Zhao et al. (2025) Fusing Enhanced Flux Measurements and Multi-Source Satellite Observations to Improve GPP Estimation for the Qinghai–Tibet Plateau Based on AutoML Techniques

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

Journal: Remote Sensing
Year: 2025
Date: 2025-12-30
Authors: Mengyao Zhao, Ying Yang, Guoyong Weng, Wei He, Hua Yang, Ngoc Tu Nguyen, J. Wang, Shuai Liu, Jiayi Chen, Xinhui Lei, Teng Ma, Ziyi Huang, Peipei Xu
DOI: 10.3390/rs18010130

Research Groups

Not explicitly mentioned in the provided text.

Short Summary

This study developed a data-driven ensemble machine learning model (AutoML-GPP) for the Qinghai–Tibet Plateau (QTP) by integrating eddy covariance data with multi-source remote sensing observations, providing improved gross primary productivity (GPP) estimates that outperform existing global products for the region. The model estimated a mean annual total GPP of 374.20 Tg C yr⁻¹ for the QTP from 2002 to 2018, showing a slight upward trend.

Objective

To develop a data-driven gross primary productivity (GPP) model for the Qinghai–Tibet Plateau (QTP) using machine learning and multi-source remote sensing observations.
To provide improved GPP estimations for the QTP, offering new insights into carbon cycling and climate–vegetation interactions.

Study Configuration

Spatial Scale: Qinghai–Tibet Plateau (QTP).
Temporal Scale: 2002 to 2018 for regional upscaling; 65.2 site years of eddy covariance data used for model development.

Methodology and Data

Models used: Eleven machine learning algorithms from two automated machine learning (AutoML) platforms (H2O AutoML and FLAML) were evaluated to construct an ensemble model named AutoML.
Data sources: 65.2 site years of eddy covariance data from 19 flux sites, multi-source remote sensing observations. Model validation was performed against extracted site-level GPP values from global GPP products (FLUXCOM X-base, GOSIF, and FluxSat).

Main Results

The AutoML model demonstrated strong performance at the site-level across various alpine ecosystems (meadow, steppe, wetland, shrub), achieving a coefficient of determination (R²) up to 0.95 and a root mean square error (RMSE) as low as 0.42 g C m⁻² d⁻¹.
AutoML-GPP showed overall superior or equivalent performance compared to global GPP products (FLUXCOM X-base, GOSIF, and FluxSat) when validated against flux observations.
Regional upscaling estimated a mean annual total GPP of 374.20 Tg C yr⁻¹ for the QTP from 2002 to 2018.
A slight upward trend in GPP of 0.08 Tg C yr⁻¹ was observed over the study period.
Spatially, higher GPP values were predominantly found in the eastern QTP.
The model effectively captured climate-induced interannual anomalies in the QTP’s GPP, with anomalies linked to climate extremes in 2008, 2010, and 2015, coinciding with GOSIF-GPP and FluxSat GPP, and outperforming FLUXCOM X-base.

Contributions

Development of a robust data-driven ensemble machine learning model (AutoML-GPP) specifically tailored for GPP estimation on the Qinghai–Tibet Plateau.
Provision of improved and more accurate GPP estimates for the QTP compared to existing global products.
New insights into the regional carbon cycle and climate–vegetation interactions on the QTP, including spatial patterns, temporal trends, and responses to climate extremes.
Integration of a comprehensive eddy covariance dataset with multi-source remote sensing observations to address GPP estimation uncertainties in a data-scarce region.

Funding

Not explicitly mentioned in the provided text.

Citation

@article{Zhao2025Fusing,
  author = {Zhao, Mengyao and Yang, Ying and Weng, Guoyong and He, Wei and Yang, Hua and Nguyen, Ngoc Tu and Wang, J. and Liu, Shuai and Chen, Jiayi and Lei, Xinhui and Ma, Teng and Huang, Ziyi and Xu, Peipei},
  title = {Fusing Enhanced Flux Measurements and Multi-Source Satellite Observations to Improve GPP Estimation for the Qinghai–Tibet Plateau Based on AutoML Techniques},
  journal = {Remote Sensing},
  year = {2025},
  doi = {10.3390/rs18010130},
  url = {https://doi.org/10.3390/rs18010130}
}

Original Source: https://doi.org/10.3390/rs18010130