Measuring root system traits of wheat in 2D images to parameterize 3D root architecture models

Abstract

Background and aims The main difficulty in the use of 3D root architecture models is correct parameterization. We evaluated distributions of the root traits inter-branch distance, branching angle and axial root trajectories from contrasting experimental systems to improve model parameterization. Methods We analyzed 2D root images of different wheat varieties (Triticum Aestivum) from three different sources using automatic root tracking. Model input parameters and common parameter patterns were identified from extracted root system coordinates. Simulation studies were used to (1) link observed axial root trajectories with model input parameters (2) evaluate errors due to the 2D (versus 3D) nature of image sources and (3) investigate the effect of model parameter distributions on root foraging performance. Results Distributions of inter-branch distances were approximated with lognormal functions. Branching angles showed mean values <90°. Gravitropism and tortuosity parameters were quantified in relation to downwards reorientation and segment angles of root axes. Root system projection in 2D increased the variance of branching angles. Root foraging performance was very sensitive to parameter distribution and variance. Conclusions 2D image analysis can systematically and efficiently analyze root system architectures and parameterize 3D root architecture models. Effects of root system projection (2D from 3D) and deflection (at rhizotron face) on size and distribution of particular parameters are potentially significant. Abbreviations β, root segment angle to the horizontal ∆β, reorientation angle of an individual root segment D e , diffusion coefficient of a solute in soil ibd, inter-branch distance IRC, inter-root competition μ, mean value σ, standard deviation of the random deflection angle (tortuosity) sg, sensitivity to gravitropism std, standard deviation θ, branching angle in the vertical plane

Phenomenal: An automatic open source library for 3D shoot architecture reconstruction and analysis for image-based plant phenotyping

“…Reciprocally, linking models with Phenomenal can greatly improve models. FSPM have been of limited use, due to cost of parameterization/acquisition of data (Landl et al, 2018). HTP platforms inverse the problematic and open the way to calibrate FSPM for different species and/or different genotypes.…”

Simulated Plant Images Improve Maize Leaf Counting Accuracy

“…This is particularly true for larger crop plants, such as maize and sorghum, where the logistics of growing and imaging plants in a controlled setting rapidly become challenging as the number of individuals being grown and imaged increases. As discussed above, simulated data has been shown to aid in improving the accuracy and reducing the biases of models trained to score plant traits from images [16,[19][20][21]. However, the creation of accurate plant morphological models to create these simulated datasets can be challenging.…”

“…A deep learning approach trained entirely on simulated data was shown to accurately count fruits in images collected from real-world plants [19]. The use of simulated training data has also been shown to increase the accuracy of both leaf counting in arabidopsis [16,20] and the estimation of characteristics of three-dimensional root systems from two-dimensional images [21]. Outside of plant science, simulated training data has been used to train neural nets to recognize different stages of galaxy development in astronomical datasets [22].…”

Call for participation: Collaborative benchmarking of functional-structural root architecture models. The case of root water uptake

“…Although the parameterisation of 3D models using a set of parameters derived from 2D images has some limitations, it has been shown to be a simple and efficient strategy allowing the simulation of realistic 3D root systems (Landl et al, 2018). Our reference dataset contains two distinct sets of images: (1) images of lupin roots grown for 11 days in an aeroponic setup (Lobet et al, 2011), and (2) images of maize roots grown for 8 days on filter papers (Hund et al, 2009).…”