Code and Source Data for "Knowledge-Guided Machine Learning can improve C cycle quantification in agroecosystems"

Datasets for code and Source Data for the study "Knowledge-Guided Machine Learning can improve C cycle quantification in agroecosystems" https://doi.org/10.1038/s41467-023-43860-5. All files belong to Licheng Liu and Zhenong Jin at University of Minnesota. deposit_code_v2.zip contains packaged codes and sample runs for KGML-ag-Carbon training, validation and implementations. Source Data.zip contains data for generating the figures inside the study. 

Note: We used Pytorch 1.6.0 (https://pytorch.org/get-started/previous-versions/, last access: 21 Oct 2023) and Python 3.7.11 (https://www.python.org/downloads/release/python-3711/, last access: 21 Oct 2023) as the programming environment for model development. Statistical analysis, such as linear regression, was conducted using Statsmodels 0.14.0 (https://github.com/statsmodels/statsmodels/, last access: 21 Oct 2023) In order to use a GPU to speed-up the training process, we installed the CUDA Toolkit 10.1.243 (https://developer.nvidia.com/cuda-toolkit, last access: 21 Oct 2023). 

To use the full kgml_lib function, please create a new environment with the same python and libs above.