EUMETSAT’s first Meteosat Third Generation Sounder satellite (MTG‑S1) is planned to be launched in July 2025, carrying the next-generation hyperspectral infrared instrument (InfraRed Sounder, IRS). This instrument will measure infrared radiances from the Earth with high spectral resolution and, from its geostationary orbit, will give us detailed information on atmospheric structures both spatially and temporally over Europe and Africa. In preparation for MTG-IRS, the Geostationary Interferometric Infrared Sounder (GIIRS) onboard China’s second Fengyun‑4 Series satellite (FY‑4B) has been assessed, for monitoring and assimilation in ECMWF’s numerical weather prediction (NWP) system. FY‑4B GIIRS was launched in 2021 and observes east Asia. It is the first operational hyperspectral infrared instrument on a geostationary platform. Its nominal characteristics are very close to future hyperspectral infrared instruments on geostationary satellites, including MTG-IRS. Therefore, experiences gained by investigating FY‑4B GIIRS can help to prepare us for MTG‑S IRS. Exploitation of data from GIIRS presents some challenges, some of which are specific to the instrument. However, we show here that these observations can have a positive impact on forecast scores, improving temperature and humidity in the lower troposphere over the GIIRS domain.
FY-4B GIIRS data assessment
FY‑4B GIIRS is generally performing better than its predecessor onboard the first Fengyun‑4 Series satellite (FY‑4A). The mid-wave infrared band of FY‑4A GIIRS had several issues which limited its potential use in NWP. These have been greatly improved for FY‑4B thanks to work by the China Meteorological Administration (CMA). However, there are still some challenging artefacts in the FY‑4B GIIRS data, including systematic differences in the pixels across the detector, and horizontally-correlated observation errors in high-peaking temperature channels. These can be addressed by using only a subset of the 128 pixels and by selecting a subset of channels with relatively small horizontal correlations.
In addition, CMA performs two yaw flip manoeuvres to FY‑4B in March and September to prevent solar intrusion into the sensor. In the yaw flip manoeuvre, the satellite rotates about the axis pointed toward the centre of Earth. When the manoeuvre takes place, the biases of channels with wavenumbers less than 742.5 cm–1 suddenly jump with a change of more than 0.5 K. This is possibly due to thermal emission from the scanning mirrors. MTG‑S will also have yaw flip manoeuvres, and EUMETSAT will monitor the internal temperature of the satellite after the launch of MTG‑S to try to mitigate this.
Impact assessment in ECMWF’s 4D-Var system
The impact of FY‑4B GIIRS data in ECMWF’s 4D‑Var data assimilation system was tested under clear‑sky conditions in Cycle 49r1 of the Integrated Forecasting System (IFS) for a summer season period, when the yaw flip manoeuvre was not performed. A carefully chosen subset of channels sensitive to atmospheric temperature, surface temperature, ozone and water vapour was used, and only a subset of pixels which have consistent statistics. Otherwise, the assimilation methodology is similar to that used for the other hyperspectral infrared sounders in the IFS system, including the cloud detection method, the use of an observation error covariance matrix with appropriate inter-channel correlations, and variational bias correction.
Globally, the impact on temperature, humidity and ozone forecasts is neutral, but it is positive over the GIIRS observational domain. Some improvements in the short-range forecast of temperature, humidity, wind and ozone are found with respect to independent observations, as shown in the figure. Medium-range forecast scores in the lower troposphere are also promising; the assimilation of FY‑4B GIIRS results in statistically significant forecast improvements for temperature and humidity as verified against own analyses in the lower troposphere over the GIIRS observational domain until forecast day 4. The root-mean-square error of forecasts decreases by up to about 1%.
Summary and outlook
The assimilation of FY‑4B GIIRS has a neutral-to-positive impact on forecast skill, and passive monitoring of these data will begin soon as part of our NWP SAF (Satellite Application Facility) activities. However, before considering operational assimilation of FY‑4B GIIRS, additional assimilation experiments will be run for a period including a yaw flip manoeuvre, and also a winter season period will be considered to confirm the assimilation of FY‑4B GIIRS has robust impact. This investigation of FY‑4B GIIRS has given us some interesting insights about hyperspectral infrared sounders on geostationary platforms and helps prepare the way to assimilate MTG‑S IRS when the data become available.