Improved remote sensing reference evapotranspiration estimation using simple satellite data and machine learning.

Area of Science:

• Hydrology
• Remote Sensing
• Agricultural Science

Background:

• Accurate reference evapotranspiration (ET0) estimation is vital for water resource management and agricultural planning.
• Limited meteorological data availability poses a significant challenge for traditional ET0 estimation methods.

Purpose of the Study:

• To develop and validate an improved remote sensing-based model for ET0 estimation.
• To assess the model's performance using readily available remote sensing data.

Main Methods:

• Utilized remote sensing data: land surface temperature (LST), day of year, and latitude.
• Employed a machine learning approach, specifically the random forest algorithm.
• Validated the model against the FAO-PM ET0 method at 567 meteorological stations in China and 120 global sites.

Main Results:

• The developed model achieved excellent performance, with R² of 0.97, RMSE of 0.40 in China and an average R² of 0.97, RMSE of 0.43 globally.
• Demonstrated high accuracy, robustness, and generalization capabilities.
• Outperformed or matched existing methods, particularly in data-scarce regions.

Conclusions:

• The remote sensing-based ET0 model offers a simple, accurate, and reliable solution for large-scale, high-resolution ET0 estimation.
• This approach holds significant potential for enhancing irrigation planning, water resource management, and climate change studies.
• Facilitates improved agricultural practices and ecosystem modeling globally.