Advancing Automated Detection in Neutron Scattering Imaging for Improved Root Structure Analysis of Rhizobox-Grown Salix Cuttings

In this work, we present a sequential procedure to analyze non-destructive neutron scattering imaging, which allows to improve the signal-to-noise ratio and maximize the identification of root structure architecture of plant cuttings grown in rhizoboxes containing humid sand sediment.Plant cuttings play a pivotal role in water bioengineering applications, contributing significantly to enhancing waterway ecohydraulic processes, ecomorphodynamic mechanisms, flood protection strategies, and river restoration techniques. Despite their crucial role, various dynamics related to in-channel and bank vegetation responses to hydromechanical forces remain insufficiently understood. In this study, we conducted flume experiments to systematically examine the impact of hydrodynamic forces and subsurface flow on the root structure development of riparian plant cuttings (Salix species). Specifically, we addressed to role of such actions on triggering dynamotropism and hydropatterning mechanisms, which both lead to (upstream) asymmetrical plant root growth and spatially-variable soil reinforcement. The plant cuttings were cultivated in rhizoboxes filled with sand, which did undergo four different treatments aiming at single out which one better controls dynamotropism and hydropatterning. After a 6-week growth period, we employed a Neutron scanner to scan the boxes and detect the intricate root structures non-destructively. Analysis of the scans revealed significant challenges in distinguishing roots from the background due to factors such as the remaining water content at the pore level, sand texture, and variations in root dimensions. The high spatial variability of these parameters significantly influenced the accuracy of root detection. To address these challenges, we have developed a process utilizing advanced tools to enhance the probability of successfully detecting roots within the sand matrix. The automatization of such a procedure will allow to non-destructively analyse a large number of samples for statistically significant correlation analyses.