Model code for simulations in "Potential vegetation changes in the permafrost area over the Tibetan Plateau under future climate warming"

This archive contains two parts:

1. CryoGridLite model for simulating the thermal regime of permafrost on the Tibetan Plateau from 1979 to 2100.  CryoGridLite is a lightweight and fast-calculated version that inherits from the CryoGrid3 Model (Westermann et al., 2016) and the CryoGrid Community Model (Westermann et al., 2023). Two scientific articles have been published using CryoGridLite to simulate the permafrost thermal dynamics in the pan-Arctic areas.

   Nitzbon et al. (2023). First quantification of the permafrost heart sink in the Earth's climate system.

   Langer et al. (2024). The evolution of Arctic permafrost over the last 3 centuries from ensemble simulations with the CryoGridLite permafrost model.
   
   
   Compared to Nitzbon et al. (2023) and Langer et al. (2024), this version implemented the surface energy balance module to provide the upper boundary condition of the model and applied a 'bucket' scheme to compute the dynamics of soil water content. Parameters and model setup can be specified in the main.m and loadExperimentSetting.m. To start the program, run the script CGLite_Launcher.m. The directory .\input can be used to store the forcing, soil stratigraphy, and initial soil temperature file (Here are 100 grid cells of input data provided). The default output directory is .\output\. 
2. Machine learning algorithms (LightGBM and XGBoost) for predicting the Normalized Difference Vegetation Index (NDVI) on the Tibetan Plateau from 2019 to 2050. This toolkit is designed to facilitate the analysis of climate impacts on vegetative growth patterns, with a focus on permafrost regions. By leveraging advanced machine learning techniques including LightGBM, XGBoost, and Ridge Regression, the toolkit enables researchers to predict and analyze the Normalized Difference Vegetation Index (NDVI) as an indicator of vegetation changes under different climate change scenarios. This toolkit contains the following:

          LightGBM.py and XGBoost.py Machine learning algorithms for NDVI predictions based on climate and land surface data.

          Calculate_trend_p_value.py Mann-Kendall significance test to identify trends in time-series data.

          ridge_regression.py Discover the most influential factors affecting NDVI changes.

          Install all the necessary Python libraries (numpy, pandas, scikit-learn, xgboost, lightgbm, optuna, pymannkendall, and xarray) using pip or conda.

The model is further described in the following article which has been submitted to the Journal of Geophysical Research: Earth Surface: Chen, R., Nitzbon, J., Schneider von Deimling, T., Stuenzi, S. M., Chan, N.-H., Boike, J., and Langer, M.: Potential vegetation changes in the permafrost area over the Tibetan Plateau under future climate warming, Journal of Geophysical Research: Earth Surface [preprint].