Donald Danforth Plant Science Center, Missouri

The Automated controlled-environment phenotyping at the Donald Danforth Plant Science Center Bellwether Foundation Phenotyping Facility

The Bellwether Foundation Phenotyping Facility is a climate controlled 70 m2 growth house with a conveyor belt system for moving plants to and from fluorescence, color, and near infrared imaging cabinets. This automated, high-throughput platform allows repeated non-destructive time-series image capture and multi-parametric analysis of 1,140 plants in a single experiment. You can read more about the Danforth Plant Sciences Center Bellwether Foundation Phenotyping Facility on the DDPSC website.

The Scanalyzer 3D platform at the Bellwether Foundation Phenotyping Facility at the Donald Danforth Plant Science Center consists of multiple digital imaging chambers connected to the Conviron growth house by a conveyor belt system, resulting in a continuous imaging loop. Plants are imaged from the top and/or multiple sides, followed by digital construction of images for analysis.

• RGB imaging allows visualization and quantification of plant color and structural morphology, such as leaf area, stem diameter and plant height.

• NIR imaging enables visualization of water distribution in plants in the near infrared spectrum of 900–1700 nm.

• Fluorescent imaging uses red light excitation to visualize chlorophyll fluorescence between 680 – 900 nm. The system is equipped with a dark adaptation tunnel preceding the fluorescent imaging chamber, allowing the analysis of photosystem II efficiency.

The LemnaTec software suite is used to program and control the Scanalyzer platform, analyze the digital images and mine resulting data. Data and images are saved and stored on a secure server for further review or reanalysis.

Experiments LT1A (TM015) and LT1B (TM016)

Duration: 10 days on LemnaTec platform

Experimental Design:

• 3 replicates of 190 BAP lines were grown in a randomized complete block design

• Watering regimes = 30% FC and 100% FC

• Drought conditions were imposed 10 days after planting

• Plants were imaged daily for 10 days (11-20 DAP) and sampled at 20 days after planting

• Experiment was repeated twice to phenotype the full BAP (Reps 1A and 1B)

About this book

This book describes the TERRA-REF data collection, computing, and analysis pipelines. The following links provide quick access to:

About TERRA-REF

The ARPA-E-funded Transportation Energy Resources from Renewable Agriculture Phenotyping Reference Platform (TERRA-REF) program aims to transform plant breeding by using remote sensing to quantify plant traits such as plant architecture, carbon uptake, tissue chemistry, water use, and other features to predict the yield potential and stress resistance of 300+ diverse Sorghum lines.

The data storage and computing system provides researchers with access to the reference phenotyping data and analytics resources using a high performance computing environment. The reference phenotyping data includes direct measurements and sensor observations, derived plant phenotypes, and genetic and genomic data.

Our objectives are to ensure that the software and data in the reference data and computing pipeline are interoperable, reusable, extensible, and understandable. Providing clear definitions of common formats will make it easier to analyze and exchange data and results.

Users

This documentation is intended to enable a wide variety of user-groups to use and contribute to the datasets and tools contained in the TERRA-REF platform, including:

Contact Info and More

• Email:

  • David LeBauer (Data / Computing Lead): dlebauer@arizona.edu

  • Nadia Shakoor (Project Director): nshakoor@danforthcenter.org

  • Todd Mockler (Principal Investigator): tmockler@danforthcenter.org

Combining advanced sensing with novel analytical approaches to accelerate breeding

The ARPA-E-funded Transportation Energy Resources from Renewable Agriculture Phenotyping Reference Platform (TERRA-REF) program aims to transform plant breeding by using remote sensing to quantify plant traits such as plant architecture, carbon uptake, tissue chemistry, water use, and other features to predict the yield potential and stress resistance of 400+ diverse Sorghum lines.

Conducting high-throughput phenotyping to connect discoveries to field performance

Breeding is currently limited by the speed at which phenotypes can be measured, and the information that can be extracted from these measurements. Current instruments used to quantify plant traits do not scale to the thousands or tens of thousands of individual plants that need to be evaluated in a breeding program. The TERRA-REF field scanner system scans over 1 acre of plants, collecting thousands of daily measurements throughout the growing season that are used to determine plant phenotypes and inform breeding decisions.

The field level phenotypic data combined with the genomic data is helping us to identify the differences between each line and the reference genome sequence for sorghum. We are using bioinformatics and quantitative genetics to characterize the observed genetic variation and identify genomic regions controlling biomass, plant architecture, and photosynthetic traits.

Using large-scale genome sequencing to drive phenotype-genotype associations and gene discovery

There is enormous potential for sorghum crop improvement. There are 50,000 sorghum accessions in the U.S. germplasm collection and most are unused and unstudied. TERRA-REF is analyzing a sorghum bioenergy association panel (BAP) that includes diverse sweet and biomass lines from all five sorghum races. The BAP captures geographic, racial, and genomic diversity.

TERRA-REF has already sequenced 384 of the lines with an average sequence coverage of 20x per line. Genome-wide association studies (GWAS) are now underway.

Providing reference quality data as a community resource

TERRA-REF is developing a data storage and computing system that provides researchers with access to all of the ‘raw’ data and derived plant phenotypes (traits). Data from sensors at a variety of locations across the US will be transferred to one location.

The reference data will facilitate data sharing and re-use of data by providing metadata, provenance for derived data sets, and standardized data processing workflows. It will include geospatial infrastructure for efficiently querying and transforming key datasets and tools that enable researchers to access, archive, use, and contribute data products. The technical documentation for this data pipeline is detailed in this book.

The data storage and computing system provides researchers with access to the reference phenotyping data and analytics resources using a high performance computing environment. The reference phenotyping data includes direct measurements and sensor observations, derived plant phenotypes, and genetic and genomic data.

Our objectives are to ensure that the software and data in the reference data and computing pipeline are interoperable, reusable, extensible, and understandable. Providing clear definitions of common formats will make it easier to analyze and exchange data and results.

Overview

Phenotyping

The TERRA-REF project is collecting phenotype data from 852 sorghum genotypes grown in three locations:

• Maricopa Agricultural Center (MAC), Maricopa, Arizona

• Kansas State University (KSU), Ashland, KS

• Donald Danforth Plant Science Center, Missouri

Genotyping

Whole genome resequencing is being carried out on ~400 sorghum accessions to understand the landscape of genetic variation in the selected germplasm and enable high-resolution mapping of bioenergy traits with genome wide association studies (GWAS). Additionally, ~200 sorghum recombinant inbred lines (RILs) will be characterized with ~400,000 genetic markers using genotyping-by-sequencing (Morris et al., 2013) for trait dissection in the RIL population and testcross hybrids of the RIL population.

Whole-genome resequencing

Experimental Design:

• 384 BAP samples were sequenced to an average depth of ~25x.
• Shotgun sequencing (127-bp paired-end) was done using an Illumina X10 instrument at the HudsonAlpha Institute for Biotechnology.

• Data were aligned against the BTx623 reference genotype

Genotyping-by-sequencing

Experimental Design:

See Also

The Maricopa field site is located at the the University of Arizona Maricopa Agricultural Center and USDA Arid Land Research Station in Maricopa, Arizona.

Overview of Experiments During the TERRA REF project

(1) IL = Recombinant Inbred Lines, BAP = Bioenergy Association Panel, SAP = Sorghum Association Panel (2). Uniformity Trial = same lines planted in strips across field (3) In season 9 a second field ‘North’ was added, and separate trials were conducted

Season 1 sorghum (April - July 2016)

Three hundred thirty one lines were planted in Season 1.

Planting maps

• Under the scanner system

Planting Design

Under scanner system

Overview

This user manual is divided into the following sections:

What data is available?

• Raw output from sensors deployed on the Lemnatec field scanner

  • Additional data from greenhouse systems, UAVs, and tractors have not been released, but can be accessed through our beta user program

• Manually-collected fieldbooks and associated protocols

• Derived data, including phenomics data, from computational approaches

• Genomic pipeline data

Environment conditions data is collected using the Vaisala CO2, Thies Clima weather sensors as well as lightning, irrigation, and weather data collected at the Maricopa site.

Data formats follow the Climate and Forecast (CF) conventions for variable names and units. Environmental data are stored in the Geostreams database.

Data sources

• WeatherStation coordinates are 33.074457 N, 111.975163 W

• EnvironmentLogger is on top of the gantry system and is moveable.

• Irrigation is managed at the field level. There are four regions that can be irrigated at different rates.

Data access

Level 1 data

Level 1 meteorological data is aggregated to from 1 Hz raw data to 5 minute averages or sums.

netCDF: 5s (12 per minute) observations

On Globus or Workbench you can find these data provided in both hourly and daily files. These files contain data at the original temporal resolution of 1/s. In addition, they contain the high resolution spectral radiometer data.

sites/ua-mac/Level_1/envlog_netcdf

• hourly files: YYYY-MM-DD_HH-MM-SS_environmentallogger.nc

• daily files: envlog_netcdf_L1_ua-mac_YYYY-MM-DD.nc

Geostreams: 5 minute observations

Data can be accessed using the geostreams API or the PEcAn meteorological workflow. These are illustrated in the sensor data tutorials.

Here is the json representation of a single five-minute observation:

Data can be accessed using the geostreams API or the PEcAn meteorological workflow.

These are illustrated in the sensor data tutorials.

Here is the json representation of a single five-minute observation from Geostreams:

[
   {
      "geometry":{
         "type":"Point",
         "coordinates":[
            33.0745666667,
            -111.9750833333,
            0
         ]
      },
      "start_time":"2016-08-30T00:06:24-07:00",
      "type":"Feature",
      "end_time":"2016-08-30T00:10:00-07:00",
      "properties":{
         "precipitation_rate":0.0,
         "wind_speed":1.6207870370370374,
         "surface_downwelling_shortwave_flux_in_air":0.0,
         "northward_wind":0.07488770951583902,
         "relative_humidity":26.18560185185185,
         "air_temperature":300.17606481481516,
         "eastward_wind":1.571286062845733,
         "surface_downwelling_photosynthetic_photon_flux_in_air":0.0
      }
   },

Variable names and units

Raw Data

Data is available via Globus or Workbench:

• /ua-mac/raw_data/co2sensor

• /ua-mac/raw_data/EnvironmentLogger

• /ua-mac/raw_data/irrigation

• /ua-mac/raw_data/lightning

• /ua-mac/raw_data/weather

Sensor information:

• Vaisala CO2 Sensor collection

• Thies Clima Sensor collection

Computational pipeline

Environmental Logger

• Description: EnvironmentalLogger raw files are converted to netCDF.

Known Issues

Known issue: the irrigation data stream does not currently handle variable irrigation rates within the field. Specifically, we have not yet accounted for the Summer 2017 drought experiments. See terraref/reference-data#196 for more information.

When the full field is irrigated (as is typical), the irrigated area is 5466.1 m2 (=215.2 m x 25.4 m)

In 2017:

• Full field irrigated area from the start of the season to August 1 (103 dap) is 5466.1 m2 (=215.2 m x 25.4 m).

• Well-watered treatment zones from August 1 to 15 (103 to 116 dap): 2513.5 m2 (=215.2 m x 11.68 m) in total, combined areas of non-contiguous blocks

• Well-watered treatment zones from August 15 - 30 (116 to 131 dap): 3169.9 m2 (=215.2 m x 14.73 m), again in total as the combined areas of non-contiguous blocks

Github Issues

• computing-pipeline#252 Developmet discussions

• Downwelling Spectral radiometer calculations reference-data #30

https://terraref.github.io/tutorials/accessing-trait-data-in-r.html

The following table lists available TERRA-REF data products. The table will be updated as new datasets are released. Links are provided to pages with detailed information about each data product including sensor descriptions, algorithm (extractor) information, protocols, and data access instructions.

See also

• Sensor calibration

• Fieldbooks and Protocols

• Data standards

• Geospatial information

You can access genomics data in one of the following locations:

The data is structured on both the TERRA-REF strorage (accessible via Globus and Workbench) and CyVerse Data Store infrastructures as follows:

Whole-genome resequencing

Raw data

Derived data

Data derived from analysis of the raw resequencing data at the Danforth Center (version1) are available as gzipped, genotyped variant call format (gVCF) files and the final combined hapmap file.

Genotyping-by-sequencing (GBS)

Raw data

Derived data

Combined genotype calls are available in VCF format.

KSU Genomics Pipeline / GBS/RIL analysis

Raw Data

• genomics/raw_data/ril/gbs

  • H5JYFBCXY_1_fastq.txt

  • H5JYFBCXY_2_fastq.txt

  • Key_ril_terra

Derived Data

• genomics/derived_data/ril/gbs/kansas_state/version1/imp_TERRA_RIL_SNP.vcf

Summary

Fluorescence intensity data is collected using the PSII camera.

Each measurement produces 102 bin files. The first (index 0) is an image taken right before the LED are switched on (dark reference). Frame 1 to 100 are the 100 images taken, with the LEDs on. In binary file 102 (index 101) is a list with the timestamps of each frame of the 100 frames.

Right now the LED on timespan is 1s thus the first 50 frames are taken with LEDs on the latter 50 frames with LED off.

Raw data access

Fluorescence intensity data is available via Clowder and Globus:

• Globus paths:

  • /raw_data/ps2top

  • /Level_1/ps2_png/

See also

Herritt, Matthew T., et al. "Chlorophyll fluorescence imaging captures photochemical efficiency of grain sorghum (Sorghum bicolor) in a field setting." Plant Methods 16.1 (2020): 1-13.

Summary

Hyperspectral imaging data is collected using the Headwall VNIR and SWIR sensors. In the Nov 2017 Beta Release only VNIR data is provided because we do not have the measurements of downwelling spectral radiation required by the pipeline.

Please see the README hyperspectral pipeline README for more information about how the data are generated and known issues.

Hyperspectral Algorithm and pipeline

See Hyperspectral extractor

Data Access

Raw Data

Raw data is available in the filesystem, accessible via Globus in the following directories:

• VNIR: /sites/ua-mac/raw_data/VNIR

• SWIR: /sites/ua-mac/raw_data/SWIR

These files are uncalibrated; see the hyperspectral pipeline repository for information on how these can be processed.

Level 1 data access

Hyperspectral data is available via Clowder, Globus #Terraref endpoint, the TERRA REF Workbench, and our THREDDS server:

• Clowder:

  • VNIR Hyperspectral NetCDFs

  • SWIR Collection: Level 1 data not available

• Globus and Workbench:

  • VNIR: /sites/ua-mac/Level_1/vnir_netcdf

  • SWIR: Level 1 data not available

• Sensor information:

  • Headwall SWIR

  • Headwall VNIR

For details about using this data via Clowder or Globus, please see Data Access section.

Level 2 data access

Level 2 data are spectral indices computed at the same resolution as Level 1. These can be found in the same Level 1 directories as their parents, but the files are appended *_ind.nc.

To get a list of hyperspectral indices currently generated https://terraref.ncsa.illinois.edu/bety/api/v1/variables?type=Reflectance%20Index

traits::

Level 3 data access

Hyperspectral Indices

The following indices are computed and provided as both Level 2 data at full spatial resolution and as Level 3 (plot level) means.

Citations can be found Morris, Geoffrey P., Davina H. Rhodes, Zachary Brenton, Punna Ramu, Vinayan Madhumal Thayil, Santosh Deshpande, C. Thomas Hash et al. "Dissecting genome-wide association signals for loss-of-function phenotypes in sorghum flavonoid pigmentation traits." G3: Genes, Genomes, Genetics 3, no. 11 (2013): 2085-2094.

See also

• Sensor calibration

• Hyperspectral data pipeline

• Geospatial information

Summary

Infrared heat imaging data is collected collected using the FLIR SC615 thermal sensor. These data are provided as geotiff image raster files as well as plot level means.

• Algorithms are in the flir2tif directory of the Multispectral extractor repository; see the readme for details.

• Sensor information: FLIR Thermal Camera collection

Level 1 Data Access

Filesystem (Globus and Workbench)

• ua-mac/Level_1/ir_geotiff

Clowder

To be created https://github.com/terraref/computing-pipeline/issues/391

Level 3 Data

Plot level summaries are named 'surface_temperature' in the trait database. In the future this name will be used for the Level 1 data as well. This name from the Climate Forecast (CF) conventions, and is used instead of 'canopy_temperature' for two reasons: First, because we do not (currently) filter soil in this pipeline. Second, because the CF definition of surface_temperature distinguishes the surface from the medium: "The surface temperature is the temperature at the interface, not the bulk temperature of the medium above or below." http://cfconventions.org/Data/cf-standard-names/48/build/cf-standard-name-table.html

Raw data access

Thermal imaging data is available via Clowder and Globus:

/ua-mac/raw_data/flirIrCamera

For details about using this data via Clowder or Globus, please see Data Access section.

Known Issues

• Data are unavailable for Season 4 (summer 2017 sorghum) and season 5 (winter 2017-2018 wheat).

  • Work to recover these data is ongoing; see terraref/reference-data#190

  • Problem description terraref/reference-data#182

See also

• Geospatial information

Meteorological data will use Climate Forecasting 'standard names' and 'canonical units' conventions. CF is widely used in climate, meteorology, and earth sciences.

Here are some examples (note that we can change from canonical units to match the appropriate scale, e.g. "C" instead of "K"; time can use any base time and time step (e.g. hours since 2015-01-01 00:00:00 UTC, etc. But the time zone has to be UTC, where 12:00:00 is approx (+/- 15 min). solar noon at Greenwich.

• standard_name is CF-convention standard names (except irrigation)

• units can be converted by udunits, so these can vary (e.g. the time denominator may change with time frequency of inputs)

Running The Pipeline

Before the Running

The pipepline is developed in Python, so a Python Interpreter is a must. Other than the basic Python standard librarys, the following third-party libraries are required:

• netCDF4 for Python

• numpy

Other than official CPython interpreter, Pypy is also welcomed, but please make sure that these third-party modules are correctly installed for the target interpreter. The pipeline can only works in Python 2.X versions (2.7 recommended) since numpy does not support Python 3.X versions.

Cloning from the Git:

git clone https://github.com/terraref/computing-pipeline.git
cd  computing-pipeline/scripts/environmental_logger
git checkout master

The extractor for this pipeline is developed and maintained by Max in branch "EnvironmentalLogger-extractor" under the same repository.

Get the Environmental Logger Pipeline to Work

To trigger the pipeline, use the following command:

python ${environmental_logger_source_path}/environmental_logger_json2netcdf.py ${input_JSON_file} ${output_netCDF_file}

Where:

• ${environmental_logger_source_path} is where the three environmental_logger files are located

• ${input_JSON_file} is where the input JSON files are located

• ${output_netCDF_file} is where the users want the pipeline to export the product (netCDF file)

Please note that the parameter for the output file can be a path to either a directory or a file, and it is not necessarily to be existed. If the output is a path to a folder, the final product will be in this folder as a netCDF file that has the same name as the imported JSON file but with a different filename extension (.nc for standard netCDF file); if this path does not exist, environmental_logger pipeline will automatically make one.

Calculation

The calculation in the Environmental Logger is mainly finished by the module environmental_logger_calculation.py under the support of numpy.

LemnaTec Scanalyzer 3D platform at the Donald Danforth Plant Science Center

Phenotype data is derived from images generated by the indoor LemnaTec Scanalyzer 3D platform at the Donald Danforth Plant Science Center using PlantCV. PlantCV is an image analysis package for plant phenotyping. PlantCV is composed of modular functions in order to be applicable to a variety of plant types and imaging systems. PlantCV contains base functions that are required to examine images from an excitation imaging fluorometer (PSII), visible spectrum camera (VIS), and near-infrared camera (NIR). PlantCV is a fully open source project: https://github.com/danforthcenter/plantcv. For more information, see:

Project website: http://plantcv.danforthcenter.org

Full documentation: http://plantcv.readthedocs.io/en/latest

Publications:

• https://doi.org/10.7717/peerj.4088

• https://doi.org/10.1016/j.molp.2015.06.005

To learn more about PlantCV, you can find examples in the terraref/tutorials repository, which is accessible on GitHub and in the TERRA REF workbench under tutorials/plantcv

• an ipython notebook demonstration of PlantCV plantcv/plantcv_jupyter_demo.ipynb.

For the TERRA-REF project, a PlantCV Clowder extractor was developed to analyze data from the Bellwether Foundation Phenotyping Facility at the Donald Danforth Plant Science Center. Resulting phenotype data is stored in BETYdb.

Computational pipeline

PlantCV extractor

• Description: Processes VIS/NIR images captured at several angles to generate trait metadata. The trait metadata is associated with the source images in Clowder, and uploaded to the configured BETYdb instance.

• Output CSV: /sites/danforth/Level_1/<experiment name>

Input

• Evaluation is triggered whenever a file is added to a dataset

• Following images must be found

  • 2x NIR side-view = NIR_SV_0, NIR_SV_90

  • 1x NIR top-view = NIR_TV

  • 2x VIS side-view = VIS_SV_0, VIS_SV_90

  • 1x VIS top-view = VIS_TV

• Per-image metadata in Clowder is required for BETYdb submission; this is how barcode/genotype/treatment/timestamp are determined.

Output

• Each image will have new metadata appended in Clowder including measures like height, area, perimeter, and longest_axis

• Average traits for the dataset (3 VIS or 3 NIR images) are inserted into a CSV file and added to the Clowder dataset

• If configured, the CSV will also be sent to BETYdb

Data access

Level 1

• BETYdb: https://terraref.ncsa.illinois.edu/bety

For details about accessing BETYdb, please see Data Access section and a tutorial on accessing phenotypes from the trait database on the TERRA REF Workbench in traits/04-danforth-indoor-phenotyping-facility.Rmd.

Raw Data

• Clowder: Bellwether Phenotyping Facility Space

• Globus and Workbench:

  • /sites/danforth/raw_data/<experiment name>

Summary

3D point cloud data is collected using the Fraunhofer 3D laser scanner. Custom software installed at Maricopa converts .png output to the .ply point clouds. The .ply point clouds are converted to georeferenced .las files using the 3D point cloud extractor

Level 1 data products are provided in both .las and .ply formats. Raw sensor output (PLY) is converted to LAS format using the ply2las extractor; ply2las extractor code is available on GitHub.

For each scan, there are two .ply files representing two lasers, one on the left and the other on the right. These are combined in the .las files.

Sensor information

• Fraunhofer 3D scanner collection

For details about using this data via Clowder or Globus, please see Data Access section.

Data access

Data is available via Clowder, Globus, and Workbench.

• Clowder: Laser Scanner 3D LAS

• Globus or Workbench File System:

  • LAS /raw_data/laser3D_las

  • PLY /raw_data/laser3D

Known issues

• The position of the lasers is affected by temperature. We added a correction for temperature to adjust for this effect. See terraref/reference-data#161

Several different sensors include geospatial information in the dataset metadata describing the location of the sensor at the time of capture.

Coordinate reference systems The Scanalyzer system itself does not have a reliable GPS unit on the sensor box. There are 3 different coordinate systems that occur in the data:

• Most common is EPSG:4326 (WGS84) USDA coordinates

• Tractor planting & sensor data is in UTM Zone 12

• Sensor position information is captured relative to the southeast corner of the Scanalyzer system in meters

EPSG:4326 coordinates for the four corners of the Scanalyzer system (bound by the rails above) are as follows:

• NW: 33° 04.592' N, -111° 58.505' W

• NE: 33° 04.591' N, -111° 58.487' W

• SW: 33° 04.474' N, -111° 58.505' W

• SE: 33° 04.470' N, -111° 58.485' W

In the trait database, this site is named the "MAC Field Scanner Field" and its bounding polygon is "POLYGON ((-111.9747967 33.0764953 358.682, -111.9747966 33.0745228 358.675, -111.9750963 33.074485715 358.62, -111.9750964 33.0764584 358.638, -111.9747967 33.0764953 358.682))"

Scanalyzer coordinates Finally, the Scanalyzer coordinate system is right-handed - the origin is in the SE corner, X increases going from south to north, and Y increases from east to the west.

In offset meter measurements from the southeast corner of the Scanalyzer system, the extent of possible motion for the sensor box is defined as:

• NW: (207.3, 22.135, 5.5)

• SE: (3.8, 0, 0)

Scanalyzer -> EPSG:4326 1. Calculate the UTM position of known SE corner point 2. Calculate the UTM position of the target point, using SE point as reference 3. Get EPSG:4326 position based on UTM

MAC coordinates Tractor planting data and tractor sensor data will use UTM Zone 12.

Scanalyzer -> MAC Given a Scanalyzer(x,y), the MAC(x,y) in UTM zone 12 is calculated using the linear transformation formula:

ay = 3659974.971; by = 1.0002; cy = 0.0078;
ax = 409012.2032; bx = 0.009; cx = - 0.9986;
Mx = ax + bx * Gx + cx * Gy
My = ay + by * Gx + cy * Gy

Assume Gx = -Gx', where Gx' is the Scanalyzer X coordinate.

MAC -> Scanalyzer

Gx = ( (My/cy - ay/cy) - (Mx/cx - ax/cx) ) / (by/cy - bx/cx)
Gy = ( (My/by - ay/by) - (Mx/bx - ax/bx) ) / (cy/by - cx/bx)

MAC -> EPSG:4326 USDA We do a linear shifting to convert MAC coordinates in to EPSG:4326 USDA

Latitude: Uy = My - 0.000015258894
Longitude: Ux = Mx + 0.000020308287

Sensors with geospatial metadata

• stereoTop

• flirIr

• co2

• cropCircle

• PRI

• scanner3dTop

• NDVI

• PS2

• SWIR

• VNIR

Available data All listed sensors

"gantry_system_variable_metadata": {
      "time": "08/17/2016 11:23:14",
      "position x [m]": "207.013",
      "position y [m]": "3.003",
      "position z [m]": "0.68",
      "speed x [m/s]": "0",
      "speed y [m/s]": "0.33",
      "speed z [m/s]": "0",
      "camera box light 1 is on": "True",
      "camera box light 2 is on": "True",
      "camera box light 3 is on": "True",
      "camera box light 4 is on": "True",
      "y end pos [m]": "22.135",
      "y set velocity [m/s]": "0.33",
      "y set acceleration [m/s^2]": "0.1",
      "y set decceleration [m/s^2]": "0.1"
    },

stereoTop

"sensor_fixed_metadata": {
      "cameras alignment": "cameras optical axis parallel to XAxis, perpendicular to ground",
      "optics focus setting (both)": "2.5m",
      "optics apperture setting (both)": "6.7",
      "location in gantry system": "camera box, facing ground",
      "location in camera box x [m]": "0.877",
      "location in camera box y [m]": "2.276",
      "location in camera box z [m]": "0.578",
      "field of view at 2m in X- Y- direction [m]": "[1.857 1.246]",
      "bounding Box [m]": "[1.857     1.246]",
    },

cropCircle

"sensor_fixed_metadata": {
      "location in gantry system": "camera box, facing ground",
      "location in camera box x [m]": "0.480",
      "location in camera box y [m]": "1.920",
      "location in camera box z [m]": "0.6",
    },

co2Sensor

"sensor_fixed_metadata": {
      "location in gantry system": "camera box, facing ground",
      "location in camera box x [m]": "0.35",
      "location in camera box y [m]": "2.62",
      "location in camera box z [m]": "0.7",
    },

flirIrCamera

"sensor_fixed_metadata": {
      "location in gantry system": "camera box, facing ground",
      "location in camera box x [m]": "0.877",
      "location in camera box y [m]": "1.361",
      "location in camera box z [m]": "0.520",
      "field of view x [m]": "1.496",
      "field of view y [m]": "1.105",
    },

ndviSensor

"sensor_fixed_metadata": {
      "location in gantry system": "top of gantry, facing up, camera box, facing ground",
      "location in camera box x [m]": "0.33",
      "location in camera box y [m]": "2.50",
    },

priSensor

"sensor_fixed_metadata": {
      "location in gantry system": "top of gantry, facing up, camera box, facing ground",
      "location in camera box x [m]": "0.400",
      "location in camera box y [m]": "2.470",
    },

SWIR

"sensor_fixed_metadata": {
      "location in gantry system": "camera box, facing ground",
      "location in camera box x [m]": "0.877",
      "location in camera box y [m]": "2.325",
      "location in camera box z [m]": "0.635",
      "field of view y [m]": "0.75",
      "optics focal length [mm]": "25",
      "optics focus apperture": "2.0",
    },

field scanner plots

There are 864 (54*16) plots in total and the plot layout is described in the plot plan table.

The boundary of each plot changes slightly each planting season. The scanalyzer coordinates of each plot are transformed into the (EPSG:4326) USDA coordinates using the equations above.

Public Domain Data

Data will be released with a CC0 license, meaning that they are in the public domain. The CC0 license allows wide use of these data and while it does not legally bind users to acknowledge the source data users are expected cite our data and research in publications, presentations, and other products.

Data are in the public domain to enable broad and unrestricted re-use. However, any derived pubications should cite the dataset:

Citing TERRA REF data:

Software and Algorithms

Unpublished Data

We plan to make data from the Transportation Energy Resources from Renewable Agriculture Phenotyping Reference Platform (TERRA-REF) project available for use with attribution.

Timing of Future Data Releases

Planned future releases include Sorghum Season 9, data from experiments at Kansas State University, and data from the Danforth Indoor Phenotyping Facility.

Additional seasons can be requested as needed. We can provide the raw data and software required to process it. We can also collaborate with you to process the data, but this will typically require new funding sources.

Contacts

• Todd Mockler, Project/Genomics Lead (email: tmockler@danforthcenter.org)

• David LeBauer, Computing Pipeline Lead (email: dlebauer@arizona.edu)

• Nadia Shakoor, Project Director (email: nshakoor@danforthcenter.org)

Summary

The willingness of many scientists to cooperate and collaborate is what makes TERRA REF possible. Because the platform encompasses a diverse group of people and relies on many data contributors to create datasets for analysis, writing scientific papers can be more challenging than with more traditional projects. We have attempted to lay out ground rules to establish a fair process for establishing authorship, and to be inclusive while not diluting the value of authorship on a manuscript. Please engage with the TERRA REF manuscript writing process knowing you are helping to forge a new model of doing collaborative scientific research.

Copyright, Attribution, and Conditions of Use:

We are making data available early for users under the condition that manuscripts led within the team not be scooped. In these cases, people who wish to use the data for publication prior to official open release date of November 2018 should coordinate co-authorship with the person responsible for collecting the data.

Overview of the TERRA REF authorship process:

Inclusive but not gratuitous

Our primary goals in the TERRA REF authorship process are to consistently, accurately and transparently attribute the contribution of each author on the paper, to encourage participation in manuscripts by interested scientists, and to ensure that each author has made sufficient contribution to the paper to warrant authorship.

Steps:

1. Read these authorship policies and guidelines.

3. Proposed ideas are reviewed by the authorship committee primarily to facilitate appropriate collaborations, identify potential duplication of effort, and to support the scientists who generate data while allowing the broader research community access to data as quickly and openly as possible. The authorship committee may suggest altering or combining analyses and papers to resolve issues of overlap.

4. Circulate your draft analysis and manuscript to solicit Opt-In authorship.

   For global analyses, the lead author should circulate the manuscript to the Network by submitting a email to the TERRA REF team.

   For analyses of more limited scope, the lead author should circulate the manuscript to network collaborators who have indicated interest at the abstract stage, those who have contributed data, and any others who the lead author deems appropriate.

   In both cases, the subject line of the email should include the phrase "OPT-IN PAPER"; This email should also include a deadline by which time co-authors should respond.

   The right point to share your working draft and solicit co-authors is different for each manuscript, but in general:

   1. sharing early drafts or figures allows for more effective co-author contribution. While ideally this would mean circulating the manuscript at a very early stage for opt-in to the entire network, it is acceptable and even typical to share early drafts or figures among a smaller group of core authors.

   2. circulating essentially complete manuscripts does not allow the opportunity for meaningful contribution from co-authors, and is discouraged.

5. Potential co-authors should signal their intention to opt-in by responding by email to the lead author before the stated deadline.

6. Potential co-authors should inform the lead author of any additional candidates for co-authorship who should be considered.

7. Lead authors should is responsible for making sure that any who have made contributions warranting co-authorship have actively opted in or out (authors should not be excluded due to a missed email or a misunderstanding of the scope of the manuscript and their contributions). The goal is to ensure that the author list is inclusive and consistent.

8. Lead authors should keep an email list of co-authors and communicate regularly about progress including sharing drafts of analyses, figures, and text as often as is productive and practical.

9. Lead authors should circulate complete drafts among co-authors and consider comments and changes. Given the wide variety of ideas and suggestions provided on each TERRA REF paper, co-authors should recognize the final decisions belong to the lead author.

10. Final manuscripts should be reviewed and approved by each co-author before submission.

11. All authors and co-authors should fill out their contribution in the authorship rubric and attach it as supplementary material to any TERRA REF manuscript. Lead authors are responsible for ensuring consistency in credit given for contributions, and may alter co-author's entries in the table to do so.

    Note that the last author position may be appropriate to assign in some cases. For example, this would be appropriate for advisors of lead authors who are graduate students or postdocs and for papers that two people worked very closely to produce.

12. The lead author should carefully review the authorship contribution table to ensure that all authors have contributed at a level that warrants authorship and that contributions are consistently attributed among authors. The lead author should also ensure that all contributions that warrant co-authorship.

    • Has each author made contributions in at least two areas in the authorship rubric?

    • Did each author provide thoughtful, detailed feedback on the manuscript?

    • Have all qualified contributors actively opted in or out of co-authorship?

Authors are encouraged to contact the TERRA REF PI (Mockler) or authorship committee (Jeff White, Geoff Morris, Todd Mockler, David LeBauer, Wasit Wulamu, Nadia Shakoor) about any confusion or conflicts.

Co-authorship

Authorship must be earned through a substantial contribution. Traditionally, project initiation and framing, data analysis and interpretation, software or algorithm development, and manuscript preparation are all authorship-worthy contributions, and remain so for TERRA REF manuscripts. However, TERRA REF collaborators have also agreed that collaborators who lead a site from which data are being used in a paper can also opt-in as co-authors, under the following conditions: (1) the collaborators' site has contributed data being used in the paper's analysis; and (2) that this collaborator makes additional contributions to the particular manuscript, including data analysis, writing, or editing. For co-authorship on opt-out papers, each individual must be able to check at least two boxes in the rubric in addition to contribution to the writing process. These guidelines apply equally to manuscripts led by graduate students.

Co-author Expectations

Each author is expected to meet all of the following conditions:

• Substantial contributions to conception and design, acquisition of data, or analysis and interpretation of data, and

• Drafting the article or revising it critically for important intellectual content, and

• Final approval of the version to be published, and

• Agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Types of Author Contributions

Author Order

By default, we will follow the conventions of the scientific community that is the target audience of the journal in which the article is published. This should typically follow:

• First is lead author

• Last is the supervisor of the lead author.

• if > 1 lead or senior authors these will be listed first and last, respectively, and identified in the author contributions section of the acknowledgements.

• All other contributors are listed alphabetically.

Publications committee

Members: David LeBauer, Todd Mockler, Geoff Morris, Duke Pauli, Nadia Shakoor, Wasit Wulamu

The publications committee ensures communication across projects to avoid overlap of manuscripts, works to provide guidance on procedures and authorship guidelines, and serves as the body of last resort for resolution of authorship disputes within the Network.

Acknowledgments

Please use the following text in the acknowledgments of TERRA REF manuscripts:

Keywords

Please use "TERRA REF"; as one of your keywords on submitted manuscripts, so that TERRA REF work is easily indexed and searchable.

Overview

TERRA-REF data can be accessed through many different interfaces: Globus, Clowder, BETYdb, CyVerse, and CoGe. Raw data is transfered to the primary compute pipeline using Globus Online. Data is ingested into Clowder to support exploratory analysis. The Clowder extractor system is used to transform the data and create derived data products, which are either available via Clowder or published to specialized services, such as BETYdb.

Tutorials (Recommended!)

We have developed tutorials to provide users with both 'quick start' vignettes and more detailed introductions to TERRA REF datasets. Tutorials for accessing trait data, sensor data, and genomics data are organized by directory ("traits", "sensors", and "genomics").

The tutorials assume familiarity with or willingness to learn Python and / or R, and provide the greatest flexibility and access to available data.

Globus: Browse and Transfer Files

Raw data is transferred to the primary TERRA-REF file system at the National Center for Computing Applications at the University of Illinois.

Use Globus Online when you want to transfer data from the TERRA-REF system for local analysis.

Transferring data using Globus Connect:

To access data via Globus, you must first have a Globus account and endpoint.

6. Select source

   • Endpoint: #Terraref

   • Path: Navigate to the subdirectory that you want.

   • Select (click) a folder

   • Select (highlight) files that you want to download at destination

   • Select the endpoint that you set up above of your local computer or server

   • Select the destination folder (e.g. /~/Downloads/)

7. Click 'go'

8. Files will be transfered to your computer

Requesting Access to unpublished data in TERRA-REF BETYdb:

To request access to unpublished data, send your Globus id to David LeBauer (dlebauer@email.arizona.edu) with 'TERRAREF Globus Access Request' in the subject.

1. fill out the terraref.org/beta user form

2. email dlebauer@email.arizona.edu with your globusid to request access.

BETYdb: Trait Data and Agronomic Metadata

BETYdb contains the derived trait data with plot locations and other information associated with agronomic experimental design.

Accessing data in R

Requesting Access to unpublished data in TERRA-REF BETYdb:

3. email dlebauer@email.arizona.edu for your account to be approved.

Using SQL and PostGIS with Docker (Advanced Users)

The fastest and most comprehensive way to access the database using SQL and other database interfaces (such as the R package dplyr interface described below, or GIS programs described in . You can run an instance of the database using docker, as described below

This is how you can access the TERRA REF trait database. It requires that you install the Docker software on your computer.

psql

R

GIS software

Clowder: Sensor Data and Metadata Browser

Data organization in Clowder

Data is organized into spaces, collections, and datasets, collections.

• Spaces contain collections and datasets. TERRA-REF uses one space for each of the phenotyping platforms.

• Collections consist of one or more datasets. TERRA-REF collections are organized by acquisition date and sensor. Users can also create their own collections.

• Datasets consist of one or more files with associated metadata collected by one sensor at one time point. Users can annotate, download, and use these sensor datasets.

Requesting Access to unpublished data in Clowder:

3. email dlebauer@email.arizona.edu for your account to be approved.

CyVerse: Genomics Data

TERRA-REF genomics data is accessible on the CyVerse Data Store and Discovery Environment. Accessing data through the CyVerse Discovery Environment requires signing up for a free CyVerse account. The Discovery Environment gives users access to software and computing resources, so this method has the advantage that TERRA-REF data can be utilized directly without the need to copy the data elsewhere.

You can also find these in the CyVerse discovery environment in the TERRA-REF Community Data folder: /iplant/home/shared/terraref.

CoGe: Genomics Data

The following protocols have been contributed by TERRA-REF team members:

• Field Scanner - Coming 2017

• Genomics - Coming 2017

The TERRA hyperspectral data pipeline processes imagery from hyperspectral camera, and ancillary metadata. The pipeline converts the "raw" ENVI-format imagery into netCDF4/HDF5 format with (currently) lossless compression that reduces their size by ~20%. The pipeline also adds suitable ancillary metadata to make the netCDF image files truly self-describing. At the end of the pipeline, the files are typically [ready for xxx]/[uploaded to yyy]/[zzz].

Installation

Software dependencies

Pipeline source code

Once the pre-requisite libraries above have been installed, the pipeline itself may be installed by checking-out the TERRAREF computing-pipeline repository. The relevant scripts for hyperspectral imagery are:

Setup

The pipeline works with input from any location (directories, files, or stdin). Supply the raw image filename(s) (e.g., meat_raw), and the pipeline derives the ancillary filename(s) from this (e.g., meat_raw.hdr, meat_metadata.json). When specifying a directory without a specifice filename, the pipeline processes all files with the suffix "_raw".

shmkdir ~/terrarefcd ~/terrarefgit clone git@github.com:terraref/computing-pipeline.gitgit clone git@github.com:terraref/documentation.git

Run the Hyperspectral Pipeline

shterraref.sh -i ${DATA}/terraref/foo_raw -O ${DATA}/terrarefterraref.sh -I /projects/arpae/terraref/raw_data/lemnatec_field -O /projects/arpae/terraref/outputs/lemnatec_field

The Maricopa field site is located at the the University of Arizona Maricopa Agricultural Center and USDA Arid Land Research Station in Maricopa, Arizona. At this site, we have deployed the following phenotyping platforms.

Twelve sensors are attached to the gantry system. Detailed information for each sensor including name, variable measured, and field of view are described in the table below, with links to more detailed specifications.

Abstract

Automated VIS and NIR imaging in a controlled growth environment

Materials

• ProMix BRK20 + 14-14-14 Osmocote pots

• pre-filled by Hummert Sorghum seed

Equipment

• Conviron Growth House

• LemnaTec moving field conveyor belt system

• Scanalyzer 3D platform

Procedures

Planting

• Plant directly into phenotyping pots 

Chamber Conditions

Pre-growth (11 days) and Phenotying (11 days)

• 14 hour photoperiod

• 32oC day/22oC night temperature

• 60% relative humidity

• 700 umol/m2/s light

Watering Conditions

• Prior to phenotyping, plants watered daily

• The first night after loading, plants watered 1× by treatment group to 100% field capacity (fc)

• Days 2 – 12, plants watered 2× daily by treatment group (100% or 30% FC) to target weight

Automation

• Left shift lane rotation within each GH, during overnight watering jobs

• VIS (TV and 2 x SV), NIR (TV and 2 x SV) imaging daily

Recipes

• Field capacity = 200% GWC (200 g water/100 g soil), based upon extensive GWC testing done by Skyler Mitchell

• Target weight (fc) = [(water weight at % fc) + [(average weight of carrier/saucer) + (dry soil weight) + (pot weight)]

• Water weight at 100% fc = dry soil weight * (%GWC/100)

• Water weight at 30% fc = water weight at 100% fc * 0.30

References

This section describes sensor calibration processes and how to access additional information about specific calibration protocols, calibration targets, and associated reference data.

LemnaTec Field Scanalyzer

Calibration protocols

Calibration targets

The following calibration targets are available:

• Aluminum 3D test object

Sensor Calibration

Environmental sensor calibration

The environmental sensor has been calibrated by LemnaTec. The output of the spectrometer is raw counts, users will need to use the calibration files to convert to units of µW m-2 s-1, taking into account the bandwidth of the chip (0.4nm) if converting to µmol m-2 s-1.

Hyperspectral calibration

Sources:

For the SWIR and VNIR sensors, factory calibration is repeated each year using the calibration lamp provided by Headwall. To convert the hyperspectral exposure image to reflectance requires the wavelength-dependent, factory calibrated reflectance of the spectralon at all VNIR and SWIR wavelengths and a good image of a spectralon panel from each camera. This includes periodic measurements of a white spectralon reflectance panel run with 20ms exposure to match panel calibration.

Dark reference measurement:

• VNIR

  • Dark measurement for VNIR camera is taken at exposure times 20, 25, 30, 35, 40, 45, 50, 55ms.

  • Data is in the same hypercube format with 180-200 lines, 955 bands, and 1600 pixel samples.

  • Measurement was done using Headwall software, so there is no LemnaTec json file.

  • The name of the folder is the exposure time. "current setting exposure" is showing the exposure time in ms.

  • Custom workflow to process the calibration files.

• SWIR;

  • Dark counts handled internally, so no calibration files are necessary.

White reference measurement:

• VNIR

  • White measurement for VNIR camera is taken at exposure times 20, 25, 30, 25, 40, 25, 50, 55 ms.

  • The name of the folder is the exposure time. Data are 1600 sample, 955 bands and 268-298 lines. White reference is located in the lines between 60 to 100 and in the samples between 600 to 1000.

  • The white reference scans was done at around 1pm ( one hour after solar noon). I don’t see the saturation with 20ms and 25ms exposure time.

  • For the calibration, this needs to be subtracted from the dark current in the same sample, band and exposure time.

  • In the following file, I stored an extra file named "CorrectedWhite_raw". This file includes only a single white pixel( one line, one sample) in 955 bands for each exposure time. Data is stored in the similar format but it doesnot include any extra files like frameIndex, image, header ,..

    https:\/\/drive.google.com\/file\/d\/0ByXIACImwxA7dVNHa3pTYkFjdWc\/view?usp=sharing

    Let me know if you have issue with opening the files.

Stereo 3D height scanner

LemnaTec applied calibration matrix to the 3D scanners.

UAV calibration

• There are calibrated reference panels and blackbody images taken with UAV sensors before and\/or after the each flight mission.

• There are also 4 white,grey and black panels laid on the ground during the flight. Knowing the proprieties of these targets would helps us radiometrically correct the UAV images.

• What are the reflectance properties of calibrated reference panels for multispectral camera?

• What are the thermal properties of reference target for thermal camera?

• What are the reflectance properties of the reference panels laid on the ground during the flight?

• Is there any other ground truth data collected during the flight for aerial data processing, such as surface reflectance, temperature and other environmental data? These type of data would be helpful for further atmospheric correction.

• There are two sets of reference reflectance panels: one that PDS uses, it is small, PDS will need to provide the specs; the second set consists of 4 8m x 8m canvas tarps, nominally 4%, 8%, 48% and 64% reflectance across vnir bands.

• We have data from an ASD spectrometer on many but not all flight days that can be used to give the most accurate actual reflectances for each. Kelly Thorp can provide the numbers. The tarps are old and the dark targets are more reflective than nominal and light targets darker than nominal.

• The thermal target is a passive black body, I dont know the surface emissivity, it is around 0.97. There are thermistors in the back of the metal plate to provide physical temperature of the body. The black body is stored in a wood box, insulated, to dampen thermal variations. Id guess it is accurate to 2C.

• There is a met station on farm for air temperature, humidity, wind speed, wind direction, solar radiation. we have a sun photometer that can be used for atmospheric water vapor content but currently dont deploy it routinely.

Halogen spectrum

Spectral response data

Relative spectral response data is available for the following sensors:

• NDVI

• PRI

• PAR

Calibration data

Abstract

Materials

• barcode scanning protractor

• barcode scanning ruler

• digital caliper

• drying oven

• forage chopper

• hand shears

• infrared thermometer

• juice extractor

• leaf punch

• meter stick

• paper bags

• scale

• spray paint

Equipment

Procedures

References

Photosynthesis / leaf chemistry from hyperspectral data references:

This section includes the following:

• Data Standards

• Directory Structure

• Data Backup

• Data Transfer

• Data Processing Pipeline

• Data Product Creation

• Sensor Calibration