Phenomic data-facilitated rust and senescence prediction in maize using machine learning algorithms.

Area of Science:

• Agronomy
• Plant Pathology
• Remote Sensing

Background:

• Traditional methods for assessing maize southern rust and senescence are subjective and labor-intensive.
• There is a need for objective, high-throughput methods for monitoring crop health and development.

Purpose of the Study:

• To evaluate unoccupied aerial system (UAS)-based high-throughput phenotyping (HTP) for measuring maize southern rust and senescence.
• To develop predictive models for early detection of these traits using vegetation indices.

Main Methods:

• Collected high-resolution aerial imagery of maize hybrids over two growing seasons using UAS.
• Extracted 36 vegetation indices (VIs) from mosaicked images.
• Applied machine learning regressions and temporal best linear unbiased predictors (TBLUPs) to predict southern rust and senescence.

Main Results:

• Machine learning models achieved high prediction accuracies (92-98%) and low RMSE for rust and senescence.
• UAS-derived VIs identified early quantitative phenotypic indicators for maize senescence and southern rust.
• Positive correlations were found between grain filling time and yield.

Conclusions:

• UAS-based HTP provides an accurate and efficient alternative to traditional methods for monitoring maize health.
• Early detection of southern rust and senescence is possible through UAS-acquired VIs.
• Findings have practical implications for enhancing precision agricultural practices and optimizing crop yield.