AHI-CALIOP Collocated Data for Training and Validation of Cloud Masking Neural Networks

Collocated data between AHI at 2km resolution (nadir) and CALIOP 1km cloud product v4.20 used for training and validating cloud identification neural networks. The main training and validation data from 2019 is stored in monthly directories, whilst the collocated dataset used to compare the NN, JMA and BoM cloud mask performances is the file "superdf.h5". All collocated data is stored as .h5 files and was built using the Python Pandas package. In this archive, the data has been stored as compressed directories for each month or as a single compressed file in the case of "superdf.h5" using tar with bzip2 compression or just bzip2 compression respectively.

本数据集为星下点2公里分辨率的先进葵花成像仪（Advanced Himawari Imager, AHI）与1公里分辨率CALIOP云产品v4.20的匹配数据，用于训练和验证云识别神经网络。2019年的主要训练与验证数据按月度目录存储；而用于对比神经网络（Neural Network, NN）、日本气象厅（Japan Meteorological Agency, JMA）及澳大利亚气象局（Bureau of Meteorology, BoM）云掩码性能的匹配数据集则存储为文件"superdf.h5"。所有匹配数据均以.h5文件格式存储，且通过Python的Pandas包构建。在本归档文件中，月度数据采用结合bzip2压缩的tar格式打包为压缩目录，而"superdf.h5"则仅通过bzip2压缩为单个压缩文件。