Data associated with "The importance of terrain and climate for predicting soil organic carbon is highly variable across local to continental scales"

The zipped folder contains the processed soil datasets including covariates, soil depths, and SOC concentrations for training the deep learning models described in the paper "The importance of terrain and climate for predicting soil organic carbon is highly variable across local to continental scales". 

"soil_profile/" contains a table including the geolocations of all the soil profiles in this study. "patch_data/" and "point_data/" contain the covariates to feed the models with patch input and point input respectively. "depth/" contains the upper and lower depths of the soil samples. "y/" contains the target variable - SOC concentration of the soil samples. The data files with suffix "_1" is a small subset of their counterparts without "_1" (10 % in sample size) used for model hyperparameters tuning.