FUNCTION-VALUED TRAITS FOR CHARACTERIZING AND COMPARING THE SPATIAL CONFIGURATION OF PLANT ROOT SYSTEMS
Date
2024-07-24
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
ORCID
0009-0009-5363-9792
Type
Thesis
Degree Level
Masters
Abstract
Plants are essential and powerful ecosystem engineers. Compared with the above-ground parts of plants, root systems have been studied less frequently as accessing them is arduous and time-consuming. Root traits act as drivers of plant and ecosystem functioning, and root phenotyping is a fundamental procedure in a plant breeding program that helps to identify crop varieties with better root traits. The improved crop plants are tolerant to abiotic stresses, e.g., heat, drought, and nutrient deficiency, playing a vital role in sustainably combating climate change and feeding a rapidly growing global population.
The widely used numeric-valued trait (NVT) approach computes and collects numerical values, such as maximum depth, maximum width, and tip count, from 2D and 3D root system models from various types of images. However, capturing the full complexity of the spatial distribution of the plant’s entire root system by NVTs is always challenging. Additionally, root systems are highly variable in spatial configurations, even among replicates of the same genotype grown in controlled laboratory settings, causing difficulties in making statistical inferences among biological replicates by the NVT approach.
This thesis describes an alternative based on function-valued traits (FVT) to address the abovementioned problems. In particular, the FVTs are related to probability distributions estimated from the image background (i.e., the area not occupied by roots) rather than the foreground (i.e., the pixels corresponding to the plant root tissue per se). Each of these three FVTs is associated with a marked point process and encodes information about the spatial configuration of roots at all length scales. In turn, these probability distributions can be analyzed and compared by using methods of functional data analysis (FDA) to assess systematic differences between plant root systems due to developmental, genetic, and environmental effects.
The practical utility of the FVT approach is demonstrated using collections of images of phosphorus-efficient and phosphorus-inefficient genotypes of sorghum grown in pouches in the presence of sufficient and growth-limiting concentration of phosphorus at 7, 10 and 14 days after transplantation. This is a challenging dataset because the images from distinct sets of biological replicates are easily distinguished “by eye,” but these visually apparent are only poorly reflected in the conventional statistical tests using NVTs. In contrast, the FDA-based statistical tests show that the three FVTs described here can readily distinguish among replicates with visually distinct spatial configurations. FVT pipelines have promising applications in characterizing other 2D and 3D branching structures, including but not limited to leaf veins, above-ground portions of plants such as trees, river systems, and blood vessels.
Description
Keywords
root system architecture, plant phenotyping, plant breeding, spatial configuration of roots, function-valued traits, marked point process, Monte Carlo estimation, lattice random walk, first contact process, survival analysis, functional data analysis
Citation
Degree
Master of Environment and Sustainability (M.E.S.)
Department
School of Environment and Sustainability
Program
Environment and Sustainability