Zhang et al. (2026) Climate-driven modeling and future projection of grassland aboveground biomass on the Qinghai-Xizang plateau using an interpretable Transformer framework

Identification

• Journal: Ecological Indicators
• Year: 2026
• Date: 2026-09-12
• Authors: Ruoqi Zhang, Qisheng Feng, Yonghui Zhang, Chengwen Yang, Ni Chong, J.J. Mai, Tiangang Liang
• DOI: 10.1016/j.ecolind.2026.115485

Research Groups

• State Key Laboratory of Herbage Improvement and Grassland Agro-ecosystems
• Key Laboratory of Grassland Livestock Industry Innovation, Ministry of Agriculture and Rural Affairs
• Engineering Research Center of Grassland Industry, Ministry of Education
• College of Pastoral Agriculture Science and Technology, Lanzhou University

Short Summary

This study proposes an interpretable Transformer framework to predict grassland aboveground biomass (AGB) on the Qinghai-Xizang Plateau. The model identifies precipitation as the dominant climate driver and projects future AGB trajectories under different emission scenarios.

Objective

• To develop a climate-driven monthly-scale AGB prediction framework for grasslands of the Qinghai-Xizang Plateau.
• To identify dominant climate driving variables and key temporal windows using multiple interpretability methods.
• To analyze spatial differentiation patterns of AGB change under different emission scenarios across multiple spatial scales.

Study Configuration

• Spatial Scale: The study focuses on the Qinghai-Xizang Plateau, a region in southwestern China with an area of approximately 2.57 × 106 km2.
• Temporal Scale: The study uses monthly climate sequences from 2003 to 2023 and projects future AGB trajectories from 2024 to 2060 under different emission scenarios.

Methodology and Data

• Models used: Transformer, LSTM, GRU, XGBoost, and Random Forest.
• Data sources:
  • Historical meteorological data (2003–2023) from the National Earth System Science Data Center.
  • Future climate scenario data (2015–2100) from the CMIP6 future temperature and precipitation projection dataset.

Main Results

• The Transformer model with the pre + tmx configuration achieved an R2 of 0.6330 and RMSE of 331.38 kg DW/ha on the test set.
• Precipitation was identified as the dominant climate driver, followed by maximum temperature, minimum temperature, and mean temperature.
• April was found to be the most important month for AGB prediction.

Contributions

• This study provides a fine-scale characterization of intra-seasonal month-to-month climate forcing dynamics and their cumulative cross-month effects on grassland AGB.
• The interpretable Transformer framework enables direct empirical testing of whether attention weights align with attribution-based estimates of climate variable importance.

Funding

• This research was funded by the State Key Laboratory of Herbage Improvement and Grassland Agro-ecosystems, the Key Laboratory of Grassland Livestock Industry Innovation, Ministry of Agriculture and Rural Affairs, and the Engineering Research Center of Grassland Industry, Ministry of Education.

Citation

@article{Zhang2026Climatedriven,
  author = {Zhang, Ruoqi and Feng, Qisheng and Zhang, Yonghui and Yang, Chengwen and Chong, Ni and Mai, J.J. and Liang, Tiangang},
  title = {Climate-driven modeling and future projection of grassland aboveground biomass on the Qinghai-Xizang plateau using an interpretable Transformer framework},
  journal = {Ecological Indicators},
  year = {2026},
  doi = {10.1016/j.ecolind.2026.115485},
  url = {https://doi.org/10.1016/j.ecolind.2026.115485}
}