How Can the New Chinese Satellite Improve Typhoon Forecasts?

FengYun‑4B (FY‑4B) is China's second-generation satellite equipped with an infrared hyperspectral sensor called GIIRS. This instrument captures infrared radiation across many wavelengths, providing a detailed atmospheric “snapshot.” Researchers tested the impact of GIIRS data on two recent typhoons: Doksuri and Gaemi. They compared three assimilation scenarios: using IASI data (a European satellite), AMSU‑A data (a microwave radiometer), and finally FY‑4B/GIIRS data. The results show that when the model ingests GIIRS observations, wind forecasts at 300 hPa and 500 hPa become more stable, with errors staying within ±6 m/s compared to the NCEP FNL reference. However, the model tends to underestimate vertical velocity, where AMSU‑A performs better. Regarding the storm track, GIIRS‑based forecasts stay within 50 km of the observed path for the entire 48‑hour window, outperforming the other two setups. In practice, the new sensor gives the model a clearer view of upper‑level winds, improving its ability to predict where the cyclone will move. These findings pave the way for broader use of GIIRS data in forecasting extreme weather, making alerts more accurate and timely.

Assimilation of GIIRS radiometric data is performed using 4D‑Ensemble variational (4D‑EnVar) techniques, which merge infrared radiation observations with the numerical forecast model. Thanks to its spectral resolution (~0.5 cm⁻¹) and spatial resolution (4 km at nadir), GIIRS provides more detailed thermodynamic profiles than IASI (coarser spectral resolution) and AMSU‑A (limited sensitivity to surface temperatures and higher layers). Incorporating these profiles improves the representation of temperature and humidity gradients, key drivers of vorticity and thus tropical cyclone dynamics. In tests on Doksuri and Gaemi, analysis of wind fields at 300 hPa and 500 hPa showed a ~30 % reduction in standard deviation compared to IASI/AMSU‑A cases, indicating higher consistency with radiosonde observations. However, the underestimation of vertical velocity (ω) suggests that while GIIRS excels at thermal components, it may not capture small‑scale ascent associated with intense convection. A possible improvement is to combine GIIRS with microwave data (e.g., AMSU‑A) to balance thermal and dynamical information. Additionally, using a more recent background analysis (e.g., ERA5) could reduce systematic biases, enhancing agreement with NCEP FNL reanalyses. In summary, GIIRS marks a significant advance in real‑time atmospheric observation, but its full potential will be realized by integrating it with other multispectral observations within a hybrid assimilation framework.