Budget: 4000000

Seasonal prediction improvement with an Earth System Model (SPIESM)

IC3's Climate Forecasting Unit (CFU) obtained close to 4 million CPU hours on Lindgren, the PDC Tier-1 HPC in Sweden, in the first climate forecasting project ever awarded with PRACE resources. 

This project contributes to the RUCSS project funded by the Spanish Science and Investigation Ministry. In RUCSS we aim at testing the seamless approach for climate modelling with the EC-Earth Earth System Model (ESM) to constrain the sources of uncertainty in both short-term climate prediction and climate-change projections by increasing the understanding of the climate system. In this project, detailed analysis of climate simulations with different time horizons will be carried out using similar metrics to better understand the development of the systematic errors in EC-Earth with the hope of reducing the risk of overconfidence in both climate predictions and long-term projections. Among the processes that will be investigated are the water vapour feedback in the extratropics, the climate variability at the surface of the tropical Pacific and the extraordinary summer warming and drying observed and foreseen over Southern Europe.

The basic premise of the seamless approach is that there are fundamental physical processes in common to both seasonal and decadal forecast, as well as climate-change time scales. If essentially the same ESM using a similar ensemble system can be validated both probabilistically and from a physical point of view on time scales where validation data exist, that is, on daily, seasonal and decadal time scales, then users will have the possibility of a) modifying the probabilistic estimates of regional climate-change, b) gaining insight into the ESM limitations to reduce the systematic error and c) improve the realism of the unavoidable physical parameterizations used in dynamical models. The seamless approach makes no distinctions about the relevance of particular processes in an ESM as a function of the time scale of the target problem, the correct representation of all physical processes affecting all types of climate simulations.

This initiative is both innovative and ambitious. A limited number of seasonal forecasting research and operational groups already exists in Europe (EUROSIP, an operational system to which Met Office, Météo-France and the ECMWF contribute), Canada (with the multi-model operational system run by Environment Canada), USA (NCEP  with their Climate Forecast System), Australia (BoM runs the POAMA system), and Korea (APCC gathers quasioperational multi-model seasonal forecast information). At the same time, a great interest in climate modelling has risen in Europe and some skill in seasonal predictions of summer temperature has been found over Southern Europe, as well as for other regions where there are European interests (e.g. South America and areas of Africa).