11 November 2015
The earth's atmosphere is chaotic. Methods to analyse its current state and predict future states are not perfect. One approach is ensemble forecasting, which predicts a representative sample of possible future states of the atmosphere and then, produces a measure of the uncertainty of these forecasts.
Ensemble forecast systems run a forecast model multiple times to produce this representative sample of forecasts. The forecast outcomes differ because of slightly perturbed initial conditions or slightly different model formulations. The actual future state of the atmosphere should be within the spread of ensemble forecasts and this spread is related to the uncertainty of the ensemble forecast.
Figure 1. Ensemble forecast system with initial condition uncertainty and forecast model uncertainty. The forecast uncertainty is represented by the dark blue shading.
The European Centre for Medium-Range Forecasts (ECMWF), of which Ireland is a founding member state, use a 51-member ensemble forecast system to predict future states of the global atmosphere. The ECMWF ensemble forecast system produces ten-day forecasts and these forecasts are extended beyond ten days on a coarser model grid. In 1995 ECMWF used a 63km global model grid and archived 14TB of data per year. Currently ECMWF use a 31km model grid for their ensemble forecasts and archive about 100TB of data per day. Through model grid refinement and improved science the quality of the ECMWF ensemble system has greatly improved over the past 20 years.
In recent years Met Éireann has contributed to the running of a limited area model based ensemble system called GLAMEPS. GLAMEPS is jointly developed by the HIRLAM and ALADIN consortia. The ensemble system uses 52 members to carry out short range forecasts focused on high-impact weather, such as severe thunderstorms, over Europe.
Through the use of ensemble forecasting techniques forecasters can make use of high quality forecast data and the associated uncertainty.