PyEOGPR: A Python package for vegetation trait mapping with Gaussian Process Regression on Earth observation cloud...

Area of Science:

• Earth Observation
• Machine Learning
• Vegetation Science

Background:

• Quantifying vegetation traits from satellite data is crucial for environmental monitoring and agroecology.
• Existing methods often require significant local processing and lack uncertainty estimates.
• Gaussian Process Regression (GPR) offers a probabilistic approach with uncertainty quantification.

Purpose of the Study:

• To introduce the PyEOGPR Python package for accessible GPR-based vegetation trait quantification.
• To enable the use of validated GPR models within cloud platforms like Google Earth Engine and openEO.
• To facilitate large-scale vegetation analysis and mapping with uncertainty estimates.

Main Methods:

• Development of the PyEOGPR Python package integrating probabilistic GPR models.
• Application of GPR models to Sentinel-2 and Sentinel-3 satellite data for vegetation trait retrieval.
• Demonstration of landscape and global scale vegetation trait mapping with uncertainty quantification.

Main Results:

• PyEOGPR provides access to 27 validated GPR models for common and challenging vegetation traits, including canopy nitrogen content.
• The package enables efficient, large-scale vegetation trait mapping without local data downloads.
• Generated maps showcase landscape-scale trait distribution using Sentinel-2 and global trait distribution using Sentinel-3 data.

Conclusions:

• PyEOGPR democratizes access to advanced GPR models for vegetation analysis in cloud environments.
• The package enhances the reliability of vegetation trait retrieval through uncertainty estimates.
• PyEOGPR improves the efficiency of Earth Observation data processing for environmental monitoring and sustainable agroecology.