High-Throughput Phenotyping for Agronomic Traits in Cassava Using Aerial Imaging.

Area of Science:

• Plant Science
• Agricultural Technology
• Genetics and Breeding

Background:

• Large-scale plant phenotyping is crucial for crop selection and improvement.
• Unmanned aerial vehicles (UAVs) offer a promising tool for high-throughput phenotyping.
• Evaluating aerial imaging for cassava breeding programs requires understanding correlations between vegetation indices and agronomic traits.

Purpose of the Study:

• To estimate correlations between agronomic data and vegetation indices (VIs) from UAVs at various altitudes.
• To select effective prediction models for evaluating aerial imaging in cassava breeding.
• To assess the potential of UAV-based phenotyping for accelerating crop improvement.

Main Methods:

• Acquisition of various vegetation indices (VIs) using multispectral and RGB sensors at different flight heights (20m, 40m, 60m).
• Analysis using mixed models to derive best linear unbiased predictors and heritability.
• Development of prediction models for agronomic traits using selected VIs and machine learning algorithms (GLMSS, KNN).

Main Results:

• Aerial imaging accurately estimated plant height (r=0.99) across flight heights, with lower altitudes providing less bias.
• Multispectral sensors showed stronger correlations than RGB for vigor, shoot yield, and fresh root yield.
• Heritability of VIs ranged from moderate to high (0.51-0.94).
• Best prediction models (GLMSS, KNN) were identified for plant vigor and dry matter content, with predictive ability varying by trait and flight height.

Conclusions:

• UAV-based phenotyping demonstrates significant potential for rapid and efficient selection in plant breeding.
• Flight height and sensor type influence the accuracy and predictive ability of aerial imaging for specific traits.
• This technology can accelerate breeding programs by providing timely and comprehensive agronomic data.