Widespread decline in critical soil moisture threshold of plant water stress across global drylands

Python version 3.11, R version 4.4.1, and Matlab2024a were used.
1. drydown_identification.m was executed to identify soil dry-down events and to save the corresponding datasets. Then, SM_threshold_estimation.R was run to estimate soil moisture thresholds.
2. Figure1.py was used to perform trend analysis of soil moisture thresholds and to generate Figure 1.
3. Figure2.R was executed to analyze trends in the critical soil moisture threshold across plant functional types, climate zones, and aridity classes, and to generate Figure 2.
4. Figure3.R was used to conduct machine-learning–based and multivariate linear regression attribution analyses, and to generate Figure 3.
5. Figure4.py was used to analyze long-term vegetation water-stress exposure and changes in the fraction of water-limited days, and to generate Figure 4.
6. Figure5.py was executed to evaluate the ability of Earth System Models to capture observed θcrit trends and their aridity dependence, and to generate Figure 5.