In Situ Root System Architecture Extraction from Magnetic Resonance Imaging for Application to Water Uptake Modeling

An automated method for root system architecture reconstruction from 3D volume datasets obtained from magnetic resonance imaging is developed and validated with a 3D semi-manual reconstruction using virtual reality and a 2D reconstruction using SmartRoot (Lobet et al., 2011). It is tested on the basis of a MRI image of a 25 days old lupin grown in natural sand with a resolution of 0.39× 0.39× 1.1 mm. The automated reconstruction algorithm was inspired by methods for blood vessel detection in MRI images. It describes the root system by a hierarchical network of nodes which are connected by segments of defined length and thickness, and allows also the calculation of root parameter profiles as root length, surface and apex densities. The obtained root system architecture (RSA) varies in number of branches, segments, and connectivity of the segments, but does not vary in the average diameter of the segments (0.137 cm for semi-manual and 0.143 cm for automatic RSA); total root surface (127 cm2 for semi-manual and 124 cm2 for automatic RSA); total root length 293 cm for semi-manual and 282 cm for automatic RSA), and total root volume (4.7 cm3 for semi-manual and 4.7 cm3 for automatic RSA). The differences in performance of the automated and semi-manual reconstruction is checked by using the root system as input for water uptake modelling with the Doussan model (Doussan et al., 1998). Both systems work well and allow for continuous water flow. Slight differences in the connectivity appear leading to locally different water flow velocities, which are below 30%.