An ensemble weather forecast is a set of forecasts that presents the possible range of future weather developments. The main advantage over a single forecast is that it provides information on the uncertainty associated with the forecast, by indicating a range of potential outcomes. The main drawback is that ensemble forecasts are computationally more expensive to produce than single forecasts. For that reason, from 1992 until a few weeks ago we produced high-resolution forecasts (HRES) as well as coarser-resolution ensemble forecasts (ENS) in the medium range. Most recently, the HRES was run at a horizontal resolution of 9 km and the ENS at 18 km. But in a far-reaching upgrade of our Integrated Forecasting System (IFS) to Cycle 48r1 on 27 June, the resolution of our medium-range ENS forecasts was increased to 9 km as well. The upgrade marks an important change in emphasis in favour of ensemble forecasts. As set out in the article on this year’s Using ECMWF’s Forecasts event, usage of ensemble forecasts still varies among those who took part in the event, but there was broad backing for our emphasis on ensemble-based outputs.
The increase in ENS resolution brought a number of improvements to our forecasts, which are detailed in this Newsletter. But IFS Cycle 48r1 went beyond this change in many ways. It brought many other improvements in data assimilation, to establish the best possible initial conditions for forecasts, and in the forecast model. It also introduced major changes to extended-range ensemble forecasts: their number of ensemble members has increased from 51 to 101, and they are now run daily instead of twice weekly. At the same time, we upgraded the forecasting system of the EU’s Copernicus Atmosphere Monitoring Service (CAMS), which we implement.
The importance of ensemble forecasts can also be seen in the article on the development of El Niño conditions in this Newsletter. Our own seasonal forecasting system, SEAS5, as well as the systems brought together by the Copernicus Climate Change Service (C3S) implemented by ECMWF, uses ensemble forecasts to describe the possible range of future temperature anomalies in the equatorial Pacific Ocean. Another article describes how well our seasonal forecasting system predicted the winter 2022/23 in the face of energy security concerns across the continent in relation to the war in Ukraine. Collaboration in training features prominently, with articles on the OpenIFS User Meeting in Barcelona and support for the ICON training carried out by the German Meteorological Service.
This Newsletter also describes the results of tests carried out at ECMWF of external machine-learning forecasts of extreme weather. They show that machine-learning models can successfully forecast extreme weather situations. These forecasts are currently not available as ensemble forecasts, but this is surely only a matter of time.