ECMWF have developed a C++ code control layer to run the data assimilation in a single executable. This is known as the Object Orientated Prediction System (OOPS). This has previously been tested for data assimilation in simplified models. In collaboration with partners in Meteo-France and the ALADIN and HIRLAM consortia the Integrated Forecast System (IFS) code has been refactored over the last two years to allow IFS to be called and controlled from OOPS, with a view to using OOPS operationally. The initial goal is to replicate the current atmospheric 4D-Var capability with OOPS. Once this is achieved OOPS will provide a flexible framework that will enable introduction of new data assimilation developments, including the saddle point weak constraint data assimilation formulation.
OOPS will also facilitate implementation of other key points in our DA strategy. For example it can facilitate the implementation of a 4D-Var solution that does not require the inverse of the B-matrix when the first guess is arbitrary, as is the case for the overlapping window configuration or asymmetric EDA, where the perturbed members could use the accurate control analysis as a first guess. Use of a hybrid B with the extended (alpha) control vector is also being tested. OOPS is also being applied to other components of the IFS, such as the NEMOVAR ocean data assimilation, and will be applied to the land DA system in due course. During the process of this major technical development project the opportunity has been taken to perform a modernisation of the IFS code base, and this allowed some errors and issues in the IFS code to be found and corrected.
A key component of ECMWF's data assimilation strategy is to move to introduce more coupling. A key first step was to run an experimental re-analysis system including ocean-atmosphere coupling in the outer loop. This has been applied both to the 20th century and satellite era re-analysis under the EU Framework-7 ERA-CLIM2 project. It has been possible to demonstrate how atmospheric observations such as scatterometer winds can influence the ocean down to 300m depth. It has also be shown that, at least in this framework, coupling can significantly improve weather forecast skill. The coupled assimilation approach is now being prepared for operations, and potentially future operational re-analyses under the Copernicus Climate Change Service (C3S). Whilst initial coupling experiments focus on the ocean-atmosphere exchange, ECMWF run a successful land surface data assimilation system, that has been used to do pioneering work on SMOS radiance assimilation, as well as demonstrating operationally the impact of ASCAT soil moisture on the root zone, through the data assimilation process. Improvements have also been made to the snow data assimilation. In the future it is expected that the land surface DA will also be coupled to the atmosphere.
The new ECMWF re-analysis being run by the EU-funded Copernicus Climate Change Service (C3S), ERA5, has already released datasets and continues towards completion of full production for the satellite era. It is also proposed to extend ERA5 back in time to the 1950s. ERA5 has many advantages over the older ERA-Interim: it has higher horizontal and vertical resolution, uses a more skilful ECMWF IFS cycle, that allows use of all current satellite data, many of the more recent instruments were not supported by ERA-Interim (e.g. IASI, CrIS and ASCAT). Furthermore ERA5 uses the all-sky framework for microwave humidity radiances, replacing the 1D+4D system used for microwave imagers and clear-sky for microwave sounders used in ERA-Interim. It also has a more advanced Ensemble component for the wavelet B calculations. ERA5 will be described in more detail in the presentation.
Alongside these major infrastructure and science projects a number of important science improvements continue to be made to the data assimilation methodology and use of observations, and several others are being researched. Recent improvements include the application of the wavelet background error formulation to be extended to humidity, using more satellite observations in cloudy and rainy areas, using the all-sky assimilation framework, and increased use of satellite data over land.