The subtropical high (hereinafter “STH”) is a fundamental large-scale weather system that governs tropical and subtropical regions. Its structural variations are closely related to regional heavy rainfall, summer heatwaves, and other extreme weather events. Accurately characterizing both the intensity and spatial features of the subtropical high is crucial for weather forecasting and short-term climate prediction. Currently, traditional weather observations and analyses are inadequate for the real-time, refined identification of STH structures. In this study, we introduce an STH identification model based on the Fengyun-4 (FY-4) satellites (FY_STH), utilizing a machine learning model that synthesizes brightness temperature data from the Advanced Geosynchronous Radiance Imager (AGRI) onboard FY-4 with satellite-retrieved outgoing longwave radiation (OLR) measurements. The model utilizes a piecewise eXtreme Gradient Boosting (XGBoost) classification decision tree ensemble to identify the subtropical high, and Bayesian optimization to refine its structure and parameters for optimal performance. The ERA5 reanalysis data serve as the benchmark for defining the subtropical high, and systematic evaluations of the FY_STH model are conducted, with sounding data used for validation. The results demonstrate that FY_STH consistently delineates the STH influence area in different seasons, achieving an average accuracy of 0.8, a false alarm rate of 0.3, a critical success index (CSI) of approximately 0.6, a dice coefficient of about 0.7, and an area under the curve (AUC) of the ROC (Receiver Operating Characteristic) around 0.9. In comparison with the CMA-GFS forecasts, FY_STH exhibits superior overall performance, with all evaluation metrics showing that its results are accurate and reliable. FY_STH is then employed to capture the relationship between Typhoon Gaemi and the STH over the western Pacific during 19–28 July 2024. It is revealed that variations in the trajectory of Typhoon Gaemi are strongly associated with shifts in the STH position, underscoring that real-time satellite monitoring facilitates a timely and precise understanding of typhoon trajectory variations and enhances the forecast accuracy based on the interaction between the STH and the typhoon.

Cloud effective radius (CER) is a fundamental microphysical property of clouds, critical for understanding cloud formation and radiative effects. Satellite spectral imagers, widely utilized in passive remote sensing, facilitate the monitoring of cloud characteristics, including CER, over extended time periods and spatial scales. Various observational methods have been employed to evaluate satellite cloud property products; however, in situ measurement evaluations of CER remain limited, particularly for products over China. This study utilized aircraft observations provided by the China Meteorological Administration Weather Modification Centre to evaluate the CER retrieved by Fengyun-4A Advanced Geosynchronous Radiation Imager (AGRI) and Himawari-8 Advanced Himawari Imager (AHI). Three flights were selected from the aircraft dataset for evaluation, involving flights through non-precipitating stratiform clouds with stable, high-quality measurements. Rigorous data selection and collocation procedures were employed to ensure a comprehensive comparison. Satellite retrievals from heterogeneous cloud fields were excluded, and representative in-cloud aircraft measurements were identified through multi-parameter filtering. The flight trajectory was adjusted to account for horizontal cloud movement corresponding to time differences between observations from different platforms. Additionally, in situ measurements from different vertical layers were adjusted to a comparable position near the cloud top. Results indicate that CER retrieved from satellites is generally overestimated compared to in situ measurements. For AGRI, the average difference (AD) is 2.90 µm, with a root mean square difference (RMSD) of 3.53 µm. For AHI, the AD is 2.92 µm, and the RMSD is 3.59 µm. To enhance future validation and evaluation of remote sensing results, factors such as instrument calibration, flight patterns, and cloud conditions will be carefully considered. Increasing the number of cases should further reduce errors associated with individual instances, enabling more precise assessments.