Dataset in a machine learning benchmark for reconstructing high-resolution three-dimensional coastal fields from coarse surface data

This dataset is for a machine learning benchmark for statistical spatial downscaling of three-dimensional coastal fields from surface fields. It contains fields data (horizontal currents, salinity and water temperature) in a coastal region (8E-9E, and 53.75N-54.25N) in the German Bight. The data contain both low-resolution (ll8.00_53.75_n8_8.zip) and high-resolution data (ll8.00_53.75_n128_128_horizontalVelX.zip, ll8.00_53.75_n128_128_salinity.zip and ll8.00_53.75_n128_128_temperature.zip) in the year of 2020. The low-resolution directory contains subdirectories 'horizontalVelX', 'horizontalVelY', 'salinity', 'temperature' and 'out2d'. Each sub-directory contains the surface fields in folder 'iv20' except 'out2d'. In 'out2d', the nc files contain fields including sea surface height. Each high-resolution directory contains 3 sub-directories, with each representing the surface (iv20), middle (iv10) and bottom (iv0 or iv1) layers. The name of the nc files corresponds to the date, e.g., 20200102.nc corresponds to data in January 2, 2020. Each nc file contains 24-hour gridded data. The resolution is 8 by 8 for low-resolution data and 128 by 128 for high-resolution data.