Ice clouds are mostly composed of different ice crystal habits. It is of great importance to classify ice crystal habits seeing as they could greatly impact single‐scattering properties of ice crystal particles. The single‐scattering properties play an important role in the study of cloud remote sensing and the Earth's atmospheric radiation budget. However, there are countless ice crystals with different shapes in ice clouds, and the task of empirical classification based on naked‐eye observations is unreliable, time consuming and subjective, which leads to classification results having obvious uncertainties and biases. In this paper, the images of ice crystals observed from airborne Cloud Particle Imager in China are used to establish an ice crystal data set called Ice Crystals Database in China, which consists of 10 habit categories containing over 7,000 images. We propose an automatic classification model of ice crystal habits, called TL‐ResNet152, which is a deep convolutional neural network based on the newly developed method of transfer learning. The results show that the TL‐ResNet152 model could achieve reliable performance in ice crystal habits classification with the accuracy of 96%, which is far more accurate than traditional classification methods. Achieving high‐precision automatic classification of ice crystal habits will help us better understand the radiation characteristics of ice clouds.

Toward the Kashi region in northwest of China, the Moderate Resolution Imaging Spectroradiometer (MODIS) aerosol optical depth (AOD) retrievals from Collection 6 (C6) MYD, Collection 6.1 (C6.1) MOD, and C6.1 MYD during 2016–2017 are compared with ground‐based measurements from the Sun‐sky Radiometer Network (SONET), and the first comprehensive evaluation of the Dark Target (DT) and Deep Blue (DB) retrievals with a 10‐km spatial resolution in the latest C6.1 MYD AOD data set during 2016–2018 is presented. In general, C6.1 MYD AOD products (both of DT and DB algorithm) are the most effective in Kashi of the three collections, and there is an overall underestimation of DB AOD, while DT AOD that slightly outperformed DB AOD in Kashi is overestimated on the whole. As to the factors that influence the accuracy of MODIS AOD, for DB algorithm, the overestimations of the surface reflectance and Single Scattering Albedo that DB aerosol model assumed can cause underestimation of DB AOD retrievals over Kashi, while the ones for DT algorithm are opposite. Besides, the coarse dust particles with lower veracity are predominant in Kashi region, which illustrated that the errors of particle size assumption in C6.1 MYD DT and DB algorithms will make large inversion error of MODIS AOD. Moreover, whatever DB or DT algorithm, the accuracy of AOD is diminished as the aerosol loading increases. The more realistic aerosol models and surface characterizations are necessary during the process of generating the MODIS aerosol retrievals in Kashi region.

Atmospheric aerosols play a crucial role in regional radiative budgets. Previous studies on clear-sky aerosol direct radiative forcing (ADRF) have mainly been limited to site-scale observations or model simulations for short-term cases, and long-term distributions of ADRF in China have not been portrayed yet. In this study, an accurate fine-resolution ADRF estimate at the surface was proposed. Multiplatform datasets, including satellite (MODIS aboard Terra and Aqua) and reanalysis datasets, served as inputs to the Santa Barbara Discrete Atmospheric Radiative Transfer (SBDART) model for ADRF simulation with consideration of the aerosol vertical profile over eastern China during 2000–2016. Specifically, single-scattering albedo (SSA) from the Modern-Era Retrospective Analysis for Research and Application, Version 2 (MERRA-2) was validated with sun photometers over eastern China. The gridded asymmetry parameter (ASY) was then simulated by matching the calculated top-of-atmosphere (TOA) radiative fluxes from the radiative transfer model with satellite observations (Clouds and the Earth's Radiant Energy System, CERES). The high correlation and small discrepancy (6–8 W m⁻²) between simulated and observed radiative fluxes at three sites (Baoshan, Fuzhou, and Yong'an) indicated that ADRF retrieval is feasible and has high accuracy over eastern China. Then this method was applied in each grid of eastern China, and the overall picture of ADRF distributions over eastern China during 2000–2016 was displayed. ADRF ranges from −220 to −20 W m⁻², and annual mean ADRF is −100.21 W m⁻², implying that aerosols have a strong cooling effect at the surface in eastern China. With the economic development and rapid urbanization, the spatiotemporal changes of ADRF during the past 17 years are mainly attributed to the changes of anthropogenic emissions in eastern China. Our method provides the long-term ADRF distribution over eastern China for the first time, highlighting the importance of aerosol radiative impact under climate change.

The geographical distributions of summertime cirrus with different cloud top heights above the Tibetan Plateau are investigated by using the 2012–2016 Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) data. The cirrus clouds with different cloud top heights exhibit an obvious difference in their horizontal distribution over the Tibetan Plateau (TP). The maximum occurrence for cirrus with a cloud top height less than 9 km starts over the western plateau and moves up to the northern regions when cirrus is between 9 and 12 km. Above 12 km, the maximum occurrence of cirrus retreats to the southern fringe of the plateau. Three kinds of formation mechanisms – large-scale orographic uplift, ice particle generation caused by temperature fluctuation, and remnants of overflow from deep-convective anvils – dominate the formation of cirrus at less than 9 km, between 9 and 12 km, and above 12 km, respectively.