Li et al. (2025) Dominant drivers of spatiotemporal variations in carbon and water use efficiency across the Yellow River Basin revealed by interpretable machine learning

Identification

Journal: Frontiers in Plant Science
Year: 2025
Date: 2025-12-03
Authors: Guangchao Li, Wenjie Hao, Liqin Han, Min Feng, Yanjie Li, Zhaoqin Yi, Yayan Lu, Kangjia Zuo
DOI: 10.3389/fpls.2025.1632172

Research Groups

College of Geography and Tourism, Henan Normal University, Xinxiang, China
College of Life Sciences, Henan Normal University, Xinxiang, China

Short Summary

This study quantified the nonlinear spatiotemporal variations of ecosystem carbon and water use efficiency (CWUE) across the Yellow River Basin (YRB) from 1982 to 2018 and identified the spatially heterogeneous dominant driving factors using interpretable machine learning. The findings reveal that CWUE generally increased with high sustainability, primarily driven by leaf area index (LAI) for water use efficiency (WUE) and temperature for carbon use efficiency (CUE).

Objective

To uncover the features of spatial distribution characteristics, nonlinear spatiotemporal variation trends, and patterns of CWUE in the YRB, while exploring the stability and sustainability of each variation pattern across different regions within the basin.
To elucidate the influence of driving factors (including climate change, vegetation change, and human activities) on the CWUE within the YRB.
To analyze the spatiotemporal heterogeneity of these driving factors across various regions of the YRB, with an emphasis on identifying the predominant factors that influence CWUE in different areas.

Study Configuration

Spatial Scale: Yellow River Basin (YRB), China (32° N-42° N, 96° E-119° E), encompassing diverse climate zones and vegetation types.
Temporal Scale: 1982-2018 for CWUE, GPP, NPP, and ET; 2000-2018 for LAI, radiation, temperature, precipitation, sunlight, and GDP.

Methodology and Data

Models used:
- Ensemble Empirical Mode Decomposition (EEMD) for nonlinear spatiotemporal trend and pattern analysis.
- XGBoost (eXtreme Gradient Boosting) with 10-fold cross-validation and random search hyperparameter optimization for predictive modeling.
- SHAP (SHapley Additive exPlanations) explanatory model for quantifying individual feature contributions and spatial heterogeneity of driving factors.
- Hurst index for sustainability assessment.
- Coefficient of Variation (CV) for stability quantification.
Data sources: Multi-source remote sensing and meteorological reanalysis data.
- Gross Primary Production (GPP): 1982-2018, 5 km, 8-day resolution (https://www.geodata.cn)
- Net Primary Production (NPP): 1982-2018, 5 km, 8-day resolution (https://www.geodata.cn)
- Evapotranspiration (ET): 1982-2018, 5 km, 8-day resolution (https://www.geodata.cn)
- Leaf Area Index (LAI): 2000-2018, 5 km, 8-day resolution (GLASS dataset, https://www.geodata.cn)
- Radiation: 2000-2018, 1 km, 8-day resolution (https://www.geodata.cn/)
- Temperature: 2000-2018, 1 km, monthly resolution (https://data.tpdc.ac.cn/)
- Precipitation: 2000-2018, 1 km, monthly resolution (https://data.tpdc.ac.cn/)
- Sunlight: 2000-2018, 1 km, annual resolution (https://www.geodata.cn/)
- Gross Domestic Product (GDP): 2000-2018, 1 km, annual resolution (https://doi.org/10.6084/m9.figshare.17004523.v1)
- Digital Elevation Model (DEM): 2000, 30 m resolution (https://www.gscloud.cn/)

Main Results

Spatial Distribution: CWUE in the YRB exhibits a spatial pattern of higher values in the southeast and lower values in the northwest, predominantly concentrated at elevations ranging from 1000 to 1500 meters. WUE NPP and WUE GPP generally decrease with increasing latitude, while CUE increases.
Temporal Trends (1982-2018): The annual mean WUE NPP, WUE GPP, and CUE in the YRB showed increasing trends with interannual change rates of 0.008, 0.005, and 0.001, respectively.
Spatial Variation Patterns: Monotonically increasing patterns dominated for WUE NPP (42.44% of the basin) and WUE GPP (41.97%). For CUE, a "decrease then increase" pattern was predominant (42.51% of the basin).
Sustainability and Stability: CWUE in the YRB demonstrated high sustainability. WUE NPP and WUE GPP primarily exhibited moderate fluctuations (52.88% and 73.41% of the area, respectively), while CUE showed very high stability with low fluctuations (97.91% of the area).
Dominant Driving Factors (2000-2018):
- Leaf Area Index (LAI) was identified as the primary determinant for WUE NPP and WUE GPP, with both increasing as LAI increased. LAI dominated approximately 42.80% of the study area for WUE NPP and 45.35% for WUE GPP, particularly in the southern YRB.
- Temperature was the key driving factor for CUE, exerting a predominant influence in 38.88% of the study area, especially in the lower reaches of the basin.
- Temperature was the secondary driver for WUE NPP and WUE GPP. Precipitation and GDP were dominant factors for WUE NPP and WUE GPP, respectively, in specific northwestern regions.

Contributions

Provided a novel, comprehensive characterization of the nonlinear spatiotemporal dynamics of carbon and water use efficiency (CWUE) in the Yellow River Basin (YRB) using advanced interpretable machine learning techniques (EEMD, XGBoost, SHAP), moving beyond traditional linear analyses.
Identified the spatially heterogeneous dominant driving factors for WUE NPP, WUE GPP, and CUE, highlighting the critical roles of LAI and temperature across different regions of the YRB.
Offered a deeper understanding of the complex carbon-water coupling processes in a crucial ecological region under combined natural and anthropogenic pressures.
Furnished strategic guidance for ecological restoration, sustainable water resource management, and achieving carbon neutrality goals within the YRB.

Funding

Postdoctoral Fellowship Program of CPSF (GZC20230732)
Major science and technology projects of Gansu Province (24ZDGE002)
Science and Technology Research Project of Henan Province (242102321167, 252102320300)

Citation

@article{Li2025Dominant,
  author = {Li, Guangchao and Hao, Wenjie and Han, Liqin and Feng, Min and Li, Yanjie and Yi, Zhaoqin and Lu, Yayan and Zuo, Kangjia},
  title = {Dominant drivers of spatiotemporal variations in carbon and water use efficiency across the Yellow River Basin revealed by interpretable machine learning},
  journal = {Frontiers in Plant Science},
  year = {2025},
  doi = {10.3389/fpls.2025.1632172},
  url = {https://doi.org/10.3389/fpls.2025.1632172}
}

Original Source: https://doi.org/10.3389/fpls.2025.1632172