Sign in Subscribe
Climate change

Machine Learning Advances in Estimating Terrestrial Evapotranspiration

Strategic Insights

— 1 min read

A new study has been published on Machine Learning and its application in accurately estimating terrestrial evapotranspiration. Researchers from the Chinese Academy of Sciences have developed a novel approach to estimate evapotranspiration using remote sensing techniques, offering significant insights for global scale estimation.

Key Takeaways:

  • Two SIF-driven coupled ET models, ETPHEM and ETPMEM, were developed by integrating physical constraints with machine learning approaches to improve ET estimation.
  • The ETPMEM model provided the most accurate ET estimates, effectively capturing the seasonal dynamics of ET, with an overall coefficient of determination (R2) of 0.84 and root mean square error (RMSE) of 0.58 mm/day.
  • The coupled ET models demonstrated robust generalization capabilities under various extreme environmental conditions and in data-sparse regions.
  • The study results showed that the two SIF-driven coupled ET models exhibit superior performance compared to the SIF-driven semi-mechanistic ET model and align more closely with ground-based observational data.
  • The researchers suggest that the ETPMEM model can be used for global scale ET estimation, as it accurately captures the seasonal dynamics of ET and has robust generalization capabilities.
  • The study was published in the Journal of Hydrology, with additional information available through contact with the researchers at the Chinese Academy of Sciences.

Statistics:

  • Overall coefficient of determination (R2) of the ETPMEM model: 0.84
  • Root mean square error (RMSE) of the ETPMEM model: 0.58 mm/day
  • Mean absolute error (MAE) of the ETPMEM model: 0.39 mm/day
  • Kling-Gupta efficiency (KGE) of the ETPMEM model: 0.88
  • Multi-year total mean ET calculated by the ETPMEM model for all sites (1.79 mm/day)
  • Multi-year total mean ET observed at the ground stations (1.80 mm/day)

Sources:

  • NewsRx. New Findings in Machine Learning Described from Chinese Academy of Sciences (Improving Terrestrial Evapotranspiration Estimation Using Physics-guided Machine Learning Model Driven By Solar-induced Chlorophyll Fluorescence). Journal of Engineering. November 3, 2025; p 1914.
  • Elsevier. Journal of Hydrology. 661, 2025.
  • Chinese Academy of Sciences. Northwest Institute of Eco-Environment & Resources. State Key Lab Cryospher Sci & Frozen Soil Engn, Qilianshan Observat & Res Stn Cryosphere & Ecol En, Lanzhou 730000, People's Republic of China.

Read more