ECMWF started assimilating GNSS radio occultation (GNSS-RO) measurements from the FORMOSAT-7/COSMIC-2 mission (COSMIC-2 hereafter) on 25 March 2020. This joint Taiwan–United States mission is a constellation of six satellites, providing around 5,000 occultations per day between ±40 degrees latitude. The satellites were launched on 25 June 2019 and declared operational on 25 February 2020, and the data became available on the Global Telecommunication System on 16 March 2020.
What are radio occultation data?
Radio occultation data are obtained by sending signals from the Global Navigation Satellite System (GNSS) to other satellites. The path of such signals crossing the atmosphere is bent as a result of refraction, and the bending angle provides information on temperature and humidity. GNSS includes the US GPS (Global Positioning System) and the Russian GLONASS (GLObal NAvigation Satellite System).
COSMIC-2 represents a significant increase in the number occultations available for operational assimilation at ECMWF, from around 3,000 per day (of which around 1,800 come from EUMETSAT’s Metop satellites) to around 8,000. These new measurements have a large impact in the tropics, improving temperature, humidity and wind forecasts in the short and medium range.
The COSMIC-2 mission
COSMIC-2 is the successor to the FORMOSAT-3/COSMIC mission, which has provided operational GNSS-RO data since 2006. The COSMIC-2 data is processed in parallel by the University Corporation for Atmospheric Research (UCAR) and Taiwan’s National Space Organization (NSPO). A key aim of this new mission has been to improve the GNSS-RO measurement quality and penetration depth in the tropical troposphere, through a combination of a new advanced receiver design and an improved antenna. COSMIC-2 provides occultation measurements using both US GPS signals and Russian GLONASS signals, increasing the number of observations provided per satellite. This is the first routine provision of GLONASS GNSS-RO measurements for operational numerical weather prediction (NWP) applications.
Early COSMIC-2 results were presented by UCAR and NSPO at the EUMETSAT ROM SAF IROWG workshop in September 2019 (see https://www.romsaf.org/romsaf-irowg-2019/en/). For example, it was demonstrated that measurements made with the GPS and GLONASS GNSS systems were of similar quality below 35 km. The differences above this height were relatively small, when compared with assumed error statistics used to assimilate GNSS-RO data.
Use of COSMIC-2 at ECMWF
UCAR made the first COSMIC-2 non-operational data available in October 2019, initially for technical testing in NWP systems. As more data became available, they were used at ECMWF for extended impact experiments.
ECMWF has initially adopted a conservative approach to COSMIC-2 by assimilating the data in the same way as is done for other operational GNSS-RO missions. This produced good results in experiments from late September 2019 to March 2020. It is a sign of improvement when the short-range forecast is closer to data from other observing systems. Such improvements are found for almost all in-situ and satellite observation types in the tropics (see the figure for an example). Perhaps most significantly, there is a clear and consistent impact on the forecast fit to a range of observing systems that are sensitive to tropospheric water vapour.
Medium-range forecast scores are also promising. We obtain large forecast improvements for tropical stratospheric zonal (latitudinal) winds. The impact is statistically significant out to 10 days, verified against observations. In the northern hemisphere extratropics, the reduction in geopotential height forecast errors from the 500 hPa level and above is statistically significant out to 3–4 days. There are also global improvements in stratospheric geopotential height and temperature biases.
Summary and outlook
The COSMIC-2 mission represents a large increase in the GNSS-RO data available for operational NWP applications. First ECMWF impact experiments with this data have been extremely promising, and they led to the operational implementation on 25 March 2020. The impact of COSMIC-2 on the forecast fit to observing systems sensitive to tropospheric water vapour represents an important step forward for GNSS-RO.
The use of COSMIC-2 is still quite conservative, and ECMWF will try to improve on the exploitation of this data in the future. We will test extending the use of the data higher into the mesosphere. In addition, following promising work conducted recently at the UK Met Office, we will investigate the use of observation uncertainty estimates that differ for the various GNSS-RO datasets.
We thank UCAR and NSPO for making the COSMIC-2 data available for pre-operational testing. We also thank Dr Ben Ruston (Naval Research Laboratory, Monterey) for sharing his operational results with COSMIC-2 data.