Background: Breeding towards resilient rice varieties is often constrained by the limited data on root system architecture obtained from relevant agricultural environments. Knowledge on the genotypic differences and responses of root architecture to environmental factors is limited due the difficulty of analysing soil-grown rice roots. An improved method using imaging is thus needed, but the existing methods were never proven successful for rice. Here, we aimed to evaluate and improve a higher throughput method of image-based root phenotyping for rice grown under field conditions. Rice root systems from seven experiments were phenotyped based on the "shovelomics" method of root system excavation followed by manual root phenotyping and digital root analysis after root imaging. Analyzed traits were compared between manual and image-based root phenotyping systems using Spearman rank correlations to evaluate whether both methods similarly rank the phenotypes. For each trait, the relative phenotypic variation was calculated. A principal component analysis was then conducted to assess patterns in root architectural variation. Results: Several manually collected and image-based root traits were identified as having a high potential of differentiating among contrasting phenotypes, while other traits are found to be inaccurate and thus unreliable for rice. The image-based traits projected area, root tip thickness, stem diameter, and root system depth successfully replace the manual determination of root characteristics, however attention should be paid to the lower accuracy of the image-based methodology, especially when working with older and larger root systems. Conclusions: The challenges and opportunities of rice root phenotyping in field conditions are discussed for both methods. We therefore propose an integrated protocol adjusted to the complexity of the rice root structure combining image analysis in a water bath and the manual scoring of three traits (i.e. lateral density, secondary branching degree, and nodal root thickness at the root base). The proposed methodology ensures higher throughput and enhanced accuracy during root phenotyping of soil grown rice in fields or pots compared to manual scoring only, it is cheap to develop and operate, it is valid in remote environments, and it enables fast data extraction.

中文翻译：

---

土壤水稻的表型根系结构：结合人工操作和基于图像的分析的鲁棒协议

背景：从相关农业环境中获得的有关根系结构的有限数据通常限制了水稻品种的育种。由于难以分析土壤种植的水稻根系，因此关于基因型差异和根系结构对环境因素的响应的知识有限。因此需要使用成像的改进方法，但是现有方法从未被证明对稻米成功。在这里，我们旨在评估和改进在田间条件下种植的水稻基于图像的表型的高通量方法。根据根系挖掘的“ shovellomics”方法，对来自七个实验的水稻根系进行表型分析，然后进行人工根表型分析和根系成像后进行数字根分析。使用Spearman等级相关性比较手动和基于图像的根表型系统的分析性状，以评估两种方法是否对表型进行相似的排序。对于每个特征，计算相对表型变异。然后进行主成分分析以评估根结构变异的模式。结果：鉴定出几种手工收集的和基于图像的根性状，它们有可能在不同的表型之间进行区分，而其他性状则不准确，因此对水稻不可靠。基于图像的特征投影面积，根尖厚度，茎直径和根系深度可以成功地替代人工确定根系特征的方法，但是应注意基于图像的方法的较低准确性，特别是在使用较旧和较大的根系统时。结论：讨论了两种方法在田间条件下水稻根表型的挑战和机遇。因此，我们提出了一个综合协议，该协议根据水浴中的图像分析和三个特征（即侧向密度，次级分支程度和根基节根厚度）的手动评分相结合，针对水稻根部结构的复杂性进行了调整。与仅使用人工评分相比，所提出的方法可确保在田间或盆栽土壤水稻的根表型分选中获得更高的通量和更高的准确性，开发和操作便宜，在远程环境中有效，并且能够快速提取数据。两种方法都讨论了田间条件下水稻根表型的挑战和机遇。因此，我们提出了一个综合协议，该协议根据水浴中的图像分析和三个特征（即侧向密度，次级分支程度和根基节根厚度）的手动评分相结合，针对水稻根部结构的复杂性进行了调整。与仅使用人工评分相比，所提出的方法可确保在田间或盆栽土壤水稻的根表型分选中获得更高的通量和更高的准确性，开发和操作便宜，在远程环境中有效，并且能够快速提取数据。两种方法都讨论了田间条件下水稻根表型的挑战和机遇。因此，我们提出了一个综合协议，该协议根据水浴中的图像分析和三个特征（即侧向密度，次级分支程度和根基节根厚度）的手动评分相结合，针对水稻根部结构的复杂性进行了调整。与仅使用人工评分相比，所提出的方法可确保在田间或盆栽土壤水稻的根表型分选中获得更高的通量和更高的准确性，开发和操作便宜，在远程环境中有效，并且能够快速提取数据。因此，我们提出了一个综合协议，该协议根据水浴中的图像分析和三个特征（即侧向密度，次级分支程度和根基节根厚度）的手动评分相结合，针对水稻根部结构的复杂性进行了调整。与仅使用人工评分相比，所提出的方法可确保在田间或盆栽土壤水稻的根表型分选中获得更高的通量和更高的准确性，开发和操作便宜，在远程环境中有效，并且能够快速提取数据。因此，我们提出了一个综合协议，该协议根据水浴中的图像分析和三个特征（即侧向密度，次级分支程度和根基节根厚度）的手动评分相结合，针对水稻根部结构的复杂性进行了调整。与仅使用人工评分相比，所提出的方法可确保在田间或盆栽土壤水稻的根表型分选中获得更高的通量和更高的准确性，开发和操作便宜，在远程环境中有效，并且能够快速提取数据。