In this study, a thermal infrared (TIR) based convolutional neural network (TIR-CNN), originally designed for retrieving cloud optical properties from TIR radiometry, is further developed to obtain global cloud base height (CBH) and cloud thermodynamic phase during both daytime and nighttime. It employs TIR radiances, cloud optical properties retrieved by TIR-CNN, altitude, landcover, and lifting condensation level as inputs to estimate global CBH and cloud phase for both single- and multi-layer clouds. This new model is trained, validated, and tested using radar-lidar products from CloudSat/CALIPSO. It provides global CBH with root-mean-square errors of 1.19 km and 1.91 km for single- and multi-layer clouds, respectively. A cloud layer classifier is trained to provide information on the quality of retrieved CBH, with total accuracies of 82% and 85% for single- and multi-layer clouds, respectively. In addition, the new model has remarkable accuracy in identifying the cloud phase within each pixel’s vertical column, particularly in differentiating mixed-phase clouds with an ice cloud top.

This study applies the Advanced Radiative Transfer Modeling System (ARMS), which was developed to accelerate the uses of Fengyun satellite data in weather, climate, and environmental applications in China, to characterize the biases of seven infrared (IR) bands of the Advanced Geosynchronous Radiation Imager (AGRI) onboard the Chinese geostationary meteorological satellite, Fengyun–4A. The AGRI data are quality controlled to eliminate the observations affected by clouds and contaminated by stray lights during the mid–night from 1600 to 1800 UTC during spring and autumn. The mean biases, computed from AGRI IR observations and ARMS simulations from the National Center for Environmental Prediction (NCEP) Final analysis data (FNL) as input, are within −0.7–1.1 K (0.12–0.75 K) for all seven IR bands over the oceans (land) under clear–sky conditions. The biases show seasonal variation in spatial distributions at bands 11–13, as well as a strong dependence on scene temperatures at bands 8–14 and on satellite zenith angles at absorption bands 9, 10, and 14. The discrepancies between biases estimated using FNL and the European Center for Medium–Range Weather Forecasts Reanalysis–5 (ERA5) are also discussed. The biases from water vapor absorption bands 9 and 10, estimated using ERA5 over ocean, are smaller than those from FNL. Such discrepancies arise from the fact that the FNL data are colder (wetter) than the ERA5 in the middle troposphere (upper–troposphere).

Cloud optical/microphysical properties (COMPs) with a high spatiotemporal resolution are required for monitoring the evolution of convective clouds, studying aerosol-cloud-precipitation interactions, and evaluating cloud microphysics parameterization schemes in weather and climate models. This study applied the bispectral method, commonly used for COMP retrievals, to Advanced Himawari Imager (AHI)/Himawari-8 observations. The AHI daytime cloud optical thickness (τ) and cloud-top particle effective radius (Re) retrievals were then evaluated with the Collection-6.1 Moderate Resolution Imaging Spectroradiometer cloud product (MOD06) for a three-month period from June 1, 2017, to August 31, 2017. Overall, the AHI-retrieved COMP results are in good agreement with the MOD06 product. The correlation coefficients of τ and Re between the AHI retrievals and MOD06 products are 0.76-0.89 and 0.63-0.87, respectively. Overestimations are mainly observed in the 1.6-μm Re (Re1.6) and ice-phase 3.9-μm Re (Re3.9) retrievals. There are τ dependencies in the retrieval biases of τ and Re1.6, as well as angular dependencies in the retrieval biases of τ and ice-phase Re3.9. Moreover, the monthly mean composites for three COMP retrievals from the MOD06 and AHI algorithms are compared. Significant differences are only found between the two Re1.6 products. This study also finds that AHI-retrieved Re and τ values have a diurnal variation over land but not over oceans.

Overshooting top (OT), a proxy for deep convection with an intense updraft that can penetrate the tropopause, has an important influence on typhoon intensification. Using 9003 Himawari‐8 satellite images and a unique OT detection algorithm, the diurnal variation of OTs within 45 western North Pacific typhoons were analyzed. We examined the distribution of OTs in different types of typhoons in terms of both intensity and intensity change and the relationship between the OTs and typhoon intensification on a diurnal scale. Our results show that a greater OT density (OTD) occurs in strong typhoons and rapid intensification (RI) typhoons. Moreover, RI typhoons showed greater diurnal variation than non‐RI typhoons. The diurnal cycle of OTD in RI typhoons was in phase with the intensification of the typhoon, with the maximum in the early morning. Therefore, OTD can become a potentially effective indicator to estimate diurnal changes in typhoon intensity.