China's 30-meter resolution soil organic carbon prediction map at the scale of cultivated land

This paper examines the effectiveness of a SOC prediction framework that employs a cross-attention mechanism to encode and integrate continuous spatial background variables and discrete and continuous mixed agricultural utilization variables. It is worth noting that, due to the database's maximum limit of 10GB for data volume, the predicted results uploaded for this dataset are actually 100 times the original pixel values of the data, rounded to two decimal places.

After such processing, the exact accuracy of the data will inevitably be affected to some extent, but the overall distribution trend of the pixel values can still be guaranteed. The actual method of obtaining this data can be found in the detailed discussion in the paper and the relevant data in the appendix (to be published later). For further detailed discussion of this dataset, please feel free to contact Jin Guo (email: guojin020201@163.com) at any time.