential predictability from AMV teleconnections especially for extreme value anomalies over Europe is explored using re-analysis driven RCM simulations. It is evaluated how such teleconnections are represented in the RCM. For instance, the multi-year mean soil water content is correlated to the AMV index in Europe. This provides the potential to predict drought tendencies, which is relevant for agricultural applications.