High-Throughput, In-Field Screening of Photosynthetic Efficiency in Crop Plants Using an Autonomous Robot

We investigate photosynthesis under fluctuating field conditions to identify efficient and resilient varieties, and to discover the underlying beneficial alleles. Photosynthesis is mostly assessed through time-consuming manual PAM and gas exchange measurements, which limit the detection of seasonal dynamics and the screening of high genetic diversity. To begin, install the camera tripod next to the experimental field.

Place the PPFR sensor on top of the tripod and set the data logger and power bank on the ground below the tripod. Now mount the LIFT sensor assembly on the front of the robot, positioning it at a height of approximately 60 centimeters above the crop canopy. Place the laptop computer, car battery, and power inverter on top of the robot.

Connect the power inverter to the car battery. Then connect the inverter to the LIFT sensor assembly. Next, position the GNSS antenna kit on top of the robot.

Connect the antenna kit to the laptop computer and start the GnssLogger desktop client and adjust the desired settings. The quantity directly measured by the LIFT sensor is the chlorophyll fluorescence yield, which is the increase in total chlorophyll fluorescence due to the excitation beam for each flashlet. Let to calculate the quantum efficiency of the photosystem II, use the given formula.

While the 300 flashlet excitation phase lasts only 750 microseconds, the total duration of a single chlorophyll fluorescence measurement, including the relaxation phase, is approximately 21 milliseconds. Next, manually drive the robot to the beginning of the first row of plots in the experimental field using the remote controller. Activate the measuring script for both the LIFT sensor and the spectrometer in continuous mode.

Start the autonomous robot navigation at a velocity of 0.5 meters per second using the robot control website. During measurement, periodically check the plot representing the chlorophyll fluorescence yield over time in the LIFT sensor desktop client to confirm the expected shape. If signals appear too weak, adjust the sensor gain accordingly.

Periodically hold the white reference panel under the excitation beam at the height of the crop canopy while the robot is turning at the field border. Use the data table package in R to read in the LIFT transient and spectral data from all files ending in data CSV and spectral CSV. Read in the GNSS data, weather data, the GeoJSON plot maps and the experimental design file.

Extract the quantum efficiency of the photosystem II from each recorded LIFT transient. Then determine the robot's heading based on consecutive GNSS positions. Merge the GNSS data with the experimental design using the assigned plot identifiers to associate spatial locations with treatment and replication information.

Filter the white reference measurements from the spectral reflectance dataset and merge these measurements with the PPFR data using timestamps to create a lookup table that links spectral reflectance to different incident light intensities. Now apply corrections to the raw spectral reflectance data using the generated lookup table. Extract genotype-specific photosynthetic response trends from the outlier filtered chlorophyll fluorescence dataset with the model.

Estimate the slope of the genotype by PPFR interaction term using the emtrends function in R to quantify genotype-specific irradiance response. The model includes MTCI to account for variations in chlorophyll content and canopy structure. During the summer, 36 soybean breeding lines were measured.

The geo-referenced dataset revealed strong spatial patterns in both quantum efficiency of photosystem II and normalized difference vegetation index across the experimental fields due to different genotypes grown in the plots and likely additional soil and field heterogeneity. Row-wise spatial patterns were partially explained by the heading direction of the field robot causing shading of the target leaves at a specific time of the day. Linear mixed effects modeling of the season-long responses of quantum efficiency of photosystem II to increasing incident photosynthetic photon flux rate, or PPFR revealed pronounced field-level heterogeneity, explaining a substantial portion of the variance in the dataset.

The extracted genotype-specific slopes showed clear differences among breeding lines, with several lines exhibiting steeper or flatter response curves of quantum efficiency of photosystem II to increasing light intensities compared to the panel average. Three-dimensional canopy reconstructions, using the MOSS T3R algorithm, showed the potential of integrating physiological and structural phenotyping. Automated LIFT measurements throughout the season reveal physiological heterogeneity within plot canopies, and importantly, distinct photosynthetic efficiencies between genotypes.

Our approach captures real-time photosynthesis measurements non-invasively and autonomously within milliseconds, enabling rapid comparison of photosynthetic responses across full seasons and diverse genotypes. Future work will integrate LIFT with genomic, thermal and 3D canopy data to more precisely predict photosynthesis, stress resilience and crop productivity.