Melba Crawford, PhD - Purdue University

Multi-modality Remote Sensing Data Acquisition and Analysis for High Throughput Phenotyping

GB is the work horse because geometry based measurements

Having a good GPS system makes it easy for plot extraction

Plant height, crop density, panicle counting, plant structure in sorghum

LIDAR from UAV doesn’t get good penetration- ground LIDAR does better.

Wheel-Based Lidar Data for Plant Height and Canopy Cover Evaluation to Aid Biomass Prediction

Implementation of UAV-Based Lidar for High Throughput Phenotyping

Her group also uses APSIM analytic models

Aim to not include l

Sorghum Biomass Prediction Using Uav-Based Remote Sensing Data and Crop Model Simulation https://ieeexplore.ieee.org/abstract/document/8519034

Increasingly high-res positioning systems are game changers for research plots where row level is critical

Latent features whenever possible to maintain explainability outside of any models one builds

Lidar provides greater detail for plant structure. Rovers used for greater penetration

New modeling strategy that incorporates time and weather is called Recurrent Neural Network (RNN)

Requires full matrix of data for each time point

Do we use detailed rover datasets to develop robust models which you then predict via more broadly applicable UAV measurements?

Challenge and opportunity is to develop cross trained scientists that are able to talk to each other and work together


Brook T. Moyers, PhD – University of Massachusetts Boston

Of rice and men: using complex genetic populations in crop improvement


We are running out of water and rice is water-sensitive

She used the MAGIC rice population for her research developed at IRRI (1,200 lines total); she used a subset of 415-538 on her experiments

Augmented Design: where you do not replicate the lines (except checks)

Heritability of several traits dramatically reduced when incorporating kinship matrix in the MAGIC population, why? Lots of G by E effect and transgressive segregants.

Lots of GxE in trait values requiring large-scale experiments with multiple environments and appropriate replication to find candidate loci


Regina Baucom, PhD – University of Michigan

Can character displacement drive the evolution of root traits?

Why study root traits? Important for nutrient acquisition, host microbes, staple crops

How have they been studied? Ecologist → healthy diverse plant community; plant scientists → genes underlying particular architectures

Need more work looking at evolutionary forces leading to root traits

Theoretically, no 2 species can compete for the same limiting resources and co-exist

Character displacement--trait differences between otherwise similar species evolve due to competition for similar resources

Is there the potential for character displacement for root traits???

Ipomoea purpurea competing vs Ipomoea hederacea was the model of choice

Plant-plant competition is an agent of selection on roots (there was no above-ground competition)


Joshua Peschel - Iowa State University

Maize Stalk-Localized Actuation through Physical Presence of an Environmental Robot

Cool robot named Ariel

They tested no robot, robot and robot+slapper

Plant mechanical properties changes as a result of human/robot interaction to plants-a secondary effect of high throughput phenotyping

Poster #519


Mitchell Feldmann - University of California, Davis

Pedigree-Informed Genome-to-Phenome Mapping of Morphometric Traits in Strawberry

New talk title- “Quantization of Strawberry Fruit..


Alejandro Vergara - International Center for Tropical Agriculture (CIAT)

Unmanned aerial vehicle based remote sensing for monitoring Cassava (Manihot esculenta Crantz) growth and development


Afef Marzougui - Washington State University

Phenotyping Aphanomyces Root Rot Resistance in Pea using Volatile Organic Compounds

VOCs are emitted by plants in response to many environmental factors, has genetic components

Resistant pea cultivar release 3 VOCs after 15 dai

Poster #423


Ian Braun - Iowa State University

Computational methodology to enable direct computation on phenotypic descriptions

Using computable language-based descriptions of phenotypes from the literature to find novel candidate genes across and among species

Poster #100

Tara Enders - University of Minnesota

Relationships between the cold-responsive transcriptome and phenome in maize seedlings

Characterized variation in maize seedlings under cold stress

Whole plant phenotyping with RGB and hyperspectral cameras along with RNASeq data


Brooke Bruning - Australian Plant Phenomics Facility

Hyperspectral distribution maps to predict the concentration and spatial distribution of nitrogen and water content in wheat

Poster #502

Water and nitrogen can be predicted using hyperspectral phenotyping methods



Malcolm Bennett, PhD - University of Nottingham


Uncovering the hidden half of plant growth and development using root phenotyping

Root Robot excavation and imaging: harrow to dig out stalks used at Penn State. Developed 3D imaging using a multi-angle system. Huge variation in crown root angle. Deep roots showed increased NUE under low N.

Root changes the porosity around it. The soil is an important part. Thistle re-engineering the surrounding rhizosphere.

Shaping 3D Root System Architecture. Roots stop branching when they go through an air space.

The xerobranching response represses lateral root formation when roots are not in contact with water

Root branching toward water involves posttranslational modification of transcription factor ARF7

New facility: Hounsfield robotic CT scanning of roots (named after the guy who invented CT) doing “4D” imaging includes time. Challenge is software. Reconstructing root phenotypes.


Nadia Shakoor, PhD - Donald Danforth Plant Science Center

TERRA-REF Year Four: De-risking the world’s largest agricultural robot for reference-grade crop phenotyping data.


https://www.terraref.org/sample-data available


David LeBauer - University of Arizona College of Agriculture and Life Sciences

TERRA REF’s Open Software, Data, and Computing for High Throughput Phenomics



George Kantor, PhD - Carnegie Mellon University - Robotics Institute

Robotics and AI Platforms for Rapid In-Field Phenotyping

1.Mobility: Autonomous vehicles, 2. perception and 3. deep neural networks.

Use in vineyards. A utility vehicle with camera and flash works night and day. Good image for a computer not for people. Automatically count all the grapes. Do early-season yield prediction and predict harvest. Build layered maps for sorghum and maize.


Molly Hanlon, PhD - Penn State

Using x-ray fluorescence technology to estimate root depth without coring or digging


Maria Newcomb, PhD - UA Maricopa Agriculture Center

Automated Chlorophyll Fluorescence Imaging from a Field Phenotyping Platform


Cody Champion - New Mexico Institute of Mining and Technology

Multivariate Curve Resolution and Other Methods Applied to the Terra-ref Hyperspectral Dataset



Magdalena M. Julkowska, PhD - KAUST

MVApp - Lessons learned from streamlining data curation and analysis


MVApp YouTube Channel

They have tutorials

People can get code at GitHub

They welcome collaborators

No more Excel! :)


Anjali Iyer-Pascuzzi, PhD – Purdue University

New tools for image based quantification of disease traits in crops

Ralstonia solanacearum Tomato bacterial wilt bacteria and slime form biofilm that works like plug.

3D imaging of plants on a turntable (4 images of each plant at 45 degree angles)

Collaborates with machine learning expert at Purdue


Katherine Meacham-Hensold – University of Illinois at Urbana- Champaign

High-throughput field phenotyping using hyperspectral imaging reveals genetic modifications to photosynthetic capacity

Handheld devices and cameras for hyperspectral imaging

Part of the RIPE project funded by Gates Foundation

Hyperspectral cameras are Pika II and Pika NIR mounted on a push car

Partial Least Square Regression(PLSR )analysis of tobacco predicting VCmax and Jmax with pretty good accuracy in single leaves.

Look at poster # 207


Steven L. Anderson, II - Texas A&M University

Utilizing structure from motion point clouds to estimate maize (Zea mays L.) height within a field-based breeding program

Poster 201 Did some grain yield vs ht measurements.


Karina Medina Jimenez - Arkansas University

High throughput kernel phenotyping of corn hybrids grown in Arkansas

Kernel size and shape measurements. Used 120 k of each hybrid from G2F


Charles Pignon - Institute for Genomic Biology

Accelerating stomatal movement under fluctuating light

Slower response of stomata to change in light (5 min) compares to PS ( 2 sec). By making stomata perform faster water could be saved. Good amount of genetic variation for the trait, speed of stomatal conductance change. Were able to find candidate genes.


Tessa Durham Brooks, PhD - Doane University

Image analysis interventions improve computational self-efficacy and inform career paths of natural science majors in the DIVAS project


Mitchell Eithun, BA – Michigan State University

Isolating juvenile phyllotactic patterns embedded within eight years of secondary growth (sweet cherry, Prunus avium L.)


Dirk Hays - Texas A&M University


50 MT of roots/ha from perennial sorghum. Rhizomes and fine roots using GPR



Autonomously generating shape estimates of thin plant parts across scales


Randy Clark, PhD - Corteva Agriscience

Phenotyping and modeling of plant responses to environmental stresses to guide crop breeding

Integrating phenomics in commercial crop development

Phenomics comes toward the end in breeding program

Crop growth models capture traits that results in better yield

Crop Growth Model Whole Genomic Prediction model. “All comes back to phenotyping in the end”. Need to know physiology and data to generate priors for the model.

Used huge transparent chambers (glass?) in the field to measure transpiration response to vapor pressure deficit throughout the day in maize

AQUAMAX ( a maize hybrid for drought prone regions) displays higher silk elongation rate under water deficit- more silk results to more pollination and therefore more yield.


Seth C. Murray, PhD - Texas A&M University

Towards predictive phenomics in selection, grain yield using unmanned aerial systems in maize

Genomes to Fields and bench-top near infrared reflectance spectroscopy data sets

Goal is to improve plants. Focus on prediction and discovery in plant breeding program.

Poster 213 phenomic prediction from NIRS whole kernel spectral yield. NGRDI time series.

20 flight dates on G2F set in Texas irr and dryland 8 flight dates, 3 flight dates pretty good predictions. Juvenile, flowering and grain fill images adequate. Checks had different spectral patterns. Soon available in Cyverse.


Jordan Ubbens - University of Saskatchewan

Latent Space Association Analysis: Towards GWAS Directly from Images




Natalia de Leon - University of Wisconsin-Madison

The Genomes to Fields GxE project: Progress and Perspective

  • Viewing yield potential in the lense of NCGA corn contest winners (~540bu/acre) vs US average (~176bu/acre). There is a huge delta which could be improved still.
  • Contest winners like all plants to emerge within a deviation of 4-12 hours.
  • www.g2f.org
  • https://www.genomes2fields.org/
  • GEMS data format
  • P=P+E+PxE
  • What alleles are important? New alleles?
  • Germplasm enhancement of maize (GEM) project
  • Access to ex PVP lines to incorporate alleles. Can tell what’s different but not necessarily what is useful.
  • Phenotyping kernel size shape to characterize E effect on genetic lines (Edgar Spalding’s team)


Ian Stavness, PhD - University of Saskatchewan

Computational Tools for Plant Phenotyping from Outdoor Images

Plot Vision developed for field imaging

Deep Plant Phenomics (paper)



Sierra Young, PhD - North Carolina State University, Biological and Agricultural Engineering

Advancements and Challenges in Technology and Data Management Practices of Field-Based, High-Throughput Phenotyping

  • There is a tech gap between the research phenomics community and traditional plant breeders
  • Although tech is improving by being cheaper and a bit more standard, it is still a hurdle to get into
  • View as a systems approach
    • Human-machine interface
      • This is not just ‘user friendliness’, but what actions are available given a person's level of expertise
      • Human controller vs human monitor
    • System (field)  to data collection
      • Platforms to date are for very specific usecases and are not ‘one size fits all’
      • Do we even want that???
      • Should have less focus on a generic system, but have the outputs from data capture be more standard for downstream steps
    • Data transfer outside of system
      • Temporal and spatial add fun and complexity
      • Acquisition to processing is not trivial
      • Human-data interaction currently is a dealbreaker for widespread use
        • I don’t want to be swapping SD cards from cameras…
    • Decision support for human
      • The science and an art
      • If we go with black box, there needs to be trust in it
      • Level of trust is highly dependent on the desired level of interaction and transparency in the system
      • These are social science questions in part
  • It is less focusing on generic boxes (steps), but the interfaces between them
  • We need to include user(s) in the design more and more
  • Focus on going from R&D across applied phenomics


Ethan Stewart - Cornell

The application of a deep learning approach for quantitative disease phenotyping in UAV images

NCLB resulted in lost value of  ~ $ 1.9 billion/yr working on detection system

Convolutional neural network. Number and size of lesions.


Amazon’s Mechanical Turk was used to generate training data


Daniel J. Robertson, PhD, PE - University of Idaho

Deep Phenotyping for Genome to Phenome Mapping of Stalk Lodging Resistance

5-20 % yield lost; corn, wheat and rice due to stalk lodging

Stalk lodging is a complex problem and requires transdisciplinary teams working together

Received 6 mil NSF EPSCOR Track II (wow!)

Working on DNA->RNA->Protein->Metabolome->Phenome using a series of differential equations.

Implemented in a Bayesian statistical framework.

Stalk lodging depends on 3 things: 1. Anchor to soil, 2.load and 3.bending strength

Build a computational model to describe the trait.

A ‘room full of experts’ can help describe it. However a room full of experts is also very Expensive…


Zheng Xu - University of Nebraska-Lincoln

Alternative Deep Learning Approaches in Plant Phenotyping Based On Hyperspectral Images


Sen Subramanian - South Dakota State University

Quantitative 3D Imaging of cell level auxin and cytokinin response ratios in legume roots and nodules

Poster #437


Carmela Rosario Guadagno - University of Wyoming

Hydration dynamics surrounding photosystems explain high-throughput chlorophyll a fluorescence phenotyping



Jeff Gustin - University of Florida

VIGOR: A Controlled Environment-based Machine Vision Assay of Seedling Emergence from Soil

Seeds germinated in 168 cells or tubes filled with sand or soil monitored emergence with the camera. Used camera mounted at one side. Added a red cap to each cell measured movement of red cap movement.  Traps seedling in cell.

Doing QTL analysis finding corr between seed wt and germ.



Kai-Wei Yang, PhD Student - Purdue University

Prediction of Sorghum (Sorghum bicolor) Performance through Remote Sensing and Crop Modeling


Barry Martin - Flagship Pioneering

Automated phenotyping and analysis of germination and early seedling growth using cloud computing



Margaret Staton - University of Tennessee

TreeSnap, a Mobile Device App and Website Enabling Tree Phenotyping by Citizen Scientists

Is a citizen-science mobile app partnered with restoration tree breeding programs

Has cool user interface with examples, information tabs

Jitter GPS to prevent bad actors from observed public data


Dani Martinez, PhD - Cornell University

Blackbird: A next-gen imaging system for high-throughput laboratory phenotyping.

VitisGen 2 Project in grapevine focused on powdery mildew resistance

Blackbird-00- next gen prototype of the HT imaging system


William S. van der Kamp - University of Saskatechewan

Plot Vision: Image Analysis for Plant Breeders

Why we need Plot Vision? Breeders have lots of plots and images of plots and need to process these faster with relative ease.



Christopher N. Topp, PhD - Donald Danforth Plant Science Center

Multiscale X-ray imaging, computer vision, and virtual reality data display provide a new paradigm for plant developmental biology research, teaching, and outreach



Oliver Scholz  - Fraunhofer Development Center X-Ray Technology

Improving a Professor - Creating a field-based phenotyping robot based on a remote controlled sensing platform.


Sanaz Jarolmasjed - Donald Danforth Plant Science Center

Hyperspectral image analysis tools in PlantCV framework


Zbigniew Kolber - Soliense Inc.

Application of Light Induced Fluorescence Transient (LIFT) Technique in Plant Phenomics


Patrick Hudson - University of New Mexico

Passive electrical properties as reliable predictors for key aspects of plant water relations

Water potential and relative water content- two measures of water status in plants

Microneedle electrical impedance spectroscopy


Stefan Gerth - Fraunhofer Development Center X-Ray Technology

CTProcessing.net - A 3D volume processing pipeline for automated analysis of Computer Tomography measurements

Scalable, adaptable, and linkable




Pawel Krajewski - Institute of Plant Genetics, Polish Academy of Sciences

MIAPPE: current developments and applications

Deciding what information to ask everyone to include in annotating phenotypic datasets.

Using pre-defined ontologies for most variables.




More info on http://miappe.org


David Pont, PhD - Scion, New Zealand Forest Research Institute

Phenotyping for Forestry - Seeing the Forest AND the Trees

Forests often planted on the land no one else wants, so lots of issues with poor soils, steep hillsides.

Hard to process images when on side of the hill is in dark shadow and the other quite bright.

Two levels phenotyping- area based and tree based

Dungey HS, Dash JP, Pont D, Clinton PW, Watt MS, Telfer EJ. 2018. Phenotyping Whole Forests Will Help to Track Genetic Performance. Trends in plant science. 23:10, 854-864


Argelia Lorence, PhD - Arkansas State University

MicroCT scanning reveals drought-induced changes in cell size and morphology in Arabidopsis seeds

When water-limited below 50%, only you see symptoms in seeds

100 to 12.5 percent changed seed volume and surface area, cell size

All major and minor nutrients changed except Na in well watered vs water-limited seeds

@DMC_lab developed small plant imaging platform

Effects of water-limitation transgenerational


Erin Sparks, PhD - University of Delaware

Robotic phenotyping at the root-soil interface

Older the plants get when logding occurs, then it affects yield

Maize brace roots proposed to promote lodging resistance

Indeed brace roots significantly contribute to stalk flexural stiffness


Ángel Ferrero-Serrano, PhD – Penn State University

Phenotypic and genome-wide association of natural variation with the local environment of Arabidopsis

in silico GWAS using 1,001 resequenced arabidopsis lines, 131 exp. measured traits mined from previously published GWAS, and 204 geoclimatic variables from the locations the accessions were collected from.

CLIMtools website with the datasets and tools for association: http://www.personal.psu.edu/sm...

AGG3 is a heterotrimmeric G-protein  involved in cold response.

A lot of this talk seemed to come from this paper published last month. Free-to-read link from the speaker’s twitter feed: http://rdcu.be/bgNXQ

Ferrero-Serrano Á, Assmann SM. 2019 Phenotypic and genome-wide association with the local environment of Arabidopsis. Nature Ecology and Evolution. PMID:30643246, doi: 10.1038/s41559-018-0754-5


1.     Kelly from Bayer Crop Sciences

She is a breeder

How can data analytics and phenomics help in crop development?



Liang Dong, PhD - Iowa State University

Miniature sensors for digital agriculture

·       Sensors with micro technology

·       Crop sensors: In-Planta Sensor for in-situ measurement of Nitrogen with tiny needles

·       Microfluidic ion-selective soil sensor: In-field measurement of N in soil

·       Measurement water transport from roots to leaves with two or more moisture sensors

·       Leaf thickness measurement with capacitance and mechanical stress changes in sensor

·       Miniature printed soil water potential sensor


Courtney Murren, PhD - College of Charleston

unPAK-ing the Arabidopsis fitness phenome: large scale mutant screen of roots, shoots and fruits with companion CURE (Course based Undergraduate Research Experiences)


Lance Cadle-Davidson, PhD - Grape Genetics Research Unit - USDA, ARS

Computer vision for high-throughput phenotyping of powdery mildew resistance


Sara Tirado - University of Minnesota

Image-based Phenotypic Platform for Monitoring Maize Growth to Estimate End-season Productivity

·       DJI p4

·       Image stiching and pipeline

·       Stand counts and plant rotations measured at VE-V3

·       Measure flowering

·       Throughout canopy closure, plant height, color

·       Data over 4-row plots

·       Drone vs hand measured height 0.96 r2

o   RMSE at 24cm (given hand vs hand is ~10cm)

o   Overall correlation is largely driven by specific dates (individual dates more like 0.41)

o   Almost no correlation at a given time point, even with hand vs hand

§  Not much variation to drive any correlation anyhow

·       How can we better predict height at a time point?

·       Hand height ~ algo height + density + planting date + genotype

·       Can find times of height dip = blowdown storm and lodging

o   Field level genotypic effect of lodging observable

·       Temporal measurements can help inform stuff


Reisha D. Peters - University of Saskatechewan

Characterization of leaf surface phenotypes based on light interaction



Margaret Frank - Cornell University

Exploring Graft-Incompatibility with Pepmato and Tomepper


Brent E. Ewars, PhD - University of Wyoming

Coming out of the dark: night time stomatal conductance phenotyping to help improve plant growth and response to drought


Duke Pauli, PhD - The University of Arizona

Desert Phenomics: Understanding Crop Hydraulics for the Climate of the Future


Kevin Silverstein - University of Minnesota

G.E.M.S: A platform to enable data-driven agricultural innovation



Jennifer C. Quebedeaux – University of Illinois at Urbana-Champaign

Manipulating stomatal patterning to improve water use efficiency in wheat


Tyler Dowd, PhD - Donald Danforth Plant Science Center

A Greenhouse Mesocosm System for Integrated Environmental Sensing, Root Phenotyping, and New Sensor Development


Ramona L. Walls - University of Arizona

A generalized workflow for integrating biodiversity trait data using ontologies


Noah Fahlgren - Donald Danforth Plant Science Center

Machine Learning Methods in PlantCV for Leaf Tracking and More


Arthur R. Woll, PhD - Cornell University

Quantitative elemental imaging using 3D confocal x-ray fluorescence microscopy at the micron scale


Changying Li - UGA

3D plant organ mapping under field conditions


Siobhan A. Braybrook, PhD - UCLA

Quantifying diverse cell shapes: morphometrics and udulomics


Boyan Peshlov - Climate Corporation

From images to insights: Large scale crop health evaluation using high resolution imagery and machine learning


Jnaneshwar Das, PhD - Arizona State University

Robots in the Wild: Collaborative Exploration and Mapping


Sruti Das Choudhury – University of Nebraska-Lincoln

Deep Learning for Early Detection and Temporal Propagation of Drought Stress based on Hyperspectral Imagery: Dataset, Algorithm and Analysis