Partial Least Square Discriminant Analysis Based on Normalized Two-Stage Vegetation Indices for Mapping Damage from...

Area of Science:

• Agricultural Remote Sensing
• Plant Pathology
• Geospatial Analysis

Background:

• Rice co-epidemics cause significant crop losses in Asia.
• Existing remote sensing methods struggle with high-resolution disease discrimination.
• High spatial resolution data is underutilized for detailed rice disease assessment.

Purpose of the Study:

• Develop normalized two-stage vegetation indices (VIs) for characterizing rice disease progression.
• Evaluate the efficacy of combined VIs with Partial Least Square Discriminant Analysis (PLS-DA) for disease classification.
• Map and assess rice disease damage at fine spatial scales using high-resolution imagery.

Main Methods:

• Utilized bi-temporal, high spatial resolution PlanetScope imagery (3 m resolution).
• Developed and applied normalized two-stage vegetation indices to capture disease-induced biophysical changes.
• Employed PLS-DA for sub-field scale classification of rice diseases.

Main Results:

• Combined normalized two-stage VIs effectively captured changes in leaf area, pigment, and canopy morphology.
• PLS-DA achieved 75.62% overall accuracy and a Kappa value of 0.47 in classifying rice diseases.
• Successfully mapped co-epidemics of rice dwarf, blast, and glume blight in Guangxi, China.

Conclusions:

• The developed VIs and PLS-DA approach are effective for fine-scale rice disease detection and mapping.
• This method demonstrates feasibility for monitoring heterogeneous disease patterns over large areas.
• Enhances precision agriculture strategies for managing rice co-epidemics.