Prediction of wheat SPAD using integrated multispectral and support vector machines

Area of Science:

• Agricultural remote sensing
• Plant physiology
• Machine learning applications in agriculture

Background:

• Accurate chlorophyll content estimation is vital for winter wheat health monitoring and management.
• Unmanned aerial vehicle (UAV) multispectral imagery offers a promising non-destructive method for assessing crop status.
• Previous studies have not fully addressed the influence of growth stages and crop density on SPAD estimation accuracy.

Purpose of the Study:

• To develop and validate a method for rapid estimation of winter wheat Soil and Plant Analyzer Development (SPAD) values using UAV multispectral data.
• To investigate the impact of different growth stages and water stress levels on SPAD estimation accuracy.
• To identify optimal vegetation indices and machine learning models for accurate SPAD prediction.

Main Methods:

• Collected winter wheat multispectral images from natural populations over three years (2020-2022).
• Extracted vegetation indices from UAV imagery and analyzed their correlation with measured SPAD values.
• Developed a Support Vector Machine (SVM) model, incorporating feature selection, to estimate SPAD values during heading, flowering, and filling stages under varying water conditions.

Main Results:

• SPAD values were higher under normal irrigation compared to water-restricted conditions.
• Multiple vegetation indices showed significant correlations with SPAD values.
• The SVM model achieved high estimation accuracy under both normal irrigation (r=0.59-0.81) and drought stress (r=0.69-0.79), with low RMSE and RE values across different environments.

Conclusions:

• UAV-based multispectral imagery, combined with an SVM model and appropriate feature selection, provides a rapid and accurate method for estimating winter wheat SPAD values.
• The developed method is effective across different growth stages and water stress levels, aiding in precise crop management.
• This approach enhances the potential for remote sensing applications in monitoring crop physiological status and optimizing agricultural practices.