Remote sensing –Beyond images
Mexico 14-15 December 2013
The workshop was organized by CIMMYT Global Conservation Agriculture Program (GCAP) and funded by the Bill & Melinda Gates Foundation (BMGF), the Mexican Secretariat of Agriculture, Livestock, Rural Development, Fisheries and Food (SAGARPA), the International Maize and Wheat Improvement Center (CIMMYT), CGIAR Research Program on Maize, the Cereal System Initiative for South Asia (CSISA) and the Sustainable Modernization of the Traditional Agriculture (MasAgro)
Fostering Friendships - Enhancing Social Bonds in the Classroom
New remote and proximal sensing methodologies in high throughput field phenotyping
1. New remote and proximal sensing methodologies in high
throughput field phenotyping
Remote Sensing – Beyond images. Mexico City, 2013
JOSE A. JIMENEZ-BERNI. CSIRO PLANT INDUSTRY. HIGH RESOLUTION PLANT PHENOMICS CENTRE
4. Breeder’s wish list
Table 2. Essential and desirable measurements for phenotyping of yield and other traits across
multiple experiments in water-limited and high temperature environments.
Essential (core)
Timing A
Desirable
(frequency)
Plant establishment counts
(plants/m2)
Ground cover (%)
DC 12-13 (1×)
Normalized Difference
Vegetation Index
Anthesis date
DC 12-37 (2×)
Canopy temperature (°C)
Harvest index
Spike number (spikes/m2)
Plant height (cm)
Thousand grain weight (g)
Grain yield (g/m2)
Observations and scores (e.g.
incomplete plots, temperature
damage, disease, lodging, and
shattering)
DC 12-37 (3×)
DC 45-65 (every
3d)
DC 35-70 (2×)
DC 90
DC 90
Timing A
(frequency)
Canopy light interception
(μmol/m2/s)
Normalized Difference Vegetation
Index
Early biomass/leaf area/tiller number
DC 35-60 (2×)
Carbon Isotope Discrimination
DC 30-32
Anthesis biomass (g/m2)
Water soluble carbohydrates
Canopy temperature during grain
filling (°C)
DC 60-65
DC 65-70 (1×)
DC 70+ (4×)
DC 12-60 (4×)
DC 30-32
DC 90
DC 90
DC 90
as required
A
Timing according to the Zadoks decimal code (DC) for scoring stages of cereal development
(Zadoks et al. 1974)
(More info: Rebetzke et al 2012, (http://prometheuswiki.publish.csiro.au/tiki-index.php)
5. Phenomobile
-30
30
LMS400
70
Canopy
Canopy
Canopy
• 3x LiDARs (Canopy Structure)
• 4x RGB cameras (Stereo
reconstruction)
• 1x Thermal IR camera (Canopy
temperature)
• 1x Hyperspectral line scanner
(Canopy biochemistry)
• 1x Full range spectrometer
(Canopy biochemistry)
• Removable light banks
11. Acquisition of airborne thermal images
The Airframe
Sensor Integration
The Flight Controller
12.
13. Airborne thermal mosaic – ready for plot extraction
Legend [deg C]
~600 m
“Old way” h2<0.1
“New way” h2>0.6
• Capture 3 images / second
• One pass of the field ~10 sec (3 passes
required)
• Time to image entire field ~4 min
• Ideal: Simultaneous measurements at
nearly a single point in time
15. Wireless infrared thermometers
• Zigbee standard
• Selectable sampling interval
(5min)
• 3G transmission from base
station
• Real time access from Internet
• 100 sensors built in 2011
• 400 sensors built in 2012
• 160 in a single deployment
(Narrabri / Cotton)
22. Take home messages
• There are no turnkey solutions
• Why use NDVI when you can use LiDAR for direct estimation of
ground cover, plant establishment and potentially LAI or biomass?
• Use imaging sensors when possible: extract information from the
right spot, not an integrated observation
• Airborne thermography as an alternative to traditional CT
measurements: no changes in environmental conditions and
multiple measurements per plot
• Wireless sensor networks for dynamic phenotyping applications
• Never underestimate the data management component and the
requirements for data processing
23. The Plant Phenomics Team
HRPPC / CSIRO PI:
Bob Furbank
Dave Deery
Xavier Sirault
Jose Jimenez-Berni (Berni)
Tony Condon
Scott Chapman & Ed Holland
Xueqin Wang
Alyssa Weirman
Tony Agostino
Pablo Rozas-Larraondo
Peter Kuffner & Michael Salim
Scott Kwasny
Dac Nguyen
Viri Silva Perez
Richard Poire
Kath Meacham
CSIRO E-health:
Jurgen Fripp & Antony Paproki
Olivier Salvado
CSIRO Informatics:
Ali Salehi
Doug Palmer & Alex Krumphol
David Lovell
Pascal Vallotton
Changming Sun
ANU:
Murray Badger
Susanne von Caemmerer
CIMMYT
Matthew Reynolds Team
USDA : John Vogel Team
24. Thank you / Gracias
For more information:
Jose.jimenez-berni@csiro.au
http://hrppc.org.au
CSIRO PLANT INDUSTRY / HIGH RESOLUTION PLANT PHENOMICS CENTRE
Hinweis der Redaktion
Broken red line: inclination angle of Maize leafSolid red line: height profile across plotNon-destructive wish listFractional coverCrop heightLAICanopy architectureGrowth rateEars / m2HIBiomassCrop yield
The next task is to extract the individual temperatures for Nearly 400 samples per plot
SensorDB is an unique system that enables turning big data time series into knowledge.--------------------SensorDB was originally designed for handling sensor data from wireless infrared canopy temperature like those ones in the picture. These sensors monitor continuously, 24h a day and 7 days a week, every 5 minutes, the temperature of wheat varieties in field experiments. Generally speaking, the cooler the crop the more it’s transpiring. That information is very valuable in breeding programs for selecting genotypes more suitable for dryer conditions or in irrigated farms for making decisions on the irrigation schedule. At this time resolution, that is more than 100k points per year per single sensor and we may have 60 or 100 of these sensors in each experiment. In total we have around 500 of these sensors, which makes more than 50M points a year. But we also handle other sources of information into SensorDB like data from airborne themography covering thousands of these experimental plots, or data from ground vehicles like this phenomobile, grain harvesters or just handheld devices or ipads registering data from the field. The result is a torrent of data with information that we need to turn into useful knowledge for plant scientists, breeders and farmers...
SensorDB is a multi user system. Each user has his own summary with real time statistics that have been previously calculated on data insertion.SensorDB also has a very fast and dynamic search engine so it’s easy to find, select and aggregate the data that we are looking for. We just type and SensorDB guesses what we are looking for. It’s not just sensors... Excel... We also have a number of methods for importing data manually into SensorDB, like for example importing data from Excel with just copy and paste.But probably the most interesting part of SensorDB for the end user is the analysis page where the user can select from a number of pre-built analysis or just create his own personalised analysis. Visulaisation and analysis through virtual laboratory environment – next slide
In the Analysis Page is what we call the personalized virtual laboratory. Here the user can visualize different streams of data like for instance weather summaries with the evolution of the weather during the growing season. It is also possible to select and apply filters to the data so for example when we select two different genotypes and restrict the envorironmental conditions in the analysis we can explore the response of a these genotypes to the changes in relative humidity and therefore select the most suitable crop variety for certain environments. This analysis can be done across multiple users, environments and locations and as the system has more information it is possible to explore more and more scenarios. It is also possible to share the analysis with other users, so this becomes a real collaboration tool.