An evaluation of FY-3C MWRI and Assessment of the long-term quality of FY-3C MWHS-2 at ECMWF and the Met Office

In this report we present an evaluation of the long-term performance of the Feng Yun-3C (FY-3C) MicroWave Humidity Sounder-2 (MWHS-2) instrument, and a first assessment of the FY-3C MicroWave Radiation Imager (MWRI) instrument. This is done by comparing statistics of observation minus the short-range forecasts (O - B) from the ECMWF and Met Office systems. The aim is both to assess the quality of the observations, and to explore how short-range forecasts from NWP systems can be used as a reference for the calibration/validation of new satellite data.

Time series of mean O - B show some jumps in bias for most channels of MWHS-2 during 2016, which are consistent for both the Met Office and ECMWF statistics. These bias changes are shown to be correlated to changes in the instrument environment temperature on board the satellite. Channels 13 and 14 show large bias changes of order 2 - 3 K. However, for all other channels the bias changes are relatively small (∼0.3K). In addition the standard deviation of O - B is stable over time, indicating that the MWHS-2 data are generally of good quality for most channels. This instrument continues to be assimilated at ECMWF and the Met Office, with the jumps in bias corrected by the variational bias correction schemes used at both centres.

Maps of O - B for MWRI show an ascending-descending bias of ∼2K for all channels, in both the ECMWF and Met Office systems. The 10.65 GHz channels have warm biases over Europe (descending only data) which are likely to be due to Radio Frequency Interference from geostationary communications satellites. Most channels of MWRI have global biases of the order -1 to -2 K for ascending data and -3 to -4 K for descending data compared to the Met Office and ECMWF shortrange forecasts, and O - B statistics suggest an intersatellite bias between MWRI and AMSR-2 of approximately 4 - 6 K depending on the channel. The standard deviation of O - B is similar to AMSR-2 for most channels, however. Time series show that the ascending-descending bias for MWRI has increased in magnitude from 1 K to 2 K from 2014 to 2016. It would be useful to try to understand further the causes of the ascending-descending bias in future work since this could lead to improvements in the data for both FY-3C MWRI and future instruments