Mechanical phenotyping of cereal crops to assess lodging resistance

Plant mechanical failure, also known as lodging, is the cause of significant and unpredictable yield losses in cereal crops. Lodging can be divided into two distinct failure modes - stalk lodging and root lodging. Despite the prevalence and detrimental impact of lodging on crop yields, there is little consensus on how to phenotype plants for lodging-resistance and breed for mechanically resilient plants. This review focuses on the phenotyping approaches that have been employed for mechanical testing to individually assess stalk and root lodging-resistance. For each approach, the benefits, draw-backs and limitations are discussed. Lastly, this review provides future perspectives as to the next steps in this important, but still developing area of plant phenotyping.


INTRODUCTION
Cereal crops are faced with a complex mechanical challenge -they must be rigid enough to support their own weight, but flexible enough to be resilient in the face of external forces (e.g. wind). Mechanical failure of cereal crops is known as "lodging" and refers to the "permanent displacement of plants from their vertical stance" (Rajkumara, 2008). There are two types of lodging distinguished by the point of mechanical failure -stalk lodging and root lodging.
Stalk lodging refers to breakage of the stem below the height of the ear and root lodging refers to failure at the root-soil interface. A plant cannot recover from stalk lodging, which results in a total loss of yield. In contrast, the impact of root lodging on yield varies according to the developmental stage of the plant at the time root lodging occurs, with yield losses becoming more severe as plants mature (Carter and Hudelson, 1988). Together, both types of lodging are estimated to reduce cereal crop yields by up to 66% (Rajkumara, 2008) and even if harvested, Erndwein et al.,p.3 lodging can reduce grain quality (Mizuno et al., 2018). Factors underlying plant susceptibility to lodging include meteorological factors (e.g. wind, rain, and hail), field management practices (Rajkumara, 2008), plant architecture (Stamp and Kiel, 1992;Brune et al., 2018), and plant biomechanics (Robertson et al., 2016;Sekhon et al., 2019). From an agronomic perspective, stalk and root lodging often occur in the same field and are rarely differentiated. However, from a plant breeding perspective, stalk and root lodging are distinct in their failure mechanisms and the mechanical phenotyping tools being developed to assess each type are similarly distinct.
There are many articles on the topic of lodging, including review articles on lodging in cereals (Berry et al., 2004;Rajkumara, 2008), and a review of how wind forces induce plant damage (Gardiner et al., 2016). Here, we limit our focus to a review of mechanical phenotyping approaches that have been developed to quantify lodging-resistance in cereals.

PLANT FAILURE ANALYSIS OF LODGING
Understanding the failure patterns associated with lodging is critical because lodging can manifest by different causes and mechanisms, which can provide insight into the most appropriate mechanical phenotyping strategies (Robertson et al., 2015a). In general, stalk lodging failure patterns can vary spatially (location of failure within the plant) and temporally (across the plants lifespan). For example, small grains tend to buckle at the lower internodes (Mulder, 1954;Laude and Pauli, 1956;Neenan and Spencer-Smith, 1975), but in barley and oats, buckling of the middle internodes is common and failure can even occur near the peduncle (White, 1991). In contrast, maize stalk lodging tends to occur near a node, but the specific failure pattern can differ by growth stage. For example, analysis of mid-season maize stalk lodging shows that plants fail at the node in a brittle snapping pattern (Elmore et al., 2005). In contrast, late-season maize stalk lodging primarily involves a mechanism known as brazier buckling (Robertson et al., 2015a). This Erndwein et al.,p.4 knowledge has been essential to develop new testing protocols that reproduced this failure patten (Robertson et al., 2014;Robertson et al., 2015a).
Analyses of failure during root lodging have been performed under artificial lodging in winter wheat (e.g. Ennos, 1993, 1994), but there are limited analyses of naturally occurring root lodging. However, one study in wheat observed natural root failure during root lodging and found two modes of failure: roots pulling free of the soil and roots breaking (Easson et al., 1992). Indeed, we observe the same types of failure in maize roots after lodging (unpublished observations), and propose that these two modes of failure are common for root lodging.
Considering the variation in failure modes described above, quantifying lodgingresistance is not a simple or singular task. Several approaches have been developed to evaluate lodging-resistance, including artificial wind and devices that induce stalk or root failure.
The following sections review the methods that have been used to assess field crops for lodging resistance.

ARTIFICIAL WIND TO EVALUATE STALK AND ROOT LODGING
Perhaps the most intuitive approach to study lodging is the use of artificial wind, however there have been limited attempts to establish artificial wind systems. One early study tested the effectiveness of a mobile wind source (an airplane propeller driven by an automobile engine) to evaluate lodging resistance in wheat, oats, and barley (Harrington and Waywell, 1950). This study found that while the artificial wind experiments provided some value to assess lodging resistance, the large size and low-throughput of the wind source made this an unsatisfactory approach to study lodging. A more recent study built a field-based mobile wind machine to study lodging in maize (Wen et al., 2019). Results from this paper suggest again that it may be a good method to evaluate lodging-resistance, but the cumbersome nature of the machine may limit Erndwein et al.,p.5 wide-spread adoption. In commercial breeding applications, Pioneer Hi-Bred has developed a mobile wind machine to select for both root and shoot lodging-resistance (Barrerio et al., 2008).
Although limited in their application, artificial wind approaches are suggested to provide valuable information about lodging-resistance. However, the expensive of construction and cumbersome nature of the artificial wind systems have limited their application.

RIND PENETRATION
The most common approach for assessing stalk lodging resistance is measuring rind penetration resistance. This measurement involves piercing the stalk rind with a probe attached to a digital force gauge (Flint-Garcia et al., 2003a;Flint-Garcia et al., 2003b;Peiffer et al., 2013) and recording the maximum force required to penetrate the rind. This method has been used throughout most of the 20th century and dates back to at least 1935 (Khanna, 1935). However, there are conflicting reports of the utility of the rind penetration procedure to evaluate lodging resistance and it is not widely used by commercial breeding programs. Some studies show that rind penetration resistance is highly correlated with lodging resistance (Anderson and White, 1994;Dudley, 1994), while others show that rind penetration is weakly correlated with lodging resistance (McRostie and MacLachlan, 1942;Butrón et al., 2002;Gou et al., 2007;Hu et al., 2012;Robertson et al., 2017). One of the studies with weak correlation compared the results from rind penetration resistance in maize to laboratory-based stalk 3-point bending strength measurements (Robertson et al., 2017), which closely mirrors the failure pattern of stalks in the field (Robertson et al., 2014;Robertson et al., 2015b). From this analysis, the rind penetration resistance accounts for less than 20% of the variation in stalk bending strength (Robertson et al., 2017). Erndwein et al.,p.6 These conflicting results could be attributed to the fact that, from a biomechanical perspective, rind penetration resistance measurements do not quantify the effect that stalk morphological properties (e.g., diameter, cross sectional area, and rind thickness) have on stalk lodging resistance. Indeed, several studies investigating the genetic architecture of rind penetration resistance have shown that there is not a direct correlation between rind penetration and other morphological features of importance (Butrón et al., 2002;Martin et al., 2004;Gibson et al., 2010;Hu et al., 2012;Li et al., 2014;Ma et al., 2014). Further, it has been suggested that the relationship between rind penetration resistance and stalk strength is highly dependent on growing conditions such as planting density, hybrid-type, and location (Robertson et al., 2016).
A second reason for the contrasting reports on the effectiveness of rind penetration resistance may be the lack of published testing standards. For example, the geometry of the probe and the rate of force application may impact measurements, but are rarely mentioned in papers that utilize the rind penetration method.

LONGITUDINAL AND TRANSVERSE CRUSHING TESTS
A series of papers from the early 1960s to the 1980s utilized two types of crushing tests -longitudinal and transverse -to assess the strength of maize stalks. Longitudinal crushing tests consist of top-to-bottom crushing short (~ 5 cm) segments of maize internodes using parallel plates in a hydraulic press (Zuber and Grogan, 1961;Zuber and Loesch, 1962). Rind thickness and short-span length bending tests (referred to in Zuber and Grogan, 1961 as "breaking strength") were also evaluated, and significant correlations between the mechanical tests and lodging rates were found.
For transverse crushing tests, there are two approaches to apply a lateral force to the stalk of crop stems. The first of these involves a fixture that is very similar to a 3-point bending test except that the span between supports is quite short (6 inches; Durrell, 1925;Jenkins, 1930;Jenkins and Gaessler, 1934). The short-span test induces transverse crushing of the stalk Erndwein et al., p.7 (Robertson et al., 2014;Robertson et al., 2015b) and is therefore categorized here as transverse crushing rather than as a bending test. The second approach to transverse crushing was described in Devey and Russell (1983) as "shearing the second elongated internode in a machine designed to apply a gradually increasing lateral force against the stalk until failure was effected." The details regarding the machine or the type of articulators used to perform the test were not clear, but from the information provided, this approach seems to be similar to the transverse stiffness test described in Robertson et al. 2015. In that study, maize internodes were gently "squeezed" using a universal testing machine in a manner that mimics the manual squeeze test that is used by farmers and breeders to assess stalk strength.
Subsequent studies combine crushing strength, rind thickness, and (occasionally) rind penetration resistance to perform recurrent selection for both high and low values of these various metrics (Thompson, 1963(Thompson, , 1972Zuber et al., 1980). Thompson (1964) examined the crushing strength as a function of the internode, concluding that the crushing strength and rind thickness of each internode are correlated with lodging. Thus, Thompson recommended consistency in which internode was chosen for testing. In each of these studies, recurrent selection led to noticeable changes in both the chosen metric and the associated lodging rates.
However, studies that included grain yield data often showed that increases in crushing strength and/or rind thickness were associated with reduced yield (Thompson, 1972;Devey and Russell, 1983). Of the methods described in this review article, these crushing methods are the slowest because the crushing tests involved significant sample preparation and laboratory testing equipment.

BENDING TESTS
In considering the natural failure pattern of lodged stalks, several field-based measures of stalk bending stiffness and stalk bending strength have been developed. A series of nondestructive field-based bending devices have been developed, each with slight variations to the Erndwein et al.,p.8 next. These devices can be divided into two categories based on whether they have been applied to small grains (e.g. wheat, rice, or barley) or large grains (e.g. sorghum or maize).
A primary challenge with field-based mechanical testing of small grains is the inability of a single stem to provide a sufficient amount of resistance to reliably detect with a load sensor.
To overcome this limitation, multiple plants are tested together in small grain applications. One device (referred to here as Berry's Device) was developed to study winter wheat lodging and consists of a hand-held force meter with a load cell attached to a push bar to measure the force required to push over multiple plants ( Fig. 1A; Berry et al., 2003). The measurement obtained from this device is the force applied to reach a discrete angle. This device has been modified to study the stem mechanics of wheat (Wiersma et al., 2011) and maize (Berry et al., 2000), but the weight of the battery pack and electronic components have prevented wide-spread adoption of this technology.
For larger grains, several different devices have been developed to test individual plant biomechanics in the field. The first device (referred to here as Guo's Device) was developed to measure the forces required to bend maize stalks across a set of discrete angles ( Fig.1B; Guo et al., 2018). In this device, a controller module with a strain sensor is connected by a rigid belt to a second unit fixed to the stalk. The controller module is pulled to discrete angles ranging from 0-to 45-degrees and the maximum equivalent force recorded. This force was shown to have a strong negative correlation with the incidence of stalk lodging (Guo et al., 2018). One feature that sets this apart from other devices for bending is the absence of a grounded base.
More recently, two similar devices -Stalker and DARLING (Device for Assessing Resistance to Lodging in Grains) -were developed to assess stalk biomechanics in larger cereal crops (Jo Heuschele et al., 2019;Cook et al., 2019). Both devices (The Stalker, Fig. 1C; DARLING, Fig. 1D) collect continuous force-rotation data and consist of a vertical support with a control box mounted at the top, a horizontal footplate attached by a hinge at the base, and an adjustable height load cell. To use either device, the operator places a stalk in contact with the Erndwein et al.,p.9 load cell, and a foot on the hinged base to anchor the device to the ground. With the Stalker, the operator pushes the device forward until an LED lights up at a preset angle of 45-degrees and then the device is then returned to vertical (90-degrees). In contrast, DARLING does not have a preset angle and the stalk can be either (1) bent until failure to obtain stalk bending strength, or (2) bent within the linear-elastic range of the material to obtain flexural stiffness.
Both devices are similar in design, reproduce natural stalk lodging failure modes (buckling), are light, portable, and provide reliable measurements for mechanically phenotyping cereal crops. However, the reliability of measurements with both devices depends on soil conditions and soil-type, which should ideally be kept constant throughout testing. Despite many similarities, there are some distinguishing differences. For example, the Stalker begins recording the force-rotation data when the stalk is bent to 45-degrees and is manually ended by the operator when the stalk is returned to the vertical. In contrast, DARLING records data continuously until the operator manually ends the test. Further, DARLING employs a graphical user interface (GUI) to enable real-time assessment of results, whereas Stalker has downloadable data post-test.
Both of these bending devices have provided important mechanical information about stalk lodging-resistance. For the Stalker, the force at a 45-degree rotation (termed the resistance force) was used to successfully differentiate maize populations at varying planting densities, nitrogen application timing, and plant growth regulator application (Jo Heuschele et al., 2019), all of which are known to alter lodging susceptibility. For DARLING, the maximum applied bending strength was used to successfully phenotype a populations of maize hybrids grown in three distinct environments (Sekhon et al., 2019) and found a high correlation with stalk lodging-resistance in all three environments. Erndwein et al.,p.10

VERTICAL ROOT PULLING RESISTANCE
While the failure mechanics of roots during lodging includes both uprooting and breakage, field-based approaches have focused on measuring plant anchorage independent of these failure mechanics. Vertical root pulling resistance (VRPR) is a parameter for assessing root anchorage that has been widely used in maize since the 1930s (Wilson, 1930;Zuber et al., 1971;Fincher et al., 1985). VRPR can be measured rapidly in the field and was shown to be negatively correlated with root lodging in maize (Kamara et al., 2003;Liu et al., 2011). VRPR has been less utilized in other cereal crops, particularly in the context of lodging. One series of studies used VRPR in rice as an approach to understand and select for drought-tolerance (Ekanayake et al., 1985;Ekanayake et al., 1986) and the study, described earlier, that tested the effectiveness of a mobile wind source on lodging resistance in wheat, oats, and barley, also evaluated VRPR. In these studies VRPR was determined to be highly variable and unsatisfactory to predict the tendency to root lodge (Harrington and Waywell, 1950).
An early device to measure VRPR consisted of a clamp and a scale, where the plant is lifted from the soil by manually pushing a lever; this method has proven inaccurate since it was impossible to control for lifting rate and measurements were extracted manually (Rogers et al., 1976;Arndt, 1979;Thompson, 1972Thompson, , 1982Jenison et al., 1981;Penny, 1981;Arihara and Crosbie, 1982;Peters et al., 1982). Other devices have been designed to reduce manual error and measure VRPR using tractor hydraulics, but this approach has proven too heavy and cumbersome for wide-spread measurements (Zuber, 1968;Donovan et al., 1982;Kevern and Hallauer, 1983;Melchinger et al., 1986).
To overcome the limitations identified from these early methods to measure VRPR, two devices were designed that are both portable, relatively accurate, and easy to use (Dourleijn et al., 1988;Fouéré et al., 1995). One device (referred to here as Dourleijn's Device; Fig.2A) uses an electric powered motor and pulley system to pull the plants out of the soil at a constant rate (Dourleijn et al., 1988). The maximum pulling force is recorded as a post-test on an attached Erndwein et al.,p.11 scale. A second device (referred to here as Fouéré's Device; Fig.2B) is anchored by nails into the soil and the stalk is symmetrically placed between the anchor feet (Fouéré et al., 1995). A force sensor then transmits a horizontal pushing force to the stalk and records the resistance force. Mechanical data is recorded as moment-angle relationships and the maximum force applied to pull the root system out of the soil is then extracted as the VRPR. This device represented several improvements upon previous devices including the use of fork prongs to prevent root system damage, nails to anchor the device to the soil, and automated recording of force measurements.

ROOT FAILURE MOMENT
A major drawback of the vertical root pulling systems is that they do not replicate how a plant fails during root lodging. An attempt to improve upon VRPR and replicate root lodging conditions was made with the introduction of a device to measure root failure moment (Rfm).
This approach was originally designed for sunflower (Sposaro et al., 2008), but more recently applied to maize . The devices to measure Rfm consist of a push bar attached to the plant stem with a steel cable at a specific height, a base protractor and an offset pulley system to pull the plant over (Fig. 2C). The Rfm is then calculated as the force when the stalk is pulled to perpendicular multiplied by the attachment height of the push bar. While not widely used, Rfm was shown to be positively correlated with root traits and planting densities, which are known to impact lodging resistance .

HANDHELD PROSTRATE TESTING
An approach to measure root anchorage in small grains has been the use of a commercially available, handheld prostrate testing device (Daiki Rika Kogyo Co., Ltd, Saitama, Japan). In this system, the prostrate device is attached perpendicular to multiple plant stems (10-15), the plants are displaced to a 45-degree angle and the pushing resistance is recorded Erndwein et al.,p.12 ( Fig. 2D). This approach has been applied to winter wheat (Xiao et al., 2015), canola (Wu and Ma, 2016), and rice (Kashiwagi and Ishimaru, 2004). Interestingly, this approach is very similar to the bending tests used to assess stalk lodging, varying only in the placement of the device lower on the stem and closer to the soil surface. While this approach is often presented as a measure of root anchorage, one study notes that is difficult to separate this measure as indicative of root lodging distinct from stalk lodging (Xiao et al., 2015).

MECHANICAL MODELS OF LODGING
Efforts to mechanically phenotype cereal crops have faced both conceptual and physical challenges. Conceptually, many of the approaches that have been used to assess mechanical properties have not been connected to the underlying failure mechanism. Whereas physically, there are limitations imposed by devices, technology, and the physical structure of the plants themselves. For example, small grain stems are so thin and slender that it is difficult to obtain accurate bending strength data of individual plants.
One way to address these challenges is the use of computational models. These models allow researchers to explore hypotheses and carry out "computational experiments" that could not be accomplished using purely empirical approaches, because every aspect of a computational model can be independently manipulated. For example, computational models of root/soil interactions have been used to gain new insights on the root anchorage of pine trees Yang et al., 2017;Yang et al., 2018) and the factors influencing root lodging in maize (Brune et al., 2018).
Computational models can be subjected to sensitivity analyses to develop hypotheses about which factors have the greatest influence on the behavior of interest (e.g. von Forell et al., 2015). These models can be used to suggest and/or test new phenotyping approaches, and to gain valuable insight on the system as a whole. In short, computational models can be used to Erndwein et al.,p.13 develop new phenotyping strategies that are conceptually linked to the behavior of interest (strength, stiffness, etc.). These insights could then be used to design new ways to circumvent the inherent physical limitations associated with various crops and current measurement technologies.

DISCUSSION
To date there has been greater progress in understanding the mechanics associated with stalk lodging than the mechanics associated with root lodging. Most cereal stalks fail by brazier buckling when lodging. Thus, devices that measure bending strength are likely to be the most appropriate for evaluating stalk-lodging resistance. Root lodging is more difficult to simulate due to a lack of devices that apply a rotation moment at the base of the plant. Devices that measure root failure moment are likely the most appropriate, but are currently cumbersome and low-throughput. It should also be noted that several patents exists for devices to measure crop lodging resistance. However, the majority of devices in such patents are not reported in scientific literature (e.g. (胡建东 et al. 2014; 侯瑞锋 et al. 2012; 胡建东 et al. 2017; 胡建东 et al. 2016)) and the efficacy of these devices is difficult to assess.
As technologies to assess plant mechanics in a field setting continue to be developed, there is an urgent need to focus on reproducibility and a complete understanding of the mechanics of plant failure. Several of the approaches outlined above suffer from a lack of reproducibility between labs and/or devices. This lack of reproducibility comes in part from a failure to understand how plants fail during lodging. For example, there is little conceptual relationship between rind puncture resistance (pushing a needle-like instrument laterally through the outer tissues of the stalk) and wind induced failure, which typically manifests as buckling, snapping, or splitting (Robertson et al., 2015a). Rind penetration is one of the most rapid ways of mechanically phenotyping cereal crops, but the lack of connection with the actual failure Lastly, there remains a lack of connection between the mechanical measures described here and the underlying biology. Linking these measures to plant anatomy, architecture, and composition is the key next phase of research in mechanical phenotyping. For example, selection for increased crushing resistance comes with decreased yields, which could be attributed to selection for smaller thicker plants. Understanding how mechanical measures vary with the underlying biology is important to select for plants with improved lodging-resistance without compromising other traits. Figure 1. Devices for measuring stalk bending strength. Berry's device (A) was developed to study winter wheat lodging and consists of a hand-held force meter with a load cell attached to a push bar that measures the resistance force to push over in multiple plants. Guo's device (B) features a hand-held two-component circuit block system and measures the forces required to bend maize stalks across a set of discrete angles. One component, a controller module, contains a strain sensor, single-axis angle sensor, microcontroller, power supply module, a signal acquisition circuit, and a radio frequency transceiver. The second component consists of another radio frequency transceiver and single axis sensor. The two components are connected by a rigid belt, and the controller is pulled to discrete angles to measure the maximum equivalent force (Feq), which is used to assess lodging resistance. The Stalker (C) and the Device for Assessing Resistance to Lodging in Grains (DARLING) (D) were developed to assess stalk biomechanics in larger cereal crops and more closely recreate natural failure Erndwein et al.,p.25 patterns during lodging. Both devices consist of a vertical support with a control box mounted at the top, a horizontal footplate attached by a hinge at the base, and an adjustable height load cell attached. For the Stalker, plants are displaced to a 45-degree rotation and the maximum force is recorded as the resistance force. For DARLING, plants are displaced until failure and the maximum applied bending strength is recorded. Fouéré's device (B) consists of a main frame, handle, adjustable force sensor, angle sensor, a two pronged steel fork with anchoring nails, and control head with electronic display and keys.

FIGURES
This device uses a force sensor to transmit a horizontal pushing force to the stalk and an electronic control system automatically records the resistance force. Sposaro's device (C) was originally developed for sunflower and later applied to maize to improve upon VRPR devices, Erndwein et al.,p.26 and better replicate root lodging failure modes. With this device, a push bar is attached to the plant stem, while a base protractor and an offset pulley system are used to pull the plant over.
Root failure moment (Rfm) can then be calculated. For smaller crops (canola, wheat, rice) a commercially available prostrate testing device (D) can been used. The device attaches to an adjustable mounted plate attached to the plant. Plants are displaced to a 45-degree angle and the pushing resistance is recorded.