Analysis of root growth from a phenotyping data set using a density-based model

Abstract

Major research efforts are targeting the improved performance of root systems for more efficient use of water and nutrients by crops. However, characterizing root system architecture (RSA) is challenging, because roots are difficult objects to observe and analyse. A model-based analysis of RSA traits from phenotyping image data is presented. The model can successfully back-calculate growth parameters without the need to measure individual roots. The mathematical model uses partial differential equations to describe root system development. Methods based on kernel estimators were used to quantify root density distributions from experimental image data, and different optimization approaches to parameterize the model were tested. The model was tested on root images of a set of 89 Brassica rapa L. individuals of the same genotype grown for 14 d after sowing on blue filter paper. Optimized root growth parameters enabled the final (modelled) length of the main root axes to be matched within 1% of their mean values observed in experiments. Parameterized values for elongation rates were within ±4% of the values measured directly on images. Future work should investigate the time dependency of growth parameters using time-lapse image data. The approach is a potentially powerful quantitative technique for identifying crop genotypes with more efficient root systems, using (even incomplete) data from high-throughput phenotyping systems.

Keywords: Density-based models; kernel-based non-parametric methods; model validation; optimization; root system architecture; time-delay partial differential equations..