Developing high yielding varieties adapted to changing environmental conditions and new agronomic management practices is an urgent priority to match the predicted demand for food and biomass in 2050. To identify a new commercial variety and optimise its productivity, a typical breeding program has to screen the performance of thousands of genotypes under a variety of environmental and management conditions. Only through a quantitative analysis of plant phenotypes in response to the environment and management practices (P=GxExM) will a geneticist be able to generate the link to the genotype and identify the causal polymorphisms in the genome that can be used in the breeder’s selection process. While significant progress has been made by public research institutions to develop high-throughput phenotyping tools and sensor networks to digitise plants and measure dynamically the environment, the automated quantitative analysis of the phenotype, i.e. extracting information from the raw data and deriving knowledge from it, has become a major bottleneck and is today preventing wide adoption of these tools in breeding companies or Agribusiness industries.
My talk will present our approach to building an Australian analytics infrastructure to address this bottleneck while illustrating at the same time our effort to translating the research tools into products for use by the Agribusiness sector. The opportunities to apply modelling approaches to integrate the information extracted at multiple temporal and spatial scales from state-of-the-art phenotyping technologies will also be explored in a move towards a revolution in plant biology.
1. Shifting plant breeding into overdrive
Dr Xavier Sirault (& teams)
Director – High Resolution Plant Phenomics Centre, Australian Plant Phenomics Facility
Senior Research Scientist and Research Team Leader, CSIRO Agriculture & Food
Vice-Chair International Plant Phenomics Network (IPPN)
CSIRO AGRICULTURE AND FOOD, CANBERRA, AUSTRALIA
2nd Asia Pacific Plant Phenotyping Conference, Nanjing, 23-25 March 2018
Nanjing International Conference Hotel
2. Understanding Pt = G × Et × M
to shift plant breeding into overdrive
Assaying accurately each plant variety for a large
number of traits simultaneously in a large number
of environments
►requires phenotyping systems that integrate
multiple sensing modalities to probe plant function
and performance simultaneously
Current issues are:
- transforming the industrial-scale amounts of plant “breeding” data collected into
information/intelligence (this include assimilating information into computer simulation models e.g. to
calculate new phenotypes – cf. Scott Chapman);
- rendering the information useful for its integration with other large-scale datasets (e.g. genomics);
- Increasing uptake of current technologies into global breeding programs (vs. pre-breeding) so that
genetic progress for yield and quality can keep pace with the required global food production/food
demand
G: Genotype
E: Environment
P: Phenotype
3. The APPF facilitates advances in plant phenomics research and plant breeding to accelerate the
development of improved crops and agricultural practices that support the sustainability of Australia’s
environment, the growth of its economy and the wellbeing of its people.
Vision:
Multi-scale phenotyping infrastructure platforms
www.plantphenomics.org.au
www.phenosmart.org.au
http://podd.org.au
Australia-wide analytics and data discovery infrastructure platforms
phenoSMART is a registered trademark of CSIRO in one or more territories in the world.
Analytics Data discovery
4. Research and
Discovery
Testing and Evaluation
Market release
and uptake
Trait discovery Pre-breeding Improving agronomy
R e s e a r c h e r s
Breeding new varieties Developing agro-chemicals Precision ag-technology
I n d u s t r y
G r o w e r s
High resolution
phenotyping, e.g.
resolving plant
structure and light
capture
Role of APPF in the Agri-Business pipeline
Yield gap
Horticulture
New Crops
Pastures
5. Challenge of field phenotyping…
Over 1M yield plots of wheat sown every year in Australia
6. In-field capability: the Phenomobile®
LiDAR Stereoscopic system
RTK-GPS / Encoder combo Hyperspectral sensors
IR sensors
Adjustable
height
Traits measured @plot scale and dynamically
Canopy structure: Early vigour, leaf area index, biomass, canopy height, early vigour, stay green/
senescence profile, flowering time (in dev env.) and awn presence (in dev env.)
Canopy biochemistry: carotenoids, xanthophylls, anthocyanins, stem carbohydrates
Canopy function: stomatal behaviour, photosynthesis, stay green / senescence index
Yield parameters: plant counts (in dev. Env.), spikes per plot (in dev. Env)
phenomobile is a registered trademark of CSIRO in one or more territories in the world.
Deery et al. 2014
Rebetzke et al. 2016
Jimenez et al. 2018
7. Translational and disruptive field technologies
Rebetzke et al. 2016
Jimenez et al. 2018
Patented and licensed
Gecko
In-situ digitisation of agriculture commodities – C4 and C3 species (wheat, barley, canola, corn, sorghum,
strawberry, grapes, cotton, vegetables)
8. • Data Processing and Analysis
• Virtualisation of Analytics (e.g. global footprint)
• Visualisation of phenomics data
• Output of traits
phenoSMART : Turning data into information
Developed as a Market place to provide a collaboration platform for Oz-wide use and diffusion of new
algorithms
www.phenosmart.org.au
9. Analytics web-platform for in-field phenotyping
Single row or whole plot analysis
(e.g. in F2-enrichment applications for carrier
selection cf. Bonnett et al. 2005)
Location
Details
Metadata import
10. 10 |
High vigour, Low RUE line
Low vigour, High RUE line
Stay green/senescence and LAD
12.0 t/ha
7.2 t/ha
11. Results available
Jimenez et al. 2018
Sirault et al. (unpublished)
Validation plant height
0.955 quantile
Crop height
Dynamics of plant height (modelling)
12. Aerial thermography – rapid evaluation
of water stress at scale
Olive orchard (Spain)
(Jimenez et al)
multispectral and LiDAR
cameras
- nutrient status
High Resolution Aerial RGB –
growth dynamic (height), lodging,
ground cover, flowering (sorghum),
DEM
Translational and disruptive aerial technologies
Irrigated
Irrigated
Irrigation scheduling, Mutant population screening
>500k plots screened in 2017 Chapman et al. (2014)
13. • Works on raw frame imagery (no mosaicking)
• Hotspot thresholding algorithm (biomass cuts)
• Output to CSV file format yielding plot-by-plot statistics
Improving heritability and QTL discovery with aerial
thermography (>7-14m)
Plot value extractionPhysical units
Match experiment’s layout
Yanco, September 2012
(RIL population segregating for Transpiration
Efficiency - TE)
From h2~ 0.1 -> h2~0.7
See Chapman et al. (2014) for algorithm description
14. Application in sugar cane
Scott Chapman and Bangyou Zheng
(CSIRO Ag & Food)
Micasense RedEdge camera
Combination with other airborne-derived datasets:
multispectral imagery
Average NDVI for each interval (10cm)
of the canopy among genotypes
Canopy N profiles in sugarcane
15. Integration of UAV and crop simulation models
TotalbiomassLAIGroundcoverZadoks
Long season
1. UAV - monitor ground cover
2. Calibrate simulation model with
UAV ground cover
3. Compute seasonal changes in
biomass and water use pattern.
This is VERY hard to measure directly
0
0.2
0.4
0.6
0.8
1
-1500 -1000 -500 0 500 1000
Watersupply/demandratio Thermal time pre and post anthesis (°Cd)
Probe G
"Better G"
Chapman and Zheng | Insights for high
precision in UAV phenotyping
16. Outcomes: Improved research efficiency..?
16 |
‘Aims - Research efficiency is improved by 20-50% due to enhanced tools & analytics’
Canopy temperature heritability
(1000 field plots)
0.09
0.68
Old way (handheld IR guns)
New - ‘HeliPOD’
(100 QTL)
Plot biomass measurements
(1000 field plots)
New - ‘Phenomobile Lite’
[Key driver of new Industry investment (‘QTL Discovery and Interactions’ project)
1 person – 2 hr
5 people – 80 hr
Old way (manual cuts)
(< 5 QTL)
17. APPF – a national service facility
• APPF is a national open-access facility, funded by the Australian Commonwealth
Government (NCRIS)
• APPF aims to make plant phenotyping accessible to all Research Institutions and Agri-
business industry in Australia (national relevance but global impact)
Coordinate the delivery of services and infrastructure nation-wide via Private-Public
partnerships and via technology transfer to industry
• Expert staff with experience in plant growth, operation of phenotyping equipment,
interpretation and handling of the data to guide you along the way
The APPF staff includes plant scientists, physicists, engineers (mechatronics,
software), web developers, horticulturalists, biometricians and modellers.
• APPF provides a structure to integrate and diffuse phenotyping products and analytics
pipeline developed by Australian Research Institutions
• Projects can be run as full service or in collaboration with the researcher / client
(research hotel model)*
• Provides specialised training to Research Scientists in Academia or Industry via
immersive training programs (e.g. Seasol International, BCS, TARI, IRRI, CRIDA, etc…)
* cost recovery model applies
18. Acknowledgments
Director - Dr. Xavier Sirault
Dr Michael Schaefer, Dr Robert Coe
and team (Science application and
Logistics)
Dr Warren Creamers and team
(Software Engineering)
Mr Peter Kuffner and team
(Hardware Engineering)
Ms Marni Tebbutt and team
(Business management)
Phenomobile and phenoSMART are registered trademarks of CSIRO in one or more territories in the world.
Interested in collaborating or sending students?
Please get in touch xavier.sirault@csiro.au
HRPPC staff (a few staff missing!)
Colleagues / Collaborators:
Dr Scott Chapman,
Dr Bangyou Zheng,
Dr Christophe Pradal (INRIA),
Dr Mike Bange
Dr Greg Rebetzke
Dr Richard James
Prof Robert Furbank (ANU, CoE TP)
Prof Graeme Hammer (UQ, CoE TP)
And many students
19. CSIRO AGRICULTURE AND FOOD
Thank you
Questions?
Dr Xavier Sirault
E-mail: xavier.sirault@csiro.au
Director High Resolution Plant Phenomics Centre
Vice Chair International Plant Phenotyping Network
Research Team Leader – Automated Phenotyping and Analytics
20. 20 |
Personal Food Computer (from MIT)
-> mechatronics/mechanical engineering
-> plant biology and environment
Raspberry Pi with stereoscopic system (CSIRO / HRPPC)
-> mechatronics engineering and micro-controller programming
-> 3D printing
Education and Outreach program
PhenoSMART (CSIRO / HRPPC)
-> Computational Vision and programming
-> Functional Data Analysis / Fitting of curves…