Zhu et al. (2025) A Novel Framework Based on Data Fusion and Machine Learning for Upscaling Evapotranspiration from Flux Towers to the Regional Scale

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

Identification

• Journal: Remote Sensing
• Year: 2025
• Date: 2025-11-25
• Authors: Ping Zhu, Qisheng Han, Shenglin Li, Hao Liu, Caixia Li, Yaoming Ma, Jinglei Wang
• DOI: 10.3390/rs17233813

Research Groups

Not specified in the provided text.

Short Summary

This study developed an integrated framework combining data fusion and machine learning to estimate spatiotemporally continuous evapotranspiration (ET) at a 30 m field scale, demonstrating high accuracy in both homogeneous and heterogeneous landscapes.

Objective

• To propose an integrated upscaling framework that combines data fusion and machine learning to enable spatiotemporally continuous evapotranspiration (ET) estimation at the field scale (30 m × 30 m).

Study Configuration

• Spatial Scale: Field scale, 30 m × 30 m resolution.
• Temporal Scale: Daily.

Methodology and Data

• Models used: Data fusion techniques, machine learning (specifically a one-dimensional convolutional neural network (1D CNN)), footprint model, SHapley Additive exPlanations (SHAP).
• Data sources: MODIS satellite data, Landsat satellite data, China Land Data Assimilation System (CLDAS) datasets, meteorological data, optical-microwave scintillometers (OMS) for evaluation.

Main Results

• The integrated framework successfully generated daily 30 m resolution land surface temperature (LST) and vegetation indices.
• For relatively homogeneous croplands, a 1D CNN using both remote sensing and meteorological data performed best, achieving a correlation coefficient (R) of 0.90, a bias of −0.14 mm/d, a mean absolute error (MAE) of 0.46 mm/d, and a root mean square error (RMSE) of 0.66 mm/d.
• For heterogeneous urban-agricultural landscapes, a 1D CNN using only remote sensing data outperformed other models, with R of 0.93, bias of −0.14 mm/d, MAE of 0.66 mm/d, and RMSE of 0.88 mm/d.
• SHAP analysis identified Land Surface Temperature (LST) and the two-band enhanced vegetation index (EVI2) as the most influential drivers in the models.

Contributions

• Proposes an innovative integrated upscaling framework combining data fusion and machine learning for field-scale (30 m) spatiotemporally continuous ET estimation.
• Addresses the limitations of coarse spatial resolution in existing ET upscaling approaches, enabling utility for precision agricultural water management.
• Provides a robust methodology for ET modeling and spatial extrapolation in heterogeneous regions.

Funding

Not specified in the provided text.

Citation

@article{Zhu2025Novel,
  author = {Zhu, Ping and Han, Qisheng and Li, Shenglin and Liu, Hao and Li, Caixia and Ma, Yaoming and Wang, Jinglei},
  title = {A Novel Framework Based on Data Fusion and Machine Learning for Upscaling Evapotranspiration from Flux Towers to the Regional Scale},
  journal = {Remote Sensing},
  year = {2025},
  doi = {10.3390/rs17233813},
  url = {https://doi.org/10.3390/rs17233813}
}