A best-practice guide to predicting plant traits from leaf-level hyperspectral data using partial least squares...

Area of Science:

• Plant Science
• Ecology
• Biotechnology

Background:

• Partial least squares regression (PLSR) is a statistical method for correlating datasets.
• PLSR can estimate leaf traits from hyperspectral optical reflectance, enabling non-destructive plant phenotyping.
• Current lack of consensus in PLSR methodology complicates interpretation and application.

Purpose of the Study:

• To propose best practices for using PLSR to predict plant traits from leaf-level hyperspectral data.
• To provide recommendations for data collection and model applicability.
• To offer a unified approach for PLSR in plant sciences.

Main Methods:

• Literature review and consensus building on PLSR methodology.
• Development of best practice guidelines for data collection and model application.
• Creation of an accompanying R script tutorial for PLSR model development.

Main Results:

• Established best practices for PLSR application in predicting plant traits from spectral data.
• Provided clear recommendations for data collection and model interpretation.
• Demonstrated PLSR model development with a practical R script.

Conclusions:

• Standardized PLSR methods enhance the reliability of plant trait prediction from hyperspectral data.
• Adoption of proposed best practices will improve consistency and reduce errors in plant phenotyping.
• Unified approaches in PLSR will advance crop breeding and ecosystem monitoring.