AUTHOR=Qin Yi , Steven Andrew D. L. , Schroeder Thomas , McVicar Tim R. , Huang Jing , Cope Martin , Zhou Shangzhi TITLE=Cloud Cover in the Australian Region: Development and Validation of a Cloud Masking, Classification and Optical Depth Retrieval Algorithm for the Advanced Himawari Imager JOURNAL=Frontiers in Environmental Science VOLUME=7 YEAR=2019 URL=https://www.frontiersin.org/journals/environmental-science/articles/10.3389/fenvs.2019.00020 DOI=10.3389/fenvs.2019.00020 ISSN=2296-665X ABSTRACT=

This paper presents a cloud masking, cloud classification and optical depth retrieval algorithm and its application to the Advanced Himawari Imager (AHI) on the Himawari-8/9 satellites using visible, near infrared and thermal infrared bands. A time-series-based approach was developed for cloud masking which was visually assessed and quantitatively validated over 1 year of daytime data for both land and ocean against the level 2 Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) 1 km cloud layer product (version 4.10). An overall hit rate (the proportion of pixels identified by both sensors as either clear or cloudy) of 87% was found. However, analysis revealed that, when partially cloudy conditions were experienced, the small footprint of the CALIOP sensor (70 meters beam size sampling every 330 meters along the ground track) had a major impact on the hit rate. When partially cloudy pixels are excluded a hit rate of ~98% was found, even for thin clouds with optical depth less than 0.25. A two-way confidence index for the cloud mask was developed which could be used to reclassify the pixels depending on applications, either biasing toward clearness or cloudiness. On the basis of the cloud masking, classification and optical depth retrieval was performed based on radiative transfer modeling. Small modeling error was found, and inspection of typical cloud classification examples showed that the results were consistent with cloud texture and cloud top temperatures. While difficult to validate retrieved cloud properties directly, an indirect quantitative validation was performed by comparing surface-level solar flux computed from the retrieved cloud properties with in-situ measurements at 11 sites across Australia for up to 3 years. Excellent agreement between calculated and measured solar flux was found, with a mean monthly bias of 2.96 W/m2 and RMSE of 8.91 W/m2, and the correlation coefficient exceeding 0.98 at all sites. Further assessment was conducted by comparing seasonal and annual cloud fraction with that of ISCCP (International Satellite Cloud Climatology Project) over Australia and surrounding region. It showed high degree of resemblance between the two datasets in their total cloud fraction. The geographical distribution of cloud classes also showed broad resemblance, though detailed differences exist, especially for high clouds, which is probably due to the use of different cloud classification systems in the two datasets. The products generated from this study are being used in several applications including ocean color remote sensing, solar energy, vegetation monitoring and detection of smoke for the study of their health impacts, and aerosol and land surface bidirectional reflectance distribution function (BRDF) retrieval. The method developed herein can be applied to other geostationary sensors.