Advancing structural-functional modelling of root growth and root-soilinteractions based on automatic reconstruction of root systems fromMRI

根系是土壤-植物相互作用的主要场所，根系构型和根系吸收特性对土壤勘探至关重要。作物育种和管理对地下过程和根部特征越来越感兴趣，这可能会通过减少水和化肥的投入而导致更可持续的农业实践。核磁共振成像是一种非侵入性成像方法，用于获取有关根-土壤系统的结构和功能信息。与数学模型相结合，这将增强我们对根-土壤相互作用的理解。为了有效地从这些体积测量中包含的大量信息中提取关于根系的结构信息，需要用于自动分割和重建根系的新技术。特别是，根段的连通性对于模拟根-土壤系统中的水流非常重要。在这个项目中，我们将利用核磁共振技术同时监测土壤种植的羽扇豆根系构型和不同时间的土壤水分。在人工标注根系的帮助下，将开发一种学习完全自动分割根系的方法。选定样本的CT图像将用于比较和评估新的从MRI图像中分割牙根系统的算法。从重建的根，重要的根结构特征，如基区和顶区、节间距离、曲率和分支角，将被用来改进根生长模型的参数，例如通过改进建议的取向角分布、分支角和轴向生长参数。在水分充足和水分有限的条件下，将评估根系发育和土壤水分条件之间的反馈。这些发现将被用来推进植物根系吸水的数学模型。将使用COMSOL多物理技术对重建的根部结构进行显式3D水流模拟，并根据测量的土壤水分进行评估。然后，这将作为测试不同升级方法的基准，这些方法产生一个有效的从土壤中吸收根部水分的汇项，可以被合并到水流模型中，例如R-SWMS。该项目的主要成果将是一个全自动的根系重建算法，并在此基础上建立一个先进的过程理解和根系水分动力学的数学模型。

Root system architecture and root uptake properties are critically important for soil exploration as roots are the major sites of soil-plant interactions. Crop breeding and management have increased interest in below-ground processes and root traits that may lead to more sustainable agricultural practices through reduced input of water and fertilisers. MRI is a non-invasive imaging method used to obtain both structural and functional information about the root-soil system. In combination with mathematical modelling, this will enhance our understanding of root-soil interactions. New techniques for automatic root segmentation and reconstruction of root systems are required to efficiently extract structural information about the root system from the vast amount of information contained in those volumetric measurements. In particular the connectivity of the root segments is important for modelling water flow in the root-soil system. In this project, we will simultaneously monitor root system architecture of soil-grown lupine plants and soil moisture at different times using MRI. With the help of manually annotated root systems, a method for learning fully automated segmentation of roots will be developed. {\textmu}CT images of selected samples will be used for comparison and evaluation of the new segmentation algorithm for root systems from MRI images. From the reconstructed roots, important root architectural features, such as basal and apical zone, internodal distance, curvature, and branching angles will be derived and used to refine the parameters of a root growth model, e.g. by improving proposal tropism angular distributions, branching angle, and axial growth parameters. Feedbacks between root system development and soil moisture conditions under water-sufficient and water-limited conditions will be evaluated. These findings will be used to advance mathematical models of water uptake by plant roots. Explicit 3D simulations of water flow will be performed on reconstructed root architectures using Comsol Multiphysics and evaluated based on measured soil water contents. This will then serve as a benchmark for testing different upscaling methods that result in an effective sink term for root water uptake from soil that can be incorporated into water flow models such as R-SWMS. Main results of this project will be a fully automated root system reconstruction algorithm and, based on this result, an advanced process understanding and mathematical model of root-water dynamics.