IRBT2CPP: A training and testing dataset for machine learning training, ranging from infrared brightness temperature to cloud physical properties

The IRBT2CPP is a dataset covering latitudes from -70° to 70° and longitudes from -180° to 180°, containing infrared brightness temperature (IRBT) data corresponding to cloud physical properties (CPP) from 2000 to 2023 for use in machine learning or other inversion algorithms. The sample IRBT includes two channels at 6.7 µm and 11 µm, with data sourced from GridSat-B1. The labeled CPP includes cloud phase (CLP), cloud top height (CTH), cloud optical thickness (COT), and cloud effective radius (CER) from the MODIS twin satellites Terra (MOD) and Aqua (MYD). The sample IRBT has a spatial resolution of 0.07° and a temporal resolution of 3 hours. All CPP labels are interpolated using nearest-neighbor interpolation to a 0.07° grid consistent with the sample IRBT. The satellite ID and corresponding satellite zenith angle are also obtained from GridSat-B1 or through subsequent calculations. In addition, there are some meteorological field elements sourced from ERA5, which are obtained through bilinear interpolation to a unified 0.07° grid using Climate Data Operators.Detailed information:All files are saved in npz format and standardized to 64×64 squares. They can be read into the CPP tag and IRBT sample respectively via “modis” and “gridsat”.The data dimensions of the IRBT sample are (20, 64, 64), where 20 represents:Latitude, Longitude, Irwin (11 μm), Irwvp (6.7 μm), SatID_IR (satellite ID for Irwin data source), SatID_WV (satellite ID for Irwvp data source), SZA_IR (satellite zenith angle for Irwin data source), SZA_WV (satellite zenith angle for Irwvp data source), Skt (surface skin temperature), Tcwv (total column water vapor), Soil (soil type), Temp1000 (1000 hPa air temperature), Temp850, Temp500, Temp300, Humi1000 (relative humidity at 1000 hPa), Humi850, Humi500, Humi300, Mask (used to assess data quality, 0 indicates normal, 1 indicates abnormal).The data dimensions of the CPP tag are (7,64,64), where 7 represents:Cloud phase (CLP), Cloud top height (CTH), Cloud effective radius (CER), Cloud optical thickness (COT), Cloud top temperature (CTT), Latitude, Longitude The file names and directory structure are organized in the following format:```Matching results for Terra (MOD) (2000-2023):IRBT2CPP.tar.gz/MOD2Gridsat/{year}/{month}{day}/{hour}/MOD06toGridsat.{month}{day}.{hour}.cropped.{i}.{j}.npzMatching results for Aqua (MYD) (2002-2023):IRBT2CPP.tar.gz/MYD2Gridsat/{year}/{month}{day}/{hour}/MYD06toGridsat.{month}{day}.{hour}.cropped.{i}.{j}.npz* i, j are the numbers assigned to a MODIS image cropped into 64*64 pieces.``` ## change list:V1: init commit.