Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant...

112
Speaker Abstracts General Session I Thursday, February 15, 2018 | 8:15 AM – 12:30 PM Plant phenotyping: overcoming the bottleneck by integrated approaches Roland Pieruschka, PhD, Forschungszenrum Jülich Forschungszentrum, Jülich, Germany Quantitative analysis of structure and function of plants has become a major bottleneck in research and applied use of plants. Approaches targeting relevant traits are needed to quantitatively address key processes and understand the dynamic interactions between genetic constitution, molecular and biochemical processes with physiological responses leading to the development of phenotypes. In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions for accurate measurement of biomass, structure and functional properties of plants across different scales and developmental stages. For instance, I will present the use of automated systems for the cultivation and imaging of model and crop species and demonstrate phenotyping pipelines across scales under controlled and filed conditions. In the second part of the presentation I would illustrate the role of plant phenotyping networks, summarize the recent activities such as the EU funded project EPPN2020 that enables European scientists to access plant phenotyping facilities across Europe and, the ESFRI listed project EMPHASIS that aims at long- term sustainable development of the plant phenotyping infrastructure in Europe. Finally, the International Plant Phenotyping Network, a non-profit association integrates the plant phenotyping community as a global communication hub. From Phenotypes to Mechanisms: Approaching Root Growth Control Using Systems Genetics Wolfgang Busch, PhD, Salk Institute for Biological Studies Salk Institute for Biological Studies

Transcript of Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant...

Page 1: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Speaker Abstracts

General Session I Thursday, February 15, 2018 | 8:15 AM – 12:30 PM

Plant phenotyping: overcoming the bottleneck by integrated approaches

Roland Pieruschka, PhD, Forschungszenrum Jülich

Forschungszentrum, Jülich, Germany

Quantitative analysis of structure and function of plants has become a major bottleneck

in research and applied use of plants. Approaches targeting relevant traits are needed to

quantitatively address key processes and understand the dynamic interactions between

genetic constitution, molecular and biochemical processes with physiological responses

leading to the development of phenotypes.

In this presentation, I will use case studies to demonstrate how plant phenotyping

infrastructure can be used to address relevant biological questions for accurate

measurement of biomass, structure and functional properties of plants across different

scales and developmental stages. For instance, I will present the use of automated systems

for the cultivation and imaging of model and crop species and demonstrate phenotyping

pipelines across scales under controlled and filed conditions. In the second part of the

presentation I would illustrate the role of plant phenotyping networks, summarize the

recent activities such as the EU funded project EPPN2020 that enables European

scientists to access plant phenotyping facilities across Europe and, the ESFRI listed

project EMPHASIS that aims at long- term sustainable development of the plant

phenotyping infrastructure in Europe. Finally, the International Plant Phenotyping

Network, a non-profit association integrates the plant phenotyping community as a global

communication hub.

From Phenotypes to Mechanisms: Approaching Root Growth Control Using

Systems Genetics

Wolfgang Busch, PhD, Salk Institute for Biological Studies

Salk Institute for Biological Studies

Page 2: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

What is the basis for the profound variation of phenotypes and within a single species?

Uncovering the relevant genetic variants and the molecular mechanisms which these

variants affect, would have tremendous implications for a large number of application

ranging from breeding to precision medicine. Using the root system of the model plant

Arabidopsis thaliana, we have approached this problem using a systems genetics

approach that integrates high throughput phenotyping, genome wide association

mapping and functional genomic approaches. We have discovered multiple novel

mechanisms that underlie the natural variation of root traits and have explored the

relation of these gene variants and root traits with climate and soil parameters. Among

the most outstanding mechanisms is a signaling module of Leucine-Rich-Receptor-Like-

Kinases in which natural genetic variation determines root growth responses to low iron

levels. Interestingly, these genes are also involved in defense responses. Overall, our

work demonstrates that systems genetics approaches harnessing existing natural genetic

variation, phenomics as well as modern post-genome-era approaches, allow us to

understand genetic and molecular mechanisms that underlie phenotypic variation and

most likely contribute to local adaption.

Integrating sensor technology and ground platforms: Case studies in

proximal sensing and field phenomics in desert environments

Pedro Andrade-Sanchez, PhD, University of Arizona

University of Arizona

This presentation will provide a brief review of ground-based sensor platforms used for

plant trait characterization, with particular emphasis on systems applied to research

under the dry desert conditions of the US Southwest and irrigated agriculture. A

description of approaches for continuous monitoring of sensor platform position,

measurements of plant spectral and thermal response along with plant geometry will be

included in this talk. Plant canopy height characterization will be presented in more detail

as a case study. Canopy height can be interpreted as one axis of canopy volume, therefore

interpretation of electronic displacement sensor signals is an efficient way to characterize

plant geometry. Canopy height in sorghum is important because of many factors,

maximum plant height usually shows strong relations with net productivity. In cereals,

rapid changes in canopy height are potential indicators of panicle initiation and onset of

grain filling, although this association likely varies with photoperiod and genetics. Since

canopy height determines the working distance between sensors and the crop surface,

accurate measurement of heights is also of value for proximal sensing. Mechanical

actuation is an integral component of sensor platforms that allow adjustments in vertical

frame position. This way, the sensor height may be raised as the crop grows to maintain

a fixed working distance, and height data may be used as covariates in analyses of

Page 3: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

proximal sensing datasets. Plant height also affects crop management, especially in

relation to lodging and mechanical harvest.

Fusion of multi-sensor imagery and machine learning for inspecting and

grading of agricultural products

Todd De Zwaan, PhD, LemnaTec Corporation

Solmaz Hajmohammadi – LemnaTec

Substantial improvements in plant breeding and crop management to feed a projected

population of 9-billion is one of this century’s grand global challenges. Early and accurate

detection and diagnosis of plant diseases, even before specific symptoms become visible

are a key factor in crop yield. This can be achieved by the development of high-resolution

systems equipped with multiple sensors measuring beyond the visible light spectrum.

This is a data intensive approach and demands analytical methods that can cope with the

resolution, size and complexity of the signals from these sensors. This approach also

needs high-throughput capabilities to measure more complex phenotypic information at

higher volumes in production environments.

In recent years, impressive results have been achieved in image detection and

classification that extended the market of computer vision applications in agriculture.

However, any nontrivial machine learning algorithm needs a high-quality dataset. A

result of the ever increasing development of sensors for rapid and non-destructive

assessment of plants is the ability to fuse the output of these sensors to create higher order

datasets. Multi-sensor fusion aims to integrate data collected at different temporal,

spectral and spatial scales to deliver more knowledge content than could be achieved by

each sensor independently.

Phenotyping occurs at laboratory, greenhouse, and field scales. Therefore, the demand

for platforms in each of these settings that have multi-sensor capabilities is high.

LemnaTec is addressing this demand with software and hardware systems that assess

phenotypes of plants and their organs from millimeter to meter scale in laboratory,

greenhouse and field settings. This talk will focus on the sensor fusion methodology in

different platforms using 2D and 3D datasets, and highlight applications of machine

learning tools for segmentation and quality monitoring using hyperspectral imaging.

Page 4: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Rapid gas exchange in the phenomic era

David Hanson, University of New Mexico

Joseph Stinziano – University of Western Ontario

Phenotyping for photosynthetic gas exchange parameters is limiting our ability to select

plants for enhanced photosynthetic carbon gain and to assess plant function in current

and future natural environments. This is due, in part, to the time required to generate

estimates of the maximum rate of ribulose-1,5-bisphosphate carboxylase oxygenase

(Rubisco) carboxylation, the maximal rate of electron transport, and Rubisco activation,

from the response of photosynthesis to the CO2 concentration inside leaf air spaces. To

relieve this bottleneck, we developed a method for rapid photosynthetic carbon

assimilation CO2 responses utilizing non-steady-state measurements of gas exchange.

Using high temporal resolution measurements under rapidly changing CO2

concentrations, we can collect traditional gas exchange parameters in around 2 minutes.

This is a small fraction of the time previously required for even the most advanced gas

exchange instrumentation. We present how we have applied this method to diurnal

changes in physiology as well as responses to light, CO2, and temperature.

Portable, scalable, high throughput geospatial analyses with Singularity

containers on cloud and high performance computing.

Tyson Swetnam, PhD, BIO5 Institute, University of Arizona

Reproducible science with geographic information systems (GIS) on cloud, high

throughput computing (HTC), and high performance computing (HPC) requires portable,

scalable, workflows as part of the Research Object. Here we present a method for running

free and open-source software for GIS; i.e. Geospatial Data Abstraction Library (GDAL),

Geographic Resources Analysis Support System (GRASS), and System for Automated

Geoscientific Analyses (SAGA), in tandem with a workflow management system,

Makeflow, on cloud and HPC using Singularity containers. Our example workflow

involves the computation of daily and monthly sum solar irradiation using an OpenMP

version of the GRASS r.sun algorithm. A single virtual machine (VM) masters the

workflow, with remote workers connected over Internet2 started on cloud, HTC, and/or

HPC platforms, all using the same Singularity container. The workflow is currently

deployed on the OpenTopography.org cyberinfrastructure, where users can select any

location on the terrestrial earth surface using national or global digital elevation model

(DEM) data to calculate global irradiation and daily hours of sunlight. Our workflow links

with OpenTopography via the Opal2 toolkit for wrapping this particular scientific

application as a Web service from a XSEDE Jetstream VM. The workers are launched on

demand on XSEDE Comet HPC and Open Science Grid HTC. Importantly, because the

workflow is containerized with Singularity, it can be re-deployed on any combination of

Page 5: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

local desktop, cloud, or HTC / HPC by simply pulling the code from our GitHub repository

and following a few basic setup instructions. Containerized workflows such as ours that

take an open science approach, as part of the Research Object, will allow for future

reproducible geospatial science on cyberinfrastructure.

Quantifying Nanoscale Biomechanical Properties of the Plant Cuticular

Waxes

Farzad Hosseinali, Texas A&M University, Biological & Agricultural Engineering

Department

The potential applications of Atomic Force Microscope (AFM) in quantifying the

biomechanical properties of plants tissue and membranes, such as the cuticle of tomato

fruits, have been introduced before. However, previous studies on the application of the

AFM in the surface characterization of cotton fiber were mainly focused on the AFM

capabilities in producing high-resolution topography images of either fiber surface or its

cross–section. In fact, cotton fiber cells are covered with a thin cuticular

membrane. The cuticle is mostly made of lipids, alcohols, and fatty acids (collectively

called ‘cotton wax’). The waxy layer can be 10 to 300 nm thick and

imparts hydrophobicity to the fiber surface. The main objective of this study was to

characterize and compare the surface nanomechanical properties of cotton fibers using

various modes of the AFM. Surface topography and friction images of the fibers were

obtained with conventional contact mode. The nanomechanical property images, such

as adhesion and deformation, were obtained with Bruker’s newly developed

high-speed force-volume technique, PeakForce QNM®. The differences in nanoscale

friction, adhesion, and deformation signals can be attributed to fiber surface

hydrophobicity and stiffness, which in turn depend on fatty acids’ hydrocarbon

chain length, film viscosity, and the waxy layer thickness.

Phylogenetic signal in subgenus Rubus (bramble, blackberry) leaflet shape

using geometric morphometrics

Juniper Kiss, Aberystwyth University

Plant phenotypic plasticity and different ways of genetic recombination during clonal and

sexual reproduction make the identification of some plant species difficult. Although DNA

barcoding has revolutionised species identification, polyploidy, hybridisation and

apomixis pose challenges to this process. Subgenus Rubus (brambles, blackberries) is one

of the most taxonomically challenging groups of dicots and their morphology based

classification has not been entirely consistent with their molecular phylogeny. The

definition of bramble species is controversial and is often reliant on leaf and leaflet

characters. Here, we combined geometric morphometrics with molecular analysis.A total

Page 6: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

of 230 leaves from 115 specimens were imaged from different environments (woodland,

sandy beach, saltmarsh, grassland) in the UK. We conducted a three-loci molecular

analysis using ITS(internal transcribed spacer) region of nrDNA and two cpDNA regions,

maturase K (matK) and trnL–trnF for 23 representative leaf samples. We analysed the

shape of five-foliate and three-foliate leaves using landmark-based image analysis. Using

Principal Component Analysis (PCA) and Canonical Variate Analysis (CVA), the leaflet

shapes clustered according to the different environments. Discrimination Analysis (DA)

also confirmed that most of the group mean shapes were highly significantly different (P

< 0.001) at different locations, while it was more obscure when analysed for differences

in between bramble series. Using squared-change parsimony, the molecular phylogeny of

the haplotypes was projected into the leaflet morphospace. Permutation tests suggested

the phylogenetic signal in leaflet arrangement morphology to be statistically significant

(P < 0.05). These results suggest that each haplotype has different shapes in different

environments, while the overall shape differences of haplotypes could be explained by

their phylogeny. We suggest a statistically robust approach to combine morphometric

analysis with molecular data to understand the variability of leaflet shape which could

affect the morphology-based classification of Rubus.

Novel Machine Vision Phenotyping of Maize NAM Plants Reveals

Modulation Effect by Priming Depending on the Cold Temperature

Gokhan Hacisalihoglu, PhD, Florida A&M University

Seedling emergence is an important factor for yield, particularly under challenging

planting conditions. In the US corn belt, maize is planted in early spring, as soon as soil

temperatures are permissive to germination. At that time, temperatures often drop below

normal, which can delay or even kill the seedling. Seed pre-treatments have been shown

to improve germination in cold conditions in crops such as rice and cabbage, but are

largely unpublished in maize. To assess the effects of pre-treatments on early maize cold

tolerance, twenty-seven inbred parents of maize Nested Association Mapping (NAM)

population were primed using a synthetic solid matrix and then tested for cold tolerance

using a soil-based emergence assay. Primed kernels were incubated at 10°C for 5 days,

and then transferred to 24°C for emergence. DSLR cameras were used to capture images

every 30 min to obtain emergence profiles of each seedling. Emergence time was

determined from the time-lapsed images and multiple measures including final

emergence percentage, time to 50% emergence, and emergence rate were extracted for

each genotype. The cold treatment reduced total emergence of several genotypes.

However, priming pre-treatment protected the sensitive genotypes allowing nearly full

emergence. We also used single-kernel near infrared reflectance spectroscopy to

determine seed density, weight, oil, protein, and starch for the kernels prior to planting.

By combining kernel characteristics and emergence time, we found small, but highly

significant correlations between the kernel and early seedling performance.

Page 7: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

A real-time, non-invasive, low-cost monitoring system for plant

phenotyping under stress

Jaderson Armanhi, University of Campinas

Phenotypic data are essential to understanding plant responses to environmental

changes. Conventional instruments to assess plant physiological status are often invasive

or destructive, such as pressure chambers and tensiometers, or designed to provide a

single data point, such as the infrared gas analyzer (IRGA) and the porometer. Although

these methods are reliable, they do not provide continuous monitoring of plant response

to environmental stresses, which might result in losses of relevant information regarding

the true physiological status and plant adaptation mechanisms. Real-time phenotyping

technologies are usually costly, and most platforms are restricted to phenotyping

facilities. Therefore, the development of low-cost phenotyping options is exceptionally

convenient for small-scale studies and experimental setups under growth chamber and

greenhouse conditions. Here we propose a simple and real-time monitoring system for

the remote study of plant physiology using low-cost and easy-to-handle electronic

components. Our system provides the constant monitor of leaf temperature, vapor

pressure deficit (VPD), soil moisture, water loss, as well as the air temperature, relative

humidity and light intensity. An integrated RGB camera was used to record plant

response over time and a modified camera was used to capture near-infrared images for

NDVI measurements. All sensors and cameras are connected to a microcontroller

Raspberry Pi that receives and processes signals and images through custom and

automated scripts. Real-time data are sent to an online server that plots graphs and

creates time-lapse movies on a webpage. Sensors and methods are currently being

validated in experiments designed to evaluate drought stress response in maize. By

providing temporal high-resolution data and imaging, our small-scale system has the

potential to bring valuable information on plant phenotyping in a low-cost manner.

High throughput monitoring anthesis progression of field-grown maize

plants

Seyed Mirnezami, Graduate Research Assistant, Iowa State University

The tassel is the male organ of the maize plant. Sufficient pollen production is crucial for

the production of hybrid seed. Good seed set requires both sufficient daily production of

pollen, but also pollen shed on enough days to ensure a good “nick” with the receptivity

of female inbreds. Traditional approaches for phenotyping anthesis progression are time-

consuming, subjective, and labor intensive and are thus impractical for phenotyping large

populations in multiple environments. In this work, we utilize a high throughput

phenotyping approach that is based on extracting time-lapse information of anthesis

progress from digital cameras. The major challenge is identifying the region of the interest

(i.e. the location of tassels in the imaging window) in the acquired images. Camera drift,

Page 8: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

different types of weather, including fog, rain, clouds, and sun and additionally, occlusion

of tassels by other tassels or leaves complicated this problem. We discuss various

approaches and associated challenges for object detection and localization under noisy

conditions. We illustrate a promising deep-learning approach to tassel recognition and

localization that is based on Region with Convolutional Neural Network (R-CNN). It is

able to reliably identify a diverse set of tassel morphologies. We subsequently extract

time-dependent tassel traits from these localized images.

Dimensionality: Curse or blessing?

David Houle, Florida State University

The ability to acquire phenome-level data is wonderful – hundred dimensional vectors of

data! Our limited human brains, however, think in just three or possibly four,

dimensions, and we like our stories simple. All that raises the question: Once you have

phenomic data, what do you do with it? Many biologists still study phenome level effects

as a very large set of one-to-one mappings between each trait and the experiment or SNP,

depending on the application. Part of the reason that biologists are often stuck in this

mode is the claimed “curse of dimensionality,” where measuring more and more on the

same number of subjects is supposed to become less and less useful. I suggest a taking a

geometric approach where the object of study is not the effects of the SNP or experiment

on each of the hundreds of phenotypes we might measure, but the total length of the effect

vector, and its direction in phenotype space. This mode of thinking follows naturally from

a fully multivariate analysis of the phenotype. A second frontier is the incorporation of

multivariate predictors, for example of a genome-full of polymorphism, or a history of

environmental influences on the phenome itself. For such applications, we need to

introduce biologically-motivated structure to the analysis, for example using

regularization. By choosing the analysis to match our questions, we can escape the curse

of dimensionality, and indeed turn dimensionality to a blessing.

Sensor Systems for Monitoring Horticultural Crops: Challenges and

Opportunities

Reza Ehsani, UC Merced

UC-Merced

Collecting site- or plant-specific data under field conditions has been a challenging and

costly task for researchers and growers. Growers need data for efficient use of crop inputs

and more efficient control of pests and diseases. Scientists need these data for selecting

the best plant in their breeding program or evaluating the effectiveness of different field

practices and treatments. Canopy size and density, growth rate, early detection of pests

and diseases, rootstock and tree age, root size, root density, yield estimation, and yield

Page 9: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

monitoring are examples of data that can be used in a plant production system. In recent

years, there has been a significant progress in the area of Unmanned Aerial Vehicles

(UAVs), multi-band and hyperspectral cameras, wireless sensors, and low-power low-cost

electronic components, and data processing units. These advances resulted in better

sensor systems. This presentation will provide an overview of the sensor system

technologies for tree crops that are commercially available or being developed.

Microneedles as wearable sensors for monitoring plant stress

Philip Miller, Sandia National Labs

Kaitlyn Read – UNM; Dave Hanson – UNM; Ronen Polsky – Sandia National Labs;

Patrick Hudson – UNM

Microneedles are microscale devices primarily used for minimally invasive drug delivery

in humans however our group has utilized them as wearable electrochemical sensors for

health monitoring applicaitons. Recently, we've adapted these sensors for plant stress

monitoring in sorghum and shown that these devices are well tolerated and easily

interfaced with several plant tissue types (leaves, root crowns, and stalks) with little

immune reponse from the plant. Three sensors systems are being developed utilizing our

electromolding method. In this technique, metal microneedles are made from

electroplating into predefined molds that are made from replicating structures fabricated

with two-photon polymerization utilizing laser direct write. With this method arrays of

both hollow and solid microneedles are possible. Hollow microneedles are being adapted

for turgor pressure sensors and solid microneedles for impedance probes and as

metabolite sensors. Initial results show impedance measurements can tract plant drought

stress and recovery via monitoring impedances at low frequencies (0.1-10kHz) between

microneedles in sorghum leaves and a probe in the soil. Microneedle metabolite sensors

have been designed to detect glucose and arrays of microneede sensors allow for

multiplexed detection for spatial mapping. A portable system for remote data logging of

impedance, metabolites, and turgor pressure has begun greenhouse testing and initial

results indicate the system is capable of measuring impedance and environmental

conditions autonomously.

Challenges and opportunities of using satellite imagery to derive insights

for Precision Agriculture applications

Xiaoyuan Yang, PhD, The Climate Corporation

Driving efficiency in agricultural production depends on a number of parameters that are

highly variable in space and time. Digital tools for precision agriculture, targeted at

tailoring subfield decisions, require timely, accurate and scalable input to generate

Page 10: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

insights. Remote sensing technology is a viable solution to achieve this goal. Various

public and private satellite platforms become available with the capability of acquiring

global observation daily or tracking field history for the past 30 years. Customized narrow

spectral bands can be designed for specific applications such as understanding nitrogen

or water limitations to yield. In addition, the unique value of active remote sensing, e.g.

radar and lidar, are being adopted. The Climate Corporation delivers farmer facing field

health imagery in Climate FieldViewTM. Here we will review the challenges and

opportunities of using satellite imagery to derive agronomic insights. Sensor calibration,

preprocessing, cloud/shadow detection, and atmospheric correction etc. are necessary to

provide high quality image input at scale. Selecting proper combinations of spatial,

temporal and spectral resolution to address a specific problem is critical as tradeoffs often

exist in image sources. Combining imagery with weather, soil, and farm management

practices data, image based information layers can be produced applying advanced data

science algorithms. These layers encapsulate the information for decision making, such

as crop monitoring, scouting, stress detection, management zoning and yield prediction.

Moreover, scalable storage/computation platform is in need for running analytics on

enormous amount of image data.

Concurrent I: Robotics Thursday, February 15, 2018 | 2:30 PM – 5:30 PM

Human-Robot Interaction for High Performing Teams in Field Applications

Brittany Duncan, PhD, University of Nebraska, Lincoln

University of Nebraska

This talk will discuss the role of human-robot interaction in field-based robot

deployments and be focused on three individual research areas: integration of robots into

high performing teams, improved teleoperation, and necessary autonomy for improved

team performance. Specific research questions that will be addressed include: 1) What

role does the use of aerial vehicles play in shared decision making with high performing

and potentially distributed teams? 2) How can interfaces and interactions amplify the

current reach of the end users? and 3) What adaptations are necessary within the

autonomy to augment user perceptions in field-based environments? This discussion will

be of interest to researchers and practitioners in agriculture and robotics communities,

as well as those in the fields of human factors, artificial intelligence, and the social

sciences.

Design and Evaluation of a Field-Based High-Throughput Phenotyping

Robot for Energy Sorghum

Sierra Young, Iowa State University

University of Illinois

Page 11: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

This article describes the design and field evaluation of a low-cost, high-throughput

phenotyping robot for energy sorghum. High-throughput phenotyping approaches have

been used in isolated growth chambers or greenhouses, but there is a growing need for

field-based, precision agriculture techniques to measure large quantities of plants at high

spatial and temporal resolutions throughout a growing season. A low-cost, tracked mobile

robot was developed to collect phenotypic data for individual plants and tested on two

separate energy sorghum fields in Central Illinois during summer 2016. Stereo imaging

techniques determined plant height, and a depth sensor measured stem width near the

base of the plant. A data capture rate of one acre, bi-weekly, was demonstrated for

platform robustness consistent for various environmental conditions and crop yield

modeling needs, and formative human-robot interaction observations were made during

the field trials to address usability. This work is of interest to researchers and practitioners

advancing the field of plant breeding because it demonstrates a new phenotyping

platform that can measure individual plant architecture traits accurately (absolute

measurement error at 15% for plant height and 13% for stem width) over large areas at a

sub-daily frequency.

Small Vision Sensors for Phenomics

Sanjeev Koppal, PhD, University of Florida

University of Florida

Biological vision performs amazing visual tasks with negligible power consumption.

Insect eyes for example, allow for optical flow, obstacle avoidance, target tracking,

navigation and even object recognition using micro-watts of power. If robotic drones had

this kind of low power vision, we could imagine massive impact on agriculture and

phenomics. However, despite the fantastic strides in computer vision in recent years,

delivering such high-performance and real-time capability, within tiny power budgets, is

still a distant dream. The reason is that core computer vision algorithms usually follow a

predictable pattern: large amounts of high-resolution imagery and video are combined

with massive amounts of computation. While this achieves spectacular results in many

domains, a new approach is required for the coming wave of next generation miniature

devices. These are micro and nano-scale devices, with feature sizes less than 1mm, that

will soon impact fields as diverse as geographic and environment sensing, agricultural

control and monitoring, energy usage and crop health. This talk is about our work in

solving the core problems that will enable computer vision on miniature platforms.

Allowing these small devices to reliably sense their surroundings has the potential for a

major transformation in phenomics and related fields.

Page 12: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

PlantCV Tools for Hyperspectral Imaging of Abiotic Stress

Malia Gehan, PhD, Donald Danforth Plant Science Center

To tackle the challenge of producing more food and fuel with fewer inputs a variety of

strategies to improve and sustain crop yields will need to be explored. These strategies

may include: mining natural variation of wild crop relatives to breed crops that require

less water; increasing crop temperature tolerance to expand the geographical range in

which they grow; and altering the architecture of crops so they can maintain productivity

while being grown more densely. These research objectives can be achieved with a variety

of methodologies, but they will require both high-throughput DNA sequencing and

phenotyping technologies. A major bottleneck in plant science is the ability to efficiently

and non-destructively quantify plant traits (phenotypes) through time. PlantCV

(http://plantcv.danforthcenter.org/) is an open-source and open development suite of

image processing and analysis tools that could initially analyze images from visible, near-

infrared, and fluorescent cameras. Here we present new PlantCV analysis tools associated

with the development of a hyperspectral and 3D imaging platform aimed at the

identification of early abiotic stress response.

Phenotyping tree shape in the field using computer vision and robotics

Amy Tabb, PhD, USDA-ARS-AFRS

United States Department of Agriculture – Agricultural Research Service

Phenotyping of tree shape is a challenging problem, not least of which because the

traditional metrics of tree shape – height, width, branch number, branch angle, branch

diameter, and branch length, may not be particularly characteristic of the structural

differences that are evident to humans between phenotypes. We describe ongoing work

to develop a robot vision system that captures the above metrics of fruit tree shape

autonomously and accurately, as well as complete tree reconstructions for use in novel

shape descriptors. I will demonstrate how the system operates in field settings, and

describe its constraints and possible applicability to other species.

Electrical Capacitance Tomography (ECT) to Monitor Root Health and

Development and Possible Application in Phenotyping

Daniel Sabo, PhD, Georgia Tech Research Institute

Ga Tech Research Inst

It is becoming increasingly important in plant phenotyping to have an understanding of

root development due to its importance to the health, development, and production

quality of plants. For breeders, it is important to develop cultivars with desired rooting

Page 13: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

traits that contribute to resource use efficiency and improved yield. On the other hand,

information about root health and development would provide needed insight into plant

development and water/nutrient requirements. This means there is a universal need for

a new, nondestructive, and in situ method that monitors root health and development.

Electrical capacitance tomography (ECT) is a nondestructive technique that allows for

this desired root monitoring. We have shown that relative ECT measurements are able to

provide information for root development, insight into speed of root growth, and the

ability to distinguish healthy developing roots from stunted and dying roots,

nondestructively. ECT was also used for presymptomatic detection in bell pepper plants

and various stress effects on the roots. ECT has the ability to provide much needed

information on root health, speed of root growth, stress effects on rooting properties, and

root mass development, making it a desirable sensing technique for plant phenotyping.

Bracing for Impact: The role of aerial roots in plant stability

Erin Sparks, University of Delaware

Damage to plants that prevents them from staying upright, called lodging, can have

a significant impact on cereal crop yield. While there is a large emphasis on

reducing lodging, we understand little about how plants achieve stability. In maize

plants, aerial roots that emerge from the stem above the soil, called brace roots,

are proposed to play an important role in structural stability. Yet how brace roots

develop, integrate environmental cues and contribute to plant stability remains a

poorly understood area of plant biology. Research in our lab focuses on questions

regarding the development and function of maize brace roots. Specifically, we are

taking a structural engineering approach to define the contribution of brace roots

to plant stability. From structural engineering, we know that there are two key

features to building stable structures: the arrangement of the building materials

and the mechanical properties of the building materials. To extrapolate these

features into plants, we have developed a field-based crawling robot for brace root

phenotyping to define the arrangement of building materials. In addition, we are

subjecting brace roots to tension and compression testing to define the mechanical

properties of the building materials. This information is being integrated into structural

engineering models to determine the contribution of brace roots to plant stability. These

experiments are among the first to define the diversity of brace root architecture and

mechanical properties in maize, which is critical to understanding the significance of

these specialized roots in plant stability.

Page 14: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Concurrent II: New Sensors Thursday, February 11, 2017 | 2:30 PM – 5:30 PM

Spatial and spectral data for improved hyperspectral phenotyping

James Janni, PhD, DuPont Pioneer

DuPont Pioneer

Phenotypes have been characterized using both the spectral and spatial information

provided by Pioneer's automated hyperspectral imaging towers. A stressed phenotype

and leaf nitrogen estimations will be used to demonstrate the sensitivity for high

throughput plant characterization. Spatial variation of spectral response will be explored

for increased precision. Spectral indices and the inversion of the PROSPECT model will

be included.

Computer vision and hyperspectral approaches to document temperature

stress responses in maize seedlings

Tara Enders, University of Minnesota

Susan St Dennis – University of Minnesota; Nathan Miller – University of Wisconsin;

Liz Sampson – University of Minnesota; Edgar Spalding – University of Wisconsin;

Nathan Springer – University of Minnesota; Cory Hirsch – University of Minnesota

Yields of maize may be reduced substantially within the next century due to global climate

change. Understanding how maize varieties respond to temperature extremes will be

instrumental in developing varieties that can withstand future abiotic stresses while still

producing high yield. We are documenting the variation of morphological traits, color,

and hyperspectral signals in maize seedlings in response to abiotic stresses in multiple

maize genotypes over time. Morphological measurements, such as plant height, width,

and area, can help characterize the impact of stresses on growth rates. Color data from

RGB images allows for quantification of physiological changes to stress, such as leaf

necrosis, which varies substantially among maize genotypes. Hyperspectral data may

capture valuable information about how genotypes respond to stress that is unable to be

captured using RGB imaging and could provide early detection of stress responses prior

to other manifestations. Documenting multiple traits across genotypes and growth

conditions will uncover the dynamics of maize responses to changing temperatures and

allow for the discovery of genomic loci that could provide improved tolerance.

Page 15: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Tissue specific electrical impedance as a potential screening tool

Kaitlyn Read, University of New Mexico

Patrick Hudson – University of New Mexico; Philip Miller – Sandia National

Laboratories; David Hanson – University of New Mexico

Electrical Impedance Spectroscopy (EIS) is a commonly used noninvasive method to

predict root dimensions, tissue damages, and other physiological parameters. These

methods typically rely on measuring through an electrically variable medium (ie soil,

hydroponic fluid, and epidermal layers), or destructively removing part of the plant. Here

we demonstrate the utility of microneedles to apply EIS methods to specific organs and

tissues in planta. Microneedles were placed on both the adaxial and abaxial surfaces of a

sorghum (Sorghum bicolor) leaf midrib, to measure water storage and water transport

tissues, respectively. An 18-gauge needle was placed 1 cm below the leaf-stalk junction to

function as the signal receiver for both microneedle placements. A handheld LCR meter

supplied a voltage of 0.6V AC, and measured impedance and phase angle at four different

frequencies. Microneedle impedance values were compared to planar metal transducers

as a control, which didn’t penetrate the plant tissue, and impedance values across all

frequencies tested were significantly lower with the microneedle devices. After in planta

EIS measurements were concluded, a fully expanded leaf was removed. Water storage

and water transport tissues were dissected, and EIS measurements were repeated in the

isolated tissues. Impedance was significantly lower in water transport tissue compared to

water storage tissue, both in planta and in isolation. One week after in planta

measurements, leaves showed no adverse response to microneedle applications, other

than superficial callose deposition at the injection site. Our results show that microneedle

EIS can distinguish specific tissues in a non-destructive fashion, and offer a novel

opportunity for high resolution, real-time plant monitoring.

Phenomics of stomata and water use efficiency in C4 species

Andrew Leakey, PhD, University of Illinois at Urbana-Champaign

Andrew D.B. Leakey1*, John Ferguson1, Nathan Miller2, Jiayang Xie1, Charles Pignon1,

Gorka Erice1, Timothy Wertin1, Nicole Choquette1, Maximilian Feldman3, Funda

Ogut4, Parthiban Prakash1, Peter Schmuker1, Anna Dmitrievna1, Dylan Allen1,

Elizabeth A. Ains

University of Illinois at Urbana-Champaign

Water use efficiency (WUE), which is physiologically distinct from drought tolerance, is a

key target for improving crop productivity, resilience and sustainability. This is because

water availability is the primary limitation to crop yield globally and irrigation uses the

largest fraction of our limited and diminishing freshwater supply. The exchange of water

Page 16: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

and CO2 between a leaf and the atmosphere is regulated by the aperture and pattern of

stomata. Mechanistic modeling indicates that stomatal conductance could be reduced or

stomatal movements accelerated to improve water use efficiency in important C4 crops

such maize, sorghum and sugar cane. While molecular genetics has revealed much about

the genes regulating stomatal patterning and kinetics in Arabidopsis, knowledge of the

genetic and physiological control of WUE by stomatal traits in C4 crops is still poor.

Understanding of natural diversity in stomatal traits is limited by the lack of high-

throughput phenotyping methods. To this end two novel phenotyping platforms were

developed. First, a rapid method to assess stomatal patterning in three model C4 species

grown in the field – maize, sorghum and setaria has been implemented. Here the leaf

surface is scanned in less than two minutes with a modified confocal microscope,

generating a quantitative measurement of a patch of the leaf surface. An algorithm was

designed to automatically detect stomata in 10,000s of these images via a training of a

pattern-recognition neural network approach. Second, a thermal imaging capture

strategy, to rapidly screen the kinetics of stomatal closure in response to light has been

developed. We are gaining insight on the underlying genetics governing stomatal stomatal

patterning through quantitative trait loci and genome wide association studies in addition

to phenotypic evaluations of sorghum with transgenically modified expression of stomatal

patterning genes. These multifaceted approaches are complemented by a recently

established field facility for comprehensive evaluation of leaf, root and canopy WUE traits

under Midwest growing conditions.

Phenomics at Scale: Driving Advances in Plant Breeding with Insights from

Diverse Sensor Platforms

Nadia Shakoor, PhD, Donald Danforth Plant Science Center

Donald Danforth Plant Science Center

With the rapid advancement and implementation of robust and high quality genetic and

genomic technologies, the functional analysis of new genomes is currently limited by the

quality and speed of high throughput phenotyping. Ongoing advances in genomics and

high throughput phenotyping creates multiple layers of valuable information that can be

exploited to rapidly advance breeding. In recent years, major contributions from

government and private organizations have been invested in the creation and use of high

throughput tools to speed the development and deployment of phenotyping and breeding

technologies to benefit researchers and farmers. The TERRA-REF program and the

Sorghum Genomics Toolbox, funded by the Department of Energy’s ARPA-E program

and the Bill and Melinda Gates foundation, are employing cutting-edge technologies to

sequence and analyze crop genomes, along with deploying various scales of imaging

platforms (e.g, UAS, tractor-based and indoor and outdoor field scanner systems) to

capture millions of phenotypic observations across growing seasons and diverse

Page 17: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

environments to accelerate crop breeding efforts by connecting those phenotypes to

genotypes.

For example, breeding for cold temperature adaptability is vital for the successful

cultivation of bioenergy sorghum at higher latitudes and elevations, and for early season

planting to extend the growing season. Through the TERRA-REF project, we have

successfully resequenced 400 bioenergy sorghum lines and carried out high throughput

phenotyping to identify candidate genes and alleles that enhance biomass accumulation

of sorghum grown under early season cold stress. Genome wide association studies

(GWAS) of temporal growth data identified potential genes and time specific quantitative

trait loci (QTL) controlling response to early cold stress, permitting an investigation into

the temporal genetic basis of cold stress response at different stages of plant development.

Heliaphen, an outdoor high-throughput phenotyping platform designed to

integrate genetics and crop modeling

Florie Gosseau, LIPM, Universite de Toulouse, INRA, CNRS, Castanet-Tolosan, France

Florie Gosseau – LIPM, Universite de Toulouse, INRA, CNRS, Castanet-Tolosan,

France; Nicolas Blanchet – LIPM, Universite de Toulouse, INRA, CNRS, Castanet-

Tolosan, France; Louise Gody – LIPM, Universite de Toulouse, INRA, CNRS, Castanet-

Tolosan, France; P

Heliaphen is an outdoor high-throughput phenotyping platform allowing automated

management of growth conditions and monitoring of plants during the whole plant cycle.

A robot moving between plants growing in 15L pots monitors plant water balance and

phenotypes plant or leaf morphology, from which we can compute more complex traits

such as the response of leaf expansion (LE) or plant transpiration (TR) to water deficit.

Here, we illustrate the platform capacities on two practical cases: a genetic association

study for yield-related traits and a simulation study, where we use measured traits as

inputs for a crop simulation model. For the genetic study, classical measurements of

thousand-kernel weight (TKW) were done under water stress and control condition

managed automatically on a sunflower bi-parental population. The association study on

TKW in interaction with hydric stress highlighted five genetic markers with one near to a

gene differentially expressed in drought conditions from a previous experiment. For the

simulation study, we used the SUNFLO crop growth model to assess the impact of two

traits measured in the platform (LE and TR) on crop yield in a large population of

environments. We conducted simulations in 42 contrasted locations across Europe and

21 years of climate data. We identified ideotypes (i.e. genotypes with specific traits values)

that are more adapted to specific growing conditions, defined by the pattern of abiotic

stress occurring in these type of environments.

Page 18: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

This study exemplifies how phenotyping platforms can help with the identification of the

genetic architecture of complex response traits and the estimation of eco-physiological

model parameters in order to define ideotypes adapted to different environmental

conditions.

Beyond Orthomosaics: Multi-Image Spectral Analysis of Agricultural UAV

Imagery

Travis Gray, University of Saskatchewan

William van der Kamp – University of Saskatchewan; Travis Gray – University of

Saskatchewan; Steve Shirtliffe – University of Saskatchewan; Hema Duddu – University

of Saskatchewan; Kevin Stanley – University of Saskatchewan; Ian Stavness –

University of Saskatchewan

Imagery from Unmanned Aerial Vehicles (UAVs) is frequently used for crop assessment

and phenotyping in agricultural research fields. In particular, spectral indices, such as

NDVI, are commonly employed to estimate traits such as the health, growth stage, and

biomass in crops. In virtually all such studies, many overlapping images from a UAV flight

are first stitched into a single orthomosaic image, and then spectral indices are derived

from the orthomosaic. But this method necessarily discards (or aggregates) much of the

original pixel information. For example, a single plot may appear in 10 or more individual

images from the UAV, but appear only once in the final orthomosaic. In this paper, we

show that an index value extracted from individual calibrated images of the same plot can

deviate by 10% or more from the index value of the corresponding orthomosaic segment.

This could have important consequences for fine-grained comparison of indices. To

address this problem, we propose alternative approaches to estimating indices, each of

which more directly incorporates all of the individual UAV images. We evaluate these

approaches, and compare them with the orthomosaic approach, by analysing weekly

index values from plant breeding experiments for lentil, wheat, and canola crops in

comparison to relevant manually-measured phenotypes.

Technology Session Thursday, February 15, 2018 | 5:30 PM – 7:10 PM

Tackling the physiological phenotyping bottleneck with low-cost, enhanced-

throughput, do-it-yourself gas exchange and ceptometry

William Salter, , School of Life and Environmental Sciences, Sydney Institute of

Agriculture, The University of Sydney

Page 19: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

William Salter – The University of Sydney; Matthew Gilbert – University of California,

Davis; Andrew Merchant – The University of Sydney; Thomas Buckley – University of

California, Davis

High throughput phenotyping platforms (HTPPs) are increasingly adopted in plant

breeding research due to developments in sensor technology, unmanned aeronautics and

computing infrastructure. Most of these platforms rely on indirect measurement

techniques therefore some physiological traits may be inaccurately estimated whilst

others cannot be estimated at all. Unfortunately, existing methods of directly measuring

plant physiological traits, such as photosynthetic capacity (Amax), have low throughput

and can be prohibitively expensive, creating a bottleneck in the breeding pipeline. We

have addressed this issue by developing new low-cost enhanced-throughput phenotyping

tools to directly measure physiological traits of wheat (Triticum aestivum). Our eight-

chamber multiplexed gas exchange system, OCTOflux, can directly measure Amax with

5-10 times the throughput of conventional instruments, whilst our handmade

ceptometers, PARbars, allow us to monitor the canopy light environment of many plots

simultaneously and continuously across a diurnal cycle. By custom-building and

optimizing these systems for throughput we have kept costs to a minimum, with

OCTOflux costing roughly half that of commercially available single-chamber gas

exchange systems and PARbars costing approximately 95% less than commercial

ceptometers. We recently used these tools to identify variation in the distribution of Amax

relative to light availability in 160 diverse wheat genotypes grown in the field. In a two-

week measurement campaign we measured Amax in over 1300 leaves with OCTOflux and

phenotyped the diurnal light environment of 418 plots using 68 PARbars. These tools

could be readily modified for use with any plant functional type and also be useful in

validating emerging HTPPs that rely on remotely sensed data to estimate photosynthetic

parameters.

Application of Crop Phenotyping to Computation Agronomy at CiBO

Jasenka French, PhD, Cibo Technologies

Cibo Technologies

At CiBO Technologies, we use software and science to solve problems across the whole

agriculture value chain. We are addressing challenges of sustainability, climate change,

and food security by unifying big data and advanced analytics with a fundamental

understanding of the complexities of agriculture.

Crop phenotyping and environment characterization via imaging technologies is an

important part of enabling massive simulations and inference problems that CiBO is

Page 20: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

building the framework to address. We will discuss CiBO’s approach and some challenges

in this demanding quest.

CT image-based Segmentation and Reconstruction of Root Systems by

Machine Learning and Computational Methods

Zheng Xu, PhD, University of Nebraska-Lincoln

Camilo Valdes – Florida International University; Stefan Gerth – Fraunhofer Institute

for Integrated Circuits IIS; Jennifer Clarke, Sleep – University of Nebraska-Lincoln

Computed Tomography (CT) scanning technologies have been widely used in many

scientific fields, especially in medicine and materials research. A lot of progress has been

made in agronomic research thanks to CT technology. CT image-based phenotyping

methods enable high-throughput and non-destructive measuring and inference of root

systems, which makes downstream studies of complex mechanisms of plants during

growth feasible. An impressive amount of plant CT scanning data has been collected, but

how to analyze these data efficiently and accurately remains a challenge.

We present new computational and machine learning methods for better segmentation

and reconstruction of root systems from 3D CT scanning data. We propose new

approaches within the category of voxel thresholding methods. Considering special

characteristics of root systems, we propose our methods based on two new local-feature

statistics, i.e., proportion and weighted proportion. We found that methods based on our

two new statistics can calibrate root system magnitudes faster than traditional vessel-

based approaches while preserve similar levels of performance. In addition, we propose

and evaluate machine learning approaches in root-system segmentation and

reconstruction from CT-images, in particular simulation-assisted machine learning

approaches. We illustrate and compare different approaches using both simulated and

real CT scanning data from Fraunhofer Institute for Integrated Circuits IIS.

PlantEye F500: combine 3D and multispectral information in one sensor

Grégoire Hummel, PhD, CEO, Phenospex B.V.

PlantEye is a high-resolution 3D laser scanner that computes a robust and validated set

of morphological plant parameters fully automatically. A core feature of PlantEye is that

it can be operated in full sunlight without any restrictions - crucial for plant phenotyping

under field conditions or if you follow a “sensor-to-plant-concept”. Phenospex has now

developed a new sensor generation, which combines the actual features of PlantEye on

the fly with a 4-channel multispectral camera in the range between 400 – 900nm. This

Page 21: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

unique hardware-based sensor fusion concept allows us to deliver spectral information

for each data point of the plant in X, Y, Z-direction and we can compute parameters like

NDVI, color index and many other vegetation indices. This new sensor generation opens

a wide range of new possibilities in plant phenotyping and increases its efficiency.

RhizoVision-Crown: An open hardware and software phenotyping platform

for root crowns using a backlight, a machine vision camera, and a new C++

image analysis program

Larry York, PhD, Noble Research Institute

Anand Seethepalli – Noble Research Institute; Haichao Guo – Noble Research Institute;

Marcus Griffiths – Noble Research Institute

Root crown phenotyping, or shovelomics, has become increasingly popular to evaluate

the root systems of crops from the field. Generally, a root crown is excavated and soil is

removed. Earlier methods used a combination of rating and manual measurements such

as number of axial roots, angles of axial roots, lateral root branching density and length,

and diameters. However, imaging followed by image analysis has become increasingly

popular because the techniques are faster and more precise. However, no standard for

imaging has emerged and reproducibility of imaging conditions is difficult, which leads

to images that may be hard to segment and analyze. Here, we describe a new phenotyping

platform that combines custom hardware and software to optimize the imaging of root

crowns and allows fast image analysis with software that is easy to install. The

RhizoVison-Crown hardware platform consists of an extruded aluminum tubular

structure with a 2 ft LED ceiling light panel on one side and a monochrome machine

vision camera on the other. Root crowns are affixed to top panels using a clip and inserted

into a fixed position for imaging, which makes switching crowns fast and ergonomic. The

hardware can be constructed for < $1000. Image acquisition settings are set so the output

of the camera is a quasi-segmented black root crown on a white background that is easily

fully segmented using simple thresholding. A barcode scanner is used to trigger the image

acquisition and the images are stored with the encoded identity using custom software.

The image analysis software is based on OpenCV and written in C++. Extracted features

include root length, area, convex hull, width, depth, diameters, and angles. Examples of

its operation and use in several species will be discussed. This relatively inexpensive and

reproducible system may allow more opportunities for researchers to conduct root

research.

Image Analysis using CyVerse

Blake Joyce, PhD, CyVerse, BIO5 Institute, University of Arizona

Page 22: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Blake Joyce – CyVerse, BIO5, University of Arizona

Phenotyping is poised to surpass genotyping as the next 'big data' challenge. UAVs,

tractors, and remote senors are capable of producing terabytes of data daily for individual

hectares of land. Now that data acquisition is becoming easier, analysis at scale will be the

next bottleneck. CyVerse can provide data storage and management to research teams,

specialized image analysis through CyVerse BisQue, and on-demand cloud computing

through CyVerse Atmosphere. We'll give an overview of image analysis using CyVerse.

Additionally, ongoing development will be discussed for phenotyping-related analyses

like machine learning image feature recognition, analyses running on GPUs, and

integration with geographical information system (GIS) analyses.

Phenotyping for Plant Breeding using 3D Sensors and a Generic 3D Leaf

Model

Oliver Scholz, Fraunhofer Development Center X-Ray Technology

Franz Uhrmann – Fraunhofer Development Center X-Ray Technology; Katharina Pieger

– Fraunhofer Development Center X-Ray Technology; Dominik Penk – University of

Erlangen-Nuernberg; Guenther Greiner, Prof. – University of Erlangen-Nuernberg

We present a setup to objectively assess sugar beet plant traits using multiview

stereoscopy with color cameras with the goal of generating relevant phenotyping

parameters to aid the breeder. The setup was tested in a greenhouse environment as well

as in the field. The assessment is performed using a generic leaf model adapted to the

specific requirements of the sugar beet breeder. The evaluation yields global plant

parameter including plant height, leaf area, leaf count, etc. as well as per-leaf parameters

for each leaf of the plant. The leaf model is based on semantic geometric parameters,

which can directly be interpreted by the breeder.

High Throughput Photosynthesis Characterization of C3 Plants

James Bunce, PP Systems

Dr. James Bunce, Ph.D. in Plant Physiology – PP Systems; Andrew Lintz, B.S. in

Mechanical Engineering – PP Systems

Single point measurements of leaf gas exchange provide basically only the parameters net

photosynthesis, stomatal conductance, and instantaneous leaf water use efficiency. From

analysis of assimilation rate vs. internal CO2 concentrations (A vs. Ci) curves, four or five

additional leaf parameters are obtained for plants with C3 carbon metabolism, which

allow estimation of photosynthesis over a range of conditions. However, determining A

vs. Ci curves conventionally requires at least 20 minutes per leaf, compared with about 2

Page 23: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

minutes for single point measurements, which greatly limits through-put. PP Systems

has developed a method of linearly ramping CO2 rapidly in their CIRAS-3 Portable

Photosynthesis System, which provides a complete A vs. Ci curve in 5 minutes per leaf.

Two initial steps are required: storing the changes in analyzer sensitivity with background

CO2, and collecting data from ramping of CO2 with an empty chamber. These two steps

need only be done once per day. The 5 minute total measurement period per leaf includes

a 2 minute initial equilibration period followed by 3 minutes of ramping up of CO2

concentration until the rate of change of A with CO2 becomes small, i.e. until CO2

becomes nearly saturating to A. With the CIRAS-3 system post-processing of the gas

exchange data is very simple: the apparent “A” of the empty chamber is subtracted from

the “A” value obtained with a leaf in the chamber at each time point of the CO2 ramping

period. This provides the actual A value at each time point, and the Ci is obtained from

this actual A, stomatal conductance, and external CO2 as in the conventional calculation

of Ci. Because of the rapid change in CO2, we have seen no significant change in stomatal

conductance during the CO2 ramps.

In situ phenotyping of root system architecture

Eric Rogers, Doctor of Philosophy, Hi Fidelity Genetics

Root system architecture (RSA) plays a pivotal role in plant fitness and yield yet

remains an untapped resource for crop development. Breeders have primarily

ignored root phenotyping due to an absence of suitable in situ phenotyping

methods. We have developed a device that can detect root growth in real time in

the field, called a RootTracker. The RootTracker runs on battery power and sends

readings wirelessly to a nearby microcomputer that stores and uploads the data to

the cloud for analysis. In its current form factor, the RootTracker can identify

individual roots, and can distinguish growth angles and growth rates. We have

demonstrated long-term deployment of RootTrackers in three maize field sites.

These trials showed we could distinguish root growth rates between two different

maize cultivars. We are currently working to refine our design and add additional

soil property sensors that would be useful to farmers and breeders.

Leveraging Sensors, Probes and Drones to Enable Data Driven Decisions

for Growers

Bruce Schnicker, The Climate Corporation

Technologies and tools that agricultural scientists only dreamed of accessing even as

recently as 10 years ago are now a standard component of our agricultural data acquisition

platform. Sensors, probes, drones, cameras, and a plethora of other technologies are

Page 24: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

routinely used at The Climate Corporation to accelerate the development of value added

models to our digital platform. These technologies are leveraged across multiple testing

formats, including Research & Development, Commercial, Growers, and our own

internally managed research farms. Data and insights from soil, weather, equipment, and

the plants themselves are required to enable development and deployment of predictive

models for agricultural use. Successful utilization of these data layers results in products

that enable growers to make data driven decisions to enhance their performance and

profitability.

General Session II Friday, February 16, 2018 | 8:30 AM – 12:30 PM

Automation and robotics for high-throughput phenotyping and precision

horticulture and agriculture

Rick Zedde, Wageningen University & Research

Wageningen University

Automated plant phenotyping offers plant scientists, breeders and growers a powerful

tool to gather vast amounts of growth data to understand and optimise plant performance

and productivity. For effective use in industry these tools need to be fast, accurate and

objective. Robots developed for phenotyping purposes can be translated to horticultural

production locations and solutions for precision agriculture might benefit phenotyping

purposes. This cross-disciplinary interaction launched as a catalyst a range of novel

technological developments.

Integrating truly transdisciplinary approaches in forming a novel pipeline

between questions and solutions addressing crop stress.

Diane Rowland, Doctor of Philosophy, University of Florida

D.L. Rowland1, A. Zare2, Y. Tseng1, S. Zou2, X. Guo2, H. Sheng2, B. Zurweller1, and R.

Gloaguen1

University of Florida

Critical breakthroughs in crop stress will require transdisciplinary approaches.

Transdisciplinarity follows the concept of “Convergent Research” from the National

Science Foundation: utilizing deep integration across disciplines by immersion in cross

disciplinary language, techniques, strategies, and constant team interaction – efforts all

aimed at one compelling problem. This contrasts “interdisciplinary” research, where

Page 25: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

groups remain in disciplinary silos, approaching research objectives in isolation with

team members meeting periodically to simply report ongoing results. Achieving

transdisciplinarity takes a period of integration and education among disparate

disciplines, and confronting existing research paradigms. Research groups within the

Center for Stress Resilient Agriculture and the Machine Learning and Sensing Laboratory

have formed a transdisciplinary group aimed at solving critical problems related to crop

stress. Efforts include group study of primary literature in both crop science and machine

learning and application of these concepts in an integrated approach to address specific

research challenges. We will present two specific examples where this approach has

brought success in addressing deeper problems not able to be solved through one

disciplinary tactic alone. The first is the utilization of hyperspectral images for

determining peanut maturity by sensing mesocarp color indicative of pod development.

This approach can address risk for aflatoxin, a mycotoxin known to be responsible for

significant levels of liver cancer across the globe, and which is produced when the crop

experiences drought stress. The second is the utilization of root modeling and stochastic

optimization approaches for determining the relationship between root architecture and

root function – an approach that challenges the current paradigm about rooting depth

and branching serving as adequate surrogates for predicting root function by directly

testing the validity of this assumption. These case study approaches show the success of

taking a truly transdisciplinary approach and the significant advances that can be

achieved when doing so.

CIAT Phenomics Platform: Aiming at improving Eco-efficiency of crops in

the changing global climate

Michael Selvaraj, PHD, International Center for Tropical Agriculture

Nowadays analyzing the phenotype is frequently slower and more expensive than

genomics due to the difficulties of measuring plant behavior at different levels and under

different environements. Thus phenotyping becomes the limiting factor for plant biology

and crop improvement. Our knowledge on the link between genotype and phenotype is

currently hampered by insufficient capacity of the plant science community to analyze the

existing genetic resources for their interaction with the environment. Advances in

developing plant phenotyping methods and tools are therefore essential for success in

characterizing shoot and root phenes to design next generation crops and forages as key

components for climate-smart or eco-efficient agriculture. Constraints in field

phenotyping capability limit our ability to dissect the genetics of quantitative traits,

particularly those related to yield, biotic and abiotic stress tolerance. The development of

effective field-based phenomics platforms remains to be a bottleneck for future advances

in genetic gain for yield and nutritional quality. However, progress in remote sensing

technology and high-performance computing are paving the way. The CIAT field

Phenomics platform at CIAT-HQ is a state-of-the-art, high-tech facility comprised of

Page 26: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

automated rainout shelters (for drought screening) and low nitrogen field plots (for

Nitrogen use efficiency screening) integrated with multi spectral imaging and Terrestrial

Laser Scanning (TLS) system mounted on phenotowers, roof of rainout shelters and

unmanned aerial vehicles (UAV). This automated, high-throughput platform allows

repeated non-destructive image capture and multi-parametric analysis of small to

medium sized field plots at multiple time points. CIAT phenotyping platform is also

developing the capacity to estimate root yield in cassava using Ground Penetrating Radar

(GPR) technology. The mounting of multi-spectral camera to a drone (UAV) can

potentially harness the full capability of proximal sensing in a reliable, flexible, and

efficient system that operates spatially at small to bigger plots. Combining this approach

with environmental characterization as (Climate, soil and management status of the

crop), with GPS positioning to spatially locate the proximal sensing data and with

automated image analysis thus appears capable of delivering a robust field based

phenomics platform.

Quantitative imaging and dynamic models of plant stem cells

Ross Sozzani, PhD, NCSU

R. Sozzani1, M. A. de Luis Balaguer1; N. Clark1, 2; A. Fisher1;

1. Department of Plant and Microbial Biology, North Carolina State University;

2. Biomathematics Graduate Program, North Carolina State University;

The stem cells in the tip of the Arabidopsis root form all the root tissues by undergoing

rounds of coordinated cell division while maintaining their undifferentiated state. A

better understanding of the transcription factors that maintain the stem cells, and control

each stem cell’s identity, would give us more insight into how the growth and development

of the root is initiated. While a number of transcription factors involved in root stem cell

maintenance have been described, a comprehensive view of the transcriptional signature

of the stem cells is lacking. We have generated a model of the transcriptional mechanisms

underlying the identity and maintenance of the Arabidopsis root stem cells that links

known and newly predicted factors involved in these processes. We have gained

quantitative insights into these key factors by accurately measure molecular dynamics,

such as intercellular trafficking and protein-protein interactions, which allowed us to

generate models that capture the behavior of the system. This model led to a map of

genetic interactions that orchestrate the transcriptional regulation of stem cell

maintenance.

Page 27: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Plant Root Quantitative Analysis

Fuqi Liao, MA, The Noble Research Institute

Root quantitative analysis has become more and more important in plant research.

Currently, many biologists manually measure the root length with help of software (for

example, WinRhizo), but it is limited to small numbers of roots and costs a lot of time.

Many research projects (for example, GWAS) need to quantify a large number of root, and

require software to detect and process root images automatically and finish analysis in

short time. In many situations, precise measurements could be made by software, instead

of manual analysis. Here, we report development of a series of software, which automated

root image analysis with parallel computing on High Performance Computing (HPC)

cluster. The software have been applied to thousands of images of Arabidopsis and

Medicago truncatula roots. With many parameters, which quantifying the roots, the

automated feature of the software allows analysis of thousands of root samples within a

short period of time. The software can detect nearly 90% of the root hairs, whose length

is less than one millimeter, and measures the root hair lengths. For roots, which were

grown for two weeks in the lab, the software can detect tips of lateral roots, and analyze

the lengths of main root, and multi-layers of lateral roots. The software also quantifies the

roots from the field with many parameters, including root area, root angle, and total

length, etc. The series of software include many algorithms to provide precise detection

and measurement of the root’s features. In addition, it includes statistical methods, such

as ANOVA and 95% CI, to find significant difference among root groups with genotypes

such as natural variants and developmental mutants. The artificial neural network will

soon be added to classify root groups.

Field Based Phenotypic Platform for Characterizing Maize Growth and

Development

Sara Tirado, University of Minnesota

Advances during the past decade have allowed researchers to link genomic information

to phenotypic information and through

this enhance crop productivity. Further progress in the ability to link genotypes,

environments, and phenotypes has been limited by the accuracy and consistency of

measuring traits of agronomic importance on a large field-based scale. Current field-

based phenotyping efforts are time and labor intensive and therefore hinder the

development of large-scale trait datasets at multiple timepoints throughout a plant’s life

cycle. This brings a need for developing fully automated, inexpensive procedures for

objectively measuring plant traits in field settings. We have developed a procedure for

utilizing RGB drone imagery to extract phenotypic traits of importance, including stand

count, plant height, canopy closure, and growth rate. This platform was used to

Page 28: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

characterize a maize association mapping population that consists of 500 diverse inbred

lines grown in replicate in Saint Paul, MN in 2017. In total five flights were conducted

throughout the growing season. Results of repeatability across replicates and variation in

growth rates as measured through plant height and canopy closure will be presented. The

application of this technology can be used to deepen our understanding of how genetic

variation and environmental influences shape the traits of corn plants in the field.

The genetic and mechanistic bases of photosynthetic cold tolerance in

legume, cowpea (vigna unguiculata (l.) walp.) via high throughput

environmental phenotyping

Donghee Hoh, MSU-DOE Plant Research Laboratory

Increasing crop production will require improvements in the efficiency, robustness, and

sustainability of photosynthesis. Among the most critical abiotic stresses that impact

photosynthesis is temperature. Cowpea (Vigna unguiculata (L.) Walp.), an important

protein source worldwide especially in developing countries was chosen as a model crop,

is especially sensitive to high temperature during reproductive development result in

severe crop losses by causing male sterility and fruit abortion. One proposed approach to

minimizing the impact of heat stress is to plant earlier than the normal to avoid extreme

heat stress later in the season. This strategy involves planting at colder temperatures,

which is generally known to retard germination and emergence in Cowpea. A major

concern is how the chilling stress impacts plant performance/yield. The goal of this

research is to identify genes and mechanisms related to chilling stress tolerance. Cowpea

has potentially sufficient genetic variation that we can use to identify quantitative trait

loci (QTL) that can be used to guide plant breeding, and test mechanistic models to

explain this sensitivity. Two sets of recombinant inbred lines (RILs) were phenotyped

using the Dynamic Environment Phenotype Imager (DEPI) and MultispeQ under control,

cold temperature conditions, under lighting conditions that simulate typical daylight. We

found considerable natural variation in the photosynthetic parameters which were used

to identify QTLs specific to cold tolerance that can be exploited for quantitative trait locus

(QTL) mapping and subsequent breeding efforts. Phenotyping and QTL data showed cold

sensitivity is strongly associated with increased photoprotective responses (qE and qI)

that is strongly linked to variations in the chloroplast ATP synthase activity, leading us to

propose a model for low-temperature photodamage that involves control of proton motive

force-induced damage to photosystem II. We also identified potential genetic control

element that could explain the adaptation of certain variants to low temperature.

Page 29: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Deciphering the association of phenome and gene expression postulating

salt tolerance mechanism in a rice landrace, Horkuch

Sabrina Elias, , University of Dhaka and University of Nebraska Lincoln

Rice production in the salty cultivated soil cannot meet the demand of the overgrowing

population as rice plant and excess salt has a rival relationship. Changes in climate is

worsening the scenario by introgression of excess salt in more and more cultivable lands.

But rice, as a major staple food need to maintain the balance in the production requiring

the need of high yielding salt-tolerant rice. Understanding the mechanism of salt tolerant

landraces with adaptive capability to withstand the harsh environment can give insights

on potential candidate genes for conferring tolerance. In order to do so, the reciprocal

cross of a salt tolerant landrace Horkuch and high yielding but sensitive variety, IR29 has

been analyzed. A set of the reciprocal F2:3 population was genotyped using DArTSeq™

for discovering SNP markers to construct linkage map and manual phenotyped under salt

stress. We have identified expression QTLs (eQTL) combining the genotyping and

RNAseq data of a subset under 150mM salt stress. An image-based non-destructive

automated and continuous phenotyping over 3 weeks of salt stress was carried out on a

selected F3 and F5 sub-populations followed by QTL identification for the digital traits

and relative growth rates from visual image data. Instead of endpoint records, image

analysis over days gave us longitudinal data, which could separate the early and late

responses to salt stress. Combining the phenomics and eQTL data, early growth indices

were found to be enriched with transport, osmotic response etc and the later stages were

enriched with genes associated with growth, carbohydrate metabolism, organ

development etc. The phenome data along with the expression data could give a

comprehensive scenario regarding potential candidates involved in tolerance mechanism.

Computational Classification of Phenologs across Biological Diversity

Ian Braun, Iowa State University

Phenotypic diversity analyses are the basis for research discoveries that span the

spectrum from basic biology (e.g., gene function and pathway membership) to applied

research (e.g., plant breeding). Phenotypic analyses often benefit from the availability of

large quantities of high-quality data in a standardized format. Image and spectral

analyses have been shown to enable high-throughput, computational classification of a

variety of traits across a wide range of phenotypes. However, equivalent phenotypes

expressed across individuals or groups that are not anatomically similar can pose a

problem for such classification methods. In these cases, high-throughput, computational

classification is still possible if the traits and phenotypes are documented using

standardized, language-based descriptions. In the case of text phenotype data, conversion

to computer-readable “EQ” statements enables such large-scale analyses. EQ statements

are composed of entities (e.g., leaf) and qualities (e.g., length) drawn from terms in

Page 30: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

ontologies. In this work, we present a method for automatically converting free-text

descriptions of plant phenotypes to EQ statements using a machine learning approach.

In each description, words related to entities and qualities are identified using the

CharaParser annotation tool (Cui, 2012). A classifier identifies potential matches between

these words and terms from a set of ontologies, including GO (gene ontology), PO (plant

ontology), and PATO (phenotype and trait ontology), among others. The features used by

this classifier include semantic, syntactic, and context similarity metrics between words

and ontology terms. This classifier is trained and tested using a dataset of manually

converted plant descriptions and EQ statements from the Plant PhenomeNET project

(Oellrich, Walls et al., 2015). The most likely matching terms identified by the classifier

are used to compose EQ statements. Any obtained results of these automated conversions

in terms of precision and recall will be presented. Potential use across datasets to enable

automated phenolog discovery are discussed.

Machine vision phenotyping platform for seedling growth and morphology

Cory Hirsch, PhD, University of Minnesota

The ability to link genotypes and phenotypes can be used to improve plant productivity

and our understanding of plant biology. Our ability to obtain genomic information

efficiently and accurately has advanced greatly, while phenotyping methods have largely

remained laborious, subjective, and/or expensive. Towards alleviating these barriers, we

have developed a user friendly and affordable platform to acquire highly standardized

RGB images, while relying on minimal equipment and space in a laboratory setting.

Currently, we have developed algorithms to extract numerous growth and morphological

traits including plant height, width, stem diameter, pixel area, and center of mass. The

traits our algorithm extract correlate well with both traditional hand measurements and

measurements using manual image analysis techniques. We are leveraging available

storage, application deployment, and compute resources through the infrastructure at

CyVerse to allow accessibility to almost any researcher. This platform is already being

used by multiple research groups at multiple Institutions and has been optimized for daily

collection of images of multiple plants to easily look at plant development. We have used

this method to monitor growth rate variation among different maize genotypes subjected

to temperature stress and also to measure variation in heterosis for seedling growth in a

panel of inbred and hybrid genotypes.

Machine learning approaches in Soybean Phenomics: Predicting Seed

Yield, Oil and Protein in Contrasting Production Systems

Kyle Parmley, Iowa State University

Page 31: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Genetic improvement of soybean [Glycine max (L.) Merr.] has permitted the expansion

of soybean across a broad geographic region. Past breeding efforts have attempted to

develop highly stable cultivars to deploy across all production systems, but these

genotypes may evade an advantageous genotype by management (G x M) interaction, i.e.,

row width spacing. The development of these production system targeted cultivars will

require continual improvement of yield per acre of soybean, which in turn will be

dependent on the modification of physiological traits. Advances in remote sensing

technologies have enabled rapid measurements of these traits on a temporal and spatial

scale, and therefore are becoming increasingly adopted in advanced breeding systems.

The objective of this study to develop yield prediction models using machine learning

approaches. We used two independent studies with 32 genotypes of the SoyNAM panel

with contrasting treatments: row width spacing (38 and 76 cm) and seeding density (123,

345, 568 x 103 seeds ha-1) from nine environments in replicated tests. Physiological trait

data of hyperspectral reflectance, leaf area index, canopy temperature, light interception,

and chlorophyll content were collected at three time points during the growing season.

Robust in-season prediction models identified informative explanatory varaiables for

seed yield, oil and protein predictions, which and will aide in breeding applications for

contrasting production systems. Preliminary results indicate prediction accuracies were

also similar for remote sensing tools with moderate and high throughput capability

thereby decreasing the temporal requirement for data acquisition. The application of

these approaches enable a mechanistic understanding of yield drivers in contrasting

production systems and enable more informative decision making capability.

Advances in sensing for high-throughput in-field and postharvest crop

phenotyping

Sindhuja Sankaran, PhD, Washington State University

Washington State University

Phenomic advancements to evaluate crop interaction with environment is a rapidly

developing area of research. In last 5 years, applications of sensing technologies in

breeding programs have increased dramatically, where multiple sensors at different

scales have been utilized to assess different traits from yield potential, biotic and abiotic

stress tolerance in field conditions to postharvest crop traits such as seed quality.

Washington State University and USDA-ARS units in the state of Washington have

several active breeding programs from cereal grains to specialty crops. Some of these

program is using sensing tools for high-throughput and accurate phenotyping as a part of

their ongoing crop improvement efforts. This talk will discuss some of these

developments in phenomic research, data analytics associated with sensing, and

applications for infield crop and postharvest crop trait evaluation.

Page 32: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

A Bayesian approach to quantitative genetics for high-dimensional traits

Daniel Runcie, PhD, University of California Davis

Statistical models for Genome-Wide Association Studies, QTL analysis, and Genomic

Prediction, are the foundation of modern quantitative genetics and crop improvement.

Driven by the explosion of whole-genome genotype data, recent improvements to these

models allow for analyses of millions of markers at a time. However, similar advances for

modeling large phenotype datasets is lacking. New phenotyping technologies collect

thousands of observations on each individual plant or line – changes in morphology

through time, molecular phenotypes such as gene expression or metabolite levels, or

performance measures across multiple environments. Jointly modeling these high-

dimensional traits can provide insight into developmental and physiological mechanisms

that link genotype and phenotype. We propose a robust and efficient method for modeling

the genotype-phenotype relationship of high-dimensional traits. The key idea underlying

our model is that groups of traits will be highly correlated due to genetic and

developmental pleiotropy. We leverage these correlated modules to prioritize the most

important signals in big data. We will demonstrate how our method provides powerful

and interpretable estimates of genetic architecture using two high-dimensional datasets:

a time-series analysis of growth curves, and a dataset of genome-wide gene expression.

Network design principles of plant shoot architectures

Saket Navlakha, PhD, The Salk Institute for Biological Studies

Saket Navlakha; Arjun Chandrasekhar; Adam Conn; Ullas Pedmale; and Joanne Chory

Salk Institute for Biological Studies

Transport networks serve critical functions in biological and engineered systems, and

their design requires trade¬offs between competing objectives. Plants need to optimize

their architecture to efficiently acquire and distribute resources while also minimizing

costs in building infrastructure. To understand how plants resolve this design trade¬-off,

we used 3D laser scanning to map hundreds of shoot architectures of tomato, tobacco,

and sorghum plants grown in several environmental conditions and through multiple

developmental time-points. Using a graph¬-theoretic algorithm that we developed to

evaluate design strategies, we find that plant architectures lie along the Pareto front

between two simple objectives — minimizing total branch length and minimizing nutrient

transport distance — thereby conferring a selective fitness advantage for plant transport

processes. The location along the Pareto front can distinguish among species and

conditions, suggesting that during evolution, natural selection may employ common

network design principles despite different optimization trade-¬offs. We also find similar

trade¬-offs sculpting the shape of neural dendritic and axonal arbors, suggesting shared

properties of branching processes in two very different biological systems.

Page 33: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Using mathematics to dissect and quantify the plant form, above and

belowground

Mao Li, PhD, Donald Danforth Plant Science Center

Donald Danforth Plant Science Center

The genome encodes the entire growing plant form, both above and below ground. Yet

conventional phenotypic measures usually consider only isolated parts of the plant to

understand genotype-to-phenotype relationships. Manually dissecting each individual

part is either labor intensive or missing integrative information, limiting our

understanding of genetic and environmental conditioning of the plant form. Using

mathematical approaches such as persistent homology based methods, combined with

imaging technologies and statistical tools could automatically dissect and

comprehensively quantify plant morphology at different scales. We describe some

examples of application such as sorghum panicle, grape rachis, root, and leaf.

Concurrent III: Integrating phenotypes through modeling Friday, February 16, 2018 | 2:30 PM – 5:30 PM

Coupling terrestrial LiDAR measurements of tree architecture with high-

resolution biophysical models to provide insights into plant-environment

interactions

Brian Bailey, PhD, University of California, Davis

Recent advances in the development of high-resolution phenotyping methods coupled

with detailed biophysical models has opened the door to a new frontier in the study of

plant-environment interactions. Methods have been developed to digitize the

architecture of roots and shoots, which can be used as inputs for highly detailed models

that can help to connect the dots between measurements. Despite these recent advances,

many challenges remain. The focus of this work is on addressing the challenge of

phenotyping and modeling plant systems at high resolution across scales from leaves to

canopies.

New phenotyping methods are presented that use terrestrial LiDAR scanning data to

rapidly measure the leaf area and angle distributions of large plant such as trees, which

are ultimately used to perform leaf-by-leaf reconstructions that can be directly fed to leaf-

resolving models. The method uses a statistical backfilling approach that is applicable

even in dense trees where the majority of foliage or woody area is not visible to the

scanner. This detailed structural data is used to provide inputs for leaf-resolving models

of radiation, microclimate, evapotranspiration, and photosynthesis. Model complexity is

afforded by performing calculations in parallel using graphics processing units (GPUs),

which allows for simulations that can resolve an unprecedented range of scales spanning

leaves to canopies.

Page 34: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Linking solar induced fluorescence with photosynthetic variability in crops

at the leaf and plot scales.

Caitlin Moore, University of Illinois

Katherine Meacham – University of Illinois; Guofang Miao – University of Illinois;

Taylor Pederson – University of Illinois; Evan Dracup – University of Illinois; Xi Yang –

University of Virginia; Kaiyu Guan – University of Illinois; Carl Bernacchi – Univ

The measurement of solar induced fluorescence (SIF) of chlorophyll has emerged as a

useful tool for monitoring plant photosynthesis. Application of SIF at the regional scale

from satellite remote sensing has delivered promising results, with strong links found

between SIF and gross primary productivity. Our ability to use SIF as a tool to monitor

photosynthesis has the potential to enhance agricultural advancement by facilitating the

identification of better performing individuals at a faster rate. However, this kind of high-

throughput phenotyping is usually achieved at the plot and/or leaf scale and there

remains an understanding gap as to what extent SIF can capture plot and leaf scale

variation in photosynthetic activity. We tested the ability of SIF to capture photosynthetic

variability at the plot scale in a C4 biomass sorghum field and at the leaf scale in a C3

tobacco experiment grown under field conditions in Central Illinois, USA. To do this, we

built a portable SIF system to collect high-resolution measurements of SIF and coupled

these with measurements of leaf-level gas exchange and photosynthetic performance

indicators (i.e. Vcmax & Jmax), in situ chlorophyll fluorescence, leaf chlorophyll content,

spectral reflectance indices and leaf area. This presentation will discuss not only the

results from our field experiments, but also some of the lessons we have learned along the

way towards developing SIF into a high-throughput phenotyping tool for use at the leaf

and plot scales.

Put the carbon back into the soil: 3D root phenotyping for improved carbon

sequestration

Suxing Liu, University of Georgia

Suxing Liu – University of Georgia; Alexander Bucksch – University of Georgia

Carbon rich soil ensures the fertile soils and the agricultural productivity of plants that

sequester atmospheric carbon available as CO2 into the soil. Deeper rooting crops are the

key to increase carbon content below ground and to improve soil quality and agricultural

output. Root phenotyping is crucial since it provides avenues to quantify deeper root traits

of important crops like maize. However, due to the opaque nature of soil, dense and highly

occluded maize root system, quantifying these traits such as whorl number, distance and

number of crown roots is very challenging.

Page 35: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

We developed a 3D reconstruction method that completely digitizes the maize root

architecture. The classic structure from motion algorithm was not able to reconstruct

detailed inner roots architecture without manual registration, we extends this method to

produce a dense point could root model based on images collections taken only from the

outside of roots. Our extensions allow automated reconstruction and measurement of the

inner occluded root system.

We demonstrate the quality of our method on two maize genotypes with six replicates for

each genotype. Our 3D root model reconstruction method is a first promising step

towards automated quantification of highly occluded maize root system. And it enable the

discovery of genes associated with deeper rooting by molecular biologists and pave a

promising way to increase soil carbon sequestration in crops.

A modular, community-driven framework for developing high-throughput

plant phenotyping tools

Noah Fahlgren, Donald Danforth Plant Science Center

Malia Gehan – Donald Danforth Plant Science Center; Arash Abbasi – Donald Danforth

Plant Science Center

Systems for collecting image data in conjunction with computer vision techniques are a

powerful tool for increasing the temporal resolution at which plant phenotypes can be

measured non-destructively. Computational tools that are flexible and extendable are

needed to address the diversity of plant phenotyping problems. To address these needs,

we developed PlantCV, an open-source framework for analyzing high-throughput plant

phenotyping data. The goal of the PlantCV project is to develop a set of modular, reusable,

and repurposable tools for plant image analysis that are open-source and community-

developed. PlantCV was originally developed to analyze data from the Bellwether

Phenotyping Facility at the Donald Danforth Plant Science Center, but the set of available

features has grown as the set of users and use cases have diversified. Here we focus on

several recent developments within the PlantCV project, including 1) Robust and

automated methods to test and review code and documentation through the GitHub

platform to manage diverse community activity; 2) Utilization of containerization

technologies and integration with the Open Science Grid and Amazon Web Services

computing platforms to improve the scalability and deployability of PlantCV; and 3)

Machine learning applications for plant feature analysis. We are using these tools, in

conjunction with high-throughput imaging platforms, as a method to dissect the genetics

of plant traits for bioenergy crops such as sorghum and camelina. PlantCV is open source

and is available at http://plantcv.danforthcenter.org.

Page 36: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Non-linear plant phenotyping pipelines: how can structural models and

machine learning can help us analyse large plant image datasets

Guillaume Lobet, PhD, Forschungszentrum Juelich

Institute of Bio- and Geosciences Agrosphere (IBG-3)

Many structural root models have been developed, either generic or for specific species,

and these have repeatedly been shown to faithfully represent the root system structure,

as well as being able to output ground-truthed data for every simulation and image,

independent of root system size. Here we will show that structural root models can be

used in combination with image analysis pipelines to assess and improve their overall

performance. First, we will show that an in-depth analysis of root image analysis pipelines

using such models reveals strong limitations in their ability to measure complex root

systems. Secondly, we will present an innovative strategy that combines root models and

machine-learning algorithms (random-forests), that can increase the measurement

accuracy.

Gravimetric phenotyping of canopy conductance in wheat and maize

reveals novel mechanisms, traits and genetic loci involved in drought

tolerance in the field

Walid Sadok, University of Minnesota

Walid Sadok – University of Minnesota; Bishal Tamang – University of Minnesota;

Remy Schoppach – University of Minnesota

Canopy conductance plays a critical role in crop drought tolerance. Particularly in

Mediterranean-type environments where crops grow on stored soil moisture, this process

controls how much water is being traded for CO2 and the soil moisture budget available

for the crop. As a result, phenotyping canopy conductance is currently a major target of

breeding programs worldwide, but is still notoriously problematic in the field. To

circumvent those limitations, we developed a gravimetric approach to phenotyping

whole-plant canopy conductance under controlled conditions where conductance is

measured as the slope of whole-plant transpiration rate (measured via balances

connected to loggers) in response to increasing vapor pressure deficit. Using this

approach, we phenotyped a wheat mapping population revealing QTL for canopy

conductance most of which co-localized with QTL of grain yield measured in 11 locations

in Australia and Mexico. Further, we identified a major QTL that was mapped to genes

suggesting the involvement of root hydraulics in controlling canopy conductance, which

was confirmed in independent experiments. We also discovered substantial genetic

variability in nocturnal transpiration in this population that was controlled by several

QTL, most of which co-localized with yield QTL, a relationship that we confirmed via

Page 37: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

simulation modelling. In maize, we recently deployed this approach on 25 parents of the

maize Nested Association Mapping panel (McMullen et al. 2009). Our investigation

revealed substantial variability in canopy conductance and nocturnal transpiration

among those lines. We also discovered a circadian control of nocturnal transpiration that

was genotype-dependent, which seemingly modulates the level of canopy conductance

during the day. Those findings illustrate ways gravimetric phenotyping of water use in

crops can lead identifying novel traits that can inform breeding programs while

illuminating novel biological insights.

References:

Tamang and Sadok (2018). Env. Exp. Bot. In press

Schoppach et al (2016). J. Exp. Bot. 67, 2847-2860

Use of biophysical first principles to select plant traits and the instruments

and analyses to measure and explain them

Brent Ewers, University of Wyoming

Predictive understanding of genotypic variation in growth and response to drought

requires quantifying a constellation of plant traits such as leaf and root physiology and

allocation to both of those organs and reproduction. High-throughput phenotyping of

these traits that connect to emerging –omics approaches are limited by instrumentation

and data analyses. Moreover, engineering new instruments is often hampered by lack of

clarity in what traits to measure and how to analyze the data. We suggest a new approach

in which both high and low throughput traits that best lead to improved predictions are

hypothesized from biophysical model outputs. Our biophysical, first principles model

suggests that traits related to quantum yield of photosynthesis, water transport

limitations from roots to leaves, and allocation to roots, leaves and reproduction are most

important in predicting growth from the crop plant Brassica rapa. We test these

instruments and data analyses across a diverse panel of crop accessions and response to

drought. We found that chlorophyll fluorescence measurements quantified quantum

yield in both high and low throughput instruments with direct comparisons not

statistically different from unity. We assessed root hydraulics and root area through low

throughput means and found proxies derived from electrical conductivity, biomass

allocation, and image analysis allowed the model to predict growth. Other tests include a

temporally dense experiment of Brassica rapa response to drought that was analyzed with

a Bayesian network model that combined gene expression modules from RNA-Seq with

our measured traits. The resulting probability tables illustrated that the gene models

provided more predictive power over the temporal changes in the traits than the network

of the biophysical traits themselves. Moreover, the gene ontologies from these modules

Page 38: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

describe hypotheses for further experiments such as testing how night transpiration

decline in drought is related to up regulation of nitrate transporters.

Genome-to-Phenome Mapping by Metabotyping in Brachypodium

distachyon: Exploring Genotypic Diversity for Biomass Accumulation and

Shoot-Root Allometry

Christer Jansson, Pacific Northwest National Laboratory

The present study aims to explore genotypic diversity for biomass accumulation and

shoot-root allometry in natural accessions of the annual C3 grass Brachypodium

distachyon (Brachypodium) under two contrasting watering regimes, well-watered and

episodic drought (henceforth referred to as control and drought conditions, resepctively),

and to what extent genotypic and phenotypic diversity correlate with metabolite profiles.

External phenotypes like biomass accumulation and shoot-root allometry represent

emergent properties, and as such are informed by internal phenotypes, e.g., biochemical

and physiological properties, in interaction with the environment. We argue that bridging

the gap between genotype and external phenotype can be facilitated by a two-step process,

whereby linkages are established between genotype and internal phenotype and between

internal and external phenotypes. In considering internal phenotypes, we point to

metabolomics as an emerging tool to provide insight into how genotypic diversity affects

phenotypic variation in plants, although it should be noted, that the genetic control of

plant metabolomes remains all but unknown, and that even in a system such as E. coli

interactions between gene variants and metabolite profiles are poorly understood. The

method for genome-to-phenome mapping, envisioned here with a moderate-sized set of

genotypes (30 accessions) and powerful statistical processing, is to proceed in two steps

by establishing a composite internal phenotype, e.g., with genotype-associated

metabotypes containing clusters of detected metabolite features) and biomass-associated

metabotypes containing profiles of detected metabolite features and their relative

abundance. The internal phenotype will then serve as a proxy for the external phenotype.

Our results suggest that this phased strategy holds great promise for further exploration

with increased number of accessions and biological replicates and with continued

development of computational signal discovery algorithms to allow for integration of

independently generated sets of metabotypes.

Concurrent IV: Crop Biology Friday, February 16, 2018 | 2:30 PM – 5:30 PM

Insights into the genotype-by-environment interaction enabled through

phenomics

Candice Hirsch, PhD, University of Minnesota

Page 39: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Candice N. Hirsch1, Zhi Li1, Sara B. Tirado1,2, Michael R. White3, Celeste Falcon3,

Nathan D. Miller4, Edgar P. Spalding4, Natalia de Leon3, Shawn M. Kaeppler3, Patrick

S. Schnable5, Nathan M. Springer2

The ability to measure variation in plant genomes is becoming routine due to advances in

sequencing technology and analysis methodologies. Our ability to understand the

interaction of genomic variation with the environment is still largely limited by the

phenotypic traits that can be measured with regards to number of traits, accuracy of trait

measurements, and the frequency at which traits can be measured. As with sequencing

technologies, we are seeing rapid advancements in sensor technologies, platforms to

deploy those sensors, and analysis pipelines that are now allowing for plants to be

measured in a range of environmental conditions to allow for new insights to be made in

the interaction of genotype and environment. We have deployed a previously developed

pipeline that enabled the evaluation of a large number of traits on ears, cobs, and kernels

to further our understanding of the genotype-by-environment interaction. This pipeline

has been used to study the impact of environmental variation on yield potential and yield

component traits, the impact of environment on observed heterosis, and the impact of

environment on small introgressions in the genome (i.e. those that would be present

during the introgression of a large effect QTL or backcrossing in a transgene). Further

advances in our ability to measure plants in real time across diverse environments will

continue to advance biological insights that can be made into the interaction of the

genome and environment.

Genetic control of soybean (Glycine max L. Merr.) shoot architecture

Kamaldeep Virdi, PhD, University of Minnesota

Gary Muehlbauer – University of Minnesota; Robert Stupar – University of Minnesota;

Aaron Lorenz – University of Minnesota; Austin Dobbels – University of Minnesota;

Suma Sreekanta – University of Minnesota; Jeffrey Roessler – University of Minnesota

Soybean shoot architecture is a complex phenotype, which can be partitioned into various

measurable traits. Variation in shoot architecture likely influences canopy light

interception, photosynthesis, and source-sink partitioning efficiency, and thus is related

to overall grain yield. Here, shoot architecture-related traits are defined as branch angle,

branch number, branch length, leaf shape and size, petiole length, petiolule length,

canopy coverage, days to flowering, maturity, determinancy, and light penetration

through the canopy. A detailed phenotypic characterization of these shoot architecture-

related traits, their contributions to overall shoot architecture, and their genetic control

is imperative to fully exploit the yield potential of soybean. We examined a set of 400

diverse maturity group 1 soybean accessions to study the natural variation for shoot

architecture-related traits, to identify relationships between traits, and to use association

Page 40: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

mapping to identify loci that are associated with shoot architecture traits. The panel was

genotyped with 32,650 SNP markers. To collect phenotype data for shoot architecture-

related traits, we used a combination of high-throughput (drone and ground-based

imagery) and conventional phenotyping platforms. Significant QTL associated with

branch angle, leaf length/width ratio, petiolule length, days to flowering, maturity and

stem termination were detected. In most cases, these QTL overlapped with previously

detected genes or QTL. For example, a leaf length/width ratio QTL on chromosome 20

was found to be coincident with the location of a previously isolated Narrow Leaf (Ln)

gene. Interestingly, we detected a branch angle QTL located on chromosome 19 that

overlaps with QTL associated with canopy coverage and light penetration, suggesting

branch angle is an important determinant of canopy coverage.

Field Phenotyping of Grain Crops Response to Agronomic Inputs

Steven Shirtliffe, PhD, Department of Plant Sciences, University of Saskatchewan

Hema Duddu – Department of Plant Science, University of Saskatchewan; Menglu

Wang – Department of Plant Science, University of Saskatchewan; Seungbum Ryu –

Department of Plant Science, University of Saskatchewan; Scott Noble – Department of

Mechanical Eng

High throughput phenotyping of grain crops has become a dynamic research areas with

applications in plant breeding and genomics. Field based phenomics quantifies remotely

quantifies visual phenotypes to associate with regions of the genome. These techniques

also have the potential to revolutionize agronomic research. We will present several

examples of current research utilizing remotely gathered imagery including, response to

nitrogen fertilizer, simulated hail damage, herbicide tolerance and plant spatial

arrangement. Phenotypes were extracted from this imagery with a variety of techniques

ranging from vegetation index analysis, classic image analysis, and deep learning. UAV

based field imagery offers a wealth of data to “ground truth” phenotypes for use in remote

crop diagnostics and precision agriculture. Furthermore this research is directly

applicable classic phenomic research as the response to crop inputs is the manageable

part of genotype by environment interaction. Finally, as the range of phenotypes

expressed to agronomic inputs is usually greater than that within elite germplasm, more

precise calibration of phenotypes is possible. Synergies exist between phenotyping genetic

and agronomic effects in crops < ./p>

Page 41: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Forward Phenomics of oat Panicles

Abdullah A Jaradat, USDA/ARS & University of Minnesota

There is a growing need for adapted and more productive germplasm to expand oat

production, optimize its yield, improve groat quality, and satisfy farmers and consumers

demand, especially in the Upper Midwest of the US. Oat germplasm, representing

different eco-geographical origins and breeding status, was characterized and evaluated

using field- and laboratory-based forward phenomics. Whole plots were phenotyped at

successive growth stages during three growing seasons using aerial and hand-held

imagery and sensors. Digital and high-throughput data were captured and compiled on

(1) color space descriptors of whole plots during key growth stages, (2) growing degree

days to panicle emergence and maturity; (3) phenotypic and structural traits of panicles

and spikelets; and (4) groat quality. A relational database was mined and statistically

analyzed to (1) cluster the oat germplasm using unsupervised hierarchical clustering

method; (2) identify traits with positive or negative, direct or indirect effect on the panicle

phenome; (3) identify a minimum set of traits which can discriminate between

structurally and agronomically different panicle phenotypes; (4) express groat weight per

plant as a function of stochastic panicle architecture and (5) quantify the effect of panicle

architecture traits that have implications for groat quality. A dynamic custom profiling

procedure was instrumental in quantitatively assessing the importance of, and visually

adjusting structural panicle traits to predict and optimize groat agronomic and quality

traits.

Data fusion with light detection and ranging and images to map and count

bolls in upland cotton

Alison Thompson, PhD, USDA-ARS

Andrew French – USDA-ARS; Michael Gore – Cornell University; Alex Conrad – USDA-

ARS

Terrestrial Light Detection and Ranging (LIDAR) has the potential to provide accurate

boll counts and maps of boll distribution over a defoliated cotton canopy. Using a field-

portable LIDAR unit and GPS receiver, cotton plants can be scanned plot-by-plot,

returning point cloud data that can be aggregated and mapped. Since 2013 small plot

LIDAR experiments have been conducted at Maricopa, AZ to develop acquisition and data

processing systems. Results, achieved using multiple mobile platforms including high

clearance tractors, proximal sensing carts, and semi-robotic rail carts, have shown that

cotton bolls can be mapped and counted provided, platform travel speeds are slow (~0.5

m/s), GPS locations are accurately synchronized with the LIDAR, and cotton plants are

fully defoliated. However, obtaining consistent results has been difficult due to slow

LIDAR scan rates, leading to spatial sampling gaps and inaccurate sensor positioning.

Page 42: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Experiments in 2017 were undertaken to reduce these problems by adopting a data fusion

approach. We deployed a LIDAR unit on a motorized proximal sensing cart capable of

scanning at rates exceeding 200 Hz, in parallel with 3 RGB cameras with multiple view

angles to generate high resolution point cloud data sets. The data sets were

complementary, where the LIDAR provided mm-level accuracies and the camera data

provided spatially contiguous observations. Data collected in November 2017 over ~0.5

ha-1 of variable density upland cotton and corresponding validation measurements of boll

counts and boll distribution will be presented.

UAS Phenotyping in Soybean Breeding and Phenomic Inference

Katy Rainey, Purdue University

Katy Rainey, PhD – Purdue University

Abstract: Phenomics and data-driven selection in crop breeding encompass applying new

and differently-used data to increase selection efficiency of crop varieties. To begin to

understand the potential applications of novel phenotypes, an initial stage of phenomic

inference is helpful. Defining features of the phenomic inference stage are high

phenotypic variance for yield potential and control of the confounding effects of crop

development. For “new” traits, this stage provides a first report of genetic architecture,

quantitative properties such a genetic correlation to yield, and initial remote sensing

prediction equations. Mixed model statistical procedures are the foundation for achieving

these aims, and for characterizing novel phenotypes. Analytical innovations are needed

in several areas, including predictive analytics and new approaches to genetic inference

of development. We present case studies from a soybean breeding pipeline where we use

UAS-acquired RGB imagery to select multiple agronomic traits. We are also developing

analytical tools from operations, to analysis, to decision making for UAS RGB and

multispectral imagery.

Whole-plant stress performance analysis: A new tool for functional

phenotyping

Menachem Moshelion, The Hebrew University of Jerusalem

Abiotic stress factors limit crop yields and long-term food security. It is widely accepted

that phenotyping tools are needed to bridge the gap between the plant molecular profile

and plant performance, particularly under stress conditions.

The need for physiological phenotyping of the whole plant led us to develop a new and

effective functional-phenotyping platform, as well as real-time analytical software to

efficiently extract knowledge from the collected data. This platform enables the

Page 43: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

simultaneous and continuous monitoring of water relations in the soil–plant–

atmosphere continuum of numerous plants under dynamic environmental conditions.

This system provides a simultaneously measured, detailed physiological response profile

for each plant in the array over time periods ranging from a few minutes to the entire

growing season, under normal, stress and recovery conditions and at any phenological

stage. Three probes for each pot in the array and a specially designed algorithm enable

the detailed characterization of whole-plant transpiration, biomass gain, stomatal

conductance, whole-plant water-use efficiency and relative water content under dynamic

soil and atmospheric conditions. The system has no moving parts and can be used in many

experimental settings.

Our algorithms are based on multi-factorial analysis of variance and the scoring of all

physiological traits. This comparative, performance-analysis approach enables the rapid

selection of plants with the desired physiological behaviors, which is very important in

pre-breeding for stress tolerance.

The screening of many plant species, using our system and conventional gas-exchange

tools, has confirmed the accuracy of the system and its diagnostic capabilities. In addition,

the quantitative physiological traits measured have been shown to be closely correlated

with total dry biomass. In the future, this platform could help with early yield prediction

and enhance functional-breeding programs.

General Session III Saturday, February 17, 2018 | 8:30 AM – 12:30 PM

Multiscale modelling of plant-soil interaction

Tiina Roose, MSc, DPhil (PhD), University of Southampton

University of South Hampton

In this talk I will describe a state of the art image based model of the soil-root interactions,

i.e., a quantitative, model of the rhizosphere based on fundamental scientific laws. This

will be realised by a combination of innovative, data rich fusion of structural imaging

methods, integration of experimental efforts to both support and challenge modelling

capabilities at the scale of underpinning bio-physical processes, and application of

mathematically sound homogenisation/scale-up techniques to translate knowledge from

rhizosphere to field scale. The specific science question I will address with these

techniques is how to translate knowledge from the single root scale to root system, field

and ecosystem scale in order to predict how the climate change, different soil

management strategies and plant breeding will influence the soil fertility.

Page 44: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

The Plant Phenology Ontology: A new informatics resource for large-scale

integration of plant phenology data

Robert Guralnick, University of Florida

University of Florida

Plant phenology — the timing of plant life-cycle events, such as flowering or leafing out —

plays a fundamental role in the functioning of terrestrial ecosystems, including human

agricultural systems. Because plant phenology is often linked with climatic variables,

there is widespread interest in developing a deeper understanding of global plant

phenology patterns and trends. Although phenology data from around the world are

currently available, truly global analyses of plant phenology have so far been difficult

because the organizations producing large-scale phenology data are using non-

standardized terminologies and metrics during data collection and data processing. To

address this problem, we have developed the Plant Phenology Ontology (PPO). The PPO

provides the standardized vocabulary and semantic framework that is needed for large-

scale integration of heterogeneous plant phenology data. The PPO was designed to be

applicable to (nearly) all gymnosperms and angiosperms, suitable for both single plants

and populations of plants, compatible with the data and data collection methods of

existing national or regional monitoring program in the USA and Europe, and

interoperable with existing OBO Foundry library ontologies, especially the Plant Ontology

and the Biological Collections Ontology. Here, we more fully describe the PPO, and we

also report preliminary results of using the PPO and a new data processing pipeline to

build a large dataset of phenology information from North America and Europe. We close

by discussing future Plant Phenology Ontology efforts, along with the social and technical

tools for further developing global in-situ phenology products and analyses, including

phenology sensor data.

Division plane orientation in plant cells

Carolyn Rasmussen, PhD, University of California, Riverside

One key aspect of cell division in multicellular organisms is the orientation of the division

plane. Proper division plane establishment significantly contributes to normal

organization of the plant body. To determine the importance of cell geometry in division

plane orientation, we designed a three-dimensional probabilistic mathematical modeling

approach based on century-old observations: equal daughter cell volume and the

resulting division plane is a local surface area minimum. Predicted division planes were

compared to a plant structure that marks the division site, the preprophase band (PPB).

PPB location typically matched one of the predicted divisions. Predicted divisions offset

from the PPB occurred when a neighboring cell wall or PPB was observed directly

adjacent to the predicted division site, as to avoid creating a “four-way junction”.

Page 45: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Population level modeling accurately predicted ~95% of in vivo cell divisions based on

geometry. This powerful model can be used to separate the contribution of geometry from

mechanical stress or developmental regulation in predicting plant division plane

orientation.

Automated plant architectural trait extraction from a field-based high-

throughput phenotyping platform

Maria Salas-Fernandez, PhD, Iowa State University

Iowa State University

Yield is determined by the plant’s capacity to capture light energy and utilize it to fix CO2

into complex organic compounds. This capacity is mostly determined by both the

biochemical process of carbon fixation and the arrangement of leaves throughout the

plant. The large scale investigation of leaf angle variation throughout the canopy and

other plant architecture traits that determine biomass yield is labor intensive, time

consuming, expensive and, in some cases, impossible to accomplish using manual

measurements. We have examined the accuracy of machine vision to estimate plant

architectural traits of a large set of sorghum lines under field conditions. Our approach is

based on the 3D reconstruction of stereo images collected sideways and the automatic

extraction of novel descriptors with validated biological significance. These features

included proxies of leaf angle, leaf length and biomass volume. A comparison of image-

derived descriptors and manually collected plant architecture parameters were

performed in a subset of lines. Subsequently, novel image-derived descriptors were

utilized as phenotypic measurements in a genome-wide association analysis to discover

genes/genomic regions controlling natural variation of plant architecture traits in

sorghum. These discoveries were compared to previously reported attempts to genetically

dissect plant architecture using manually collected data and provide new knowledge

about the genetic mechanisms underlying these important traits. The completely

automated processing methods developed in our study represent a new tool for plant

breeders and advance the interdisciplinary field of high-throughput phenotyping.

Machine learning in plant phenotyping: will it relieve the bottleneck?

Sotirios Tsaftaris, MSC, PhD, University of Edinburgh

University of Edinburgh

Advances in automation and imaging equipment have created a renaissance in the

measurement of phenotyping traits in plants. While previously the bottleneck was the

image acquisition currently it is extracting those measurements from imaging data in a

Page 46: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

reliable and automated fashion (to deal with high-throughput and high data

dimensionality) that is now the new bottleneck. This bottleneck is expected to increase

as we venture in the field (higher variability) even further. I will describe how the use of

machine learning can open the road towards relieving the bottleneck. I will present

solutions, stemming from our work and from others on automated leaf counting, plant

growth assessment, and trait identification. However, machine learning algorithms and

particularly new deep learning methods, need data to learn from. Thus, I will discuss the

need for curated annotated data, and guidelines and opportunities in collecting such data.

I will conclusively demonstrate with examples how open data can help address the

bottleneck.

The shape of plants to come: in situ computation and field math

Alexander Bucksch, PhD, University of Georgia

The genome encodes the entire growing plant form, both above and below ground. Yet

conventional phenotypic measures usually consider only isolated parts of the plant to

understand genotype-to-phenotype relationships. Manually dissecting each individual

part is either labor intensive or missing integrative information, limiting our

understanding of genetic and environmental conditioning of the plant form. Using

mathematical approaches such as persistent homology based methods, combined with

imaging technologies and statistical tools could automatically dissect and

comprehensively quantify plant morphology at different scales. We describe some

examples of application such as sorghum panicle, grape rachis, root, and leaf.

Measurements that matter: Ensuring quality and traceability of data for

agricultural insights

Michael Malone, PhD, Climate Corporation

Daniel Hangartner – Climate Corporation; Nicholas Dinchev-Vogel – Climate

Corporation

Farmers have a growing number of tools at their disposal to monitor their fields and

adjust their management practices accordingly. Unfortunately more data does not always

lead to better decisions: The old computing adage "garbage in, garbage out" can be equally

applied to digital and precision agriculture. Here we will discuss common challenges that

can undermine the utility of sensor data and propose a framework of best practices to

ensure that every new sensor and cloud data source provides value to customers and

companies alike.

Page 47: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Root phenotyping using X-ray technology: Automation of data

segmentation for 4D analysis

Stefan Gerth, Fraunhofer EZRT

Joelle Claußen – Fraunhofer EZRT; Norbert Wörlein – Fraunhofer EZRT; Anja Eggert –

Fraunhofer EZRT; Norman Uhlmann – Fraunhofer EZRT

During the last years, X-ray technology has been applied for the non-destructive

visualization of optical inaccessible structures in plants. Formerly, this technology was

only used for medical imaging. Nowadays, it is used as a standard tool in industrial

applications for material analysis. With X-ray computed tomography (CT) the 3D volume

information of objects can be reconstructed using X-ray projections of the object from

different points of view. A conical X-ray beam projects the plant on a 2D flat panel

detector. The resulting spatial sampling frequency is determined by the geometrical

magnification and the pixel size of the flat panel detector.

Due to the non-destructive nature of CT it is possible to track the growth of plant organs

such as cassava tubers or root systems without excavating the plants. Thus, the

belowground development of an individual plant can be observed time dependently in

natural soil. To extract the root system out of the soil, new algorithms are used for the

automatic segmentation. This allows searching for new traits with an increased

throughput compared to manual volume segmentation.

We observed the root growth of different Cassava varieties over a period of two months.

Within this timeframe, we periodically analyzed the root system and applied an

automated segmentation algorithm. Thus, we are able to characterize the volumetric

growth of storage roots time resolved and track them already at an early stage in a reliable

and reproducible manner. For example, this approach allows calculating the total volume

as a function of the depths for each of the different varieties. Due to the time resolved

experiment, this data can be used to calculate the biomass growth rates, as well. Using

this approach, we are able to segment automatically the root system in a reliable and

comparable way, even with variations in the soil humidity.

From text blobs to computable data: challenges in mining phenotypical data

from text

Hong Cui, PhD, University of Arizona

University of Arizona

In biomedical domains, ontologies have been seen as an effect approach to data

integration that brings collections of less structured data to computation. Phenotypic

Page 48: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

character data occupies a prominent position in both evolution research aiming at

building the Tree of Life and other branches of biology interested in linking genotypes to

phenotypes to answer questions such as which genes cause what diseases. However, the

vast majority of phenotypic character information is published in (semi-)natural

languages and not directly suitable for computation. Converting such information into

ontologized statements such as Entity Quality (EQ) has been underway, along with initial

signs of success in using ontologized phenotypic characters. The mining and the curation

processes, however, are not without challenges. In this talk, I will present results from our

recent text mining and curation projects and discuss these challenges. In searching for

other ways to overcome these obstacles, I encourage the community to rethink the roles

that authors of phenotypical information can play in the production of computable data

and possibly the ontologies.

Concurrent V: Microclimate effects on plant phenotype Saturday, February 17, 2018 | 2:30 PM – 5:30 PM

Uncovering the genetic basis for natural variation of root system dynamics

in Arabidopsis

Therese LaRue, Stanford University

Thérèse LaRue, Heike Lindne;Ankit Sirinivas; Guillaume Lobet; and José Dinneny

Carnegie Institute

As the interface between the soil environment and the rest of the plant, root systems play

a key role in determining plant growth. The distribution of roots, termed root system

architecture (RSA), is influenced by the surrounding environment. GLO-Roots (Growth

and Luminescence Observatory for Roots) is a new soil-based root imaging technology,

which enables detailed observations of Arabidopsis thaliana root system growth in soil.

Using this system, we are characterizing how the root systems of a diverse population of

Arabidopsis accessions grow over time and will perform a genome wide association study

to identify common alleles involved in RSA control. In parallel, modelling soil-root

interactions to predict RSA function and performance under different stress conditions

will inform us about improved RSA strategies. Together, this work will investigate how

plant root systems are distributed spatially within the soil and identify ways plants

regulate root system growth to cope with

Evaluation of Plant Environmental Stress Response using “RIPPS”, an

Automated Phenotyping System

Miki Fujita, PhD, RIKEN CSRS

Page 49: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Miki Fujita – RIKEN; Kaoru Urano – RIKEN; Takanari Tanabata – Kazusa Inst.; Saya

Kikuchi – RIKEN; Kazuo Shinozaki – RIKEN

High-throughput and accurate measurements of plant traits facilitate the understanding

of gene function. Especially, with recent advances in quantitative genomics such as QTL

or GWAS, there is a growing need for precise quantification of multiple traits in plants.

However, in the case of environmental stress responses such as drought, it is difficult to

quantify the adaptive responses because multiple environmental factors are intricately

involved in the phenotype. Therefore, precise control of growth conditions is of great

importance to evaluate plant responses to environmental stresses. Recently we have

developed an automatic phenotyping system that evaluate plant growth responses to a

wide spectrum of environmental conditions. The system named RIPPS (RIKEN Plant

Phenotyping System) controls individual soil moisture in continuously rotating 120 pots

by a combination of automatic weighing and watering systems that enable the precise

control of soil water condition is necessary for quantifying the adaptive responses to

environmental stresses such as limited water conditions. RIPPS also take image of top

and side view of the plants every two hours. In this presentation, we’ll demonstrate the

utility of the RIPPS in evaluating drought or salinity tolerance and water use efficiency.

High-throughput 3D analysis of barley shoots reveals novel QTL involved in

leaf growth under salt

Bettina Berger, Australian Plant Phenomics Facility - University of Adelaide

Ben Ward – University of Adelaide; Helena Oakey – University of Adelaide; Chris Brien

– University of Adelaide; Allison Pearson – University of Adelaide; Sonia Negrao – King

Abdullah University of Science and Technology; Rhiannon Schilling – University of

If we want to improve crop productivity in a changing climate, we need to understand

plant growth, stress responses and the underlying genetic control. While we now have

phenotyping tools to measure plant growth at high-throughput in forward genetics

screens, we often just scratch the surface by treating the plant shoot as a single object, not

taking into account its individual components.

To be able to dissect the shoot of small grain cereals, such as wheat and barley, into

individual leaves, we have developed a novel computer vision approach. With only a small

number of images we were able to recover the structure of the shoot, the number of leaves,

their length and growth over time. With few images as input, the 3D shape of the plant

may be ambiguous. To overcome this issue, we used multiple view geometry and prior

knowledge of plant structure to generate plausible models of the shoot. We then applied

this method to screen a bi-parental barley population under low and high soil salinity. In

Page 50: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

addition to previously characterised loci related to plant development, we also identified

novel QTL for individual leaf expansion under salinity.

The trait components that constitute whole plant water use efficiency are

defined by unique, environmentally responsive genetic signatures in the

model C4 grass Setaria

Max Feldman, Donald Danforth Plant Science Center

Max Feldman – Donald Danforth Plant Science Center; Patrick Ellsworth – Washington

State University; Asaph Cousins – Washington State University; Ivan Baxter – USDA-

ARS

The complex relationship between plant growth and water use is largely determined by

genetic factors that influence both the morphological and biochemical characteristics of

plants. Improving the efficiency by which plants utilize water is an important breeding

objective that can be translated to improve productivity in agriculture while

simultaneously making it a more sustainable endeavor. To assess the genetic basis of

water use efficiency and trait plasticity, we have utilized high-throughput phenotyping

platform and mass spectrometry to quantify plant size, evapotranspiration and stable

isotope composition of an interspecific Setaria italica x Setaria viridis recombinant inbred

line population in both a well-watered and water-limited environment. Our findings

indicate that measurements of plant size and water use in this system are strongly

correlated. We used a linear modeling approach to partition the traits into the predicted

values of plant size given water use and deviations from this relationship at the genotype

level. The resulting traits describing plant size, water use, water use efficiency and δ13C

are all substantially heritable, responsive to soil water potential differentials and provide

a framework to understand the components of plant water use efficiency. Biparental

linkage mapping successfully identified several pleiotropic loci that exhibit medium-to-

large effects on most traits in addition to many smaller effect loci associated with fewer

traits or specific to well-watered or water-limited environments. This study is the first

report characterizing the genetic architecture of water use efficiency in the model C4

species Setaria and mechanistically links measurement of water use efficiency with δ13C

through several common large effect QTL.

Purdue's New Automatic Phenotyping Greenhouse with Micro-climates

Removed

Jian Jin, PhD, Purdue University

Purdue University

Page 51: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Purdue University deployed a new fully automatic phenotyping greenhouse in May 2017.

This facility is featured for (1) Continuous scanning of each crop plant for up to 20

times/day; (2) Clearly removing the micro-climates impact (the variance of

environmental conditions caused by distribution of lighting, temperature, airflow and so

on across the greenhouse space); (3) Advanced hyperspectral imaging system and data

modeling for plant physiological features predictions. Dr. Jin will also share his view of

next generation plant phenotyping in the next 10 years.

Should Soil Water Availability considered in plant phenotyping for abiotic-

tolerance, and how?

Rony Wallach, Prof., Hebrew University of Jerusalem

Rony Wallach, Prof. – Hebrew University of Jerusalem, Israel

The fundamental mechanism of water flow in plants has been described for many years.

Plants are subjected to atmospheric water demand on one hand and soil water flux into

the roots that should compensate for the demand. How far plants are able to sustain their

leaf water demand is therefore largely dependent on the hydraulic properties of the soil-

root system that depends, inter alia, on the soil water content.

As long as soil hydraulic conductivity do not limit the water flow to the rhizosphere, soil

water fully compensates for the atmospheric water demand (potential transpiration rate).

However, when soil water diminishes, the soil hydraulic conductivity turns to be the

limiting factor in water capture in drought prone environments.

By using a newly developed high throughput platform that monitors at a high-frequency

the transpiration rate and soil water flux to the entire root system, the momentary balance

between input and output fluxes was tracked. The commonly used description of

transpiration rate vs. time, is replaced by transpiration rate vs. soil -water content,

yielding a piecewise linear curve. The break point between the two linear lines denotes

the soil water content from which it becomes a limiting factor under the given

environmental conditions, noted as qcrit.

Combining these results with independently measured soil hydraulic properties enables

to transform the change in soil water content to the change in soil hydraulic conductivity,

to be a measure of soil water availability. This approach enables to phenotype different

plants and cultivars for drought tolerance and a succeeding recovery. In addition, given

that periods of drought vary in length, timing and intensity, the current approach enables

to relate different plant and roots traits with the different types of drought. The

application of these results to plant phenotyping, screening, and water management (e.g.

irrigation) will be discussed.

Page 52: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

A Machine-Vision Seedling Emergence Assay

Nathan Miller, University of Wisconsin-Madison

Nathan Miller – University of Wisconsin-Madison; Jeff Gustin – University of Florida;

Mark Settles – University of Florida; Edgar Spalding – University of Wisconsin-

Madison

Seedling emergence is a critical stage in the establishment of a successful crop.

Germination and robust seedling establishment are selected traits during the

development of new varieties but with inefficient, largely manual methods. We have

developed an in-lab, soil-based machine vision emergence platform that automatically

measures the emergence profile, including rate, percent, duration, and time to 50%. The

assay is scalable, can accommodate the application of chemical or environmental

treatments, and can be used with different soil types. Each modular unit has a camera

which monitors 168 seedlings and can fit in an environmentally controlled chamber.

Currently the system includes 12 modular units that can monitor 2,016 kernels at a time.

Streaming images are stored and processed by CyVerse’s cyberinfrastructure using a

custom application. This publicly available software measures percent emergence with

2% False Negative Rate and time to 50% emergence within 30 minutes for maize

seedlings. In initial experiments investigating sources of variation in emergence profile,

we found that time to 50% emergence in Mo17 and B73 were remarkably consistent across

seed lots grown in multiple seasons. Comparing 4 field seasons, genetic differences

between these two inbreds explained 33% of the variance while field season explained

only 1.5%, suggesting time until emergence is under genetic control. A larger screen of

40 diverse genotypes ranged from 103 hours to 142 hours with an average variation of 15

hours within a genotype. Our current target is screening 250 RILs of the IBM mapping

population, which we expect to complete in 8 weeks. While current studies focus on

maize, the assay could be used with any large seed with no modification needed.

Concurrent VI: Graduate training in phenomics: an interdisciplinary

adventure

Saturday, February 17, 2018 | 2:30 PM – 5:30 PM

An Introduction to Deep Learning in Plant Phenotyping Without Agonizing

Pain

Jordan Ubbens, MSc, University of Saskatchewan

Jordan Ubbens – University of Saskatchewan; Ian Stavness – University of

Saskatchewan

Page 53: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Deep learning has long dominated the computer vision field as the most powerful family

of techniques for many image-based vision tasks. Due to these successes, reviews have

called for the penetration of deep learning into the plant phenotyping field. The Deep

Plant Phenomics (DPP) software platform is an open-source Python package which offers

deep learning functionality for various image-based phenotyping tasks. In this talk we

demonstrate how to use the pre-trained networks in the DPP package to perform

vegetation segmentation and leaf counting, as well as how to train and deploy your own

regression models for measuring phenotypic traits of interest. Advanced topics will be

covered including how to design a network using the tool, as well as how to monitor and

interpret the training process. We will also give a brief overview of the state of deep

learning in plant phenotyping and discuss current issues unique to applications in plant

phenotyping.

Teaching students to use supercomputers for phenomics

Eric Lyons, Dr, University of Arizona

Nirav Merchant – University of Arizona

Plant phenomics is transforming plant research. However, a new generation of skills are

required to manage and analyze the hundreds of terabytes of data generated by high-

throughput phenotyping technologies. Applied Concepts in Cyberinfrastructure is a

project based class that takes students from diverse educational backgrounds and teaches

them how to use supercomputers to solve real world problems. Each year, the course

identifies a group of researchers to be the class’ client with a problem in scaling their

computational analyses. Over the course of the semester, the class learns practical

theories behind using different types of high performance, distributed, cloud, and

through computing systems; best practices for using those systems, managing code, and

documentation; and how to work as a team and a team of teams to deliver a solution for

the client. Over the 6 years this course has been taught, the class has worked on problems

from distributed annotation of plant genomes to modeling groundwater movement to

predicting future disease vectoring mosquito abundance to butterfly range maps to

exoplanet detection, with many of these projects becoming peer-reviewed publications.

Each year brings new technologies, new domain areas, and new (unforeseen) challenges.

This talk will discuss the lesson learned by the instructors and how plant phenomics can

benefit by this approach for developing new, scalable solutions for analyzing high-

throughput data.

P3, the Predictive Plant Phenomics Graduate NSF Research Traineeship

(NRT) at Iowa State University

Carolyn Lawrence-Dill, Iowa State University

Page 54: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Carolyn Lawrence-Dill – Iowa State University; Theodore Heindel – Iowa State

University; Julie Dickerson – Iowa State University; Patrick Schnable – Iowa State

University

Modern engineering and data analysis techniques make it feasible to develop methods to

predict plant growth and productivity based on genome and environment information,

however broader skillsets will be needed to unlock this potential, so student training

methods must adapt. This poster describes the structure and activities of a

National Science Foundation Graduate

Research

Traineeship (NRT) award focused on Predictive Plant Phenomics (P3). Our program aims

to increase agronomic output as highlighted by the National Plant Genome Initiative’s

current five-year plan [NST, 2014]. Ph.D. training production levels and types are not

always a good fit for addressing complex technical and societal problems such as these.

To train these scientists, the P3 NRT is using the T-training model proposed by the

American Society of Plant Biology (ASPB) and described in “Unleashing a Decade of

Innovation in Plant Science: A Vision for 2015-2025”. This approach requires that

students get broad exposure to multiple disciplines, work with industry, and develop

effective communication and collaboration skills - all without increasing the time to

graduation. This poster describes how we are working towards meeting these challenges.

Initial results show that the P3 students have more contact with faculty across

departments than single discipline graduate students and are open to learning about new

areas. However, we are still grappling with some issues like finding the best mechanism

for balancing student skills through leveling activities such as boot camps and

introductory course training early on in the program. To learn more about the P3 NRT,

visit: https://www.predictivephenomicsinplants.iastate.edu.

Developing the Pipeline of Plant Phenomics Experts at the Wheat and Rice

Center for Heat Resilience

Argelia Lorence, PhD, Arkansas State University

Argelia Lorence – Arkansas State University

The Wheat and Rice Center for Heat Resilience (WRCHR) is an NSF-funded collaborative

project among the University of Nebraska-Lincoln, Arkansas State University, and Kansas

State University. We are taking a multidisciplinary approach involving plant

physiologists, quantitative geneticists, computational biologists, biochemists, engineers,

informaticians, and precision agronomists to: 1) elucidate the physiological and genetic

basis of high night temperature resiliency of rice and wheat, 2) translate these discoveries

into genetic and phenotypic markers for public and private breeding programs, and 3)

Page 55: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

develop a broad continuum for STEM education. Trends at the global-, regional-, and

farm-level point to an increase in minimum night temperatures that is significantly higher

than the rate of increase in maximum day temperatures. Increases in night temperatures

significantly decrease grain yield and quality of rice and wheat, which together provide

over 50% of the caloric intake for humans worldwide. We are building genome to

phenome linkages using automated image-based phenomics approaches in combination

with transcriptomics and metabolomics applied to wheat and rice diversity panels. The

gene and pathways discovered from this approach will be functionally tested for their role

in improving the temperature resilience in rice and wheat. The planned approach

integrates across greenhouse and field scales, captures complex interactions between the

environment and genome during grain development at high spatio/temporal resolution,

and couples genomics and phenomics outcomes within a quantitative, model-based

framework. The broader impacts of the project include: mentoring five early career

faculty; b) an interdisciplinary graduate course for students in plant sciences, agricultural

engineering, computational biology, biochemistry, statistics, and computer sciences with

traditional and online delivery; c) a phenomics boot camp for minority undergraduates,

and d) a hands-on plant phenotyping module that will involve high school students and

teachers from the three participating states.

Reinventing Postgraduate Training in the Plant Sciences through

Modularity, Customization, and Distributed Mentorship

Natalie Henkhaus, PhD, American Society of Plant Biologists

Vanessa Greenlee – Boyce Thompson Institute; Crispin Taylor – American Society of

Plant Biologists; David Stern – Boyce Thompson Institute

The Plant Science Research Network (PSRN), funded by an NSF Research Coordination

Network grant, consists of fourteen professional societies and organizations whose

members are active in plant science research, education, and advocacy. The PSRN has

identified a set of radical recommendations for postgraduate training that emerged from

two workshops held in October 2016 and September 2017. Both workshops were

supported by scenario development, as reported elsewhere, to encourage out-of-the-box

thinking and innovative recommendations. The recommendations call for a cultural shift

that embraces and extends educational delivery trends towards self-learning and distance

learning, considers trainee well-being as an essential requirement for success, and

acknowledges the requirement for two-way communication with the public. This shift is

intended to reinforce a broadening of the STEM workforce in both diversity and numbers,

while continuing to maintain excellence in scientific training. The recommendations are

meant to catalyze pilot programs, and also to build on emergent prototypes that exist

globally. These recommendations broaden and deepen the “T-training” concept

presented in the 2013 publication, Unleashing a Decade of Innovation in Plant Science.

Page 56: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

The PSRN’s overarching objective is through engagement of the entire plant science

community; the PSRN members aim to build consensus around research, education, and

training objectives that will promote discovery, broaden participation, and have a

measurable impact on pressing challenges around food, agriculture, the environment,

and human health. Learn more by joining the Plant Science Research Network on Plantae:

https://community.plantae.org/organization/plant-science-research-network/

Help! My data is a out of control! Novel services for management of

distributed phenotypic data

Ramona Walls, University of Arizona

Andrew Magill – Texas Advanced Computing Center; Ming Chen – University of

Arizona; James Carson – Texas Advanced Computing Center; Maria Esteva – Texas

Advanced Computing Center

The integration and management of data for plant phenotyping studies presents multiple

challenges. Many phenotype datasets are big, have multiple contributors, contain

components at different stages of completion, and are stored across different platforms.

Phenotype data are structured in myriad ways, have metadata and identifiers (local or

global) that need to be managed pre- and post-publication, and need to link to data and

metadata for other objects such as specimens, accessions, projects, and publications.

Manually performing data management actions for datasets containing hundreds or

thousands of files is tedious at best, impossible at worst. Solving the genotype-to-

phenotype challenge requires that data be discoverable, trustworthy, interpretable, and

accessible (at least to the creators), no matter where they are located or what stage of

completion they are in. Therefore, a new generation of data management tools is needed

for the kind of big data being generated by plant phenotyping studies. Identifier Services

(IDS) is an Early-concept Grants for Exploratory Research (EAGER) project that is

exploring technical solutions to managing large, distributed data. IDS developed a

number of proof-of-concept micro-services for scientists to register their data, organize

and describe them with metadata, and run checks for data identity, authenticity, and

integrity. IDS records the relations among data components, including those stored

across repositories and storage platforms and in different stages of completion. These

relations provide a representation of the dataset, based on a simple generic data model

that can be adjusted to represent different types of research. The data model in turn

supports visualization and management of large datasets, through tasks like bulk

metadata upload and dataset creation based on queries. Together with community data

standards, the micro-services provided by IDS trace data provenance and establish

authenticity over time, making researchers’ lives easier and supporting reproducible

science.

Page 57: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Who is Phenome 2018? Our journey delivering the digital phenotyping

revolution through a combined focus on technology and people.

Bobby Brauer, Monsanto

Jenna Hoffman – Monsanto; Matt McCown – Monsanto; Roger Weyhrich – Monsanto

Recent decades have witnessed great changes in agriculture through advances in the

system of germplasm, traits, chemistry and biologics - leveraged by growers to produce

sustainable crop yields. Improvements in technologies such as high throughput

genotyping have greatly accelerated our ability to discover these components while

innovations in phenomics are beginning to allow us to measure high quality field

phenotypes at scale. Historically, field phenotyping has been the domain of experts

walking plots with pad and pen, or more recently, experts walking plots with mobile

computers.  Environmental metrics have been of a coarse spatial resolution that is

challenging to relate to small research plots.  In addition, sensing technologies to quantify

temporally changing phenotypes and environmental factors have met with significant

cost and technological barriers to scale - both for the sensors themselves as well as for

data acquisition & retrieval.  That is all changing, however, and quickly.  We can now

entertain the imminent prospect of completely reinventing how we manage field testing,

phenotyping, and characterization of agricultural products – with technology at the front

lines.  This revolution at Monsanto is only made possible by a collaboration of

technologists, engineers, breeders, agronomists and data scientists to deliver scalable,

scientifically validated, highly usable solutions that truly enhance field R&D.  This talk

will describe some of the key principles of this revolution, its enabling technologies, and

the culture shifts that we are creating to bring about the future of high throughput, high

resolution phenotyping and field characterization.

Page 58: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Poster Abstracts

Data Computation, Analysis, Modeling

Saguaro Cactus - 1

Peter Pietrzyk, University of Georgia

Automated phenotyping of root hair traits from microscopy images

Chartinun Chutoe – Mahidol University; Patompong Saengwilai – Mahidol University;

Alexander Bucksch – University of Georgia

Improving nutrient and water uptake in crops is one of the major challenges to sustain a

fast-growing population that faces increasingly nutrient limited soils. Root hairs, which

are specialized epidermal cells, compromise up to 70% of the total root surface area.

Therefore, it is likely that root hairs are important for nutrient and water uptake from the

soil. Microscopy provides a mean to record root hairs as digital images. However,

quantifying root hairs in microscopy images remains a bottleneck because of their

geometry and their complex spatial arrangement. We present a method to automatically

quantify phenotypic traits of root hairs in digital microscopy images. Our method uses a

machine learning approach that classifies root hair, parent root and the image

background. In that way, we resolve complexities of root hairs that may cross each other

or form blobs of two or more hairs. We define metrics to distinguish complex cases

computationally. As a result, we measure the root hair traits, length, number and

orientation. We demonstrate our method on examples of rice and maize under phosphor,

nitrogen and potassium stress. Preliminary results suggest that our method reliably

distinguishes between genotypes and treatments on the basis of the extracted traits. We

believe our study paves a way towards identifying the genetic control of root hair traits

and increased agricultural production in future.

Saguaro Cactus - 2

Valerie Cross, Purdue University

Utilizing Hyperspectral Imaging to Predict Relative Water Content in

Sorghum

Valerie Cross – Purdue University; Jian Jin – Purdue University; Mitch Tuinstra –

Purdue University

Page 59: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Non-destructive, high throughput phenotyping has become one of the key goals in

agriculture research. Quickly measuring or predicting plant traits is important for

improving the breeding process and management of crops. Hyperspectral imaging has

potential to improve the speed, accuracy, and applicability of high throughput

phenotyping. We are using hyperspectral imaging to measure and predict the nitrogen

and water content of sorghum under drought and low-nitrogen conditions in a

greenhouse. This current study was developed using three genotypes of sorghum under

four different treatments, two each of water and nitrogen. The plants were imaged and

ground truth data points, including relative water content, nitrogen content, chlorophyll

content, and biomass, were collected from each plant on the day of imaging. From the

collective hyperspectral imaging data, partial least squares models were developed to

correlate the hyperspectral images with the ground truth data. This resulted in partial

least squares models that predict relative water content and nitrogen content in three

sorghum cultivars.

Saguaro Cactus - 3

Sruti Das Choudhury, University of Nebraska-Lincoln

Holistic and Component Plant Phenotyping Analysis using Visible Light

Image Sequence

Sruti Das Choudhury – University of Nebraska-Lincoln; Ashok Samal – University of

Nebraska-Lincoln; Tala Awada – University of Nebraska-Lincoln

Image-based plant phenotyping facilitates the extraction of observable traits by analyzing

large number of plants in a relatively short period of time with little or no manual

intervention. The emergence timing, total number of leaves present at any point of time

and the growth of individual leaves of a plant during vegetative stage life cycle are the

significant phenotypic expressions that best contribute to assess the plant health. Imaging

techniques have the potential to compute advanced phenotypes by considering whole

plant as a single object (holistic phenotypes) as well as its individual components, i.e.,

leaves and stems (component phenotypes), which provide valuable insights into the

physiological characteristics of the plants regulated by genotypes. We introduce three

holistic phenotypes, namely, plant aerial density, bi-angular convex-hull area ratio and

plant aspect ratio to respectively provide information on biomass, plant rotation due to

shade avoidance and canopy architecture. A novel method to automatically detect the

individual leaves and stem of a maize plant by analyzing 2-dimensional visible light image

sequences captured from the side using a graph theoretic approach is introduced. The

total number of leaves are counted and the length of each leaf is measured for all images

in the sequence for automated leaf growth monitoring. We also introduce a set of new

component phenotypes, namely stem angle, inter-junction distance, junction-tip

Page 60: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

distance, leaf-junction angle, integral leaf-skeleton area and leaf curvature, with

significance in plant science. To evaluate the performance of the proposed algorithm, a

benchmark dataset is indispensable. Being inspired by the unavailability of such a dataset,

we introduce University of Nebraska-Lincoln Component Plant Phenotyping Dataset

(UNL-CPPD) and provide ground truth to facilitate new algorithm development and

uniform comparison. Detailed experimental analyses are performed on UNL-CPPD to

demonstrate the temporal variation of the component phenotypes in maize regulated by

genotypes.

Saguaro Cactus - 4

David LeBauer, PhD University of Illinois

TERRA REF Open Software, Data, and Computing to Advance Phenomics

Max Burnette – National Center for Supercomputing Applications; Rob Kooper –

National Center for Supercomputing Applications; Craig Willis – National Center for

Supercomputing Applications; TERRA REF Team – University of Arizona, Danforth

Center, Washington University St. Louis, St. Louis University, George Washington

University, Kansas State University, USDA, University of Illinois, National Center for

Supercomputing Applications

Automated measurements have the potential to advance science and agriculture.

However, there are many technical and economic barriers to entry for scientists. Sensors

and sensing platforms are expensive and difficult to use; software used to process and

interpret these data streams can be expensive and inflexible - if it exists. We need the

ability to build, extend, and combine databases and software components into novel

pipelines.

The TERRA Phenotyping Reference Platform (TERRA-REF) team is developing modular,

reuseable phenomics. We are also creating a large and heterogeneous reference data set

for field and controlled-environment phenotyping platforms. Finally, we have cloud

based development environments that allow users to develop, evaluate, and share

algorithms.

This poster will describe our design approach that enables integeration and development

of modular, interoperable, and extensible components. We will also describe how

software components can be reused, improved, and created. We want to promote and

facilitate the sharing data and tools within the phenomics community so that scientists

cans spend more time on discovery.

Page 61: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Saguaro Cactus - 5

Tyson Swetnam, CyVerse BIO5 Institute University of Arizona

Portable, scalable, high throughput geospatial analyses with Singularity

containers on cloud and high performance computing.

Mats Rynge – Information Sciences Institute, University of Southern California; Jon

Pelletier – Department of Geosciences, University of Arizona

Reproducible science with geographic information systems (GIS) on cloud, high

throughput computing (HTC), and high performance computing (HPC) requires portable,

scalable, workflows as part of the Research Object. Here we present a method for running

free and open-source software for GIS; i.e. Geospatial Data Abstraction Library (GDAL),

Geographic Resources Analysis Support System (GRASS), and System for Automated

Geoscientific Analyses (SAGA), in tandem with a workflow management system,

Makeflow, on cloud and HPC using Singularity containers. Our example workflow

involves the computation of daily and monthly sum solar irradiation using an OpenMP

version of the GRASS r.sun algorithm. A single virtual machine (VM) masters the

workflow, with remote workers connected over Internet2 started on cloud, HTC, and/or

HPC platforms, all using the same Singularity container. The workflow is currently

deployed on the OpenTopography.org cyberinfrastructure, where users can select any

location on the terrestrial earth surface using national or global digital elevation model

(DEM) data to calculate global irradiation and daily hours of sunlight. Our workflow links

with OpenTopography via the Opal2 toolkit for wrapping this particular scientific

application as a Web service from a XSEDE Jetstream VM. The workers are launched on

demand on XSEDE Comet HPC and Open Science Grid HTC. Importantly, because the

workflow is containerized with Singularity, it can be re-deployed on any combination of

local desktop, cloud, or HTC / HPC by simply pulling the code from our GitHub repository

and following a few basic setup instructions. Containerized workflows such as ours that

take an open science approach, as part of the Research Object, will allow for future

reproducible geospatial science on cyberinfrastructure.

Saguaro Cactus - 6

Ian Braun, Iowa State University

Computational Classification of Phenologs across Biological Diversity

Ian Braun – Iowa State University; Carolyn Lawrence-Dill – Iowa State University

Phenotypic diversity analyses are the basis for research discoveries that span the

spectrum from basic biology (e.g., gene function and pathway membership) to applied

Page 62: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

research (e.g., plant breeding). Phenotypic analyses often benefit from the availability of

large quantities of high-quality data in a standardized format. Image and spectral

analyses have been shown to enable high-throughput, computational classification of a

variety of traits across a wide range of phenotypes. However, equivalent phenotypes

expressed across individuals or groups that are not anatomically similar can pose a

problem for such classification methods. In these cases, high-throughput, computational

classification is still possible if the traits and phenotypes are documented using

standardized, language-based descriptions. In the case of text phenotype data, conversion

to computer-readable “EQ” statements enables such large-scale analyses. EQ statements

are composed of entities (e.g., leaf) and qualities (e.g., length) drawn from terms in

ontologies. In this work, we present a method for automatically converting free-text

descriptions of plant phenotypes to EQ statements using a machine learning approach.

In each description, words related to entities and qualities are identified using the

CharaParser annotation tool (Cui, 2012). A classifier identifies potential matches between

these words and terms from a set of ontologies, including GO (gene ontology), PO (plant

ontology), and PATO (phenotype and trait ontology), among others. The features used by

this classifier include semantic, syntactic, and context similarity metrics between words

and ontology terms. This classifier is trained and tested using a dataset of manually

converted plant descriptions and EQ statements from the Plant PhenomeNET project

(Oellrich, Walls et al., 2015). The most likely matching terms identified by the classifier

are used to compose EQ statements. Any obtained results of these automated conversions

in terms of precision and recall will be presented. Potential use across datasets to enable

automated phenolog discovery are discussed.

Saguaro Cactus - 7

Travis Gray, University of Saskatchewan

Beyond Orthomosaics: Multi-Image Spectral Analysis of Agricultural UAV

Imagery

William van der Kamp – University of Saskatchewan; Travis Gray – University of

Saskatchewan; Steve Shirtliffe – University of Saskatchewan; Hema Duddu – University

of Saskatchewan; Kevin Stanley – University of Saskatchewan; Ian Stavness – University

of Saskatchewan; Mark Eramian – University of Saskatchewan

Imagery from Unmanned Aerial Vehicles (UAVs) is frequently used for crop assessment

and phenotyping in agricultural research fields. In particular, spectral indices, such as

NDVI, are commonly employed to estimate traits such as the health, growth stage, and

biomass in crops. In virtually all such studies, many overlapping images from a UAV flight

are first stitched into a single orthomosaic image, and then spectral indices are derived

from the orthomosaic. But this method necessarily discards (or aggregates) much of the

Page 63: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

original pixel information. For example, a single plot may appear in 10 or more individual

images from the UAV, but appear only once in the final orthomosaic. In this paper, we

show that an index value extracted from individual calibrated images of the same plot can

deviate by 10% or more from the index value of the corresponding orthomosaic segment.

This could have important consequences for fine-grained comparison of indices. To

address this problem, we propose alternative approaches to estimating indices, each of

which more directly incorporates all of the individual UAV images. We evaluate these

approaches, and compare them with the orthomosaic approach, by analysing weekly

index values from plant breeding experiments for lentil, wheat, and canola crops in

comparison to relevant manually-measured phenotypes.

Saguaro Cactus - 8

Austin Meier, Oregon State University

The Planteome: A Resource for Reference Ontologies for Plants and Plant

Genomics Database

Laurel Cooper – Oregon State University; Austin Meier – Oregon State University; Justin

Elser – Oregon State University; Pankaj Jaiswal – Oregon State University; Mari-

Angelique Laporte – Bioversity International; Elizabeth Arnaud – Bioversity

International; Seth Carbon – Lawrence Berkeley National Laboratory; Chris Mungall –

Lawrence Berkeley National Laboratory; Barry Smith – University at Buffalo; Dennis

Stevenson – New York Botanical Garden

The Planteome project (http://www.planteome.org) is a centralized online plant

informatics portal which provides semantic integration of a large and growing corpus of

plant genomics data with a suite of reference and species-specific ontologies for plants.

The Planteome reference ontologies include the Plant Ontology, Plant Trait Ontology, the

Plant Stress Ontology and the Plant Experimental Conditions Ontology, along with other

reference ontologies developed by collaborators. Species-specific trait ontologies for crop

plants are mapped to the associated terms in the reference ontologies to create an

integrated ontological network. This integration facilitates studies of plant traits,

phenotypes, diseases, gene function and expression, and genetic diversity data across a

wide range of plant species. In addition, collaboration and annotation tools are being

developed, including the Planteome Noctua platform and remotely accessible APIs. All

the Planteome ontologies are publicly available and are maintained at the Planteome

GitHub site (https://github.com/Planteome) for sharing, and tracking revisions and

issues. The associated data files are freely available for download from the project SVN

(http://planteome.org/svn), and also directly from the Annotation search page on the

Planteome portal (http://browser.planteome.org/amigo/search/annotation).

Page 64: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Saguaro Cactus - 9

Miao Liu, The Climate Corporation

Opportunities & Gaps in Science-Driven Insights in Digital Ag

Frank Dohleman – The Climate Corporation

Digital Agriculture is a new and growing area of research and development. Finding

sufficient high quality data to drive insights from measurements and models for growers

provides an exciting opportunity and enables informed, probabilistic decision making to

maximize the productivity, efficiency, sustainability and profitability of their operations.

The adoption of digital agriculture has grown dramatically in previous years, with over

100 million acres mapped in the Climate FieldviewTM platform.

What have been the biggest successes in digital agriculture to date? What are the

opportunities continue to evolve and drive value out of the digital agriculture arena to

help farmers combat the same challenges they've faced for decades? This presentation

will provide insights into the confluence of soil science, plant science and agronomy with

engineering, software development, and data science to help drive smarter, more efficient

agriculture.

Saguaro Cactus - 10

Seyed Vahid Mirnezami, PhD - Iowa State University

High throughput monitoring anthesis progression of field-grown maize

plants

Seyed Vahid Mirnezami – Iowa State University; Yan Zhou – Iowa State University;

Srikant Srinivasan – IIT Mandi; Baskar Ganapathysubramanian – Iowa State University;

Patrick Schnable – Iowa State University

The tassel is the male organ of the maize plant. Sufficient pollen production is crucial for

the production of hybrid seed. Good seed set requires both sufficient daily production of

pollen, but also pollen shed on enough days to ensure a good “nick” with the receptivity

of female inbreds. Traditional approaches for phenotyping anthesis progression are time-

consuming, subjective, and labor intensive and are thus impractical for phenotyping large

populations in multiple environments. In this work, we utilize a high throughput

phenotyping approach that is based on extracting time-lapse information of anthesis

progress from digital cameras. The major challenge is identifying the region of the interest

(i.e. the location of tassels in the imaging window) in the acquired images. Camera drift,

Page 65: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

different types of weather, including fog, rain, clouds, and sun and additionally, occlusion

of tassels by other tassels or leaves complicated this problem. We discuss various

approaches and associated challenges for object detection and localization under noisy

conditions. We illustrate a promising deep-learning approach to tassel recognition and

localization that is based on Region with Convolutional Neural Network (R-CNN). It is

able to reliably identify a diverse set of tassel morphologies. We subsequently extract

time-dependent tassel traits from these localized images.

Saguaro Cactus - 11

Ethan Stewart, PhD - Cornell

The development and application of a deep learning approach for

quantitative disease phenotyping

Chad DeChant – Columbia; Harvey Wu – Columbia; Tyr Wiesner-Hanks – Cornell; Nick

Kaczmar – Cornell; Rebecca Nelson – Cornell; Hod Lipson – Columbia; Michael Gore –

Cornell

Plant disease is estimated to cause a 13% reduction in global crop production. In order to

breed for improved crop varieties with improved disease resistance, accurate measures of

disease symptoms are required. Traditional visual assessments of disease incidence and

severity are time consuming and prone to human error. Conventional image analysis can

help to improve accuracy and throughput, but requires consistent image conditions that

are difficult to achieve in the field. The advent of deep learning algorithms has helped to

overcome these challenges through training a network that recognizes features of interest

across a diverse range of field environments. Convolutional neural networks (CNN) have

been used to classify images for the presence/absence of one or more diseases. We

previously trained a CNN to recognize foliar lesions caused by Northern leaf blight—a

serious disease of maize—in in UAV-acquired images of field-grown maize plants. We are

now extending this approach by applying instance segmentation using the Mask R-CNN

framework to segment aerial images into 3 classes: leaf, disease lesions and background.

With such an approach it will be possible to provide an aerial based assessment of NLB

disease incidence (qualitative) and severity (quantitative) throughout the maize growing

season.

Saguaro Cactus - 12

Erin Gilbert, MS - University of Minnesota

Hyperspectral phenotyping for early detection of soybean sudden death

syndrome

Page 66: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Erin Gilbert – University of Minnesota; Grace Anderson – University of Minnesota;

James Kurle – University of Minnesota; Cory Hirsch – University of Minnesota

Soybean Sudden Death Syndrome (SDS), a disease caused by Fusarium virguliforme, is

becoming increasingly common in northern latitudes. From 2003 to 2005, SDS was

estimated to be among the five most damaging soybean diseases in the United States, and

it continues to spread nearly unimpeded. Initial infection and fungal development takes

place in the soil, and causes no visible foliar symptoms until late in the growing season,

when symptoms appear during late reproductive stages as interveinal chlorosis and

necrosis accompanied by defoliation. SDS is managed by planting resistant varieties or

application of fungicidal seed treatments. Evaluation for resistance to SDS is typically

conducted as assessment of root rot severity, which requires destructive sampling, or

foliar symptom expression during early vegetative growth approximately 30 days after

planting or during late reproductive stages 60 days or more after planting. In other plant

disease systems, hyperspectral radiometry has been used to detect wavelengths that are

informative for both abiotic and biotic stresses. We are investigating the use of

hyperspectral imaging to assess the response of soybean lines to infection of soybean

seedlings by F. virguliforme to enable earlier and more rapid evaluation of varietal

resistance. Here we demonstrate the use of hyperspectral imagery to detect SDS infected

seedlings in a growth chamber experiment within 20 days of emergence. Using open

source packages, we have automated the identification of plants and pixel wavelength

data extraction. A focus of our analysis is towards understanding changes to detect the

type and timing of stresses. Data obtained through hyperspectral imagery was correlated

with foliar disease and root rot severity evaluated using traditional qualitative methods of

disease assessment.

Saguaro Cactus -13 Suxing Liu, University of Georgia

Put the carbon back into the soil: 3D root phenotyping for improved carbon

sequestration

Suxing Liu – University of Georgia; Alexander Bucksch – University of Georgia

Carbon rich soil ensures the fertile soils and the agricultural productivity of plants that

sequester atmospheric carbon available as CO2 into the soil. Deeper rooting crops are the

key to increase carbon content below ground and to improve soil quality and agricultural

output. Root phenotyping is crucial since it provides avenues to quantify deeper root traits

of important crops like maize. However, due to the opaque nature of soil, dense and highly

Page 67: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

occluded maize root system, quantifying these traits such as whorl number, distance and

number of crown roots is very challenging.

We developed a 3D reconstruction method that completely digitizes the maize root

architecture. The classic structure from motion algorithm was not able to reconstruct

detailed inner roots architecture without manual registration, we extends this method to

produce a dense point could root model based on images collections taken only from the

outside of roots. Our extensions allow automated reconstruction and measurement of the

inner occluded root system.

We demonstrate the quality of our method on two maize genotypes with six replicates for

each genotype. Our 3D root model reconstruction method is a first promising step

towards automated quantification of highly occluded maize root system. And it enable the

discovery of genes associated with deeper rooting by molecular biologists and pave a

promising way to increase soil carbon sequestration in crops<./p>

Eagle Claws Cactus - 14

Liao Fuqi, MA Noble Research Institute

Plant Root Quantitative Analysis

Fuqi Liao, MS – The Noble Research Institute

Root quantitative analysis has become more and more important in plant research.

Currently, many biologists manually measure the root length with help of software (for

example, WinRhizo), but it is limited to small numbers of roots and costs a lot of time.

Many research projects (for example, GWAS) need to quantify a large number of root, and

require software to detect and process root images automatically and finish analysis in

short time. In many situations, precise measurements could be made by software, instead

of manual analysis. Here, we report development of a series of software, which automated

root image analysis with parallel computing on High Performance Computing (HPC)

cluster. The software have been applied to thousands of images of Arabidopsis and

Medicago truncatula roots. With many parameters, which quantifying the roots, the

automated feature of the software allows analysis of thousands of root samples within a

short period of time.

The software can detect nearly 90% of the root hairs, whose length is less than one

millimeter, and measures the root hair lengths. For roots, which were grown for two

weeks in the lab, the software can detect tips of lateral roots, and analyze the lengths of

Page 68: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

main root, and multi-layers of lateral roots. The software also quantifies the roots from

the field with many parameters, including root area, root angle, and total length, etc.

The series of software include many algorithms to provide precise detection and

measurement of the root’s features. In addition, it includes statistical methods, such as

ANOVA and 95% CI, to find significant difference among root groups with genotypes such

as natural variants and developmental mutants. The artificial neural network will soon be

added to classify root groups<./p>

Data Crunching and New Analytics

Eagle Claws Cactus - 15

Amy Tabb, PhD - USDA-ARS-AFRS

Phenotyping tree shape in the field using computer vision and robotics

Phenotyping of tree shape is a challenging problem, not least of which because the

traditional metrics of tree shape – height, width, branch number, branch angle, branch

diameter, and branch length, may not be particularly characteristic of the structural

differences that are evident to humans between phenotypes. We describe ongoing work

to develop a robot vision system that captures the above metrics of fruit tree shape

autonomously and accurately, as well as complete tree reconstructions for use in novel

shape descriptors. I will demonstrate how the system operates in field settings, and

describe its constraints and possible applicability to other species.

Education & Outreach

Eagle Claws Cactus - 16

Carolyn Lawrence-Dill, Iowa State University

P3, the Predictive Plant Phenomics Graduate NSF Research Traineeship

(NRT) at Iowa State University

Carolyn Lawrence-Dill – Iowa State University; Theodore Heindel – Iowa State

University; Julie Dickerson – Iowa State University; Patrick Schnable – Iowa State

University

Page 69: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Modern engineering and data analysis techniques make it feasible to develop methods to

predict plant growth and productivity based on genome and environment information,

however broader skillsets will be needed to unlock this potential, so student training

methods must adapt. This poster describes the structure and activities of a

National Science Foundation Graduate

Research

Traineeship (NRT) award focused on Predictive Plant Phenomics (P3). Our program aims

to increase agronomic output as highlighted by the National Plant Genome Initiative’s

current five-year plan [NST, 2014]. Ph.D. training production levels and types are not

always a good fit for addressing complex technical and societal problems such as these.

To train these scientists, the P3 NRT is using the T-training model proposed by the

American Society of Plant Biology (ASPB) and described in “Unleashing a Decade of

Innovation in Plant Science: A Vision for 2015-2025”. This approach requires that

students get broad exposure to multiple disciplines, work with industry, and develop

effective communication and collaboration skills - all without increasing the time to

graduation. This poster describes how we are working towards meeting these challenges.

Initial results show that the P3 students have more contact with faculty across

departments than single discipline graduate students and are open to learning about new

areas. However, we are still grappling with some issues like finding the best mechanism

for balancing student skills through leveling activities such as boot camps and

introductory course training early on in the program. To learn more about the P3 NRT,

visit: https://www.predictivephenomicsinplants.iastate.edu.

Eagle Claws Cactus - 17

Jordan Ubbens, MSc - University of Saskatchewan

An Introduction to Deep Learning in Plant Phenotyping Without Agonizing

Pain

Jordan Ubbens – University of Saskatchewan; Ian Stavness – University of Saskatchewan

Deep learning has long dominated the computer vision field as the most powerful family

of techniques for many image-based vision tasks. Due to these successes, reviews have

called for the penetration of deep learning into the plant phenotyping field. The Deep

Plant Phenomics (DPP) software platform is an open-source Python package which offers

deep learning functionality for various image-based phenotyping tasks. In this talk we

demonstrate how to use the pre-trained networks in the DPP package to perform

vegetation segmentation and leaf counting, as well as how to train and deploy your own

regression models for measuring phenotypic traits of interest. Advanced topics will be

covered including how to design a network using the tool, as well as how to monitor and

interpret the training process. We will also give a brief overview of the state of deep

Page 70: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

learning in plant phenotyping and discuss current issues unique to applications in plant

phenotyping.

Eagle Claws Cactus-18

Zheng Xu, PhD - University of Nebraska-Lincoln

CT image-based Segmentation and Reconstruction of Root Systems by

Machine Learning and Computational Methods

Camilo Valdes – Florida International University; Stefan Gerth – Fraunhofer Institute for

Integrated Circuits IIS; Jennifer Clarke, Sleep – University of Nebraska-Lincoln

Computed Tomography (CT) scanning technologies have been widely used in many

scientific fields, especially in medicine and materials research. A lot of progress has been

made in agronomic research thanks to CT technology. CT image-based phenotyping

methods enable high-throughput and non-destructive measuring and inference of root

systems, which makes downstream studies of complex mechanisms of plants during

growth feasible. An impressive amount of plant CT scanning data has been collected, but

how to analyze these data efficiently and accurately remains a challenge.

We present new computational and machine learning methods for better segmentation

and reconstruction of root systems from 3D CT scanning data. We propose new

approaches within the category of voxel thresholding methods. Considering special

characteristics of root systems, we propose our methods based on two new local-feature

statistics, i.e., proportion and weighted proportion. We found that methods based on our

two new statistics can calibrate root system magnitudes faster than traditional vessel-

based approaches while preserve similar levels of performance. In addition, we propose

and evaluate machine learning approaches in root-system segmentation and

reconstruction from CT-images, in particular simulation-assisted machine learning

approaches. We illustrate and compare different approaches using both simulated and

real CT scanning data from Fraunhofer Institute for Integrated Circuits IIS.

Eagle Claws Cactus - 19

Natalie Henkhaus, PhD - American Society of Plant Biologists

RCN: Coordinated Plant Science Research and Education Network

Vanessa Greenlee – Boyce Thompson Institute; Crispin Taylor – American Society of

Plant Biologists; David Stern – Boyce Thompson Institute

Page 71: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

The Plant Science Research Network (PSRN), funded by an NSF Research Coordination

Network grant, consists of fourteen professional societies and organizations whose

members are active in plant science research, education, and advocacy. The PSRN has

identified a set of radical recommendations for postgraduate training that emerged from

two workshops held in October 2016 and September 2017. Both workshops were

supported by scenario development, as reported elsewhere, to encourage out-of-the-box

thinking and innovative recommendations. The recommendations call for a cultural shift

that embraces and extends educational delivery trends towards self-learning and distance

learning, considers trainee well-being as an essential requirement for success, and

acknowledges the requirement for two-way communication with the public. This shift is

intended to reinforce a broadening of the STEM workforce in both diversity and numbers,

while continuing to maintain excellence in scientific training. The recommendations are

meant to catalyze pilot programs, and also to build on emergent prototypes that exist

globally. These recommendations broaden and deepen the “T-training” concept

presented in the 2013 publication, Unleashing a Decade of Innovation in Plant Science.

The PSRN’s overarching objective is through engagement of the entire plant science

community; the PSRN members aim to build consensus around research, education, and

training objectives that will promote discovery, broaden participation, and have a

measurable impact on pressing challenges around food, agriculture, the environment,

and human health. Learn more by joining the Plant Science Research Network on Plantae:

https://community.plantae.org/organization/plant-science-research-network/

Phenomics Enabled Biology

Eagle Claws Cactus - 20

Md Rofiqul Islam, Gauhati University

Agarwood: DNA fingerprinting to Biomarkers

Md Rofiqul Islam – Gauhati University; Sofia Banu – Gauhati University

Aquilaria malaccensis is a woody plant producing agarwood in it's stem which is highly

valuable resinous fragrant deposits. Agarwood is widely used in traditional medicines,

incense and perfume. The objective of this study was to convert transcript into usable

biomarkers. To achieve this objective, cDNA-AFLP technique was used to identify

transcriptionally regulated genes in A. malaccensis. Samples of wood were collected from

plants showing infection from three different locations of Assam and cDNA were

prepared. cDNA-AFLP analysis involved selective amplification with 64 different pair of

primers that allowed the visualisation of 2760 reliable differentially expressed transcript

Page 72: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

derived fragments (TDFs).Of these 30 different TDFs were successfully cloned and 20

fragments were sequenced. Eight of which have been identified as Aquilaria transcript

using homology search by BLAST, six have been found to be directly involved in terpenoid

pathway. Primers were designed from these TDFs sequence and expression patterns in

infected and non-infected plants were studied using the real time polymerase chain

reaction. All the terpenoid genes TDFs were found to be up regulated in infected Aquilaria

plants as compared to non-infected plants. The study has identified the TDFs that are

overexpressed in infected plants which can be used as biomarkers for distinction of non-

infected from infected plants using very small wood samples from the plant thus it can be

used as a tool for reliable identification of status of infection in Aquilaria prior to

harvesting .

Eagle Claws Cactus - 21

Ramesh Katam, Florida A&M University

Physiological and Proteomic Analysis of Pistachio Rootstocks in Response to

Salinity Stress

Mohammad Akbari – University of Tabriz; Nasser Mahna – University of Tabriz

Pistachio (Pistacia vera L.), cultivated in arid and semi-arid regions, is one of the most

important nuts worldwide. However, the mechanisms underlying salinity tolerance of this

plant is not well understood. Hence, the studies were carried out both physiological and

molecular level to unravel the metabolic pathways associated with the salt tolerance

mechanisms in various cultivars. Five one-year-old pistachio rootstocks were treated with

four saline water regimes (control, 8 dS m-1, 12 dS m-1, and 16 dS m-1) for 100 days and

physiological, biochemical and proteomic analysis were carried out. Salinity decreased

the Relative water content, Total chlorophyll content and carotenoids in the leaves, and

ascorbic acid and total soluble proteins in both leaves and roots.

Results shows that three different ion exclusion strategies were observed in studied

rootstocks, (i) Na+ exclusion in UCB-1, because retained most of its Na+ in the roots (ii)

Cl- exclusion in Badami, which most of its Cl- remained in the roots (iii) and similar

concentrations of Na+ and Cl- were observed in the leaves and roots of Ghazvini, Akbari

and Kale-Ghouchi. Based on the results, rootstocks arranged from tolerant to susceptible

follows: UCB-1 > Badami > Ghazvini > Kale-Ghouchi > Akbari. High throughput

comparative proteomics of roots identified 153 upregulated and 69 downregulated

proteins in UCB-1 (as tolerant) and 340 upregulated and 18 downregulated proteins in

Akbari (as susceptible). The majority of identified proteins have the functions related to

stress responsive proteins, signal transduction, cell wall and cytoskeleton metabolism,

and transcription factor. The data suggests a strong linkage of molecular mechanism with

Page 73: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

the physiological traits in the cultivars with various salt tolerances, and lead to further

functional elucidation and genetic engineering approaches to improve salt tolerance in

plant species.

Eagle Claws Cactus - 22

Raksha Singh, University of Arkansas Fayetteville

Unraveling the complexity of callose deposition during plant immunity by

integrating two new players: AtNHR2A and AtNHR2B

Laura Ortega, Undergraduate – University of Arkansas; Clemencia Rojas, Assiatant

Professor – University of Arkansas

Callose deposition occurs throughout plant development and in response to potential

fungal, oomycete and bacterial pathogens. Callose biosynthesis is linked to glucosinolate

metabolism and requires the plant hormone ethylene and the genes PEN2 (penetration

2), a myrosinaseand PEN3 (penetration 3), an ABC transporter. Although callose

biosynthesis has been elucidated, little is known about the cellular events mediating its

deposition to the cell wall. We have identified two genes in Arabidopsis thaliana:

AtNHR2A (A. thaliana nonhost resistance 2A) and AtNHR2B (A. thaliana nonhost

resistance 2B) as new components of the plant innate immune system. AtNHR2A and

AtNHR2B play essential roles in callose deposition in response to the bacterial pathogen

Pseudomonas syringae pv tabaci. Using laser scanning confocal microscopy, we showed

that fluorescent versions of AtNHR2A and AtNHR2B: RFP-AtNHR2A and AtNHR2B-

GFP localize to the cytoplasm and to small and dynamic subcellular structures

reminiscent of the endomembrane system. Interestingly, genetic analysis of Atnhr2b x

Atnh2a double mutant together with the finding that both RFP-AtNHR2A and AtNHR2B-

GFP partially co-localize indicate that both proteins have a synergistic function in callose

deposition. Genetic analysis of the double mutants: Atpen2 X Atnhr2a, Atpen2 X

Atnhr2b, Atpen3 X Atnhr2a, Atpen3 X Atnhr2b showed that AtNHR2A and AtNHR2B

are components of the PEN2/PEN3 pathway. While the identity of the subcellular

compartments where AtNHR2A and AtNHR2B localize is still unknown, our findings

suggest that these compartments participate in the delivery of callose to the cell wall and

thus, this work provides fundamental knowledge regarding the poorly understood

phenomenon of cell wall modification during bacterial infection.

Eagle Claws Cactus - 23

Nathanael Ellis, Donald Danforth Plant Science Center

Page 74: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Identifying root architectural differences in field excavated crown roots

between density adapted populations of maize

Nathanael Ellis – Donald Danforth Plant Science Center; Kari Miller – Donald Danforth

Plant Science Center; Jode Edwards – USDA-ARS Corn Insects and Crop Genetics

Research Unit, Ames, IA; Christopher Topp – Donald Danforth Plant Science Center

Since the early 1900’s, annual gains in U.S. maize production have largely been driven by

developing inbred lines for crossing to produce hybrids. Two historically important maize

populations, Iowa Stiff Stalk Synthetic (BSSS) and Corn Borer Synthetic (BSCB), have

been recurrently selected for increased hybrid grain yield, from which many elite

commercial lines have arisen. Yield increase, over this 70-year experiment, is largely

attributed to the selection of maize improved performance in progressively dense

environments. Studies have shown structural and biological differences between

historical and modern maize for above ground traits, but only few experiments focus on

the hidden-half below in similar populations. Here, we compare historical (C0) and

modern (C17) synthetic lines and their hybrids, measuring whole crown roots (CRs) over

time to examine growth rate, dry-weight, and root architecture. RCs were analyzed with

2D root imaging software. BSSS and BSCB C0 populations had consistently larger Crown

Root Area (CRA) compared to C17 synthetic populations, in both 4” dense planting and

12” sparse planting. CRA for each hybrid line depended on cycle and density, Hybrid C0

was larger than parents when sparse but smaller when grown dense. Hybrid C17 CRA was

larger compared to parents in both sparse and dense, and slightly more than C0 in dense.

Hybrids were grown in dense mono- or polyculture rows and 2-D imaged from two angles

to identify difference in biomass distribution. More prominently at flowering, the CR

width of C17 was wider, allocating biomass away from neighbor plants and narrow

between neighbors, regardless neighbor cycle. These results confirm our preliminary data

of major root architectural changes between C17 and C0 in response to density as well as

variation in biomass distribution based on neighboring plants. We hope to narrow down

specific growth rate response at much finer scales, both temporally and spatially.

Eagle Claws Cactus - 24

Florie Gosseau, LIPM, Universite de Toulouse, INRA, CNRS, Castanet-Tolosan, France

Heliaphen, an outdoor high-throughput phenotyping platform designed for

the whole plant cycle.

Florie Gosseau – LIPM, Universite de Toulouse, INRA, CNRS, Castanet-Tolosan, France;

Nicolas Blanchet – LIPM, Universite de Toulouse, INRA, CNRS, Castanet-Tolosan,

France; Louise Gody – LIPM, Universite de Toulouse, INRA, CNRS, Castanet-Tolosan,

Page 75: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

France; Pierre Casadebaig – AGIR, ENSAT, INRA, Castanet-Tolosan, France; Nicolas

Langlade – LIPM, Universite de Toulouse, INRA, CNRS, Castanet-Tolosan, France

Heliaphen is an outdoor high-throughput phenotyping platform allowing automated

management of growth conditions and monitoring of plants during the whole plant cycle.

A robot moving between plants growing in 15L pots monitors plant water balance and

phenotypes plant or leaf morphology, from which we can compute more complex traits

such as the response of leaf expansion (LE) or plant transpiration (TR) to water deficit.

Here, we illustrate the platform capacities on two practical cases: a genetic association

study for yield-related traits and a simulation study, where we use measured traits as

inputs for a crop simulation model. For the genetic study, classical measurements of

thousand-kernel weight (TKW) were done under water stress and control condition

managed automatically on a sunflower bi-parental population. The association study on

TKW in interaction with hydric stress highlighted five genetic markers with one near to a

gene differentially expressed in drought conditions from a previous experiment. For the

simulation study, we used the SUNFLO crop growth model to assess the impact of two

traits measured in the platform (LE and TR) on crop yield in a large population of

environments. We conducted simulations in 42 contrasted locations across Europe and

21 years of climate data. We identified ideotypes (i.e. genotypes with specific traits values)

that are more adapted to specific growing conditions, defined by the pattern of abiotic

stress occurring in these type of environments.

This study exemplifies how phenotyping platforms can help with the identification of the

genetic architecture of complex response traits and the estimation of eco-physiological

model parameters in order to define ideotypes adapted to different environmental

conditions.

Eagle Claws Cactus - 25

Juniper Kiss, Aberystwyth University

Phylogenetic signal in subgenus Rubus (bramble, blackberry) leaflet shape

using geometric morphometrics

Juniper Kiss – Aberystwyth University; Dániel Knapp – Eötvös Loránd University,

Institute of Biology; Gábor Kovács – Eötvös Loránd University, Institute of Biology;

Michal Sochor – Crop Research Institute; Nigel Cooper – Anglia Ruskin University

Plant phenotypic plasticity and different ways of genetic recombination during clonal and

sexual reproduction make the identification of some plant species difficult. Although DNA

barcoding has revolutionised species identification, polyploidy, hybridisation and

Page 76: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

apomixis pose challenges to this process. Subgenus Rubus (brambles, blackberries) is one

of the most taxonomically challenging groups of dicots and their morphology based

classification has not been entirely consistent with their molecular phylogeny. The

definition of bramble species is controversial and is often reliant on leaf and leaflet

characters. Here, we combined geometric morphometrics with molecular analysis.

A total of 230 leaves from 115 specimens were imaged from different environments

(woodland, sandy beach, saltmarsh, grassland) in the UK. We conducted a three-loci

molecular analysis using ITS (internal transcribed spacer) region of nrDNA and two

cpDNA regions, maturase K (matK) and trnL–trnF for 23 representative leaf samples. We

analysed the shape of five-foliate and three-foliate leaves using landmark-based image

analysis. Using Principal Component Analysis (PCA) and Canonical Variate Analysis

(CVA), the leaflet shapes clustered according to the different environments.

Discrimination Analysis (DA) also confirmed that most of the group mean shapes were

highly significantly different (P < 0.001) at different locations, while it was more obscure

when analysed for differences in between bramble series. Using squared-change

parsimony, the molecular phylogeny of the haplotypes was projected into the leaflet

morphospace. Permutation tests suggested the phylogenetic signal in leaflet arrangement

morphology to be statistically significant (P < 0.05). These results suggest that each

haplotype has different shapes in different environments, while the overall shape

differences of haplotypes could be explained by their phylogeny. We suggest a statistically

robust approach to combine morphometric analysis with molecular data to understand

the variability of leaflet shape which could affect the morphology-based classification of

Rubus.

Eagle Claws Cactus - 26

Karsten Nielsen, University of Saskatchewan

My Lentils are Bigger than Your Lentils

Steve Shirtliffe, Ph.D – University of Saskatchewan; Hema Duddu, Ph.D – University of

Saskatchewan; Kirstin Bett, Ph.D – University of Saskatchewan; Menglu Wang, B.Sc. –

University of Saskatchewan

One of the most time-consuming components in plant breeding programs is the process

of phenotyping, or identifying traits of interest in a field environment. This data collection

process is typically carried out using human labor, and consumes considerable time and

money. Analyzed overhead images may instead be used as an equivalent or improved

source of information. A diversity panel of 324 lines of lentil (Lens culinaris Medik.),

arranged in randomized complete block design (RCBD) and consisting of three replicates,

were grown in microplots in two locations in Saskatchewan, Canada in 2016 and 2017. An

Page 77: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

additional subset of 6 varieties was grown in 3 locations in Saskatchewan in 2017 for the

purpose of collecting whole-plot biomass once every two weeks from emergence until

maturity. All trials were imaged using ground and/or aerial-based overhead imaging

techniques which produce high resolution images.

This research is ongoing. For final analysis, two-dimensional merged images

(orthomosaics) will first be produced. Then, utilizing Structure from Motion (SfM)

techniques, 3-dimensional point clouds will be constructed and utilized as an estimate of

volume. Volume estimates will be compared with biomass measurements. In partnership

with the University of Saskatchewan’s Department of Computer Science, machine

learning techniques may be utilized to correlate UAV imagery with plot biomass. This has

potential to allow rapid, non-destructive biomass estimations and predictions with high

temporal resolution.

Bonker Hedgehog Cactus - 27

Kyle Parmley, Iowa State University

Machine learning approaches in Soybean Phenomics: Predicting Seed Yield,

Oil and Protein in Contrasting Production Systems

Kyle Parmley – Iowa State University

Genetic improvement of soybean [Glycine max (L.) Merr.] has permitted the expansion

of soybean across a broad geographic region. Past breeding efforts have attempted to

develop highly stable cultivars to deploy across all production systems, but these

genotypes may evade an advantageous genotype by management (G x M) interaction, i.e.,

row width spacing. The development of these production system targeted cultivars will

require continual improvement of yield per acre of soybean, which in turn will be

dependent on the modification of physiological traits. Advances in remote sensing

technologies have enabled rapid measurements of these traits on a temporal and spatial

scale, and therefore are becoming increasingly adopted in advanced breeding systems.

The objective of this study to develop yield prediction models using machine learning

approaches. We used two independent studies with 32 genotypes of the SoyNAM panel

with contrasting treatments: row width spacing (38 and 76 cm) and seeding density (123,

345, 568 x 103 seeds ha-1) from nine environments in replicated tests. Physiological trait

data of hyperspectral reflectance, leaf area index, canopy temperature, light interception,

and chlorophyll content were collected at three time points during the growing season.

Robust in-season prediction models identified informative explanatory varaiables for

seed yield, oil and protein predictions, which and will aide in breeding applications for

contrasting production systems. Preliminary results indicate prediction accuracies were

also similar for remote sensing tools with moderate and high throughput capability

Page 78: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

thereby decreasing the temporal requirement for data acquisition. The application of

these approaches enable a mechanistic understanding of yield drivers in contrasting

production systems and enable more informative decision making capability.

Bonker Hedgehog Cactus - 28

Sara Tirado, University of Minnesota

Field Based Phenotypic Platform for Characterizing Maize Growth and

Development

Sara Tirado – University of Minnesota; Nicolai Haeni – University of Minnesota; Volkan

Isler – University of Minnesota; Candice Hirsch – University of Minnesota; Nathan

Springer – University of Minnesota

Advances during the past decade have allowed researchers to link genomic information

to phenotypic information and through this enhance crop productivity. Further progress

in the ability to link genotypes, environments, and phenotypes has been limited by the

accuracy and consistency of measuring traits of agronomic importance on a large field-

based scale. Current field-based phenotyping efforts are time and labor intensive and

therefore hinder the development of large-scale trait datasets at multiple timepoints

throughout a plant’s life cycle. This brings a need for developing fully automated,

inexpensive procedures for objectively measuring plant traits in field settings. We have

developed a procedure for utilizing RGB drone imagery to extract phenotypic traits of

importance, including stand count, plant height, canopy closure, and growth rate. This

platform was used to characterize a maize association mapping population that consists

of 500 diverse inbred lines grown in replicate in Saint Paul, MN in 2017. In total five

flights were conducted throughout the growing season. Results of repeatability across

replicates and variation in growth rates as measured through plant height and canopy

closure will be presented. The application of this technology can be used to deepen our

understanding of how genetic variation and environmental influences shape the traits of

corn plants in the field.

Bonker Hedgehog Cactus - 29

Dalal Alonazy, Florida A&M university

Strong Protein-interactions identified in Drought tolerant Peanut Leaf

Proteome

Ramesh Ramesh – FAMU

Page 79: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

In peanut water stress predisposes to pre-harvest fungal infection leading to aflatoxin

contamination. Major changes during water stress include oxidative stress leading to

destruction of photosynthetic apparatus and other macromolecules within cells. To gain

better understanding of the effects on molecular and cellular functions, two peanut

cultivars with diverse drought tolerance characteristics were subjected to water stress

(WS). Leaf samples at different stress intervals were subjected to proteome analysis using

two-dimensional electrophoresis complemented with MALDI/TOF mass spectrometry.

Ninety-six proteins were differentially expressed in response to water stress in both

cultivars. Three proteins: glutamine ammonia ligase, chitin class II and actin isoform B,

were unique to tolerant cultivar. Four proteins: serine/threonine protein phosphate PP1,

choline monooxygenase, peroxidase 43 and SNF1-related protein kinase regulatory

subunit beta-2, which play a role as cryoprotectants through signal transduction and

defense were induced in drought tolerant cultivar following WS. Several of the leaf

proteins that were over expressed in tolerant cultivar to WS were suppressed in

susceptible cultivar. Protein interaction prediction analysis suggests that more proteins

interacting in tolerant cultivar were shown to activate other proteins in directed system

response networks. Interologs of these proteins were found in Arabidopsis and we believe

that similar mechanism might exist in peanut.

Bonker Hedgehog Cactus - 30

Anique Josuttes, University of Saskatchewan, Department of Plant Science

Utilizing Deep Learning to Predict the Number of Spikes in Wheat (Triticum

aestivum)

Steve Shirtliffe – University of Saskatchewan, Department of Plant Science; Curtis

Pozniak – University of Saskatchewan; Duddu Hema Sudhakar – University of

Saskatchewan, Department of Plant Science; Menglu Wang – University of

Saskatchewan, Department of Plant Science

There is potential to use phenotypic traits as selection tools in breeding programs. The

bottleneck lies in obtaining phenotypic measurements. Using image analysis to analyze

phenotypic traits of interest may result in more accurate and representative phenotyping.

The number of wheat spikes is shown to have correlations with end yield. Sixteen diverse

Triticum aestivum varieties were seeded in three locations in 2016 and 2017. The trial is

composed of three reps in a randomized complete block design. Plots are seeded double

wide to allow for destructive phenotypic measurements to take place in one plot and yield

to be measured in the other. The total number of wheat spikes per plot were counted.

Also, measurements that contribute to spike metrics such as: plant number, tiller number,

and kernel weight, were collected. Both ground and aerial platforms were tested in the

trial to gather high resolution images.

Page 80: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Collected images have been subjected to two-dimensional merged images, also known as

orthomozaics. Individual plots that were captured from a zenith view are being analyzed

further using artificial neural networks, in hopes of detecting the number of wheat heads

on a per plot basis. This allows for potential, non-destructive, yield predictions. This

research is on-going.

Bonker Hedgehog Cactus - 31

Therese LaRue, Stanford University

Uncovering the genetic basis for natural variation of root system dynamics

in Arabidopsis

As the interface between the soil environment and the rest of the plant, root systems play

a key role in determining plant growth. The distribution of roots, termed root system

architecture (RSA), is influenced by the surrounding environment. GLO-Roots (Growth

and Luminescence Observatory for Roots) is a new soil-based root imaging technology,

which enables detailed observations of Arabidopsis thaliana root system growth in soil.

Using this system, we are characterizing how the root systems of a diverse population of

Arabidopsis accessions grow over time and will perform a genome wide association study

to identify common alleles involved in RSA control. In parallel, modelling soil-root

interactions to predict RSA function and performance under different stress conditions

will inform us about improved RSA strategies. Together, this work will investigate how

plant root systems are distributed spatially within the soil and identify ways plants

regulate root system growth to cope with

Bonker Hedgehog Cactus - 32

Carolyn Rasmussen, PhD University of California, Riverside

Division plane orientation in plant cells

One key aspect of cell division in multicellular organisms is the orientation of the division

plane. Proper division plane establishment significantly contributes to normal

organization of the plant body. To determine the importance of cell geometry in division

plane orientation, we designed a three-dimensional probabilistic mathematical modeling

approach based on century-old observations: equal daughter cell volume and the

resulting division plane is a local surface area minimum. Predicted division planes were

compared to a plant structure that marks the division site, the preprophase band (PPB).

PPB location typically matched one of the predicted divisions. Predicted divisions offset

from the PPB occurred when a neighboring cell wall or PPB was observed directly

adjacent to the predicted division site, as to avoid creating a “four-way junction”.

Page 81: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Population level modeling accurately predicted ~95% of in vivo cell divisions based on

geometry. This powerful model can be used to separate the contribution of geometry from

mechanical stress or developmental regulation in predicting plant division plane

orientation.

Bonker Hedgehog Cactus - 33

Amritpal Singh, Purdue University

Towards the Development of an Aerial Platform for High Throughput

Phenotyping of Maize

Jieqiong Zhao – Purdue University; Ali Masjedi – Purdue University; Yuhao Chen –

Purdue University; Javier Ribera – Purdue University; Yun-Jou Lin – Purdue University;

Addie Thompson – Purdue University; Edward Delp – Purdue University; David Ebert –

Purdue University; Ayman Habib – Purdue University; Melba Crawford – Purdue

University; Mitchell Tuinstra – Purdue University

Tremendous progress has been made in plant genotyping and sequencing in recent years.

Low-cost, high-throughput genotyping platforms are now available to rapidly

characterize crop germplasm collections and breeding populations. Availability of

abundant cheap molecular markers has facilitated gene/quantitative trait loci (QTL)

discovery through genome-wide association studies as well as selection of genetically

superior candidates via genomic selection. Obtaining accurate plant phenotypes however

has remained a major challenge in plant breeding. Plant phenotyping has not achieved a

similar level of accuracy and throughput as achieved in genotyping. High-throughput

phenotyping (HTP) methods using unmanned aerial vehicles (UAVs) fitted with sensors

can potentiality phenotype a large number of research plots for numerous traits. The

objective of this study was to evaluate a UAV based phenotyping system for its ability to

phenotype maize hybrids for traits that are important for crop improvement. Two

replications of 250 maize hybrids that are a part of Genomes to Fields initiative were

grown at the Agronomy Center for Research and Extension (ACRE) at Purdue University

in 2017. High resolution spatial data was collected using RGB-visible, visible/near-

infrared (VNIR), and shortwave infrared (SWIR) sensors fitted on a DJI M600 flying

platform. Ground reference data was collected for plant height, ear height, stand count,

and end of the season stay-green. Remotely sensed data will be processed to obtain the

estimates of the plant height, stand count, and stay green. Predictive ability of this UAV

based platform will be assessed by correlating the results to the ground reference data.

Successful implementation of the UAV based platform may help the private and public

maize breeding programs to accurately phenotype the maize traits and reduce laborious

phenotyping efforts.

Page 82: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Bonker Hedgehog Cactus - 34

Kamaldeep Virdi, PhD - University of Minnesota

Genetic control of soybean (Glycine max L. Merr.) shoot architecture

Gary Muehlbauer – University of Minnesota; Robert Stupar – University of Minnesota;

Aaron Lorenz – University of Minnesota; Austin Dobbels – University of Minnesota;

Suma Sreekanta – University of Minnesota; Jeffrey Roessler – University of Minnesota

Soybean shoot architecture is a complex phenotype, which can be partitioned into various

measurable traits. Variation in shoot architecture likely influences canopy light

interception, photosynthesis, and source-sink partitioning efficiency, and thus is related

to overall grain yield. Here, shoot architecture-related traits are defined as branch angle,

branch number, branch length, leaf shape and size, petiole length, petiolule length,

canopy coverage, days to flowering, maturity, determinancy, and light penetration

through the canopy. A detailed phenotypic characterization of these shoot architecture-

related traits, their contributions to overall shoot architecture, and their genetic control

is imperative to fully exploit the yield potential of soybean. We examined a set of 400

diverse maturity group 1 soybean accessions to study the natural variation for shoot

architecture-related traits, to identify relationships between traits, and to use association

mapping to identify loci that are associated with shoot architecture traits. The panel was

genotyped with 32,650 SNP markers. To collect phenotype data for shoot architecture-

related traits, we used a combination of high-throughput (drone and ground-based

imagery) and conventional phenotyping platforms. Significant QTL associated with

branch angle, leaf length/width ratio, petiolule length, days to flowering, maturity and

stem termination were detected. In most cases, these QTL overlapped with previously

detected genes or QTL. For example, a leaf length/width ratio QTL on chromosome 20

was found to be coincident with the location of a previously isolated Narrow Leaf (Ln)

gene. Interestingly, we detected a branch angle QTL located on chromosome 19 that

overlaps with QTL associated with canopy coverage and light penetration, suggesting

branch angle is an important determinant of canopy coverage.

Bonker Hedgehog Cactus - 35

Tae-Kyeong Noh, Seoul National University

Early diagnosis of plant responses to osmotic stress using spectral image

analysis

Do-Soon Kim, Seoul National University; Tae-Young Lee – Seoul National University;

Hejin Yu – Seoul National University

Page 83: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

The spectral image analysis is expected to be a key technology for high-throughput

phenotyping (HTP) or screening (HTS) but not yet fully applied to HTP/HTS. Therefore,

this study was conducted to establish spectral image analysis system with emphasis on

thermal and chlorophyll fluorescence image analysis for plants grown under osmotic

stresses. Various plants were treated with NaCl and poly ethylene glycol (PEG) at an early

growth stage to induce an abiotic stress. Plant body temperature estimated by thermal

image analysis increased with the extent of abiotic stress, giving negative correlations

with photosynthetic rate. Chlorophyll fluorescence intensity decreased with the extent of

abiotic stress, giving positive correlations with photosynthetic rate. These results suggest

that image-based parameters such as plant body temperature and chlorophyll

fluorescence intensity can replace photosynthesis rate and be used to diagnose crop

tolerance to abiotic stress. Normalization of image-based parameters were able to classify

based on osmotic stress tolerance. In conclusion, our results revealed that the spectral

imaging system established in this study could give biological significance to the crop

images in association with physiological parameters and be a part of HTP and HTS.

Bonker Hedgehog Cactus - 36

Haichao Guo, Noble Research Institute, LLC

Decreasing nodal root number in maize enhances nodal and lateral root

length while increasing shoot biomass when nitrogen is limiting

Haichao Guo – Noble Research Institute, LLC; Anand Seethepalli – Noble Research

Institute, LLC; Larry York – Noble Research Institute, LLC

The root phenome influences crop performance, yet our understanding is limited.

Previous simulations using the structural-functional model SimRoot indicated reduced

nodal root number was a promising target for maize breeding. These simulations were

partially confirmed in field and greenhouse studies relying on natural variation of nodal

root number among recombinant inbred lines. In this study, we directly tested the

hypothesis that reduced nodal root number allows increased elongation of remaining

nodal roots and their laterals, and possibly reduces the competition for resources among

roots. We assessed the influence of nodal root number with a manipulative experiment

by excising roots as they emerge from whorls on plant growth and root system

architecture. The maize genotype used was inbred line B73 and the plants were grown in

150 cm tall mesocosms in high and low nitrogen (N) media. Nodal root number

manipulation treatment levels were excising 0%, 33%, and 67% of all nodal roots as they

emerged over time. An isotopic label of 15N was injected at 143 cm depth to test deep N

capture. At 42 days after planting, the root systems were intensively phenotyped by

washing away media, separating the root system by axial root class, dividing into 30 cm

segments from top to bottom, then scanning with a photo scanner. Root scans were

Page 84: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

analyzed using a new C++ software called RhizoVision-Scan. Excising nodal roots

resulted in higher lateral per axial root length, greater total root length, and increased

deeper rooting regardless of N level. Under low N, excising nodal roots significantly

increased plant diameter and plant height, and the two excision levels resulted in an

averaged 50% increase in shoot biomass relative to the 0% excision control. These results

confirm the utility of reduced nodal root number as a promising target for maize, and

possibly other cereals.

Bonker Hedgehog Cactus - 37

David Hanson, University of New Mexico

Rapid gas exchange in the phenomic era

Joseph Stinziano – University of Western Ontario

Phenotyping for photosynthetic gas exchange parameters is limiting our ability to select

plants for enhanced photosynthetic carbon gain and to assess plant function in current

and future natural environments. This is due, in part, to the time required to generate

estimates of the maximum rate of ribulose-1,5-bisphosphate carboxylase oxygenase

(Rubisco) carboxylation, the maximal rate of electron transport, and Rubisco activation,

from the response of photosynthesis to the CO2 concentration inside leaf air spaces. To

relieve this bottleneck, we developed a method for rapid photosynthetic carbon

assimilation CO2 responses utilizing non-steady-state measurements of gas exchange.

Using high temporal resolution measurements under rapidly changing CO2

concentrations, we can collect traditional gas exchange parameters in around 2 minutes.

This is a small fraction of the time previously required for even the most advanced gas

exchange instrumentation. We present how we have applied this method to diurnal

changes in physiology as well as responses to light, CO2, and temperature.

Bonker Hedgehog Cactus - 38

Abdullah A Jaradat, USDA/ARS & University of Minnesota

Forward Phenomics of oat Panicles

There is a growing need for adapted and more productive germplasm to expand oat

production, optimize its yield, improve groat quality, and satisfy farmers and consumers

demand, especially in the Upper Midwest of the US. Oat germplasm, representing

different eco-geographical origins and breeding status, was characterized and evaluated

using field- and laboratory-based forward phenomics. Whole plots were phenotyped at

successive growth stages during three growing seasons using aerial and hand-held

imagery and sensors. Digital and high-throughput data were captured and compiled on

Page 85: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

(1) color space descriptors of whole plots during key growth stages, (2) growing degree

days to panicle emergence and maturity; (3) phenotypic and structural traits of panicles

and spikelets; and (4) groat quality. A relational database was mined and statistically

analyzed to (1) cluster the oat germplasm using unsupervised hierarchical clustering

method; (2) identify traits with positive or negative, direct or indirect effect on the panicle

phenome; (3) identify a minimum set of traits which can discriminate between

structurally and agronomically different panicle phenotypes; (4) express groat weight per

plant as a function of stochastic panicle architecture and (5) quantify the effect of panicle

architecture traits that have implications for groat quality. A dynamic custom profiling

procedure was instrumental in quantitatively assessing the importance of, and visually

adjusting structural panicle traits to predict and optimize groat agronomic and quality

traits.

Bonker Hedgehog Cactus - 39

Zoë Migicovsky, Dalhousie University

Rootstock effects on shoot system phenotypes in a ‘Chambourcin’

experimental vineyard

Daniel Chitwood – Independent Researcher; Peter Cousins – E. & J. Gallo Winery; Anne

Fennell – South Dakota State University; Zachary Harris – Saint Louis University; Laura

Klein – Saint Louis University; Laszlo Kovacs – Missouri State University; Misha

Kwasniewski – University of Missouri; Mao Li – Donald Danforth Plant Science Center;

Jason Londo – USDA-ARS; Allison Miller – Saint Louis University

Grafted species such as grapevines are an ideal model for understanding how rootstocks

can impact shoot systems phenotypes. We examined an experimental vineyard in Mount

Vernon, Missouri which includes a common scion (‘Chambourcin’) own-rooted as well as

grafted onto three different rootstocks. The vineyard also includes 3 different irrigation

treatments. Leaf shape and ionomic data was collected in 2014 and 2016, while gene

expression (RNA-seq) was collected in 2016 only. We report the results of analyses

determining the impact that varying rootstocks can have on leaf shape, ionomic profile

and gene expression. Future work will expand sampling to include additional phenotypes

and time points across three years.

Organ Pipe Cactus - 40

Sabrina Elias, University of Dhaka and University of Nebraska Lincoln

Deciphering the association of phenome and gene expression postulating

salt tolerance mechanism in a rice landrace, Horkuch

Page 86: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Sabrina Elias – University of Dhaka, University of Nebraska Lincoln; Taslima Haque –

University of Dhaka, University of Texas at Austin; Samsad Razzaque – University of

Dhaka, University of Texas at Austin; Sazzadur Rahman – Bangladesh Rice Research

Institute; Sudip Biswas – University of Dhaka; Thomas Juenger – University of Texas at

Austin; Harkamal Walia – University of Nebraska Lincoln; Zeba Seraj – University of

Dhaka

Rice production in the salty cultivated soil cannot meet the demand of the overgrowing

population as rice plant and excess salt has a rival relationship. Changes in climate is

worsening the scenario by introgression of excess salt in more and more cultivable lands.

But rice, as a major staple food need to maintain the balance in the production requiring

the need of high yielding salt-tolerant rice. Understanding the mechanism of salt tolerant

landraces with adaptive capability to withstand the harsh environment can give insights

on potential candidate genes for conferring tolerance. In order to do so, the reciprocal

cross of a salt tolerant landrace Horkuch and high yielding but sensitive variety, IR29 has

been analyzed. A set of the reciprocal F2:3 population was genotyped using DArTSeq™

for discovering SNP markers to construct linkage map and manual phenotyped under salt

stress. We have identified expression QTLs (eQTL) combining the genotyping and

RNAseq data of a subset under 150mM salt stress. An image-based non-destructive

automated and continuous phenotyping over 3 weeks of salt stress was carried out on a

selected F3 and F5 sub-populations followed by QTL identification for the digital traits

and relative growth rates from visual image data. Instead of endpoint records, image

analysis over days gave us longitudinal data, which could separate the early and late

responses to salt stress. Combining the phenomics and eQTL data, early growth indices

were found to be enriched with transport, osmotic response etc and the later stages were

enriched with genes associated with growth, carbohydrate metabolism, organ

development etc. The phenome data along with the expression data could give a

comprehensive scenario regarding potential candidates involved in tolerance mechanism.

Organ Pipe Cactus - 41

Tara Enders, University of Minnesota

Computer vision and hyperspectral approaches to document temperature

stress responses in maize seedlings

Susan St Dennis – University of Minnesota; Nathan Miller – University of Wisconsin; Liz

Sampson – University of Minnesota; Edgar Spalding – University of Wisconsin; Nathan

Springer – University of Minnesota; Cory Hirsch – University of Minnesota

Page 87: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Yields of maize may be reduced substantially within the next century due to global climate

change. Understanding how maize varieties respond to temperature extremes will be

instrumental in developing varieties that can withstand future abiotic stresses while still

producing high yield. We are documenting the variation of morphological traits, color,

and hyperspectral signals in maize seedlings in response to abiotic stresses in multiple

maize genotypes over time. Morphological measurements, such as plant height, width,

and area, can help characterize the impact of stresses on growth rates. Color data from

RGB images allows for quantification of physiological changes to stress, such as leaf

necrosis, which varies substantially among maize genotypes. Hyperspectral data may

capture valuable information about how genotypes respond to stress that is unable to be

captured using RGB imaging and could provide early detection of stress responses prior

to other manifestations. Documenting multiple traits across genotypes and growth

conditions will uncover the dynamics of maize responses to changing temperatures and

allow for the discovery of genomic loci that could provide improved tolerance.

Organ Pipe Cactus - 42

Donghee Hoh, MSU-DOE Plant Research Laboratory

The genetic and mechanistic bases of photosynthetic cold tolerance in

legume, cowpea (vigna unguiculata (l.) walp.) via high throughput

environmental phenotyping

Isaac Osei-Bonsu – MSU-DOE Plant Research Laboratory; Jeffrey Cruz – MSU-DOE

Plant Research Laboratory; Philip Roberts – University of California, Riverside; Bao-Lam

Huynh – University of California, Riverside; Oliver Tessmer – MSU-DOE Plant Research

Laboratory; Timothy Close – University of California, Riverside; Linda Savage – MSU-

DOE Plant Research Laboratory; David Hall – MSU-DOE Plant Research Laboratory;

David Kramer – MSU-DOE Plant Research Laboratory

Increasing crop production will require improvements in the efficiency, robustness, and

sustainability of photosynthesis. Among the most critical abiotic stresses that impact

photosynthesis is temperature. Cowpea (Vigna unguiculata (L.) Walp.), an important

protein source worldwide especially in developing countries was chosen as a model crop,

is especially sensitive to high temperature during reproductive development result in

severe crop losses by causing male sterility and fruit abortion. One proposed approach to

minimizing the impact of heat stress is to plant earlier than the normal to avoid extreme

heat stress later in the season. This strategy involves planting at colder temperatures,

which is generally known to retard germination and emergence in Cowpea. A major

concern is how the chilling stress impacts plant performance/yield. The goal of this

research is to identify genes and mechanisms related to chilling stress tolerance. Cowpea

Page 88: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

has potentially sufficient genetic variation that we can use to identify quantitative trait

loci (QTL) that can be used to guide plant breeding, and test mechanistic models to

explain this sensitivity. Two sets of recombinant inbred lines (RILs) were phenotyped

using the Dynamic Environment Phenotype Imager (DEPI) and MultispeQ under control,

cold temperature conditions, under lighting conditions that simulate typical daylight. We

found considerable natural variation in the photosynthetic parameters which were used

to identify QTLs specific to cold tolerance that can be exploited for quantitative trait locus

(QTL) mapping and subsequent breeding efforts. Phenotyping and QTL data showed cold

sensitivity is strongly associated with increased photoprotective responses (qE and qI)

that is strongly linked to variations in the chloroplast ATP synthase activity, leading us to

propose a model for low-temperature photodamage that involves control of proton motive

force-induced damage to photosystem II. We also identified potential genetic control

element that could explain the adaptation of certain variants to low temperature.

Organ Pipe Cactus - 43

Salme Timmusk, Uppsla BioCenter, Swedish University of Agricultural Sciences; The

Bashan Institute of Science 1730 Post Oak Court, Auburn, AL 36830, USA

Mineral nanoparticles improve plant growth promoting rhizobacterial

performance

Gulaim Seisenbaeva – Uppsala BioCenter, SLU; Dept. of Molecular Sciences; Lawrence

Behers – Uppsala BioCenter, SLU; The Bashan Institute of Science 1730 Post Oak Court,

Auburn, AL 36830, USA

We work with plant growth promoting rhizobacteria (PGPR) i.e. with the native bacterial

species which commonly are present in agricultural soils. Our strains are isolated from

harsh habitats as we have learned that the isolates from the suboptimal conditions,

coevolved with host plant over long period of time, have considerably higher potential for

plant stress alleviation. We design synthetic root biofilms and investigate their mode of

action taking into account that rhizobacterial properties are dynamic and highly

dependent on root age.

We study a novel use of nanotitania (TNs) as agents in the nanointerface interaction

between plants and colonization of PGPR. The effectiveness of PGPRs is related to the

effectiveness of the technology used for their formulation. TNs produced by the Captigel

patented SolGel approach, characterized by the transmission and scanning electron

microscopy are used for formulation of the harsh environment PGPR strains. Changes in

the seedlings biomass, root architecture and in the density of single and double inoculants

with and without TNs are monitored during two weeks of stress induced by drought salt

Page 89: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

and by the pathogen Fusarium culmorum. We show that double inoculants with TNs form

stable biofilms on plant roots. Regression analysis indicates that there is a positive

interaction between seedling biomass and TN-treated second inoculant colonization. We

conclude that the nanoparticle treatment provides an effectual platform for PGPR

rational application via design of root microbial community. How rhizosphere varies

along soil–plant system and to what extent it affects acquisition of water and nutrients?

The model system established provides a basis for systems approach using microscale

information technology for microbiome metabolic reconstruction including identifying

genes contributing to variation in phenotypic plasticity.

These new advancements importantly contribute towards solving food security issues in

changing climates.

Organ Pipe Cactus - 44

Marian Brestic, Slovak University of Agriculture in Nitra, Slovakia

Phenotyping of isogenic chlorophyll-deficient wheat mutant lines in relation

to photoprotection and photosynthetic capacity

Marian Brestic – Slovak University of Agriculture in Nitra, Slovakia; Marek Zivcak –

Slovak University of Agriculture in Nitra, Slovakia; Oksana Sytar – Slovak University of

Agriculture in Nitra, Slovakia; Klaudia Bruckova – Slovak University of Agriculture in

Nitra, Slovakia; Xinghong Yang – College of Life Sciences, Shandong Agricultural

University, Taian, China; Xiangnan Li – Northeast Institute of Geography and

Agroecology, Chinese Academy of Sciences, Changchun, China

Chlorophyll-deficient mutants are characterized by unique changes in content and

composition of light harvesting pigment protein complexes in chloroplast. In our

experiments, we examined light responses and photosynthetic capacity of chlorophyll-

less isogenic mutant lines of hexaploid bread wheat (Triticum aestivum L.) and tetraploid

durum wheat (Triticum durum L.) in comparison to parental lines (WT) in different

growth phases and different environmental conditions. Despite the strong chlorina

phenotype of young plants, the tested mutant lines expressed an ability to adapt to natural

environmental conditions, to grow and provide satisfactory grain yield in field conditions.

A detailed in vivo analysis of photosynthetic parameters enabled to characterize the

photosynthetic phenotype of mutant lines. We observed a higher expression of gene

mutations related to a typical chlorina phenotype in tetraploid durum wheat mutants

compared to the hexaploid accessions. In later growth phases, we observed partial relief

of chlorina phenotype, including photosynthetic pigment composition, CO2 assimilation

rate, plant growth and responses of PSII photochemistry. The shift of the phenotype

Page 90: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

towards the wild-type in later growth phases was more evident in bread than in durum

wheat, as well as in plants grown in growth chamber compared to plants grown outdoors.

Low chlorophyll content in leaves of mutant lines limited CO2 assimilation only in early

growth phases; in general, the photosynthetic rate per chlorophyll unit was relatively high

in all mutant lines. In the majority of mutants, we observed a limited photoprotective

capacity. Concluding all, our results show that the chlorophyll-less mutant lines of wheat

represent specific biological models with a diverse leaf traits and photosynthetic

responses, including differences in acclimation capacity and a strong interaction with

crop phenology. In this respect, the isogenic mutant lines represent valuable plant models

for different photosynthetic studies and the crop phenotyping programs. (Supported by

APVV-15-0721, VEGA-1-0831-17 and APVV-SK-CN-2015-0005).

Organ Pipe Cactus - 45

Gokhan Hacisalihoglu, PhD Florida A&M University

Seed Priming Modulates Cold Sensitivity in Maize NAM Parental Inbred

Lines

Gokhan HACISALIHOGLU – Florida Agricultural and Mechanical University,

Tallahassee, FL; J. Gustin – Univ. of Florida; Nathan Miller – University of Wisconsin; S.

Kantanka – Florida A&M University; A.M. Settles – Univ. of Florida

Seedling emergence is an important factor for yield, particularly under challenging

planting conditions. In the US corn belt, maize is planted in early spring, as soon as soil

temperatures are permissive to germination. At that time, temperatures often drop below

normal, which can delay or even kill the seedling. Seed pre-treatments have been shown

to improve germination in cold conditions in crops such as rice and cabbage, but are

largely unpublished in maize. To assess the effects of pre-treatments on early maize cold

tolerance, twenty-seven inbred parents of maize Nested Association Mapping (NAM)

population were primed using a synthetic solid matrix and then tested for cold tolerance

using a soil-based emergence assay. Primed kernels were incubated at 10°C for 5 days,

and then transferred to 24°C for emergence. DSLR cameras were used to capture images

every 30 min to obtain emergence profiles of each seedling. Emergence time was

determined from the time-lapsed images and multiple measures including final

emergence percentage, time to 50% emergence, and emergence rate were extracted for

each genotype. The cold treatment reduced total emergence of several genotypes.

However, priming pre-treatment protected the sensitive genotypes allowing nearly full

emergence. We also used single-kernel near infrared reflectance spectroscopy to

determine seed density, weight, oil, protein, and starch for the kernels prior to planting.

By combining kernel characteristics and emergence time, we found small, but highly

significant correlations between the kernel and early seedling performance.

Page 91: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Organ Pipe Cactus - 46

Trevis Huggins, PhD USDA Agricultural Research Service Dale Bumpers National Rice

Research Center

Association analysis for loci regulating grain quality traits and marker

development in the USDA rice collection

Jeremy Edwards, Scientist – USDA Agricultural Research Service; Ming-Hsuan Chen,

Scientist – USDA Agriculture Research Services; Aaron Jackson – USDA Agriculture

Research Services; Anna McClung, Plant Breeder – USDA Agriculture Research Services

Uncovering underlying genetics associated with grain quality is important to world food

security. Rice is consumed as a whole grain, therefore cooked rice texture, stickiness,

chewiness, grain dimensions and grain appearance can affect palatability and

marketability. Amylose and protein content play significant roles in determining eating

and cooking quality and affect the translucency of milled kernels. Kernel translucency is

influenced by the presence of chalk, an opaque area in the grain, which occurs when starch

granules are loosely packed in the endosperm. The minicore (MC) panel is a

representative germplasm subset of the USDA Rice Core Collection, specifically designed

to capture maximum diversity in a manageable size and is ideally suited for genome-wide

association (GWA) experiments that have a high phenotyping cost. The publically

available re-sequencing dataset of the 203 MC accessions by next generation sequencing

(NGS) produced ~3.3 million SNPs. The SNPs were used to conduct GWA analysis on the

grain traits, apparent amylose content (AAC), alkali spreading value (ASV), percent grain

chalk (Chk) and percent grain protein (Prot). Major known starch related genes, such as

soluble starch synthase IIa (SSIIa) and Waxy, were identified, as well as 11 novel grain

quality loci, seven novel chalk loci and seven novel protein loci. Further analysis of regions

surrounding significant SNPs with Perl scripts identified several overlapping

chromosomal regions associated with multiple traits. These results will be instrumental

in determining molecular markers useful in marker assisted selection for grain quality

and may provide insights into the biological processes that influence it.

Organ Pipe Cactus - 48

Arno Meiring, MA Yokohama National University

Genetic analysis of multiple elements including iodine and cesium in brown

rice grown in field and controlled conditions

Tatsuo Nakamura, PhD – Yokohama National University

Page 92: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Rice is a staple food for large parts of Asia and Africa, and serves as a model organism for

cereals. Information regarding the accumulation of elements in rice grains is essential to

expanding the understanding of plant physiology and genetics, and is a powerful tool for

developing healthier crops. As such, this study investigates element accumulation in

multiple populations across field and controlled environments. Multi-element analysis

was conducted on brown rice harvested from 69 accessions of the World Rice Core

Collection (WRC) and 50 accessions from the Rice Core Collection of Japanese Landraces

(JRC) maintained by the NIAS Genebank, as well as on 98 backcrossed inbred lines (BIL)

and 54 chromosome segment substitution lines (CSSL) derived from Nipponbare ×

Kasalath, obtained from the Japanese Rice Genome Resource Center. All populations

were cultivated both in the field using standard agronomic practices for the region, as well

as in controlled indoor conditions. Essential nutrients and other elements of interest were

supplied using a modified version of the Kimura-B growth medium. Elemental analysis

was performed using inductively coupled plasma mass spectrometry (ICP-MS) to analyze

grain content for 20 elements, including magnesium, phosphor, potassium, manganese,

iron, zinc, copper, arsenic, cadmium, iodine and cesium. Association and linkage

mapping was performed in the form of a genome-wide association study (GWAS) and

quantitative trait loci (QTL) analysis respectively, revealing potential targets for marker

assisted selection.

Organ Pipe Cactus - 49

Karina Morales, Texas A&M University

Characterizing a rice diversity panel with a 7K SNP chip and flowering time

evaluation

Stephon Warren – Texas A&M University; John Carlos Ignacio – International Rice

Research Institute; Yuxin Shi – Cornell University; Rodante Tabien – Texas A&M

AgriLife Research; Tobias Kretzschmar – International Rice Research Institute; Susan

McCouch – Cornell University; Michael Thomson – Texas A&M University

Rice (Oryza sativa L.) is an essential food crop with demands for increased yield as it

provides the daily caloric intake of over 50% of the world’s growing population. Flowering

is one of the most sensitive stages of rice growth and is highly variable among varieties

and across environments. In Texas, farmers often desire early flowering varieties as these

can avoid peak temperatures of the summer months and give sufficient time for the ratoon

crop to mature before the cold temperatures of winter begin. This experiment took place

at the Texas A&M AgriLife Research Center in Beaumont, TX where 208 rice varieties of

diverse origins were planted in spring 2017 and were grown through the summer of 2017.

Beginning approximately 50 days after planting, notes were collected once a week on

Page 93: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

flowering percentage to estimate days to 50% flowering. Each variety was genotyped using

the Illumina 7K rice SNP chip developed at Cornell University. This project aims to

identify genetic loci which contribute to extremely early and late flowering time. Upon

identifying these loci, we will use the CRISPR/Cas9 genome editing system to validate

candidate genes in diverse genetic backgrounds to gain a better understanding of how

each locus may contribute to days to heading in rice. Ultimately, the improved knowledge

on manipulating flowering time genes will lead to more precise tools to provide early

flowering in any genetic background for each target environment.

Organ Pipe Cactus - 49

Marjorie Lundgren, PhD Massachusetts Institute of Technology

C4 anatomy can evolve via a single developmental change

While the vast majority of flowering plants use C3 photosynthesis, some lineages evolved

the C4 pathway to overcome environmental limitations on carbon fixation. In most C4

plants, the carbon assimilation and reduction steps of photosynthesis differentially occur

in leaf mesophyll and bundle sheath tissue types, respectively, while C3 plants complete

both steps simultaneously within the mesophyll. Thus, leaf anatomy is tightly linked to

photosynthetic pathway, with C3 plants requiring large mesophyll volumes for

photosynthesis, and efficient C4 leaves requiring large bundle sheath volumes to

accommodate the necessary photosynthetic organelles, but little mesophyll. Because C4

photosynthesis requires specific leaf anatomy, its evolution is assumed to involve

important modifications to several anatomical traits. Indeed, C4 plants typically achieve

enlarged bundle sheath and reduced mesophyll areas compared to C3 plants, but the

underlying modifications have only been assessed through species comparisons, which

likely capture numerous changes besides those necessary for C4 functionality. The grass

Alloteropsis semialata provides a unique intraspecific continuum of closely related C3,

C3-C4 intermediate, and C4 states, allowing me to determine the minimum anatomical

changes that accompanied the transition between non-C4 and C4 phenotypes, and

distinguishing the anatomical changes that occurred only after C4 emergence. Here, I

show that only an increase in vein density, driven specifically by minor vein development,

distinguishes C4 from non-C4 plants. Furthermore, using a rare C3 x C4 F1 hybrid of A.

semialata, I show that this important minor vein phenotype is genetically determined and

can be inherited via C3 x C4 hybridization. Exploiting this intraspecific diversity, my

results show that a single developmental change is sufficient to produce functional C4 leaf

anatomy, partially explaining the recurrent origins of this complex trait and providing key

information needed for engineering the efficient C4 phenotype into C3 crops < ./p>

Page 94: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Organ Pipe Cactus - 50

Marek Zivcak, Slovak University of Agriculture in Nitra

Use of hyperspectral data to assess leaf traits of diverse wheat genetic

resources in the field

Marek Zivcak – Slovak University of Agriculture in Nitra, Slovakia; Marian Brestic –

Slovak University of Agriculture in Nitra, Slovakia; Lenka Botyanszka – Slovak University

of Agriculture in Nitra, Slovakia; Pavol Hauptvogel – National Agricultural and Food

Centre in Piestany, Slovakia

Hyperspectral analysis has been introduced as an alternative technology to characterize

the different properties of crop canopies, including applications in field phenotyping of

genetic resources. However, the open question is the reliability of hyperspectral indices

in the estimation of leaf properties when applied in a broad spectrum of genotypes

differing in plant and leaf morphology, anatomy and chemical composition of leaves. To

examine this issue, we tested the set of wheat genebank accessions with a broad

phenotypical diversity using the hyperspectral field records as well as the subsequent

analyses of phenotypic and physiological traits, such as leaf area, leaf thickness (measured

as leaf mass per area unit, LMA), leaf nitrogen content, chlorophyll and carotenoid

content, chlorophyll a to b ratio, chlorophyll to carotenoid ratio, SPAD value, etc. We

found relatively high diversity in all observed traits (thick vs. thin leaves, high vs. low

chlorophyll concentration; very small vs. very large leaves), providing good background

for correlation analyses between the hyperspectral parameters and related phenotypic

traits. We found that the parameters proposed in literature for estimation of some traits

are not useful to be used for germplasm with a large or unknown phenotypic variability.

Anyway, we found a few parameters correlating well across the entire collection of wheat

genotypes, which can be regarded as more reliable and universal, useful for the use in

phenotyping in genebank wheat collections or wheat breeding. Our results can be useful

as a background for the next activities in phenotyping for biomass improvements,

nutrient use efficiency or abiotic stress tolerance in wheat and other crops. The study was

supported by the national grants APVV-15-0721 and VEGA-1-0831-17.

Organ Pipe Cactus - 51

Kevin Falk, MSc Iowa State University

Studies of Root System Architecture in Soybean using Computer Vision

Kevin Falk, MSc – Iowa State University

Page 95: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Root system architecture (RSA) studies are tedious, susceptible to introduced variation,

measurements are time consuming and the extracted features may not translate to

meaningful outcome, i.e., increase in yield or other important traits. With the advent of

computer vision, there is a renewed interest in uncovering “the hidden half”, i.e.,

discovering trait correlations within and between genotypes and phenotypes. This study

included 300 diverse soybean accessions from a wide geographical distribution and

deployed 2-D (in controlled conditions) and stereo imaging platforms (field tests),

processing and data analytic tools to deep phenotype for important RSA traits using in-

house imaging software, ARIA. Both 2-D and stereo imaging platforms reveal tremendous

genetic variability for RSA traits for root shape, length, mass, and angle. The stereo

imaging platform developed for this study makes it possible to phenotype hundreds of

genotypes and extract numerous root system traits. In addition, the 2-D platform

developed is non-destructive, adding observations of seedling root growth and

development.

Arizona Rainbow Cactus - 53

Rishi Masalia, University of Georgia

Phenotypic and transcriptomic responses to overlap in water-related

limitation stresses in cultivated sunflower seedlings

Liana Mosley, Undergraduate – University of Georgia; John Burke, Dr. – University of

Georgia

Water limitation is considered by many to be one the most detrimental effects unto crop

yields, and can occur through a variety of abiotic stresses influencing plant water

potential. As such, understanding how crops respond to water limitation is becoming

increasingly important. One approach is to understand how plants respond to variety of

water limitation stresses and identify commonalities in response. Here, we investigate the

extent to which a single genotype of cultivated sunflower exhibits a shared phenotypic

and transcriptomic response across three water limitation stresses: a repeated soil dry

down, and two osmotic challenges, salinity (100mM NaCl) and PEG-6000 (8.25% by

volume). Phenotypically stressed plants had a decrease in total biomass with a shift

towards root allocation and increased water use efficiency consistent with previous

drought literature. Transcriptionally, we identified 1,332 unique differentially expressed

genes (DEGs) across leaf and root tissue 10 days post-treatment, with a majority of DEGs

unique to an individual stress. Across all stresses, 51 genes were shared and expressed in

the same direction relative to control. Of the treatments surveyed, PEG-6000 and salt

cluster both phenotypically and transcriptionally away from dry down, except in the

subset of shared root DEGs where stressed individuals experienced a distinct treatment

response. Moreover, of the shared 51 DEGs 2 root DEGs colocalize with previous genome-

Page 96: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

wide association results of sunflower seedlings exposed to PEG-6000 stress at the same

concentration (8.25% by volume). This shared response suggests that efforts aimed at

producing plants that are more resilient to a particular water limitation stress may convey

benefits in other stresses.

Organ Pipe Cactus - 52

Max Feldman, Donald Danforth Plant Science Center

The trait components that constitute whole plant water use efficiency are

defined by unique, environmentally responsive genetic signatures in the

model C4 grass Setaria

Max Feldman – Donald Danforth Plant Science Center; Patrick Ellsworth – Washington

State University; Asaph Cousins – Washington State University; Ivan Baxter – USDA-

ARS

The complex relationship between plant growth and water use is largely determined by

genetic factors that influence both the morphological and biochemical characteristics of

plants. Improving the efficiency by which plants utilize water is an important breeding

objective that can be translated to improve productivity in agriculture while

simultaneously making it a more sustainable endeavor. To assess the genetic basis of

water use efficiency and trait plasticity, we have utilized high-throughput phenotyping

platform and mass spectrometry to quantify plant size, evapotranspiration and stable

isotope composition of an interspecific Setaria italica x Setaria viridis recombinant inbred

line population in both a well-watered and water-limited environment. Our findings

indicate that measurements of plant size and water use in this system are strongly

correlated. We used a linear modeling approach to partition the traits into the predicted

values of plant size given water use and deviations from this relationship at the genotype

level. The resulting traits describing plant size, water use, water use efficiency and

&delta;13C are all substantially heritable, responsive to soil water potential differentials

and provide a framework to understand the components of plant water use efficiency.

Biparental linkage mapping successfully identified several pleiotropic loci that exhibit

medium-to-large effects on most traits in addition to many smaller effect loci associated

with fewer traits or specific to well-watered or water-limited environments. This study is

the first report characterizing the genetic architecture of water use efficiency in the model

C4 species Setaria and mechanistically links measurement of water use efficiency with

&delta;13C through several common large effect QTL.

Page 97: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Arizona Rainbow Cactus - 54

Chenyong Miao, PhD - University of Nebraska-Lincoln

Analysis of sorghum time-series phenotype data using nonparametric

regression and machine learning

Chenyong Miao – University of Nebraska-Lincoln; Piyush Pandey – University of

Nebraska-Lincoln; Zhikai Liang – University of Nebraksa-Lincoln; Daniel Carvalho –

University of Nebraksa-Lincoln; Xiaoyang Ye – University of Nebraska-Lincoln; Vincent

Stoerger – University of Nebraska-Lincoln; Yuhang Xu – University of Nebraska-Lincoln;

Yufeng Ge – University of Nebraska-Lincoln; James Schnable – University of Nebraska–

Lincoln

With the rapid development of high throughput phenotyping technology, time-series

phenotype data can be easily obtained in many crop species such as sorghum and maize.

Time-series phenotype data provides more information than traditional manually

measurements in the field. However, the analysis of the time-series phenotype data and

comparisons across different plant accessions also presents some challenges which are

not issues for single time point phenotypic datasets. For example, variation in flowering

time can have pleiotropic effects on many other plant traits, which can confound efforts

to characterize the distinct genetic architectures controlling non-flowering time

phenotypes in the growing season. Here we present data of 347 sorghum accessions from

the sorghum association panel (SAP). Plants were imaged using RGB and hyperspectral

cameras every two days at the UNL Greenhouse innovation center from July 7th to August

31st, which spanned flowering stages for the vast majority of these accessions. This

dataset was used to explore a number of approaches to reduce the confounding effects of

flowering time variation on quantitative genetic analyses. The first approach employed

uses manual scoring of the flowering dates to sort and compare phenotypic measurements

relative to the date of flowering rather than the date of planting. The second approach

employed builds on incorporating nonparametric curve fitting method to increase

measurement accuracy and quantify “rate of change” phenotypes. Finally, the 20,000

sorghum images scored for flowering were used to train a machine learning classifier to

distinguish images of flowering and vegetative stage of sorghum plants, which greatly

decreasing labor costs of using flowering time as the indexed traits in high throughput

phenotyping contexts.

Arizona Rainbow Cactus - 55

Fabiana Moreira, Msc - Purdue University

Improving Efficiency of Soybean Breeding with Phenomic-enable Canopy

Selection

Page 98: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Fabiana Moreira – Purdue University; Anthony Hearst – Purdue University; Keith

Cherkauer – Purdue University; Katy Martin Rainey – Purdue University

The Soybean Breeding program at Purdue University determined that average canopy

coverage (ACC) as measured by UAS (unmanned aerial systems) is highly heritable, with

a high genetic correlation with yield. In this study, we compare selection for ACC to

selection for yield in the early stages of a soybean breeding pipeline. In 2015, we collected

UAS imagery 8 times around 13 to 56 days after planting in progeny rows to determine

ACC for each plot. The best soybean lines were then selected with three parameters: Yield,

ACC and Yield given ACC (Yield|ACC), and the lines were assessed at two locations in

2016. We found that performance of lines in the selection categories to be statistically

equivalent. Additionally, ACC selected 4 of the top 10 ranked lines, and selected generally

earlier-maturing lines. We repeated the progeny row selection in the next cycle in 2016,

with 7 surveys around 13 to 56 days after planting, but in 2017 the yield selection

categories out-performed ACC selection, perhaps because canopy growth was unusually

high in 2016. Rankings were also less favorable, out of the top 10 ranked lines, ACC

selected 1. Even though, ACC did not result in the best performance lines in the second

year of selections, the cost and time involved in harvesting thousands of lines usually leads

breeders to perform visual selection and our results indicate that ACC has a role in

efficient selection of high-yielding soybean lines.

Arizona Rainbow Cactus - 56

James Ta, University of California, Davis

Camera phenotyping and path analysis to reveal indirect effects of genetic

architecture in the shade avoidance syndrome

James Ta – University of California, Davis; Daniel Runcie – University of California,

Davis

Quantitative trait loci (QTL) mapping is an important tool for understanding the

genotype-phenotype relationship and explaining genotype-by-environment interactions.

QTL mapping studies have revealed disease and yield-related QTLs that have been pivotal

in increasing agricultural production. However, many of these studies rely on single-time

point phenotypes to characterize developmental stages. Because these traits do not

capture development throughout time, single-time point analyses might not capture

transient QTLs or QTL effects that change in smaller intervals of time. Consequently,

using high-throughput phenotyping (HTP) – in conjunction with QTL mapping – can

uncover novel QTLs because these systems can fully capture multiple stages of

development. This provides more data to detect QTLs with higher statistical power, and

Page 99: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

greater potential for further modeling. For this study, we use a camera phenotyping

system, along with manually-measured traits, to investigate the genetic architecture of

the shade avoidance syndrome (SAS) in A. thaliana. The SAS is a collection of responses

that plants display when shaded by other plants. The genetic control of the SAS at the

hypocotyl stage is well-known; however, the genetic architecture underlying the SAS

beyond the hypocotyl is less understood. Because shading influences plant growth over

time, we can use HTP and modeling to reveal novel SAS QTLs throughput development

and better understand the SAS in an integrated way. We identify several development-

specific SAS QTLs, which potentially involve new SAS genes or gene modules. Using path

analysis, we show that some of the effects of these QTL on later-development traits can

be explained as indirect effects arising from direct effects earlier in development. Our

work provides greater insight into how multiple elements – genetics, environment, and

the relationships between phenotypes – integrate to influence the genotype-phenotype

map.

Sensors and Systems

Arizona Rainbow Cactus - 57

Farzad Hosseinali, Texas A&M University, Biological & Agricultural Engineering

Department

Quantifying Nanoscale Biomechanical Properties of the Plant Cuticular

Waxes

Farzad Hosseinali – Texas A&M University, Biological & Agricultural Engineering

Department; Alex Thomasson – Texas A&M University, Biological & Agricultural

Engineering Department; James Batteas – Texas A&M University, Chemistry

Department; Eric Hequet – Texas Tech University, Plant and Soil Science Department;

Jane Dever – Soil and Crop Science Department; Edward Barnes – Cotton Incorporated

The potential applications of Atomic Force Microscope (AFM) in quantifying the

biomechanical properties of plants tissue and membranes, such as the cuticle of tomato

fruits, have been introduced before. However, previous studies on the application of the

AFM in the surface characterization of cotton fiber were mainly focused on the AFM

capabilities in producing high-resolution topography images of either fiber surface or its

cross–section. In fact, cotton fiber cells are covered with a thin cuticular membrane. The

cuticle is mostly made of lipids, alcohols, and fatty acids (collectively called ‘cotton wax’).

The waxy layer can be 10 to 300 nm thick and imparts hydrophobicity to the fiber surface.

The main objective of this study was to characterize and compare the surface

nanomechanical properties of cotton fibers using various modes of the AFM. Surface

topography and friction images of the fibers were obtained with conventional contact

mode. The nanomechanical property images, such as adhesion and deformation, were

Page 100: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

obtained with Bruker’s newly developed high-speed force-volume technique, PeakForce

QNM®. The differences in nanoscale friction, adhesion, and deformation signals can be

attributed to fiber surface hydrophobicity and stiffness, which in turn depend on fatty

acids’ hydrocarbon chain length, film viscosity, and the waxy layer thickness.

Arizona Rainbow Cactus - 58

William Salter, School of Life and Environmental Sciences, Sydney Institute of

Agriculture, The University of Sydney

Tackling the physiological phenotyping bottleneck with low-cost, enhanced-

throughput, do-it-yourself gas exchange and ceptometry

William Salter – The University of Sydney; Matthew Gilbert – University of California,

Davis; Andrew Merchant – The University of Sydney; Thomas Buckley – University of

California, Davis

High throughput phenotyping platforms (HTPPs) are increasingly adopted in plant

breeding research due to developments in sensor technology, unmanned aeronautics and

computing infrastructure. Most of these platforms rely on indirect measurement

techniques therefore some physiological traits may be inaccurately estimated whilst

others cannot be estimated at all. Unfortunately, existing methods of directly measuring

plant physiological traits, such as photosynthetic capacity (Amax), have low throughput

and can be prohibitively expensive, creating a bottleneck in the breeding pipeline. We

have addressed this issue by developing new low-cost enhanced-throughput phenotyping

tools to directly measure physiological traits of wheat (Triticum aestivum). Our eight-

chamber multiplexed gas exchange system, OCTOflux, can directly measure Amax with

5-10 times the throughput of conventional instruments, whilst our handmade

ceptometers, PARbars, allow us to monitor the canopy light environment of many plots

simultaneously and continuously across a diurnal cycle. By custom-building and

optimizing these systems for throughput we have kept costs to a minimum, with

OCTOflux costing roughly half that of commercially available single-chamber gas

exchange systems and PARbars costing approximately 95% less than commercial

ceptometers. We recently used these tools to identify variation in the distribution of Amax

relative to light availability in 160 diverse wheat genotypes grown in the field. In a two-

week measurement campaign we measured Amax in over 1300 leaves with OCTOflux and

phenotyped the diurnal light environment of 418 plots using 68 PARbars. These tools

could be readily modified for use with any plant functional type and also be useful in

validating emerging HTPPs that rely on remotely sensed data to estimate photosynthetic

parameters.

Page 101: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Arizona Rainbow Cactus - 59

Oliver Scholz, Fraunhofer Development Center X-Ray Technology

Phenotyping for Plant Breeding using 3D Sensors and a Generic 3D Leaf

Model

Franz Uhrmann – Fraunhofer Development Center X-Ray Technology; Katharina Pieger

– Fraunhofer Development Center X-Ray Technology; Dominik Penk – University of

Erlangen-Nuernberg; Guenther Greiner, Prof. – University of Erlangen-Nuernberg

We present a setup to objectively assess sugar beet plant traits using multiview

stereoscopy with color cameras with the goal of generating relevant phenotyping

parameters to aid the breeder. The setup was tested in a greenhouse environment as well

as in the field. The assessment is performed using a generic leaf model adapted to the

specific requirements of the sugar beet breeder. The evaluation yields global plant

parameter including plant height, leaf area, leaf count, etc. as well as per-leaf parameters

for each leaf of the plant. The leaf model is based on semantic geometric parameters,

which can directly be interpreted by the breeder.

Arizona Rainbow Cactus - 60

Cory Hirsch, Dr. University of Minnesota

Machine vision phenotyping platform for seedling growth and morphology

Tara Enders – University of Minnesota; Nathan Miller – University of Wisconsin-

Madison; Elizabeth Sampson – University of Minnesota; Sara Tirado – University of

Minnesota; Nathan Springer – University of Minnesota; Edgar Spalding – University of

Wisconsin-Madison

The ability to link genotypes and phenotypes can be used to improve plant productivity

and our understanding of plant biology. Our ability to obtain genomic information

efficiently and accurately has advanced greatly, while phenotyping methods have largely

remained laborious, subjective, and/or expensive. Towards alleviating these barriers, we

have developed a user friendly and affordable platform to acquire highly standardized

RGB images, while relying on minimal equipment and space in a laboratory setting.

Currently, we have developed algorithms to extract numerous growth and morphological

traits including plant height, width, stem diameter, pixel area, and center of mass. The

traits our algorithm extract correlate well with both traditional hand measurements and

measurements using manual image analysis techniques. We are leveraging available

storage, application deployment, and compute resources through the infrastructure at

CyVerse to allow accessibility to almost any researcher. This platform is already being

Page 102: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

used by multiple research groups at multiple Institutions and has been optimized for daily

collection of images of multiple plants to easily look at plant development. We have used

this method to monitor growth rate variation among different maize genotypes subjected

to temperature stress and also to measure variation in heterosis for seedling growth in a

panel of inbred and hybrid genotypes.

Arizona Rainbow Cactus - 61

Mitchell Feldmann, UC Davis Strawberry Research Program

Quantitative Methods for Studying Fruit Morphology in Strawberry

Yash Bhartia – UC Davis; Scott Newell – UC Davis; Julia Harshman – UC Davis; Steven

Knapp – UC Davis Strawberry Research Program

Several phenotypic characteristics, including, shape, and external color, are determinants

of both grower- and consumer-centric fruit quality in strawberry (Fragaria × ananassa).

Strong artificial selection for superior shelf-life, increased yield of marketable fruit, and

other commercial production traits has significantly changed fruit morphology and

quality attributes to produce high yielding cultivars with large, ultra-firm fruit. The

genotype-to-phenotype networks underlying these changes have not been investigated in

depth, and genes targeted by selection have not yet been identified in strawberry.

Moreover, it remains unclear what level of phenotypic complexity is necessary and

sufficient to support genomic-based inquiries and discoveries, expand what is known

about modern germplasm, and enhance breeding practices in strawberry. Our current

work explores approaches for and challenges associated with quantifying fruit size, shape,

and color from one-dimensional (e.g. height and width), two-dimensional (e.g. area and

cross-section shape), and three-dimensional (e.g. volume and surface topology)

perspectives. We demonstrate methods for collecting, measuring, and classifying digital

images for 1-D and 2-D analyses and suggest a method for reconstructing fruit models

from multiple images for 3-D analyses. Multivariate and spatial statistics will be used to

determine parameters paramount in identifying and quantifying fruit defects,

differentiating between marketable and non-marketable fruits, and understanding fruit

phenotypes critical for markets that require long shelf-life and sustained fruit quality, as

impacted by harvesting, handling, and shipping.

Page 103: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Arizona Rainbow Cactus - 62

Mahendra Bhandari, Texas A&M University

Assessing Wheat Foliar Disease Severity Using Ground- and Aerial-based

Remote Sensing Systems

Amir Ibrahim – Texas A&M University; Qingwu Xue – Texas A&M University; Haly Neely

– Texas A&M University; Nithya Rajan – Texas A&M University; Jinha Jung – Texas

A&M AgriLife Research; Murilo Maeda – Texas A&M AgriLife Research; Juan Landivar

– Texas A&M AgriLife Research; Bryan Simoneaux – Texas A&M University; Geraldine

Opena – Texas A&M University; Anil Adhikari – Texas A&M University

Remote sensing has been widely used as an indirect approach to study the agronomic and

physiological traits of plants. Plant diseases cause significant yield reductions in wheat

(Triticum aestivum L.). Remote detection and assessment of plant diseases are important

to improve disease phenotyping in breeding programs. This study investigates the

potential use of low-cost Unmanned Aerial System (UAS), equipped with RGB and

multispectral sensors, to quantify leaf rust severity caused by the fungus Puccinia

Triticina in wheat. In addition, Green seeker was used to taking Normalized Difference

Vegetation Index (NDVI) measurements. RGB images were acquired using rotary wing

(DJI Phantom 4 pro) from the field of different wheat genotypes grown at Castroville, TX.

Different vegetation indices (VIs) and image classification approaches were applied to

separate the diseased and healthy canopies. The obtained image dataset was further

processed to generate plot level data. The relationship between VIs and visual field data

on disease severity was examined to find a suitable vegetation index that can evaluate the

leaf rust severity in wheat. Disease severity was highly correlated to excess green index

(r= -0.86), NDVI (r= -0.91, P<0.01). The results show that UAS imaging and automated

data extraction can help to obtain high throughput phenotyping data on disease severity

with higher precision. This tool has a great potential to enable rapid assessment of the

large breeding nurseries by providing high-resolution measurements from small plots

and observation rows.

Arizona Rainbow Cactus - 63

Subodh Bhandari, Cal Poly Pomona

Measurement and Validation of Plant Water and Nitrogen Stresses using

UAV-based Remote Sensing and Machine-Learning Techniques

Amar Raheja, PhD – Cal Poly Pomona; Mohammad Chaichi, PhD – Cal Poly Pomona;

Robert Green, Lecturer – Cal Poly Pomona; Dat Do, Graduate Student – Cal Poly

Page 104: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Pomona; Mehdi Ansari, Graduate Student – Cal Poly Pomona; Frank Pham, Graduate

Student – Cal Poly Pomona; Joseph Wolf, Undergraduate Student – Cal Poly Pomona;

Tristan Sherman, Undergraduate Student – Cal Poly Pomona; Kevin Gonzalez, Graduate

Student – Cal Poly Pomona; Antonio Espinas – Cal Poly Pomona

The presentation talks about the Unmanned Aerial Vehicle (UAV)-based remote sensing

and machine learning techniques to measure water and nitrogen stresses. The main

advantage of UAV-based remote sensing is the immediate availability of high resolution

data. Near infrared (NIR) images obtained using remote sensing techniques help

determine the crop performances and stresses of a large area in a short amount of time

for precision agriculture, which aims to optimize the amount of water, fertilizers, and

pesticides using site-specific management of crops. However, for widespread usages of

these techniques on a routine basis by the end users, the accuracy of remote sensing data

must be validated using the proven ground-based methods. Equally important is the

reduction in the overall cost associated with these techniques. UAVs equipped with

multispectral sensors and digital cameras are flown over lettuce and citrus plots at Cal

Poly Pomona’s Spadra farm. Different rows of lettuce plot is subject to different level of

water and nitrogen treatments. The soil moisture and nitrogen levels were determined

prior to beginning the study. The multispectral images are used in the determination of

normalized differential vegetation index (NDVI) that provides information on the health

of the plant. Machine learning classifiers are developed using the Red-Green-Blue (RGB)

images. Handheld Spectroradiometer, Water Potential Meter, and Chlorophyll Meter are

used for ground-truthing. Correlation between NDVI, chlorophyll content, and water

potential will be shown. The developed machine learning algorithm is able to predict the

plant health reasonably well. Machine learning techniques with sufficient validation have

potential to provide significantly cheaper solutions to plant health assessment using just

the digital cameras.

Arizona Rainbow Cactus - 64

Menglu Wang, Bsc Department of Plant Sciences, University of Saskatchewan

Can Satellite Imagery be Used in Phenotyping?

Menglu Wang – University of Saskatchewan; Hema Duddu – University of Saskatchewan;

Steve Shirtliffe – University of Saskatchewan; Ti Zhang – University of Saskatchewan;

Sally Vail – Agriculture Agri-Food Canada

The application of small unmanned air vehicle (UAV) in phenotyping has become

common because it can capture high resolution imagery of crops in field plots. However

the infrastructure, complexity, and regulatory requirements are impediments to

adoption. High resolution satellite imagery is available as an alternative imagery source,

Page 105: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

however the utility of this technology to phenotype crops at the small plot scale has not

been tested. The objective of this study is to compare extracted dataset from satellite and

drone to determine if satellite imagery can be used for field phenotyping of small plots.

An oat trial with different nitrogen treatments (0, 50, 100, and 150 kg ha-1) was utilized

for this study. Geoeye-1 satellite multispectral imagery (Blue, Green, Red, and Near

Infrared) with a spatial resolution of 50cm and UAV imagery captured with Micasense

multispectral Rededge camera on the same date with a ground resolution of 2cm with

similar spectral bands. Vegetative indices, including normalized difference vegetation

index (NDVI) and optimized soil adjusted vegetation index (OSAVI) were calculated for

individual and groups of field plots. Vegetation indices from satellite and UAV imagery

for grouped plots and within the groups were strongly correlated. Results from other trials

with different traits will be presented in poster. Therefore, satellite imagery has potential

for phenotyping at the small plot scale for some traits. However, atmospheric effects and

the temporal resolution will still remain issues for satellite imagery.

Arizona Rainbow Cactus - 65

Caitlin Moore, University of Illinois

Linking solar induced fluorescence with genetic variability in productivity of

biomass sorghum

Katherine Meacham – University of Illinois; Guofang Miao – University of Illinois; Taylor

Pederson – University of Illinois; Evan Dracup – University of Illinois; Xi Yang –

University of Virginia; Kaiyu Guan – University of Illinois; Carl Bernacchi – University of

Illinois

The measurement of solar induced fluorescence (SIF) of chlorophyll has emerged as a

useful tool for monitoring plant photosynthesis. Application of SIF at the regional scale

from satellite remote sensing has delivered promising results, with strong links found

between SIF and gross primary productivity. Our ability to use SIF as a tool to monitor

photosynthesis has the potential to enhance agricultural advancement by facilitating the

identification of better performing individuals at a faster rate. However, this kind of high-

throughput phenotyping is usually achieved at the plot and/or leaf scale and there

remains an understanding gap as to what extent SIF can capture plot and leaf scale

variation in photosynthetic activity. We tested the ability of SIF to capture photosynthetic

variability at the plot scale in a C4 biomass sorghum field and at the leaf scale in a C3

tobacco experiment grown under field conditions in Central Illinois, USA. To do this, we

built a portable SIF system to collect high-resolution measurements of SIF and coupled

these with measurements of leaf-level gas exchange and photosynthetic performance

indicators (i.e. Vcmax & Jmax), in situ chlorophyll fluorescence, leaf chlorophyll content,

spectral reflectance indices and leaf area. This presentation will discuss not only the

Page 106: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

results from our field experiments, but also some of the lessons we have learned along the

way towards developing SIF into a high-throughput phenotyping tool for use at the leaf

and plot scales.

Pancake Prickly Pear Cactus - 66

Jaderson Armanhi, University of Campinas

A real-time, non-invasive, low-cost monitoring system for plant phenotyping

under stress

Rafael de Souza – University of Campinas; Paulo Arruda – University of Campinas

Phenotypic data are essential to understanding plant responses to environmental

changes. Conventional instruments to assess plant physiological status are often invasive

or destructive, such as pressure chambers and tensiometers, or designed to provide a

single data point, such as the infrared gas analyzer (IRGA) and the porometer. Although

these methods are reliable, they do not provide continuous monitoring of plant response

to environmental stresses, which might result in losses of relevant information regarding

the true physiological status and plant adaptation mechanisms. Real-time phenotyping

technologies are usually costly, and most platforms are restricted to phenotyping

facilities. Therefore, the development of low-cost phenotyping options is exceptionally

convenient for small-scale studies and experimental setups under growth chamber and

greenhouse conditions. Here we propose a simple and real-time monitoring system for

the remote study of plant physiology using low-cost and easy-to-handle electronic

components. Our system provides the constant monitor of leaf temperature, vapor

pressure deficit (VPD), soil moisture, water loss, as well as the air temperature, relative

humidity and light intensity. An integrated RGB camera was used to record plant

response over time and a modified camera was used to capture near-infrared images for

NDVI measurements. All sensors and cameras are connected to a microcontroller

Raspberry Pi that receives and processes signals and images through custom and

automated scripts. Real-time data are sent to an online server that plots graphs and

creates time-lapse movies on a webpage. Sensors and methods are currently being

validated in experiments designed to evaluate drought stress response in maize. By

providing temporal high-resolution data and imaging, our small-scale system has the

potential to bring valuable information on plant phenotyping in a low-cost manner.

Page 107: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Pancake Prickly Pear Cactus - 67

Nicholas Kaczmar, Cornell University

Application of unmanned aerial vehicles for high-throughput phenotyping of

canopy traits in maize

Ethan Stewart – Cornell University; Michael Gore – Cornell University

The advent of affordable unmanned aerial vehicle (UAV) technologies and relaxation of

federal regulations make UAVs appealing tools for collecting plant phenotypic data in

field environments. Routine phenotypic measures such as plant height are important for

genetic studies and breeding programs, but are time and labor intensive to collect.

Additionally, vegetation indices like the normalized difference vegetation index (NDVI)

are typically measured by plane or satellite platforms, thus not offering the resolution

needed to assess individual field trial plots. UAVs offer the potential to gather large field

data sets at high spatial and temporal resolution. The Genomes to Fields (G2F) Initiative

is a publicly initiated and led research initiative supporting translation of maize genomic

information for the benefit of growers, consumers and society. UAVs fitted with RGB and

multispectral cameras were flown at low elevation over two G2F reps of approximately

500 maize hybrids. These flights were conducted at 6 time points per growing season at

the Cornell Musgrave Research Farm in Aurora, NY from 2015-17. Images were stitched

together and geo-referenced using ground control points of known positions to give geo-

referenced orthomosaic images of the field. Digital elevation models and 3D point clouds

were produced from the RGB stitched images, allowing mean plant height to be calculated

for each plot. In addition, plot level NDVI was calculated from the multispectral images.

Such higher resolution data sets will allow the underlying genetic basis of dynamic

phenotypes to be more extensively studied.

Pancake Prickly Pear Cactus - 68

Kaitlyn Read, University of New Mexico

Tissue specific electrical impedance as a potential screening tool

Patrick Hudson – University of New Mexico; Philip Miller – Sandia National

Laboratories; David Hanson – University of New Mexico

Electrical Impedance Spectroscopy (EIS) is a commonly used noninvasive method to

predict root dimensions, tissue damages, and other physiological parameters. These

methods typically rely on measuring through an electrically variable medium (ie soil,

hydroponic fluid, and epidermal layers), or destructively removing part of the plant. Here

we demonstrate the utility of microneedles to apply EIS methods to specific organs and

Page 108: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

tissues in planta. Microneedles were placed on both the adaxial and abaxial surfaces of a

sorghum (Sorghum bicolor) leaf midrib, to measure water storage and water transport

tissues, respectively. An 18-gauge needle was placed 1 cm below the leaf-stalk junction to

function as the signal receiver for both microneedle placements. A handheld LCR meter

supplied a voltage of 0.6V AC, and measured impedance and phase angle at four different

frequencies. Microneedle impedance values were compared to planar metal transducers

as a control, which didn’t penetrate the plant tissue, and impedance values across all

frequencies tested were significantly lower with the microneedle devices. After in planta

EIS measurements were concluded, a fully expanded leaf was removed. Water storage

and water transport tissues were dissected, and EIS measurements were repeated in the

isolated tissues. Impedance was significantly lower in water transport tissue compared to

water storage tissue, both in planta and in isolation. One week after in planta

measurements, leaves showed no adverse response to microneedle applications, other

than superficial callose deposition at the injection site. Our results show that microneedle

EIS can distinguish specific tissues in a non-destructive fashion, and offer a novel

opportunity for high resolution, real-time plant monitoring.

Pancake Prickly Pear Cactus - 69

Grégoire Hummel, PhD, CEO Phenospex B.V.

PlantEye F500: combine 3D and multispectral information in one sensor

PlantEye is a high-resolution 3D laser scanner that computes a robust and validated set

of morphological plant parameters fully automatically. A core feature of PlantEye is that

it can be operated in full sunlight without any restrictions - crucial for plant phenotyping

under field conditions or if you follow a “sensor-to-plant-concept”. Phenospex has now

developed a new sensor generation, which combines the actual features of PlantEye on

the fly with a 4-channel multispectral camera in the range between 400 – 900nm. This

unique hardware-based sensor fusion concept allows us to deliver spectral information

for each data point of the plant in X, Y, Z-direction and we can compute parameters like

NDVI, color index and many other vegetation indices. This new sensor generation opens

a wide range of new possibilities in plant phenotyping and increases its efficiency.

Pancake Prickly Pear Cactus - 70

Miki Fujita, PhD - RIKEN CSRS

Evaluation of Plant Environmental Stress Response using “RIPPS”, an

Automated Phenotyping System

Page 109: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Miki Fujita – RIKEN; Kaoru Urano – RIKEN; Takanari Tanabata – Kazusa Inst.; Saya

Kikuchi – RIKEN; Kazuo Shinozaki – RIKEN

High-throughput and accurate measurements of plant traits facilitate the understanding

of gene function. Especially, with recent advances in quantitative genomics such as QTL

or GWAS, there is a growing need for precise quantification of multiple traits in plants.

However, in the case of environmental stress responses such as drought, it is difficult to

quantify the adaptive responses because multiple environmental factors are intricately

involved in the phenotype. Therefore, precise control of growth conditions is of great

importance to evaluate plant responses to environmental stresses. Recently we have

developed an automatic phenotyping system that evaluate plant growth responses to a

wide spectrum of environmental conditions. The system named RIPPS (RIKEN Plant

Phenotyping System) controls individual soil moisture in continuously rotating 120 pots

by a combination of automatic weighing and watering systems that enable the precise

control of soil water condition is necessary for quantifying the adaptive responses to

environmental stresses such as limited water conditions. RIPPS also take image of top

and side view of the plants every two hours. In this presentation, we’ll demonstrate the

utility of the RIPPS in evaluating drought or salinity tolerance and water use efficiency.

Pancake Prickly Pear Cactus - 71

Yang Yang, Purdue University

The Expanding High-throughput Phenotyping Capabilities at Purdue

University

Mitchell Tuinstra – Purdue University; Erin Robinson – Purdue University; Chris

Hoagland – Purdue University; Jason Adams – Purdue University

To conduct high-throughput plant phenotyping, we need to establish systems capable of

processing plants of various stature in real-time, acquiring non-destructive phenotypic

data, and providing accurately adjusted and variable environmental conditions.

This poster introduces Purdue University’s high-throughput phenotyping platforms with

unique features both in the field and in the controlled-environment facilities. Purdue’s

Indiana Corn and Soybean Innovation Center, a 25,500-square-foot field phenotyping

facility, is designed for research and development of remote sensing platforms such as

UAV-based imaging systems and the PhenoRover, a ground-based mobile sensing

system. Researchers in the facility also use phenotyping equipments such as root and leaf

scanners, root washing stations, seed counters and color sorters, 3D printers and

scanners, ovens and grinders, as well as an automated threshing and shelling line for plant

and seed processing.

Page 110: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

The controlled-environment phenotyping facility (CEPF) has the unique feature of

continuously rotating 256 individual plants of up to 3-meter tall in a growth chamber of

which the key environmental parameters are precisely controlled. The autonomous

fertigation system conducts weight-based dosing of water and fertilizer, thereby providing

the capability of establishing different watering and nutrient management regimes. The

RGB and hyperspectral imaging systems provide the capability to monitor plant size and

growth rate with high temporal and spatial resolutions in different plant developmental

stages.

Overall, the phenotyping platforms make it possible to perform reproducible, large-scale

experiments that would not be possible by hand. The synergy of the CEPF and the in-field

phenotyping facility at Purdue University creates a unique data pipeline that enables

better understanding in basic plant physiological processes and empowers researches for

crop yield improvement under various environmental conditions. This synergy also offers

a unique opportunity for evaluating and improving existing phenotyping systems, as well

as developing new technologies.

Pancake Prickly Pear Cactus - 72

Daniel Runcie, PhD - University of California Davis

A Bayesian approach to quantitative genetics for high-dimensional traits

Statistical models for Genome-Wide Association Studies, QTL analysis, and Genomic

Prediction, are the foundation of modern quantitative genetics and crop improvement.

Driven by the explosion of whole-genome genotype data, recent improvements to these

models allow for analyses of millions of markers at a time. However, similar advances for

modeling large phenotype datasets is lacking. New phenotyping technologies collect

thousands of observations on each individual plant or line – changes in morphology

through time, molecular phenotypes such as gene expression or metabolite levels, or

performance measures across multiple environments. Jointly modeling these high-

dimensional traits can provide insight into developmental and physiological mechanisms

that link genotype and phenotype. We propose a robust and efficient method for modeling

the genotype-phenotype relationship of high-dimensional traits. The key idea underlying

our model is that groups of traits will be highly correlated due to genetic and

developmental pleiotropy. We leverage these correlated modules to prioritize the most

important signals in big data. We will demonstrate how our method provides powerful

and interpretable estimates of genetic architecture using two high-dimensional datasets:

a time-series analysis of growth curves, and a dataset of genome-wide gene expression.

Page 111: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Pancake Prickly Pear Cactus- 73

Jian Jin, PhD - Purdue University

Purdue's New Automatic Phenotyping Greenhouse with Micro-climates

Removed

Purdue University deployed a new fully automatic phenotyping greenhouse in May 2017.

This facility is featured for (1) Continuous scanning of each crop plant for up to 20

times/day; (2) Clearly removing the micro-climates impact (the variance of

environmental conditions caused by distribution of lighting, temperature, airflow and so

on across the greenhouse space); (3) Advanced hyperspectral imaging system and data

modeling for plant physiological features predictions. Dr. Jin will also share his view of

next generation plant phenotyping in the next 10 years.

Pancake Prickly Pear Cactus -74

Sierra Young, Iowa State University

Design and Evaluation of a Field-Based High-Throughput Phenotyping

Robot for Energy Sorghum

This article describes the design and field evaluation of a low-cost, high-throughput

phenotyping robot for energy sorghum. High-throughput phenotyping approaches have

been used in isolated growth chambers or greenhouses, but there is a growing need for

field-based, precision agriculture techniques to measure large quantities of plants at high

spatial and temporal resolutions throughout a growing season. A low-cost, tracked mobile

robot was developed to collect phenotypic data for individual plants and tested on two

separate energy sorghum fields in Central Illinois during summer 2016. Stereo imaging

techniques determined plant height, and a depth sensor measured stem width near the

base of the plant. A data capture rate of one acre, bi-weekly, was demonstrated for

platform robustness consistent for various environmental conditions and crop yield

modeling needs, and formative human-robot interaction observations were made during

the field trials to address usability. This work is of interest to researchers and practitioners

advancing the field of plant breeding because it demonstrates a new phenotyping

platform that can measure individual plant architecture traits accurately (absolute

measurement error at 15% for plant height and 13% for stem width) over large areas at a

sub-daily frequency.

Page 112: Speaker Abstracts - Plant · In this presentation, I will use case studies to demonstrate how plant phenotyping infrastructure can be used to address relevant biological questions

Pancake Prickly Pear Cactus - 75

James Bunce, PP Systems

High Throughput Photosynthesis Characterization of C3 Plants

Dr. James Bunce, Ph.D. in Plant Physiology – PP Systems; Andrew Lintz, B.S. in

Mechanical Engineering – PP Systems

Single point measurements of leaf gas exchange provide basically only the parameters net

photosynthesis, stomatal conductance, and instantaneous leaf water use efficiency. From

analysis of assimilation rate vs. internal CO2 concentrations (A vs. Ci) curves, four or five

additional leaf parameters are obtained for plants with C3 carbon metabolism, which

allow estimation of photosynthesis over a range of conditions. However, determining A

vs. Ci curves conventionally requires at least 20 minutes per leaf, compared with about 2

minutes for single point measurements, which greatly limits through-put. PP Systems

has developed a method of linearly ramping CO2 rapidly in their CIRAS-3 Portable

Photosynthesis System, which provides a complete A vs. Ci curve in 5 minutes per leaf.

Two initial steps are required: storing the changes in analyzer sensitivity with background

CO2, and collecting data from ramping of CO2 with an empty chamber. These two steps

need only be done once per day. The 5 minute total measurement period per leaf includes

a 2 minute initial equilibration period followed by 3 minutes of ramping up of CO2

concentration until the rate of change of A with CO2 becomes small, i.e. until CO2

becomes nearly saturating to A. With the CIRAS-3 system post-processing of the gas

exchange data is very simple: the apparent “A” of the empty chamber is subtracted from

the “A” value obtained with a leaf in the chamber at each time point of the CO2 ramping

period. This provides the actual A value at each time point, and the Ci is obtained from

this actual A, stomatal conductance, and external CO2 as in the conventional calculation

of Ci. Because of the rapid change in CO2, we have seen no significant change in stomatal

conductance during the CO2 ramps < ./p>