Code and datasets for: Cost-effective land carbon sink conservation in China and its limited synergy with biodiversity

This repository contains code and datasets for the paper "Cost-effective land carbon sink conservation in China and its limited synergy with biodiversity." A. Code structure The code in this study is organized into two main parts: 1. Machine learning code The “machine_learning” folder contains Python code for training machine learning models (ANN, RF, XGBoost, and LightGBM) and using the models to make predictions. The code is written in Python 3.8 and runs on Ubuntu 20.04 with CUDA 11.8. 2. Data analysis code The “data_analysis” folder contains Python code for the analyses decribed in the paper and generating figures. The code is written in Python 3.8 and can be executed on both Ubuntu and Windows environments. All code is provided in Jupyter Notebook, displaying results at each step. Input data required for the code to run are stored in “00_data\input” directory within each folder. B. Dataset Structure The datasets generated by the code are located in the “00_data\output” directory within each folder: 1. Machine learning outputs "machine_learning\00_data\output" contains predicted carbon sink capacity (indicated by NBP, unit: gC m-2 d-1) for past (2001-2015) and future (2020-2100) periods by four machine learning models. 2. Data analysis outputs "data_analysis\00_data\output" contains the main results of this study: The "CS_XGB_LGB_mean_std_2020_2100" folder contains average carbon sink capacity predictions from the high-performance models (XGBoost and LightGBM) and uncertainties (1 std) for all pixels. The pixel means are illustrated in Fig. 1 of the paper. The "CS_dCS_PD" folder contains .tif files for carbon sink capacity (CS), changes in carbon sink capacity (∆CS), and population density (PD) under future scenarios (SSP1-2.6, SSP2-4.5, and SSP5-8.5) and periods (2020-2060 and 2060-2100) . They are depicted in figs. S9-11 and used for cost-benefit analysis. The "cost_benefit" folder contains .tif files for cost-benefit analysis results mapped in space under future scenarios and periods, as shown in Fig. 2 of the paper. The "priority_levels" folder contains .tif files for cost-effective priority levels mapped in space under future scenarios and periods, as shown in Fig. 3 of the paper. The "cs_bio_priorities" folder contains .tif files for multi-scenario and -period mean priorities for carbon sink conservation (Fig. 4A), priorities for biodiversity conservation, and their synergy (Fig. 5C). "source_data_for_figures.xlsx" contains the source data for al