ECMWF's prime longterm goal is to deliver operational mediumrange weather forecasts of increasingly high quality, over the range from three to ten days and beyond. A complementary longterm goal is to establish and deliver a reliable operational seasonal forecasting capability. Overall success will depend crucially on new and improved satellite observations, on improvements in the dataassimilation system, and on improvements in the forecast models. Ensemble prediction will play a major role in attaining both goals.
Medium-range and seasonal-to-interannual weather forecasting have attained new levels of achievement and continue to offer exciting prospects for further development. A new generation of operational and research satellites is coming on stream; four- dimensional variational assimilation is established as a powerful and effective method to use all observations; numerical methods continue to provide improved accuracy and substantial economies; parametrization schemes are improving steadily through new approaches which jointly exploit field experiments, large-eddy simulations and operational data assimilation; ensemble prediction systems are providing a new dimension in probabilistic forecasting; the development of simplified Kalman filters, based on singular vectors, will benefit both the assimilation systems and the ensemble prediction systems; coupled atmosphere-chemistry models offer new ways to improve the initialisation of forecast models; coupled atmosphere-ocean models have delivered succesful ensemble forecasts of the 1997-98 El Nino event; and computer vendors are confident they can meet the requirements for computational power in an affordable manner. These developments will undoubtedly lead to further gains in medium-range and seasonal-to-interannual forecast skill.
The new satellite data essential to achieve these goals over the next decade will be provided as a result of investments of $3B or more by Europe (EUMETSAT and ESA), $3B or more by Japan and $10B or more by the USA. By early 1999, in readiness to exploit the new satellite data, ECMWF will have implemented an advanced operational Earthsystem model and data assimilation system. ECMWF's Earthsystem model will comprise the following coupled modules:
Atmosphere
-an atmospheric general circulation model
Ocean circulation -
-an ocean general circulation model ocean ice processes
-Ocean surface waves -
-ocean surface wave model
Land
-land biosphere module
-land surface, soil, hydrological and snow model Ozone
parametrized stratospheric ozone chemistry
Some modules of the model are already quite sophisticated, but others are at an early stage of development. ECMWF's advanced fourdimensional variational data assimilation system (4DVar) has been developed specifically to optimise the use of satellite data. By the end of 1999, the operational 4DVar system will be supported by a powerful new algorithm (a simplified Kalman filter) to provide flowdependent forecast error structures at the start of each 4DVar cycle.
ECMWF strategy to achieve its goals entails:
· Further development of the Earthsystem assimilation facility to make best use of all available data (especially satellite data) to provide accurate analyses, together with accurate estimates of the uncertainty of the analyses;
· Further development of the integrated highresolution Earthsystem modelling facility;
· Development of the methodology of ensemble forecasting for mediumrange and seasonal forecasting;
· Operational delivery of an enhanced range ofmeteorological and associated products;
· Maintenance and extension of the Centre's scientific andtechnical collaborations with scientific and technical institutes in the Member States and beyond.
The presentation will summarise the methods by which the models and assimilation systems have been developed, the specific areas where further gains are likely to be made, and the areas where ECMWF will focus its efforts.