Next Generation Phenotyping Technologies in Breeding for Abiotic Stress Tolerance in Maize. Mitch Tuinstra

1. Next Generation
Phenotyping Technologies
in Breeding for Abiotic
Stress Tolerance in Maize
Mitch Tuinstra, Ph.D.
Professor of Plant Breeding
Scientific Director – Institute for Plant Sciences
Purdue University

2. Micheline Pelletier/Sygma/Corb
In the next 50 years, we’re
going to have to produce more
food than we have in the last
10,000 years. We need to find
ways to employ technology and
science to increase production
to feed a hungry planet.
Norman Borlaug
Nobel Peace Prize Laureate
April 7, 2009
The Challenge ….
Food security and sustainability will depend on advances in plant-based agriculture. We
need to develop higher-yielding plants that are more nutritious, use water and nutrients
more efficiently, and can tolerate more variation in the environment.

3. Data-Driven Agriculture – Genomics
Genome sequencing has
transformed that way we
do plant breeding and
genetic research.
State-of-the-Art: BIG DATA
Normal Modified
brachytic

4. Data-Driven Agriculture – Phenomics?
State-of-the-Art: LITTLE DATA
• Phenomics is the study of plant characteristics or traits.
• Phenomic analyses are expensive.
−$325 per hybrid for a 4 location replicated yield trial with data
collected on phenology, grain yield, grain moisture, lodging,
and plant population

5. Physiology
Genetic Markers & Genomics
Control over Growth Conditions
Direct Applicability to Crop Production
Growth & Morphology
Anatomy
Yield
Biochemistry
Adapted from Bruce et al. (2002)
Monitor Crop Characteristics with
Greater Accuracy
• Study plants in state-of-the-art field and controlled
environment facilities

6. • The CEPF was developed for
plant production and high-
throughput imaging of diverse
crop species.
−Conviron Growth House for precise
lighting and temperature control
−256 plant capacity (0.5-4m)
−Argus multi-feed nutrient injection for
fertility management
−Automated weight-based irrigation
system for measuring plant water
use
Controlled Environment Phenomics
Facility (CEPF)

7. Controlled Environment Phenomics
Facility (CEPF)
• Plants in the CEPF are grown on
an automated conveyer system for
daily imaging and phenotyping
• RGB and Hyperspectral imaging
towers can accommodate crop
plants up to 4m tall for shoot-
based phenotyping
• An X-Ray CT Scanner is being
installed to enable below-ground
imaging for root-based
phenotyping

8. Greenhouse Imaging System
• A greenhouse plant imaging system has
been developed at Purdue as a test-bed
for new plant sensors and robotics
equipment.
−Top-view and side-view Middleton Spectral
Vision MSV 500 hyperspectral cameras
−Automated conveyor belt with space for 108
plants
−Flexible imaging booth for sensor testing

9. Greenhouse Imaging System
• Models for Relative Water
Content (RWC) and N Content
in 8 temperate and tropical
maize and sorghum genotypes
• Hyperspectral imaging used to
model the impacts of water
and fertility treatments based
on variation in RWC and N
content

10. • RWC and N content measurements
are time consuming, labor intensive,
and destructive
• Important to develop models that can
predict responses across genotypes
and species
• Accurate and nondestructive
measurements
Greenhouse Imaging System - Why does
it matter?

11. Indiana Corn
and Soybean
Innovation
Center (ICSC)
Field-based Phenomics Research
• Purdue has invested in a field phenomics research
capacity at the Agronomy Farm.
‐ 25,000 ft2 research facility
‐ Plant and seed processing/phenotyping lab
‐ Phenomics tool development workshop
‐ 10 Gb fiber optic connection to high performance
computing facility on campus

12. ICSC Research Facility
• 1406 acre research farm with WIFI
network in every field
• Calibration and ground validation
studies for agronomic, morphological,
physiological and biomass composition
traits
Agronomic Traits
• Plot stand
• Main stem leaf collar height
• Main stem diameter base and top
collar
• Tiller number and height
• Leaf number and leaf angle
• Leaf size distribution
• Total leaf area and leaf area index
• Plant biomass
• Flowering date
• Stem and root lodging
Physiological Traits
• Canopy temperature
• Leaf chlorophyll and nitrogen
Composition Traits
• Cellulose and hemicellulose content
• Lignin composition
• Saccharification assessment

13. ICSC Research Facility – PhenoRover
The PhenoRover accommodates
up to 11 sensors for data
collection with RGB, HS-VNIR,
LIDAR, and video.
RGB
LiDAR-Based
Point Cloud
Image-Based
Point Cloud

14. ICSC Research Facility – UAV Systems
Multiple airframes have been
developed for data collection
with RGB, LiDAR, HS-VNIR,
HS-SWIR, and FLIR sensors
RGB
Hyperspe
ctral
SWIR
Hyperspec
tral VNIR
LiDAR DSM

15. Co-aligned Sensor Data
• Geo-referencing methodologies enable precise co-
registration of multiple sensors; new opportunities for
temporal and multi-scale analyses

16. Feature Extraction and Machine Learning
Automated row/range/plot identification
Plot: 4810
1 2 3 4 5 6 7 8 9 10 11 12

17. Data Visualization Tools
Ground Reference Map
Hyperspectral
Data

18. Automated Trait Predictions
Spacing standard deviation (cm)
34 32 28 29 3632 3233 26302625
Spacing mean (cm)
28 17 8 11 2413 1210 13191111
Date: July 21, 2016
Sensor: Sony Alpha 7R
UAV: DJI S1000+
Altitude: 44 m
Velocity: 5 m s-1
Overlap: 70%
• Measurements of geometric traits: plant height,
canopy characteristics … etc.
• Trait prediction by machine learning: plant centers and
population, canopy architecture, biomass yield … etc.

19. Predictive Modeling of Biomass Yield
Sorghum Biodiversity Panel

20. Nutritional Deficiencies
Map
Automated phenotyping
• Automated crop phenotyping platforms will enable
gene discovery and optimization of crop varieties
and production systems for food, feed, fiber, and fuel
production.
Mobilizing Research for Food Security

21. Institute for Plant Sciences
Yang Yang, Chris Hoagland, Jason Adams, and Richard Westerman, Purdue
University
Greenhouse Phenotyping
Jian Jin and Valerie Cross, Purdue University
Field Phenomics
Melba Crawford, Ed Delp, Ayman Habib, David Ebert, Keith Cherkauer, Mike
Leasure, Clifford Weil, Ali Masjedi, and Neal Carpenter, Purdue University
This research was supported by grants and donations from the AgAlumni Seed Company,
AgReliant Seed, Corteva AgriSciences, ARPA-E TERRA, Sumitomo, United Sorghum Checkoff,
and Purdue University
Acknowledgements