ECMWF continue to increase the number of satellite instrument products operationally assimilated or monitored. In recent years some of the highest profile developments have been in three broad areas: hyperspectral infrared sounders; all-sky microwave sounders and imagers; atmospheric composition observations.
In the area of hyperspectral sounding we now use observations from CrIS and AIRS, as well as two IASI instruments. For many years these data were used predominantly over the ocean. This was mostly down to lack of confidence in cloud detection and characterisation over land, due to poor knowledge of land surface temperature. However the cloud (and aerosol) screening has been improved and, as was already the case for microwave observations, considerable use of hyperspectral infrared sounder data is now made over land surfaces. This was shown to be highly beneficial. Effort is beginning to prepare for the new generation of hyperspectral sounders of geostationary platforms, such as CMA's GIIRS instrument on the FY4 series and EUMETSAT's IRS on the MTG series. The data volume produced by these instruments is huge and efforts have focussed on evaluation of different approaches to efficient assimilation of their information content, with principal component, reconstructed radiances and transformed retrievals all being evaluated.
The all-sky approach allows 4D-Var to explicit account for cloud and rain impacts on satellite measurements. This has been used for many years for microwave imagers, and has been extended to microwave humidity sounders. Typically it has been found that all-sky assimilation of humidity sounder radiances doubles the impact of a given instrument. The main impact is on wind, through the 4D-Var tracer mechanism. To be able to do all-sky assimilation needs three critical components: firstly a very good linear model is needed for the 4D-Var minimisation; secondly a very accurate fast observation operator for cloud and rain is needed; thirdly a symmetric and observation error model is needed, that correctly allows for the increased observation error, including representivity error, encountered in cloudy scenes. Effort is now also focussing on adding microwave temperature sounders and even infrared sounders to the all-sky framework.
ECMWF have for many years been at the forefront of extending traditional Numerical Weather Prediction (NWP) to atmospheric composition, which is now carried out under the Copernicus Atmospheric Monitoring Service (CAMS). This has required the use of more observations providing atmospheric composition information, such as GOMAS, GOME-2, SCIAMACHY, OCO-2 etc. A particular focus has also been aerosol, and the aerosol climatology from the CAMS analysis system is now used as the climate field for aerosol in the weather forecast model, bringing very significant benefit to forecasts.
There are of course many other areas of development, including allowing for horizontal gradients for radio occultation, radiances and radiosondes, allowing for inter-channel correlated error, and efficient automatic warning for observation anomalies, just to mention a few. There is also a continuous efforts to make use of new and innovative satellite instruments, including the new GOES-16 and Himawari-8 imagers. These represent a major advance in geostationary observation. Preparations are also well underway to use data from two new ESA Earth Explorer missions, EarthCARE (cloud radar and lidar) and Aeolus (doppler wind lidar). GOES-16 is also the first geostationary satellite to carry a lightning imager, closely followed by FY-4A and in a few years time MTG will also have this capability. Research into the feasibility of assimilating lightning data through initial comparison of diagnosed lightning with these new observations is beginning.
Many of the tools needed for successful satellite data assimilation are developed collaboratively. One of the most significant is EUMETSAT's Satellite Application Facilities (SAFs). In particular the SAFs for NWP, Radio Occultation Mission (ROM) and Hydrology (H) have supported data assimilation developments. Of particular note is the NWPSAF's continuous development of the RTTOV radiative transfer model, cloud screening modules and profile training databases. RTTOV includes a PC radiative transfer capability, the RTTOVSCATT module for all-sky microwave assimilation as well as the Fastem ocean emissivity model, which has over many years been developed jointly with JCSDA, JMA and Environment Canada. The ROMSAF produces the Radio Occultation Processing Package (ROPP). A recent example of the effectiveness of ROPP has been through improved use of CMA's GNOS data, where through strong collaboration between CMA and the ROMSAF it was possible to improve ROPP processing of GNOS to comparable quality to EUMETSAT's GRAS system.