The measurement of the characteristics of living organisms is referred to as phenotyping (Singh et al., 2016). While the use of phenotyping in plant biology and genetics can be traced back at least to Gregor Mendel sorting and counting peas by shape and pod color 160 years ago, addressing current questions in plant biology, genetics, and breeding often requires increasingly precise phenotyping of a wide range of traits. Accurate phenotyping has played a role in both novel discoveries about the fundamental biology of plants and the development of improved crop varieties around the world. With the advent of inexpensive genotyping tools, crop functional genomics has entered the “big data” era, but efficient large‐scale phenotyping is still an impediment hindering plant functional genomics. The precise measurement of plant traits both throughout the growth cycle and across environments is expensive and labor intensive. A convergence of interdisciplinary efforts has led to the development of new technologies for nondestructive phenotyping in plants to measure large numbers of traits accurately with higher throughput (Close and Last, 2011). Improvements in imaging and automation, as well as in data processing and analytics, are helping to fill significant gaps in efforts to employ these new technologies to connect genetic variation with phenotypes (Yang et al., 2020). In recent years, plant phenomics research has transitioned from the development of methods and molecular genetic analysis of model plants in controlled environments toward accelerated efforts for applications in plant breeding, association studies, and stress phenotyping in crops grown under complex field conditions (Costa et al., 2018). In this special issue, “Advances in Plant Phenomics: From Data and Algorithms to Biological Insights,” we present six papers that capture plant phenomics extending to multiple scales, from field‐wide traits, to individual plots or plants, to specific gene interactions.

In the context of field‐scale image acquisition and processing, one of the first challenges that must be addressed in drone‐based imaging of agricultural fields is turning free‐flown images acquired over an area into a single mosaic image from which phenotypes can be extracted. Current methods rely mostly on the ability to locate each pixel in space, requiring costly global positioning systems (GPS) and/or inertial measurement units (IMU) to track the position of ground control points relative to the image acquisition device. These approaches are computationally taxing, demand larger data storage, and require the purchase of software licenses, leading to a high barrier of entry. Aktar et al. (2020) have developed a method called Video Mosaicking and summariZation (VMZ) to provide an alternative pipeline that is faster, less computationally demanding, and much cheaper to implement. The authors show that compared to other methods, VMZ not only works faster but also produces mosaics with superior quality. This work, demonstrated here in maize, begins to democratize drone‐based phenotyping for large‐ and small‐scale field researchers across multiple species.

Although many field‐based plant phenotypes can be aerially captured, proximal sensing techniques can also connect physical phenomena with biology. Resistance to lodging is an essential aspect of plant standability, affecting grain yield and quality. While susceptibility to lodging is known to have a genetic component, differences in the timing and severity of damaging winds and rainfall can make it challenging to quantify lodging consistently from year to year. Hence, despite the prevalence and detrimental economic impact of lodging, there is a lack of adequate protocols for assessing lodging and connecting field‐based mechanical methods to the underlying plant biology. Erndwein et al. (2020) summarize field‐based mechanical phenotyping methods currently used to assess lodging resistance in cereal crops. The authors discuss the complexity of accurately phenotyping natural lodging failures. By also focusing on devices used to measure root lodging in cereals, this review fills the gap and complements the well‐studied aspects of stalk lodging, biomechanics, and lab‐based mechanical phenotyping devices (reviewed in Berry et al., 2004; Niklas and Spatz, 2012; Shah et al., 2017). The authors provide an overview of the current status of field‐based phenotypic approaches, discuss possible reasons for the lack of reproducibility between studies and between devices, and provide suggestions for best practices. This review is accessible to those without a background in biomechanics and enables them to use the right tools or method for measuring lodging resistance in breeding programs, which is necessary to connect field‐based mechanical measurements with the underlying genetic architecture of lodging for crop improvement.

While environmental factors can radically reshape the canopy through events such as lodging, more subtle variations in plant canopy architecture play a substantial role in the efficiency of light capture and utilization, as well as water use efficiency (Peng et al., 2020). Of the many traits that contribute to overall canopy architecture, leaf angle has been the focus of particular genetic investigation as it has been linked to increased grain yield (Duvick, 2005). However, the manual measurement of leaf angle is labor intensive and relatively low throughput and therefore expensive when large numbers of data points are required. Kenchanmane Raju et al. (2020) report the development of a semi‐automated, MATLAB‐based software framework, Leaf Angle eXtractor (LAX), for the estimation of leaf angles from plant images procured using regular 6‐megapixel cameras. LAX is targeted at quantifying subtle changes in leaf angles in individual plants measured at densely spaced time points throughout development to identify spatial and temporal patterns of changes that occur at a shorter timescale of hours and days. The authors quantified leaf angle changes from multiple leaves in individual maize and sorghum plants subjected to water deprivation. LAX leaf angle measurements were able to reveal leaf angle changes coinciding with leaf wilting down to an interval of one minute in each plant. This high resolution provides spatiotemporal information, differentiating individual leaves in the plants that showed wilting from those that did not at any given time point.

One challenge inherent in high spatiotemporal information is how best to quantify differences in patterns of change over time so that informative comparisons can be made between distinct genotypes. Germination is conventionally scored as the percentage of seeds germinated at a single time point after planting. However, particularly for wild species that exhibit seed dormancy, germination can be a complex process with different timing of germination among individuals within genetically uniform populations. While a range of approaches have been proposed to summarize the complex behavior of the germination of a population of seeds, Talská et al. (2020) propose an approach based on smoothing splines and functional regression. Their approach is sensitive to both the proportions of seeds that ultimately germinate and the timing of germination across the window of observation. The authors evaluate their method against previously proposed approaches to quantifying germination behavior using simulated data and experimental data from a set of 105 wild pea accessions.

Another challenge of advanced data acquisition techniques is the deluge of data, often collinear, resulting in more predictors than observations. This challenge is particularly true with hyperspectral data, as reflection at hundreds of wavelengths can be measured on a single sample, necessitating the use of data reduction techniques. However, the potential advantages of hyperspectral imaging outweigh the challenges, as reflectance at wavelengths across different parts of the spectrum lets us “see” chemical changes in plants not yet visible to the naked eye or even standard imaging techniques. Ugarte Fajardo et al. (2020) apply partial least squares–penalized logistic regression and hyperspectral biplot techniques to the biological problem of early detection of black leaf streak disease in banana. Their methods allow for highly accurate and sensitive identification of the disease while still in the presymptomatic phase, before devastating spread and irreversible damage are caused by the fungus.

In the coming years, it seems likely that integrating new high‐throughput phenotyping methodologies to address questions relevant to plant genetics, physiology, and biochemistry will be commonplace. Nepal et al. (2020) demonstrate how a range of new phenotyping approaches developed over the past several years can be employed to address questions of gene‐by‐gene interactions. Nepal and colleagues employed a commercial imaging system to track the growth of plants under control, drought, and stress conditions, and a second technology—based on spectral analysis—that allowed rapid measurements of photosynthetic parameters. Their study employed these technologies to study the interactions of two genes, MIOX4—a myo‐inositol oxygenase, the first enzyme in the pathway for synthesizing ascorbate in plants—and AVP1, which is a pyrophosphate‐energized vacuolar membrane proton pump. The authors found that a combined overexpression line accumulates more biomass than wild type or single gene overexpression lines under drought conditions. They also show that increases in both linear electron flow and photosystem II efficiency accurately predict observed increases in seed count at maturity.

The collection of papers in this special issue represents work extending multiple scales in both biological organization and data analysis. At the biological scale, these papers encompass field‐level traits, traits among individual plants, and gene–gene interactions within a plant. Similarly, from a data analysis perspective, articles in this issue cover aspects including data acquisition and processing, algorithms for data analysis, applications in high‐throughput phenotyping for early detection of stress, and methods for differentiating growth and stress tolerance levels in gene‐by‐gene interaction studies. It seems quite likely that in coming years, just as the tools of molecular biology and, later, bioinformatics originated first in specialized disciplines, many of the phenotyping tools and analytical approaches reflected here will become standard parts of the plant biology toolkit deployed as needed to address questions of plant biology, genetics, and breeding.

衡量生物体特征的方法称为表型分析（Singh等，2016）。尽管表型在植物生物学和遗传学中的应用至少可以追溯到160年前Gregor Mendel按形状和荚果颜色对豌豆进行分类和计数，但要解决植物生物学，遗传学和育种方面的当前问题，通常需要对植物进行更精确的表型分析。性状广泛。准确的表型在有关植物基本生物学的新发现和全世界改良作物品种的发展中都发挥了作用。随着廉价基因分型工具的出现，作物功能基因组学进入了“大数据”时代，但是有效的大规模表型化仍然是阻碍植物功能基因组学的障碍。在整个生长周期和整个环境中对植物性状的精确测量都是昂贵且费力的。2011）。成像和自动化以及数据处理和分析的改进，正在帮助填补使用这些新技术将遗传变异与表型联系起来的重大空白（Yang等，2020）。近年来，植物表型学研究已从控制环境中模型植物的方法和分子遗传学分析的发展过渡到加速在复杂田间条件下种植的作物中进行植物育种，关联研究和胁迫表型研究的努力（Costa等。，2018）。在本期特刊“植物物候学的发展：从数据和算法到生物学见解”中，我们提出了六篇论文，描述了植物物候学的扩展范围，从整个领域的性状到单个地块或植物，再到特定的基因相互作用，扩展到多个尺度。

在现场规模的图像采集和处理的背景下，基于无人机的农田成像中必须解决的首要挑战之一是将在某个区域采集的自由流动图像转换为可从中提取表型的单个镶嵌图像。 。当前的方法主要依赖于在空间中定位每个像素的能力，需要昂贵的全球定位系统（GPS）和/或惯性测量单元（IMU）来跟踪地面控制点相对于图像采集设备的位置。这些方法在计算上很费力，需要更大的数据存储，并且需要购买软件许可证，从而导致较高的进入壁垒。Aktar等。（2020年）已开发出一种称为视频拼接和汇总（VMZ）的方法，以提供一种更快捷，运算量更少，实现成本更低的替代管道。作者表明，与其他方法相比，VMZ不仅工作速度更快，而且还可以产生高质量的镶嵌图。这项工作在玉米中得到了证明，它开始使跨多种物种的大型和小型实地研究人员的基于无人机的表型民主化。

尽管可以空中捕获许多基于场的植物表型，但是近端传感技术也可以将物理现象与生物学联系起来。抵抗倒伏是植物站立性的重要方面，影响谷物的产量和质量。虽然已知倒伏易感性具有遗传成分，但是在破坏性风和降雨的时间和严重性上的差异可能使逐年一致地倒伏量化具有挑战性。因此，尽管倒伏普遍存在并且对经济造成不利影响，但缺乏评估倒伏以及将基于田间的机械方法与基础植物生物学联系起来的适当协议。Erndwein等。（2020年）总结了目前用于评估谷物作物抗倒伏性的基于田间的机械表型方法。作者讨论了准确表型自然倒伏失败的复杂性。通过重点研究用于测量谷物根部倒伏的设备，本研究填补了空白，并完善了对茎倒伏，生物力学和基于实验室的机械表型检测设备的深入研究（Berry等，2004； Niklas和Spatz，2012； Shah等，2017）。作者概述了基于现场的表型方法的当前状态，讨论了研究之间以及设备之间缺乏可重复性的可能原因，并提供了最佳实践的建议。那些没有生物力学背景的人可以进行此综述，并使他们能够使用正确的工具或方法来测量育种程序中的抗倒伏性，这对于将基于田间的机械测量结果与潜在的倒伏遗传结构联系起来对于作物改良而言是必要的。

尽管环境因素可以通过诸如倒伏等事件从根本上重塑冠层，但植物冠层结构中更细微的变化在光捕获和利用的效率以及水的利用效率方面发挥着重要作用（Peng等人，2020年）。在影响整个冠层结构的许多特征中，叶角一直是特定基因研究的重点，因为它与谷物产量的增加有关（Duvick，2005）。但是，人工测量叶片角度是劳动密集型的，并且吞吐量相对较低，因此在需要大量数据点时成本很高。Kenchanmane Raju等人。（2020年）报告了基于MATLAB的半自动化软件框架Leaf Angle eXtractor（LAX）的开发情况，该框架用于从使用常规6兆像素相机获取的植物图像中估计叶片角度。LAX的目的是量化在整个发育过程中在密集间隔的时间点测量的单个植物的叶片角度的细微变化，以识别在较短的小时和天时间尺度上发生的变化的时空格局。作者定量分析了缺水的玉米和高粱植物中多片叶子的叶片角度变化。LAX叶角测量能够揭示出每棵植物中叶片萎缩至每分钟间隔1分钟的变化与叶片角度变化的一致性。这种高分辨率提供了时空信息，

高时空信息固有的挑战是如何最好地量化随时间变化的模式差异，以便可以在不同的基因型之间进行有益的比较。通常，将发芽定为种植后单个时间点发芽的种子的百分比。但是，特别是对于表现出种子休眠的野生物种而言，发芽可能是一个复杂的过程，在遗传上均一的种群中，个体之间的发芽时间不同。虽然已经提出了一系列方法来总结种子种群发芽的复杂行为，但Talská等人（2003）。（2020年）提出了一种基于样条平滑和功能回归的方法。他们的方法对最终发芽的种子比例和观察窗内的发芽时间都很敏感。作者使用来自105个野生豌豆种质的模拟数据和实验数据，对照先前提出的量化发芽行为的方法，评估了他们的方法。

先进的数据采集技术的另一个挑战是大量数据，通常是共线的，导致预测结果多于观测结果。对于高光谱数据，这一挑战尤其如此，因为可以在单个样品上测量数百个波长的反射，因此必须使用数据缩减技术。但是，高光谱成像的潜在优势胜过挑战，因为跨光谱不同部分的波长处的反射率使我们可以“看到”肉眼看不见的植物甚至标准成像技术中的化学变化。Ugarte Fajardo等。（2020年）将局部最小二乘-罚逻辑回归和高光谱双谱图技术应用于香蕉黑叶条纹病早期检测的生物学问题。他们的方法可以在仍处于症状前期的情况下，对病害进行高度准确和灵敏的鉴定，然后再由真菌引起破坏性扩散和不可逆转的损害。

在未来几年中，整合新的高通量表型分析方法来解决与植物遗传学，生理学和生物化学有关的问题似乎很普遍。尼泊尔等。（2020）证明了过去几年中开发的一系列新的表型方法可用于解决基因间相互作用的问题。尼泊尔及其同事采用了一种商业成像系统来追踪在控制，干旱和胁迫条件下的植物生长，并且采用了第二种基于光谱分析的技术，该技术可以快速测量光合参数。他们的研究使用这些技术来研究两个基因MIOX4（一种肌球蛋白）的相互作用。肌醇加氧酶，植物抗坏血酸合成途径中的第一个酶，以及AVP1，这是一种由焦磷酸盐激发的液泡膜质子泵。作者发现，在干旱条件下，组合的过表达系比野生型或单基因过表达系积累的生物量更多。他们还表明，线性电子流和光系统II效率的增加都能准确预测成熟时种子数量的增加。

本期特刊中的论文集代表了在生物学组织和数据分析方面扩展多个规模的工作。在生物学规模上，这些论文涵盖了田间水平的性状，单个植物之间的性状以及植物内部的基因-基因相互作用。同样，从数据分析的角度来看，本期文章涵盖以下方面：数据获取和处理，数据分析算法，高通量表型在早期检测压力中的应用以及在基因组中区分生长和胁迫耐性水平的方法基因相互作用研究。似乎在未来几年中，就像分子生物学的工具以及后来的生物信息学首先起源于专门学科一样，