Edited by: Yanbo Huang, United States Department of Agriculture (USDA), United States
Reviewed by: Sebastien Christian Carpentier, Bioversity International, Belgium; Quan Qiu, Beijing Research Center of Intelligent Equipment for Agriculture, China
This article was submitted to Technical Advances in Plant Science, a section of the journal Frontiers in Plant Science
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
In order to evaluate the impact of water deficit in field conditions, researchers or breeders must set up large experiment networks in very restrictive field environments. Experience shows that half of the field trials are not relevant because of climatic conditions that do not allow the stress scenario to be tested. The PhénoField® platform is the first field based infrastructure in the European Union to ensure protection against rainfall for a large number of plots, coupled with the non-invasive acquisition of crops’ phenotype. In this paper, we will highlight the PhénoField® production capability using data from 2017-wheat trial. The innovative approach of the PhénoField® platform consists in the use of automatic irrigating rainout shelters coupled with high throughput field phenotyping to complete conventional phenotyping and micrometeorological densified measurements. Firstly, to test various abiotic stresses, automatic mobile rainout shelters allow fine management of fertilization or irrigation by driving daily the intensity and period of the application of the desired limiting factor on the evaluated crop. This management is based on micro-meteorological measurements coupled with a simulation of a carbon, water and nitrogen crop budget. Furthermore, as high-throughput plant-phenotyping under controlled conditions is well advanced, comparable evaluation in field conditions is enabled through phenotyping gantries equipped with various optical sensors. This approach, giving access to either similar or innovative variables compared manual measurements, is moreover distinguished by its capacity for dynamic analysis. Thus, the interactions between genotypes and the environment can be deciphered and better detailed since this gives access not only to the environmental data but also to plant responses to limiting hydric and nitrogen conditions. Further data analyses provide access to the curve parameters of various indicator kinetics, all the more integrative and relevant of plant behavior under stressful conditions. All these specificities of the PhénoField® platform open the way to the improvement of various categories of crop models, the fine characterization of variety behavior throughout the growth cycle and the evaluation of particular sensors better suited to a specific research question.
The last three decades have witnessed a decline in the growth of yield trends (
While genomic capacity encountered a breakthrough in 2010, phenotyping capacity has become the major limitation in breeding programs aimed at building genotypes that maintain or increase crop performance under climate changes and reduced inputs (
The French Plant Phenotyping Network
The PhénoField® platform is management by the applied research institute ARVALIS and is part of the PHENOME-EMPHASIS/France project. It is an original field phenotyping platform enabling the design of a large range of drought and nutrient availability scenarios and the fine characterization of crop functioning as a response to these abiotic stresses. This accurate monitoring of both growing conditions and crop growth in the field is a key to improving the analysis of genetics × environment interactions and to identifying genotypic markers associated with favorable crop behavior. To this end, the PhénoField® platform manages a moving rainout shelter and irrigation systems that allow the application of different field drought conditions (since 2015), all the while coupled with environmental sensors to control drought stress environments. PhénoField® uses high-throughput phenotyping technologies set (validated and innovative sensors) on an automated gantry (since 2017), allowing frequent and non-invasive high-resolution measurements of the canopy. Its location at Ouzouer-le-Marché/Beauce la Romaine (41), central France, makes PhénoField® representative of irrigated crop farms of the Beauce area with the capability of studying large genotype panels of various species (bread wheat, durum wheat, corn, etc.).
2017 was the 1st year offering advanced capability on PhénoField®. During this crop season, PhénoField® carried out a bread wheat field trial in the framework of the BREEDWHEAT project
This paper presents the PhénoField® phenotyping platform. We first evaluate its capacity to control crop-growing conditions and potential biases due to the presence of mobile shelters. Related to this, a set of tools and procedures have been assessed to finely monitor and record weather data and soil water status; then, the high throughput phenotyping system is described. It includes automated sensing tools and the related data processing methods.
As an example here, results obtained during the 2017 BREEDWHEAT experiment have been analyzed to answer two questions: Is the platform able to generate the desired abiotic-stress scenarios? How is the phenotyping system able to reveal differentiated bread wheat behaviors amongst water deficit conditions, nitrogen deficit conditions or a studied genotypic panel?
The PhénoField® platform is located at Ouzouer-le-Marché/Beauce la Romaine (41) in Beauce region, one of France’s most productive agricultural areas hosting a wide variety of cultivated crops (
Location of PhénoField® platform in France, near Orléans, and aerial view of the 8 rainout shelters and gantries around wheat field trial (
Each rainout shelter covers 655 m2 (about 25 m × 25 m), and is equipped with an automaton controlling its movement. The central controller is linked to a rain contact sensor and sonic anemometer to, respectively, trigger the movement of the rainout shelters and secure the infrastructure (in case of strong wind). Each of the 8 rainout shelters is seated on three 150 m-long rails in order to move them from a garage position (when it does not rain
Four crop growth areas to ensure a correct crop rotation every year (switch between green area and yellow area) and to avoid the effects of drop shadows on the crop trial (at least 35 m).
The 8 rainout shelters protect 384 m × 6 m field trial microplots (1 m × 6 m;
The eight rainout shelters are equipped with their own individual irrigation networks, allowing precise management of the water supply in protected plots. Two booms per rainout shelter allow up to 16 different irrigation modalities. Hence PhénoField® can perform between 1 controlled irrigation modality on 384 microplots to up to 16 modalities of 24 microplots. Using the rainfall area (when rainout shelter parking is on area no. 1 or no. 4), PhénoField® allows a 768-microplot field trial. Roof gutters collect the rain water used for irrigation. This whole infrastructure allows crops to be subjected to a pre-determined duration of water stress at any desired period of their cycle.
The management of crop water stress implies precise soil characterization established with measurements of soil resistivity and water holding capacity (WHC) on the entire PhénoField® platform.
The electrical resistivity of the soil is a physical quantity related to the soil’s intrinsic characteristics (clay content, texture, water content, depth, etc.), with the higher values representing soil resistance to current flow. This magnitude of soil resistivity is measurable at high resolution and allows, for some types of soil, to extrapolate geographically located measures of water holding capacity (
The soil observed on the agricultural field is Beauce clay loam with a loamy clay texture on calcareous Beauce rock. Samples have been taken to determine the WHC/cm of the different types of soil layers based on granulometric analyses. Spatialization of soil layer thicknesses was performed at the beginning of platform construction by using the 1,100 pits opened for pouring concrete pads, from 0 to 1.5 m for each pit. In more detail, we measured four kinds of soil thicknesses: LA which corresponds to plowed horizon, S which is cambic horizon, C1 and C2 which correspond to calcaric material (C1 is cryoturbed limestone and C2 is sandy calcaric material). Based on these data combined with pF data for each horizons characterized, soil mapping was generated with krigeage models providing soil layer thicknesses but also WHC estimation at every point of the platform. As microplots are georeferenced, an estimation of the WHC was performed for each of them by computing the mean of the WHC points contained in the corresponding area. Data management was operated by PostgreSQL software, a relational database management system extended with PostGIS software to add support for geographic objects.
Evaluation of the shelters’ capacity to efficiently intercept rainfall was evaluated in 2017 by measuring precipitation along transects of crops protected by the shelters. Seven pluviometers were installed at equivalent distances (6 m) with five installed under the area protected by rainout shelters and two others installed on each side of this protected area. Pluviometers were positioned between each microplot-line.
Possible side effects on photosynthetically active radiation (PAR) and temperature were also assessed in 2017 by using, respectively, two quantum sensors (SKP215, Campbell Scientific) and two thermocouples (T109, Campbell Scientific). One of each sensor was set up in the center of the area protected from rain and the other one outside the protected area. Comparisons of air temperature and PAR inside and outside the shelter-protected area were performed by measuring the cumulative PAR and degree-day over the period of crop protection. The cumulative daily light was calculated as the sum of PAR received each day by the crop (in μmol.m–2) and the cumulative degree-day as the mean of maximum and minimum daily temperatures added up over the day with the 0 value corresponding to the 1st day of rain interception.
The meteorological conditions on the PhénoField® platform are monitored by a weather station measuring air temperature, atmospheric pressure, diffuse radiation, relative humidity, wind speed and direction in 15-min steps. The soil humidity and soil water tension at 30, 60, and 90 cm deep are recorded in control plots under each rainout shelter.
Irrigation management was realized using Irrinov® method (
It was therefore necessary to have other kind of sensors to measure soil humidity. To do that, soil humidity was measured with Time Domain Reflectometry probes (TDR-TRIME-PICO 64) installed at 30, 60, and 90 cm deep in control plots under each shelter. This type of probe has the advantage of being buried for 5–10 years without being moved. To position them deep in the soil it was necessary to make small trenches and TDR probes are known to be very sensitive to their immediate environment (air, ground contact with the pins, pebbles, etc.) so they must have been calibrated with a series of five gravimetric measurements performed every 2 months. Gravimetric water content was determined by measuring the weight of freshly collected soil (near each probe) and a soil sample oven-dried at 110°C over 48 h (see
In addition to measurements by the probes on control plots, soil nitrogen content was measured following a colorimetric method using a KCL extraction on samples taken before sowing, at the end of winter and after harvesting.
Agro-meteorological conditions are incorporated into a dynamic crop model (called ‘CHN’) used to estimate crop growth, manage crop practices and evaluate crop responses to water and nitrogen shortage (
In 2016–2017, a bread winter wheat field trial was conducted for the BREEDWHEAT project. It aimed to evaluate 22 varieties, mutual to other field experiments and known for their diversity of responses to different stresses, especially differing in behavior to nitrogen- and water-stressed conditions. Using six of the eight rainout shelters from PhénoField®, it was implemented with a double split-plot design in order to group water management treatments under rainout shelters and two nitrogen fertilization levels per rainout shelter (one per span) (see
The two water management treatments consisted of:
Well Water conditions (called “WW”) without rain interception and good irrigation practices (following the IRRINOV® method and CHN model).
Water Deficient conditions (called “WD”) with the interception of rainfall in the period between the first node and grain filling growth stages (from 22nd February to the 25th June) and irrigation occurring only to allow nitrogen uptake from fertilizer.
Each water management treatments was applied to 3 shelters and separated per span so as to have two nitrogen levels:
With optimum nitrogen supply (receiving a total 132 kg N.ha–1, called “N+”).
Without N supply (called ‘N0’).
The 22 bread winter wheat varieties were randomized under each span to evaluate their agronomic performances under these 4 modalities with 3 biological replicates (22 varieties × N+/N0 × WW/WD × 3 replicates). One control variety, APACHE, was triplicated in order (i) to perform destructive measurements, (ii) to grow above the soil tensiometers and TDR probes and (iii) to measure non-destructive variables and yield components.
The sowing was performed on 2016 October 20th and the harvest occurred on 2017 July 11th for the WD and 2017 July 18th for the WW due to differences in maturity stages. Good agricultural practices in plant protection were performed to avoid weeds, pest and disease effects on the trial. Agronomic traits were measured on each microplot:
Phenology: sowing date, emergence date, heading date, flowering date, harvest date.
Yield components: plant density (plants.m–2), spike density (spikes.m–2), dry matter grain yield (GY, t.ha–1), thousand Kernel Weight (TKW, g), grain protein content (P, %). Nitrogen Grain Quantity (Nabs, Kg.ha–1) was calculated using GY*P/5.7.
On check plot, above ground biomass and nitrogen content (based on the Dumas combustion method) were measured at flowering stage and also at maturity stage, distinguishing straw and grain to measure harvest index and nitrogen harvest index.
Statistical analyses were conducted using R studio software version 3.4.4 (
A set of eight fully automated phenotyping gantries were installed over the moving rainout shelter rails in order to acquire frequent crop canopy measurements via remote sensors, thus ensuring non-invasive measurements and the collection of a large amount of phenotyping data. Each 25 m wide gantry is able to lift a payload at a 6 m height, allowing data acquisition on any type of crop, even tall maize cultivars. These data are obtained with the sensors installed on a high throughput phenotyping bay, mounted on the gantries during experimental campaigns. It allows smoothed screening from 0.1 to 3 m.s–1 and centimetric controlled repositioning of canopy sizing from 0 to 3 m. Each sensor head can carry a set of sensors with no limit of power consumption and up to 150 kg. New sensor installation is possible thanks to its payload capacity and its agile interfacing. An open robotic operating system (ROS;
Two identical phenotyping bays are currently used on the 8 gantries to carry several types of optical sensors. The position of the sensors was optimized in order to spatially sample the area of interest and allow intra-plot borders removal. Each sensor bay had 2 measuring viewpoints: an optical head at the vertical of the vegetation (nadir) and an angular view positioned at 45°. The two sensor bays also included 4 xenon flashes to allow active measurement and standardization of daily radiation acquisition. Flashes are distributed on the vertical and inclined bays to ensure good illumination homogeneity over the camera and spectrometric field of view. An ultrasonic actuator coupled to the robotized gantries was used to estimate the height of the crop canopy and automatically set up distance to target in a closed loop control. The available sensors were (
Bay carrying the sensors with 2 angles of view. The shapes show the 2 LIDARs (in red), the 4 cameras (in purple), the 4 spectroradiometers (in yellow), the flashes (in blue) and the telemeters (in green).
RGB industrial cameras (VLG40c, Baumer, Ger; 2044 pixels * 2044 pixels for 28° optical aperture) to ensure the measurement of the fraction cover, green fraction, green plant area index and average leaf index. The resulting fields of view in the object plane measure 60 cm*60 cm corresponding to a resolution of 0.29 mm per pixel at a 1.5 m distance. The typical configuration for wheat is a set of 3 RGB cameras (two cameras viewing at 0° from vertical and one at 45°).
A VIS-NIR spectroradiometer (MMS1, ZEISS, and Ger) with a measurement range of 380–1,100 nm covered by 256 pixels feed by a large core optical fiber of numerical aperture 0.2. The resulting full field of view at a sensing distance of 1.50 m is 60 cm. It allows the quantification of the light reflected by the crop canopy and the biochemical composition of plants via vegetation indices traits. The typical configuration for wheat is 3 spectroradiometers (two sensors viewing at 0° and one at 45°).
LiDARS (LMS 400-1000, Sick, and Ger.) scan at 650 nm with detection ranging from 70 to 300 cm. This sensor allows the characterization of the 3D structure of the canopy and the estimation of plants height. The acquisition is continuous for a given microplot with a scanning frequency of 290 Hz and an angular step of 0.2°. The resulting transversal and longitudinal resolutions are, respectively, of 5 mm and 1 mm for a scanning speed of 0.3 m.s–1 at a sensing distance of 1.5 m. The typical configuration for wheat is 2 LiDARS (both viewing at 0°).
The level and stability of the sensing chain including illumination, geometric configuration, light transmission and sensor response functions were set up to optimize signal to noise ratio of low level data and were documented. Every day of acquisition, controls were performed systematically against a secondary calibration surface and tracked by the National Institute of Standards and Technology (NIST) through Spectralon® (Labsphere, NH, United States) in accordance with good practice for uncertainty management (GUM). This data were used to correct the white balance of RGB cameras and to calculate physical units of reflectance. Acquisition was optimized to maximize the sampling within the microplot and to allow full acquisition of the platform in 1 day with a two-sensors bay. During the 2017 campaign an operation speed of 0.3 m.s–1 was chosen allowing three acquisitions of each RGB image and of VIS-NIR reflectance measurements. LiDAR acquisition was carried out all over the plot area.
For RGB cameras, a white balance process was first applied to adjust intensities of the red, green, and blue channels at a same intensity on a reference gray panel. This standardization was important for a robust color based image analysis.
The first use a RGB images was the calculation of the green cover fractions (GCF) at 0° and 45°. A support vector machine algorithm trained on a reference dataset, was used to classify for each image the green and non-green pixels and determine the percentage of green elements for a given viewing angle (
The green cover fractions at 0° and 45° were used to estimate the Green Area Index (GAI) and Average Leaf Angle (ALA). Both variables were estimated by inverting a simple Poisson model using the measured gap fractions Po, calculated as (1-GCF). The model used to relate Po to GAI is:
Where θp is the viewing angle, θl is the mean leaf angle and
For spectroradiometers, a calibration measurement on a spectrally characterized reference surface was done before each data acquisition session. The reflectance at the crop level was then obtained by dividing the canopy reflectance by the calibration measurement. At the plot level, the averages of the normalized reflectances were computed.spectroradiometer Satisfactory signal to noise ratio (giving a threshold of 20) ranges from 450 to 820 nm. Then physically expressed reflectance was sampled by Gaussian filters corresponding to bands needed for calculation of the vegetation indexes. Three vegetation indexes from remote sensing literature were selected for their asserted link with different phenological aspects of the aerial part of monitored crop. The vegetation indices calculations were made with a 3 nm Full Width Half Maximum (FWHM) for all bands. The Normalized Difference Vegetation Index (NDVI; initially proposed by
The Meris Terrestrial Chlorophyll Index (MTCI) was initially proposed by
The Modified Chlorophyll Absorption Ratio Index (MCARI2) proposed by
Plant height (cm) was estimated from the analysis of the 3D point cloud generated from the combination of the LiDAR scans of height (z) and the (x,y) positioning of the sensor, recorded by the gantry’s encoders. The plot mean height was calculated using the algorithm developed by
After statistical and physical validation against expected intermediate values and validation of biophysical values in order to control non-divergence in case of inversion techniques, data were uploaded and shared through a dedicated database named PhenX (
The use of the mobile rainout shelters from 23rd February to 26th June 2017 reduced detected precipitations under the protected area for the WD environment. During these 4 months, the sum of precipitations collected by the 2 pluviometers outside the protected area reached 161–190 mm (
Precipitation levels (mm, blue bars) and yield mean (t/ha, red dots for N+ environment and yellow dots for N0) of the 4 plot-lines using their location from the beginning of the protected area (m, red dotted lines) in water-deficit condition
During the period of use of the mobile shelters, PAR and air temperature were also affected with a 49% linear decrease in the PAR and a 0.85°C global increase in the air temperature (
For each soil layers, LA: plowed horizon, S: cambic horizon, C1 and C2: calcaric material (C1: cryoturbed limestone C2: sandy calcaric material),
Relationship between water content and pF on PhénoField soil horizons; LA: plowed horizon, S: cambic horizon, C1 et C2: calcaric material (C1: cryoturbed limestone; C2: sandy calcaric material).
Mapping the soil characteristics revealed an important variability of the WHC over the site with a WHC varying from 102 to 275 mm (
Characterization of the soil water holding capacity (mm) on the PhénoField® platform with rectangles representing a span in each of the eight shelters and for the four positions.
Soil water tension at 60 cm
The linear relationship between the gravimetric measurement and TDR values allowed good calibration of the TDR probes (
A summary of rainfall, irrigation and nitrogen fertilization per month is reported (
Combining the soil mapping and the weather measurements, CHN model helped us to monitor daily soil water deficit under each rainout shelter (
Soil water deficit level for well-watered conditions
In another way, the CHN model allowed us to calculate abiotic stress factors induced on crops per replicate (
Impact of water
With regard to variety behavior, both water and nitrogen deficiency significantly reduced the yield (
Yield (t.ha–1;
Looking at plant height based on LIDAR data, both water and nitrogen stresses significantly reduced the wheat height at the beginning of the grain-filling period (
Mean of wheat height (cm)
Considering tensiometers at more than 120 cbar (
The Green Plant Area Index was calculated from the RGB cameras by using both nadir and angle view. It is an integrative trait linked to the LAI that makes possible discrimination between WW and WD modalities at the beginning of stem elongation and also between N+ and N0 modalities (
Green fraction temporal evolution is also based on RGB Cameras data.
Green fraction (%) during the growing season
Using spectroradiometer data, as for other HTP variable acquired with time sequences, the area under the curve (AUC) of the MTCI index was calculated during grain filling (between flowering and maturity stage).
Relationship between the N grain yield and the area under the MTCI curve between flowering and maturity under well-watered conditions with and without nitrogen input (WW N+ and WW N0) and water deficient conditions with nitrogen input (WD N+,
Looking at parameters of curves such as the maximum value,
Correlation between yield (t.ha–1) and plant height during grain filling.
Another way to analyze these data could be the difference between two modalities (optimal and stressed). In
Mean of yield loss related to plant height loss at maturity stage between WW N+ and WW N0 (Nitrogen stress impact in green points) and WD N+ and WD N+ (Water stress impact in blue points).
PhénoField® is a prototypical platform built in 2014. This field based phenotyping infrastructure (using rainout shelters, soil water sensors, and fine soil characterization) is offering various research possibilities on drought tolerance without other impact on the environment. Adjustments and development of methods are needed for its operation to better serve the needs of several research topics. Based on data acquired during 2017, we should be able to take into account experimental limits.
During BREEDWHEAT 2017, the control of water stress levels were allowed by the monitoring of soil water tension and soil humidity. It helps us to avoid plant water stress for WW conditions. In addition, the control system of the rainout shelters allowed automatic interception of up to 92% rainfall (
During the raining period, crop protection by shelters induced a PAR decrease and an overall air temperature increase (
Mapping soil characterization is important to understand soil heterogeneity in order to take it into account in the interpretation of agronomic variables (
PhénoField® uses high throughput field based phenotyping development to characterize responses of crop to abiotic stresses. The example of the BREEDWHEAT field trial conducted in 2017 demonstrated the capacity of the system to characterize drought and nitrogen stress impact on wheat growth, with accuracy needed to differentiate treatments like wheat varieties.
In this trial, water and nitrogen deficiency have different impact on agronomical traits but we have a significant difference between varieties on all agronomical traits (
In our case, the application of drought during the stem elongation period (
Differences linked to stress conditions were also noticed with the embedded sensors LIDARs, RGB cameras and spectroradiometers. First of all, these sensors provided temporal information on wheat varieties and crop management systems. The height calculated with the LIDARs could be a good indicator of the onset and magnitude of the plant’s stress (
In addition to temporal analyses, curves parameters whether they are directly read on a drawn curve direct or calculated from a fitted curve, could also be an alternative way of analyzing data in high throughput phenotyping systems. Nevertheless, stress indicators such as areas under the curve shown previously, appear to be more relevant than using point values to explain performance from high-throughput phenotyping data.
It seems that these variables that describe the behaviors of varieties will enrich breeding methodologies in order to accelerate genetic progress especially given the predictions that climate change will bring about more drought and heat stress in the majority of wheat environments (
As shown with 2017 dataset, the automated movement of the 8 rainout shelters demonstrated its performance to control water and nitrogen deficiency on bread winter wheat field trials without other significant impact on the trial environment. This data set is limited on genotype number to perform 3 replicates and 4 stress conditions. PhénoField® platform, with its large plot capacity, could also provide genetics’ field trials with more than 300 genotypes under sole drought conditions and enhance knowledge on physiological analyses, varieties tolerance evaluation or genomic regions controlling these complex traits. 2017 was the 1st year using field HTP with a data-processing pipeline that still has to be improved. However, it showed promising results, especially with the dynamics of sensor traits allowing the calculation of relevant indicators of abiotic stresses.
The actual set of sensors allows testing of many traits and new parameters to select the best way to discriminate modalities as varieties under nitrogen or water deficiency. The possibilities to analyze the 2017 data set are really important and we are aware that only a small part was explored in this paper. Nevertheless, our first objective was to demonstrate the PhénoField® platform’s capacity to efficiently conduct the test protocols (intensity and duration of stress). Conducting trials on field HTP induced significant soil environment genotype interactions and, mapping the spatial variation of soil characteristics (as well as WHC soil resistivity) is essential for incorporating field variability into crop management and for the interpretation of experimental results. Acquisition of field phenotyping data is now a well-established process, allowing weekly data registration from spectroradiometers, LiDARs and RGB cameras. Acquisition of measurements for the entire platform amounts to about 100 GB per day and only a minor part of it is used to calculate the height, GPAI or vegetation index presented here. In future publications, we will explore more in details curve parameters in regards to agronomic data. Moreover, since 2018, the inversion of radiative models available in PhénoField® data processing chain allows access to the chlorophyll content, which is even more relevant than the vegetation indexes in qualifying nitrogen stresses. Indeed, the calculation of the commonly used vegetation indexes only used 1% of the available spectra and exploitation of other wavelengths could highlight other physiological processes (
PhénoField® is also a platform acting as a field HTP reference used to calibrate and develop innovative phenotyping tools such as the set-up of new sensors. Thus, measurements by drones can be compared to the acquisition of gantries. Drones will allow complementary sampling to intercept diurnal variation and data acquisition faster than with embedded sensors and could be applied on multiple experimental sites. Alternatively, root phenotyping techniques such as Minirhizotrons, using scanner-based methods, could be tested in parallel with gantry measurements in order to access correlated traits between root and above-ground biomass.
Indeed, PhénoField® is connected to other HTP platforms in order to enhance knowledge on genotyping using environment interactions. As part of the PHENOME-EMPHASIS/France network, PhénoField® is highly connected to other French field platform like Pheno3C (INRA, Clermont-Ferrand), but also to controlled platform like PhenoArch (INRA Montpellier). Phénomobile which consists of field mobile HTP robots and ALPHI (A Light Innovant PHenotyping), a system using a boom and a tractor with inbuilt sensors, are also used in the same network
KB: platform management, measurement, and writing of the manuscript. FL: data analyses and writing of the manuscript. AF: measurement, system establishment, system discussion, and writing of the manuscript. CH: technical realization and measurement. MB: results and discussion. JL: results, discussion, and evaluation of the manuscript. BdS: system discussion. BP: database organization and model CHN discussion. ST: system processing data. J-PC: evaluation of the manuscript.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
We would like to thank L. Boulard for his commitment as a farmer, Y. Flodrops, E. Tremblay, J. C. Gapin, T. Joie, F. Savignard, and C. Fontaine for their contribution during the field crop management; P. Poix for the preventive and curative maintenance of PhenoField® equipments; the UMT CAPTE (lead by INRA and ARVALIS) and F. Baret research team for his research on data processing tools. We also thank L. Inchboard for revision of the English version.
The Supplementary Material for this article can be found online at: