THE PHENOTYPING MOBILE OPTICAL PLATFORM (PHENOMOP)

• Sponsoring Institution: National Institute of Food and Agriculture
• Project Status: COMPLETE
• Funding Source: OTHER GRANTS
• Reporting Frequency: Annual
• Accession No.: 1024456
• Grant No.: 2020-70410-32911
• Cumulative Award Amt.: $166,164.00
• Proposal No.: 2020-07840
• Project Start Date: Sep 15, 2020
• Project End Date: Sep 14, 2022
• Grant Year: 2020
• Program Code: [EGP]- Equipment Grants Program

Recipient Organization
UNIVERSITY OF CALIFORNIA, DAVIS
410 MRAK HALL
DAVIS,CA 95616-8671

Performing Department
Plant Sciences

Non Technical Summary
Population growth and climate change are putting increased pressure on plant breeding programs to develop new cultivars that contribute to improved food quality and quantity, while also being adapted to biotic and abiotic stressors (Cooper et al., 2014; Asseng et al., 2019). Recent advances in whole-genome sequencing of many crops have significantly improved plant breeding, butthe pace of methods for evaluating the phenotypes of genetic lines (genotypes, e.g., Macintosh vs. Granny Smith apple) is lagging behind(Spalding et al., 2016). Objective and accessible high-throughput phenotyping is essential for the rapid identification and quantification of desirable (or undesirable) plant traits in large, diverse breeding populations of nearly any agronomically important species.This knowledge can then be combined with advanced genetic tools to improve crop water, light and nutrient use efficiency, pathogen or pest resistance, and ultimately yield and quality.We propose a novel, efficient and highly versatile tool that will advance crop phenomics--high throughput acquisition and analysis of multidimensional phenotypes through space and time, to inform and promote fundamental advances in crop breeding research (Yang et al., 2020) --by combining advances in sensor technology, analytical tools, automation, and artificial intelligence, while harnessing UC Davis' exceptional breadth of talent in translational plant sciences and access to multiple test environments across the state. ThePhenotypingMobileOpticalPlatform(PhenoMOP) is a mobile, optical remote sensing platform that will enable automated, rapid assessment of a wide range of plant traits.The novel instrument is designed to serve plant breeders, agronomists, and horticulturalists, and overcomes shortcomings of other phenotyping platforms. Knowledge and technologies developed may be translated to additional testing and production systems with focused phenotyping platforms.

Animal Health Component

50%

Research Effort Categories

Basic

25%

Applied

50%

Developmental

25%

Classification

Knowledge Area (KA)	Subject of Investigation (SOI)	Field of Science (FOS)	Percent
102	7210	1081	50%
404	7210	2090	50%

Knowledge Area
102 - Soil, Plant, Water, Nutrient Relationships; 404 - Instrumentation and Control Systems;

Subject Of Investigation
7210 - Remote sensing equipment and technology;

Field Of Science
2090 - Statistics, econometrics, and biometrics; 1081 - Breeding;

Keywords

genotype

micrometeorology

phenotype

remote sensing

plant breeding

plant physiology

tower

Goals / Objectives
We seek toaddress the limitations of current high-througput phenotyping platforms (i.e. drones, gantry systems, tractor mounted systems)by building a mobile, tower-mounted suite of optical sensors that enable high frequency data collection of hyperspectral reflectance, lidar, thermal, and RGB imagery (for machine vision), coupled with information on environmental conditions. This tower can be moved to a location of interest by "sweeping" (like a mop) across the field to measure remotely-sensed proxies for plant structural, biochemical, and physiological traits. The PhenoMOP will be a self-contained trailer with a telescoping 30- meter tower (max) that can beeasily deployed, secured at any field location, and quickly redeployed to another location. It will include four spectrometers that will be connected via fiber optics to a 2D scanner, mounted atop the tower, that can be programmed to focus sequentially on each plot in the field (with a field of view of down to 20 cm2). The optical platform will include four highly stable spectrometers designed to measure solar-induced chlorophyll fluorescence (SIF) and the complete VIS-NIR spectrum (400-1650 nm), and a lidar instrument mounted on top of a weatherproof RGB (red, green, blue) + thermal camera. The tower will also include a weather station, a sonic anemometer and infrared gas analyzer to measure CO2and H2O exchange between the plants and the atmosphere. The flexibility of the PhenoMOP will allow us to measure 1010individual plots per day, providing information on plant traits at both diurnal and seasonal timescales. All parts are off-the-shelf, and will be assembled into a single, well-integrated phenotyping platform for use across any crop, including tree crops.

Project Methods
The PhenoMOP is a tower-based sensor platform that measures remotely sensed proxies for plant physiological, biochemical and structural traits from a telescoping 100ft tower mounted on a mobile trailer. The tower-trailer unit is mobile, includes an HVAC enclosure to keep computer and spectrometers at a stable temperature, and can be taken down and moved to a new field in a short amount of time (seetrailer-telescopesection). At the base of the tower will be four highly stable spectrometers: two which are custom built with a high spectral resolution for enabling SIF retrievals in the red and far-red spectrum, and two for hyperspectral measurements across the visible-near-infrared (400-1000nm) and near-infrared-shortwave-infrared (100nm-1650nm, seespectrometerssection). A single stainless-steel encased fiber optic will run from the base of the tower (in HVAC enclosure) to the top of the tower, where it will be placed in an optical enclosure equipped for reflected radiance and incoming irradiance measurements (seeoptical configurationsection).Spectral measurements will be co-located with an RGB + thermal digital camera on a pan-tilt unit, capable of scanning different plots (seeimaging systemsection). Atop the spectrometer enclosure will be a full-waveform lidar instrument for capturing fine scale structural changes with a cm2resolution (seelidarsection). Additionally, micrometeorological sensors and sonic anemometers coupled to an infrared gas analyzer will be placed at an appropriate height on the tower to make measurements of ecosystem CO2and H2O fluxes, as well as air temperature, relative humidity, precipitation, and radiation fluxes (seeflux and micrometeorologysection). In addition to being able to operate the instrument remotely, all data will be collected, stored, and processed instantaneously in the cloud (seedata collection and processingsection).?Spectrometers?Four thermally stabilized commercial spectrometers will be purchased from Ocean Optics, Inc., Florida, USA (two QEPro spectrometers, one VIS-NIR Flame spectrometer, and one NIR-SWIR Flame Spectrometer. The two QEPro spectrometers?(henceforth referred to as ' red' and ' far-red') cover a SIF retrieval wavelength range at high spectral resolution (red: λ = 670-732 nm (prototype) λ = 650-712 nm (final version), full width half maximum (FWHM) = 0.3 nm; Far-red: λ = 729-784 nm, FWHM = 0.3 nm), which encompass the two fluorescence emission peaks around 685 and 740 nm. The Flame spectrometers of the PhotoSpec prototype provides moderate resolution spectra (λ = 339-1022 nm and 1022-1650nm, 1.2nm FWHM) in order to retrieve vegetation indices (Fig. 6). The arrangement of the optical bench of the spectrometers is based on the crossed Czerny-Turner principle and the spectrum is measured by a CCD detector. The QEPro spectrometers are equipped with a 2400 groove/mm grating and a back-thinned Hamamatsu S7031-1006 detector with 1044 pixels, whereas the Flame spectrometers have a 600 groove/mm grating and a Sony ILX511B linear silicon CCD array with 2048 pixels. The detectors in the QEPro spectrometers are typically kept at − 10°C to reduce dark current. The red and far-red QEPro spectrometers are equipped with an OG590 and a RG695 optical long- pass filters, respectively, and the Flame spectrometer with a 25 μm entrance slit and an order sorting filter at the detector. The advantage of having four spectrometers is twofold: 1) it is cheaper than purchasing a 400-2500 nm field spectrometer; 2) each spectrometer can be customized to achieve the quality we need for making robust remote sensing measurements.Optical configurationThe linear ends of the quad-furcated fiber bundle (4-1) serve as entrance slits for four spectrometers (Fiber Optic Systems, Simi Valley, CA). The FWHM of 0.3 nm is determined by the fiber width of 0.06 mm, which acts as an entrance slit. The quad-furcated fiber will connect via an SMA connector to a single fiber optic that runs to the top of the tower. At the top of the tower is a weatherproof enclosure with a telescoping lens and clear window. A rotating opal diffuser can be programmed via a roto-motor to hover over the top of the telescope for capturing incident irradiance measurements necessary for retrievals of vegetation reflectance and SIF (Fig. 7, bottom right). The resulting optical data will result in a 10-40 cm footprint on the plant canopy, depending on distance viewing angle.Imaging systemAll imaging components will be co-located and stacked on top of each other (left panel, Fig. 7). We will mount atop the tower an AXIS Q872-E scanning visible and thermal camera, which is designed for outdoor conditions and has an internal cooling system. Software will be written to control the camera remotely and permit scanning routines to return to field locations of interest. The precision of the motors isextremely accurate (0.1 degrees), and we will align the camera with the spectral field of view, to provide a 'visual' of what the spectrometers are 'seeing' (top right, Fig. 7). The added advantage of the thermal imaging capabilities will enable estimates of canopy temperature, allowing us to infer canopy temperature depression, and evapotranspiration from individual plots, normalized against a 'reference' temperature in the field.LidarThe OS-2 lidar unit from Ouster (San Francisco, CA, USA) offers industry-leading combination of range, resolution, reliability, all-weather durability, size, weight and power. It is the smallest and lightest long-range high-resolution lidar on the market and can be directly integrated with the rest of the optical equipment. It has a 240 m range at 80% reflectivity and 120 m at a 10% reflectivity. The range accuracy is 0.3 cm, with a rotation rate of 10 or 20 Hz at a wavelength of 865 nm and a beam divergence of 0.09-degree FWHM. The lidar unit can capture 0.65 to 2.0 million points per second and records the range, intensity, azimuth angle and timestamp of the measurement. Waveforms are processed internally and further processed in our custom software (see below).Flux and micrometeorologyAll micrometeorology and flux measurements will come from the EddyFlux system (LI-7900EF) and basic biomet sensor package (7900-109) from LI-COR (Lincoln, NE, USA). These will include measurements of carbon and water fluxes via a sonic anemometer and a closed-path infrared gas analyzer. This will be coupled with information on environmental conditions including incoming and reflected shortwave radiation, precipitation, relative humidity, and air temperature. These measurements will integrate over individual field plots, providing ecosystem-scale CO2and H2O fluxes and environmental conditions to inform diurnal and seasonal patterns in abiotic factors that might influence plant trait response.Data collection and processingWe have an open source software written in Python for similar systems we have built over the last five years (Grossmann et al., 2018; Magney et al., 2019a). We will modify the code with processing routines previously developed by PI Magney to integrate the lidar, RGB, and thermal data. All data will be uploaded onto the cloud following data collected, will be processed and provide a summary PDF of figures that summarize results from individual plots on a daily basis. Further, because we will be able to operate the instrument remotely via a GoToMyPC connection and mobile hotspot, allowing us to troubleshoot on the fly and change scanning patterns and retrieval methods remotely. The software and processing pipeline will be made publicly available via a GitHub repository.

Progress 09/15/20 to 09/14/22

Outputs
Target Audience:The target audience is plant breeders. We have used the equipment we built using these funds to inform breeding practices at UC Davis in common bean, poplars, and almonds. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?We were able to invovle several members of the lab in the build, maintaince and deployment of PhenoMOP. Christopher Wong, postdoc at UC Davis, is the current instrument lead, and is active in all field deployments. We have also trained 1) Francis Ulep, 2) Marie Klien, and 3) Erica Edwards on the equipment and they are running it in their respective experiments. How have the results been disseminated to communities of interest?Yes, in addition to the publications mentioned previously, we have presented the equiptment at several field days in beans (2021 and 2022) as well as outreach meetings to groups like SeedCentral and the Seed Biotechnology Center at UC Davis. What do you plan to do during the next reporting period to accomplish the goals? Nothing Reported

Impacts
What was accomplished under these goals? We were able to achieve all of these goals. We have collected two seasons of field data in common bean, one in almond, and one in poplar during the reporting period. We have submitted a paper describing the instrument (which was accepted pending minor revisions). We have submitted a second paper showing how hyperspectral advances could be used to phenotype drought resilience in bean (currently under review). We have future plans to use the equipment in other experiments in different crops with several of the Co-Is in the upcoming season. The equipment is all working well, and we have a processing pipleine that enables many researchers to use the data we collect.

Publications

• Type: Journal Articles Status: Under Review Year Published: 2022 Citation: Wong, C.Y.S., Jones, T., McHugh, D., Gilbert, M., Gepts, P., Palkovic, A., Buckley, T., Magney, T.S. PhenoSpec: A mobile tower-based remote sensing system for continuous high-throughput phenotyping of vegetation physiology. Plant Methods. In revisions.
• Type: Journal Articles Status: Under Review Year Published: 2022 Citation: Wong, C.Y.S., Gilbert, N., Pierce, M., Parker, T., Palkovic, A, Gepts, P., Magney, T.S., Buckley, T. Hyperspectral remote sensing for phenotyping physiological drought response of common & tepary bean. Plant Phenomics. Under review
• Type: Conference Papers and Presentations Status: Published Year Published: 2021 Citation: Wong, C., Buckley, T., Gilbert, M., Gepts, P., Pierce, M., Palkovic, A., Jones, T., Magney, T.S. Bean physiology under drought stress: High-throughput field phenotyping using tower-based spectral reflectance and solar-induced fluorescence (SIF). American Geophysical Union Fall Meeting 2021. New Orleans, LA. Dec. 12 2021

Progress 09/15/20 to 09/14/21

Outputs
Target Audience:Scientists looking for high-throuoput phenotyping tools. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?Because this was an equiptment grant, there was not any funding available for technical support or student training. However, I was able to fund one technician, Bobby Arlen, to help with construction v 1.0 of the PhenoMOP.Through the projects listed above, in both grape and beans, one M.S. student, Devin McHugh and one postdoc, Chris YS Wong, were trained. Both individuals are funded on those specific projects but have been trained on running and operating PhenoMOP. How have the results been disseminated to communities of interest?Research results are being submitted for publication (see publications section) and presented at scientific conferences (American Geophysical Union in Dec. 2021, both by Devin McHugh and Chris Wong). The PhenoMOP was also shown at two field days, in both bean and poplar phenotype/genotype experiments at UC Davis. What do you plan to do during the next reporting period to accomplish the goals?Next year, we have to finish assembly of the PhenoMOP. This includes, installation of the short-wave infrared spectrometer, the lidar system, and the eddy covaraince flux and micrometerology measurments. In year 1, we wanted to make sure all of the base components were running smoothly. We expect the final instrument to be running by spring 2022; however, the PhenoMOP in it's current state is fullly functional and has been accomplishinig it's goals in the first two experiments outlined above. In summer 2022, we will also work more closely with Dr. Malikka Nocco to look at phenotyping in Almond Orchards, as well as Drs. Taylor and Buckley and phenotying drougght stress accross large poulations of beans and poplars.

Impacts
What was accomplished under these goals? In year 1 of the project,weaddressed the limitations of current high-througput phenotyping platforms (i.e. drones, gantry systems, tractor mounted systems)by building a mobile, tower-mounted suite of optical sensors that enable high frequency data collection of hyperspectral reflectance, thermal, and RGB imagery (for machine vision), coupled with information on environmental conditions. This tower can be moved to a location of interest by "sweeping" (like a mop) across the field to measure remotely-sensed proxies for plant structural, biochemical, and physiological traits. The PhenoMOP is ina self-contained trailer with a telescoping 30- meter tower (max) that can beeasily deployed, secured at any field location, and quickly redeployed to another location. The primary goals of year 1 were to get the custom software, hypspectral reflectance, thermal imagery,scanning mechanisms, and tower assmebly constructed. We were able to accompish this, and next summer we will focus on integration of the lidar instrument, as well as the gas analyzer and micrometeorology measurements. Additionally, we will add on the fourth spectrometer, extending into the short-wave infrared. In spring/summer 2021, we were able to deploy PhenoMOP in two different experiments: 1) in a water-stress experiment in grapevine and 2) a high-throughput phenotyping experiment in bean. In the first experiemnt, the PhenoMOP made measurements of four different blocks of grape, all under different water regimes and with coincident flux tower measurements.This research providedinsight to further understand how plants respond to variations in environmental stress, such as drought, through their spectral responses. Weexplore d the spefic differences in their individual performance and abilities to capture grape vineyard drought in California's Central Valley. Data analysis is still under way, but preliminary results suggest that SIF was able to capture the onset of water stress well before tranditional vegetation indices, such as NDVI./ In the second experiment, we aimed to utilize tower-based remote sensing for the high-throughput phenotyping of 320 bean genotypes under control and drought treatments. We measurerd300 bean and 20teparybean genotypes in control (irrigated) and drought (terminal drought) field treatment plots. The PhenoMOPmonitored 672 targets representing 175 genotypes. We also had three intensive campaigns with physiology measurements:Stomatal conductance (gs) at midday (12h), leaf water potential: predawn and midday. From this analysis, we found that: 1)Hyperspectral data used in PLSR models can be used to predict the variation of plant traits over time, during drought stress, and across genotypes; 2)Using different principal components that differentially emphasizes specific wavebands, stomatal conductance and predawn and midday leaf water potential can be predicted from the canopy scale; 3)Vegetation indices can be applied to track variation of vegetation; 4)NDVI is sensitive to vegetation structure and greenness tracking developmental growth and senescence; 5)PRI is sensitive to xanthophyll cycle and carotenoid:chlorophyll pool ratios tracking short-term and long-term dynamics in response to environment and drought stress; 6)SIF represents light energy reemissions, a highly dynamic process for regulating light energy usage and dissipation showing high variation over time and between genotypes.Our results highlights potential applications for a tower-based remote sensing instrument for continuous and automated monitoring of vegetation for high-throughput phenotyping of drought resilience across bean genotypes

Publications

• Type: Journal Articles Status: Under Review Year Published: 2021 Citation: Wong, C.Y.S., Bambach, N., Alsina, M., McElrone, A., Jones, T., Buckley, T., Kustas, W., Magney T.S. Detecting short-term stress and recovery events in a vineyard using tower-based remote sensing of photochemical reflectance index (PRI). Irrigation Science.
• Type: Journal Articles Status: Submitted Year Published: 2021 Citation: Wong, C.Y.S., McHugh, D., Jones, T., Buckley, T., Magney, T.S. Tower-based remote sensing enables continuous high-throughput phenotyping of vegetation physiology. New Phytologist
• Type: Conference Papers and Presentations Status: Accepted Year Published: 2021 Citation: Wong, C.Y.S., Buckley, T.N., Gilbert, M., Gepts, P., Pierce, M., Palkovic, M, Jones, T., Magney, T.S. Bean physiology under drought stress: High-throughput field phenotyping usingtower-based spectral reflectance and solar-induced fluorescence (SIF). American Geophysical Union Fall Meeting. Dec. 12 2021
• Type: Conference Papers and Presentations Status: Accepted Year Published: 2021 Citation: McHugh, D., Wong, C.Y.S., Bambach, N.E., Kustas, W.P., Mar Alsina, M., McElrone, A.J., Magney, T.S. Solar-induced chlorophyll fluorescence as a ground-based and remotely-sensed physiological indicator of grapevine stress. American Geophysical Union Fall Meeting. Dec. 12 2021.