The 4-km monthly global air-sea carbon flux density dataset from 2000 to 2018

We combined 24.6 million ocean observation data points, terabyte-level remote sensing images, and petabyte-level reanalysis data to create a standardized sample set containing more than 3 million records. We proposed a spatiotemporal feature-embedding machine-learning method to solve the issues of sparse data, missing data, and discontinuous changes existing in most current research. Our approach enables efficiently exploiting these massive datasets, leading to the first fully continuous and monthly global ocean  partial pressure dataset covering the years 2000 to 2018. Based on the dataset, we presented a depiction of the 4-km monthly global air-sea carbon flux density  that encompassed coastal oceans and characterized the ocean carbon budget for the period of 2000 to 2018 by integrating open datasets, including atmosphere partial pressure, wind speed, sea surface temperature, sea surface salinity.