Evaluation of FY-4A/4B GIIRS Atmospheric Temperature Profile Products Using Radiosonde Observation Data through a Per-Level Collocation Method

FY-4A and FY-4B are the first two geostationary satellites equipped with an infrared hyperspectral sounder known as the Geostationary Interferometric Infrared Sounder (GIIRS). This paper validates the FY-4A/4B GIIRS Atmospheric Temperature Profile (ATP) products by comparing them with routine radiosonde data collected from 89 stations across China between February 2023 to January 2024. To mitigate the effects of balloon drift, a collocation strategy was implemented for each pressure level to identify the nearest temperature retrieval within 1800 seconds and 0.2 degrees of the actual observation time and location of the radiosonde. This strategy enhances the agreement between radiosonde observations and satellite retrievals, particularly under clear sky condition. Utilizing the per-level collocation strategy, we conducted a comprehensive evaluation of the GIIRS ATP products and analyzed how the evaluation results vary in relation to sensors (FY-4A/4B), sky conditions (cloudy/clear), months, times (morning/evening), pressure levels, and regions. The results indicate that the FY-4A/4B GIIRS ATP products demonstrate a strong consistency with radiosonde observations, achieving high coefficients of determination (0.990–0.997), favorable biases (-0.92 to 0.06 K), and low RMSE (root mean square error) values (1.58 to 2.57 K). In comparison to cloudy conditions, the evaluation results are significantly improved and exhibit more pronounced variations across months and pressure levels under clear sky conditions. The summer months demonstrate better performance than the winter months, and the retrievals in the evening are slightly better than those in the morning. Near the top of the atmospheric boundary layer (850 hPa) and the tropopause (200 hPa), the RMSE show higher values and increased spatial heterogeneity compared to the mid-tropospheric level (500 hPa). FY-4B demonstrates significant superiority over FY-4A under clear sky conditions; however, this trend does not persist under cloudy conditions due to the large negative biases associated with FY-4B, necessitating further improvement in the retrieval method.