Data and code from: Temporal shift and climate drivers of vegetation resilience in a high-altitude national park of China

Li, Qiuran 1 ; Gao, Xiang1 ; Lin, Jiaxin 1 ; Song, Xiaojiao1 ; Chang, Shuhang1

Published Mar 27, 2026 on Dryad. https://doi.org/10.5061/dryad.jwstqjqqk

Data files

Mar 27, 2026 version files 371.87 MB

Data_code_readme.zip

371.86 MB
README.md

3.90 KB

Abstract

This dataset contains the processed analytical data and calculation scripts supporting the study on vegetation resilience in a high-altitude national park. The data include monthly and annual time series of the Normalized Difference Vegetation Index (NDVI) and key climate drivers, including precipitation (PRE), temperature (TEM), potential evapotranspiration (PET), and the Standardized Precipitation Evapotranspiration Index (SPEI) from 2002 to 2021.

The dataset consists of:

A processed tabular dataset (.csv) integrating multi-source remote sensing and meteorological data, resampled to 1 km resolution.
Analytical scripts used for NDVI threshold filtering (NDVI < 0.1), vegetation type reclassification, and statistical modeling.

These data are provided to ensure the reproducibility of the temporal shift analysis and climate-driven resilience modeling presented in the associated journal article.

Dataset DOI: 10.5061/dryad.jwstqjqqk

Dataset for vegetation resilience and climate driving factors in China

This dataset contains the processed data, results, and Python/R scripts used to analyze vegetation resilience (proxied by AR1) and its responses to climate driving factors across seven major vegetation types in China.

Files and variables

The repository is organized into folders (Fig3 through Fig11). Each directory contains specific data files (CSV or raster) and scripts required to reproduce the results.

Core variables across datasets:

x, y: Geographic coordinates (Longitude and Latitude in Decimal Degrees).
AR(1): First-order autocorrelation coefficient (dimensionless, range: -1 to 1), used as a proxy for vegetation resilience.
slope / trend_value: The rate of change per unit of time (e.g., NDVI/year or AR1/year).
p_value: Statistical significance (values < 0.05 indicate significant change).
Climate Drivers: Includes Monthly NDVI/KNDVI, Precipitation (mm), Average Temperature (°C), Potential Evapotranspiration (mm), and SPEI.
Vegetation Types: Categorized into Steppe, Meadow, Alpine vegetation, Shrub, Desert, Needleleaf forest, and Broadleaf forest.

Data and code organization (by Figure)

Fig3: Autocorrelation and piecewise trends

Data: ar1.csv, pre_2011_05_trends_parallel.csv, post_2011_06_trends_parallel.csv.
Code: R scripts (Fig3a_code.txt, Fig3b_code.txt) using raster, terra, and parallel packages.

Fig4 & Fig5: Resilience shifts and ecological zones

Data: all_trend.csv, ar1_trend_duandian_first/second.csv, trend_type_veg.csv].
Visualization: Python scripts for Pie/Donut charts and stacked bar charts using matplotlib and pandas
Key Concept: Positive AR1 slope indicates "Critical Slowing Down" (resilience loss).

Fig6: Environmental gradients

Data: Merged trend data for SPEI, Precipitation, Temperature, and PET.
Code: Python script for 2x2 multi-panel plots, handling unit conversions and "Zero-Crossing" analysis.

Fig7 & Fig8: Random Forest feature importance

Data: Model results for three periods (Full, Pre-breakpoint, Post-breakpoint) and predictor variable files.
Methodology: Pixel-by-pixel Random Forest Regression using scikit-learn and joblib for parallel processing.

Fig9: PDP and ICE curves

Code: Implementation of Partial Dependence Plots (PDP) and Individual Conditional Expectation (ICE) curves.
Logic: Uses B-spline interpolation to visualize the marginal effect of climate drivers on resilience.

Fig10 & Fig11: Stability evolution and coupling analysis

Data: KNDVI_rolling_mean.csv, merged_results2.csv.
Fig11 Logic: A nested pie chart visualizing the coupling between Resilience (Outer Ring) and Greening/Browning (Inner Ring).

Code/Software requirements

R (v4.x+): raster, terra, parallel, ggplot2.
Python 3: pandas, numpy, matplotlib, scikit-learn, scipy, joblib.

Access information

Data Source: Data was derived from multi-source remote sensing and meteorological datasets.
Zip Archive: All data and scripts mentioned above are included in Data_code_readme.zip.

Code/Software

All data processing and visualization were conducted using Python 3, leveraging standard scientific libraries such as NumPy and Pandas for data manipulation, Matplotlib and Seaborn for figure generation, and SciPy or Xarray for trend analysis and spatial data handling. Each figure-specific folder contains standalone Python scripts (.py or .ipynb) that can be executed to replicate the study's findings, provided that the required library environment is correctly configured.