Model- and Hyperspectral Imaging-based Techniques to Enhance the Root Phenotyping Pipeline

Plant phenotyping is expected to provide new impetus to crop improvement and stress resistance. Advances in imaging technology have led to high-throughput phenotyping platforms above ground, overcoming limitations in data collection. However, sufficient throughput and resolution remain challenging when targeting crop root systems - particularly when studying mature plants in situ. Thus, beside advanced sensor technology, novel modelling and statistical approaches are needed allowing for root phenotyping data to be increasingly translated to breeding relevant information.

Using European legume accessions as examples, we will show that 1) root system architecture models as a tool can overcome the inference problem of high throughput phenotyping platforms, that 2) machine learning (ML) approaches can be powerful tools to analyze trait variations between cultivars, and that 3) hyperspectral imaging is a promising tool to facilitate automatic root system segregation and even root taxa discrimination.

This contribution will thus provide 1) key root traits of peas to be phenotyped for model-assisted extrapolation from seedling towards mature root system architecture, 2) methodological guidelines on how to use powerful machine learning methods such as random forest models for enhancing the phenotypical exploration of crops such as faba bean, and 3) illustrate how hyperspectral imaging can be used to segregate legume root systems from the soil background in rhizobox studies.

Overall, the presentation will outline how the integration of advanced techniques into the phenotyping pipeline can facilitate the analysis, interpretation and subsequent use of root phenotyping data by breeders and in plant researchers alike.