PEANUT VARIETY AND QUALITY EVALUATION HARNESSING MULTISCALE DATA FOR TRAIT PREDICTION TO SUPPORT CULTIVAR RELEASE DECISIONS

• Sponsoring Institution: National Institute of Food and Agriculture
• Project Status: ACTIVE
• Funding Source: AFRI COMPETITIVE GRANT
• Reporting Frequency: Annual
• Accession No.: 1030023
• Grant No.: 2023-67013-39624
• Cumulative Award Amt.: $500,000.00
• Proposal No.: 2022-10305
• Project Start Date: May 1, 2023
• Project End Date: Apr 30, 2026
• Grant Year: 2023
• Program Code: [A1143]- Conventional Plant Breeding for Cultivar Development

Project Director
Balota, M.

Recipient Organization
VIRGINIA POLYTECHNIC INSTITUTE
(N/A)
BLACKSBURG,VA 24061

Performing Department
Tidewater AREC

Non Technical Summary
The overall objective of this proposal is to ensure release of peanut cultivars resilient to the climate change to allow producers of the Virginia and Carolinas (VC) peanut growing region to maintain competitiveness in the marketplace. This will be realized through evaluation of the developed lines within the well-established Peanut Variety and Quality Evaluation (PVQE) multi-state program with the addition of new high-throughput methods and traits for accurate selection of 1) early maturing peanut cultivars with 2) tolerance to heat and dry conditions and 3), resistance to southern corn rootworm (SCRW), using "smart" technologies and computer algorithms. This proposal fully addresses the Program Area Priorities for testing and evaluation of developed materials in pipeline for release using a well-established PVQE regional program. It has industry' s support and stakeholders' s involvement with release decisions (through the PVQE Advisory Committee, PVQEAC), marketability (shellers and processors), seed production, intellectual properties (university administrators), and has potential to improve its relevance to the breeding programs at NCSU and TAREC. Education of field-based peanut breeders at national and international level can be accommodated through collaborations of the PI with national and African peanut breeders, well established through previous NIFA/AFRI and USAID projects. Once validated, the new methods developed through this project will be implemented in future PVQE testing. This will improve and diversify the coordinated efforts of the multi-state PVQE project to help selection of improved peanut cultivars for the VC region and to meet the industry requirements for quantity and quality of peanut.

Animal Health Component

50%

Research Effort Categories

Basic

0%

Applied

50%

Developmental

50%

Classification

Knowledge Area (KA)	Subject of Investigation (SOI)	Field of Science (FOS)	Percent
201	1830	1081	50%
211	1830	1130	25%
203	1830	2020	25%

Knowledge Area
211 - Insects, Mites, and Other Arthropods Affecting Plants; 203 - Plant Biological Efficiency and Abiotic Stresses Affecting Plants; 201 - Plant Genome, Genetics, and Genetic Mechanisms;

Subject Of Investigation
1830 - Peanut;

Field Of Science
2020 - Engineering; 1130 - Entomology and acarology; 1081 - Breeding;

Keywords

aerial imagery

peanut variety and quality evaluation

peanut maturity

southern corn rootworm

Goals / Objectives
The overall goal of this proposal is to ensure the release of climate-resilient peanut cultivars to the producers of the Virginia and Carolinas (VC) region to allow them maintain competitiveness in the marketplace. This will be realized through evaluation of the advanced breeding lines within the well-established Peanut Variety and Quality Evaluation (PVQE) multi-state program with the addition of new high-throughput methods and traits for accurate selection ofearly maturing peanut cultivars withtolerance to heat and dry conditions, andresistance to southern corn rootworm (SCRW), using "smart" technologies and AI algorithms. Once validated, these new methods will be implemented in future PVQE testing. Specifically, our goal is toimprove and diversify the coordinated efforts of the multi-state PVQE project to help selection of improved peanut cultivars for the VC region and to meet the industry requirements for quantity and quality of peanut. To achieve this goal, the current proposal has three specific objectives:Objective 1. Develop and validate models to estimate peanut maturity for implementation in future PVQE testing.Objective 2. Develop and validate models to estimate peanut heat and drought tolerance for implementation in future PVQE testing.Objective 3. Develop and validate a greenhouse/lab assay and models to estimate pod SCRW injury for implementation in future PVQE testing.

Project Methods
Open field PVQE trials will be used for Objectives 1, 2 and 3 including replicated trials of 25 to 30 breeding lines and check cultivars at 5 locations in the VC region: Tidewater Agricultural Research and Extension Center (TAREC), Suffolk, VA; Upper Coastal Plain Research Station (UCPRS), Rocky Mount, NC; Edisto Research and Education Center (EPEC), Blackville, SC; and in farmers' fields near Williamston, NC and Council, NC. Two-row plots, 11 m long with 0.91-m spacing between rows, will be arranged in a randomized complete block design (RCBD). Planting will be in early to mid-May at a seeding rate of 13 seeds m-2. To complete Objective 2, in addition to open field PVQE trials, rainout shelter trials will be conducted at TAREC each year to impose drought and heat stress during seed development. Data from each trial will be collected manually and aerially throughout the growing season according to established protocols for each objective. At optimum maturity plots will be dug, plants will be windrow dried for a few days, combine-picked, and yield and farmer stock grading will be used to determine the support price and crop economic value according to the USDA standards. To complete Objective 3, additional experiments will be performed in the greenhouse. Briefly, peanut genotypes will be grown in 5-gal pots in the greenhouse at TAREC and infested with SCRW larvae reared from laboratory colonies that we already initiated at TAREC. Second instar larvae will be used and allowed to feed and develop naturally. Following three weeks of larvae feeding, plants will be removed from the pots. Pods will be visually observed for SCRW injury.Under Objective 1, pod maturity will be determined manually at 100, 125, 135, and 140 days after planting (DAP) by digging three plants per plot from all open field trials to obtain a 150-pod sample and assess their mesocarp color. After sorting the pods by color in 'white', 'yellow', 'orange', 'brown', and 'black' color groups, Peanut Maturity Index (PMI) will be calculated as the rate of orange, brown and black pods from the total pods. At the same time, flight missions will be conducted to collect plot images with an unmanned aerial system (UAS) (Matrice 300 v2, DJI, Los Angeles, CA) with a six-band imaging sensor on-board (Micasense Altum, Micasense Inc., Seattle, WA). Collected & geo-referenced images will be stitched with Pix4D Mapper software (Pix4D, Inc., Lausanne, Switzerland) using Ag multispectral template to obtain seamless reflectance orthomosaics of Blue, Green, Red, Red-Edge, and Near-Infrared (NIR) wavebands. Stitched reflectance will be imported to a geographical information system software (QGIS ver. 3.24, Open Source) and broadband vegetation indices (VIs) will be calculated, including NDVI, Green-NDVI, Red-Edge-NDVI, Modified Nonlinear Index (MNLI), Ratio Vegetation Index (RVI), Transformed Difference Vegetation Index (TDVI), and Soil Adjusted vegetation Index (SAVI) using the "raster calculator" toolbox. These indices will be used to train and validate models to estimate the PMI from manual measurements.Under Objective 2, leaf wilting will be assessed every other week after 50% canopy closure, i.e. approximately mid-July, on open field trials and rainout shelters. At the time of wilting observations, flight missions with multispectral (visible and NIR wavebands) camera will be conducted and images processed as described at Obj. 1 for open field trials. For rainout shelters, thermal imagery data will be collected in addition to reflectance in visible and NIR wavebands. Extracted vegetation indices as described at Obj. 1 will be used to train and validate models to estimate leaf wilting and plant stress. Weather information will be collected under each shelter using all-in-one microclimate weather sensors connected to a data logger with cellular connectivity. In this way we will follow a holistic approach to not only identify the most stress tolerant peanut breeding lines but also identify the key weather parameters responsible for the onset of drought stress.Under Objective 3, PVQE open field trials at all locations (described above) will be used for field direct assessment of peanut pod injury by SCRW feeding. At the optimum maturity and on the same day with plot digging, 50-seed samples will be collected from each plot and rated by percentage of scarred (outside pod fed upon) and penetrated (pod penetrated and seed injured). In addition to the greenhouse pot testing, this activity will identify breeding lines resistant to SCRW injury. In addition, after rating the injury of the greenhouse SCRW larvae fed pods, the injured and non-injured pods will be placed on a paper scanner and images will be scanned at 600 dpi. We expect to collect over 2000 images with 50 pods (injured and non-injured) per collected image. Collected images will be labelled for non-injured and different levels of pod injury using open source software (Imagetagger). Labelled images will be annotated further to induce random noise and illumination variations to obtain a further robust imagery dataset (~10,000). The annotated images will be split in the ratio of 7:2:1 (for training:validation:testing). This activity will result in development of DL models for SCRW pod damage evaluations using a controlled lab assay. Classical statistical analysis of the General Liner Model will be used with the majority of phenotyping and agronomic data. Non-linear models, ML and DL models will also be used. The DL models will include the Convolution Neural Network, AlexNet, GoogLeNet, InceptionV3, ResNet50, and DarkNet53. A range of statistical and ML models will be applied to the image data including the Stepwise Linear Regression, Partial Least Square Regression, and LASSO Regression to identify the best predictors of the PMI using reflectance and key VIs. Lastly, we will use ML models including the K-nearest neighbor, Support Vector Machine, Logistic Regression, Long Short-Term memory, and MLP and RBF based ANN models. All formulations and analysis will be conducted in Python platform using TensorFlow, Keras, Scikit-learn, and PyTorch libraries. Images collected from 70% of the field plots in YRs 1 & 2 of the project will be used for model training and the remaining 30% for model validation. Additional validation and models' refinement for robustness will continue in YR 3 of the project with independent datasets from all experimental sites. The validation results will be reported using coefficient of determination (R2), root mean square error, and mean bias error.During the course of this project three graduate students will receive science-based knowledge and graduate training. Extension and outreach activities will also be performed annually and will include field tours, growers' meetings and meetings of the PVQE Advisory Committee (PVQEAC) as the advisory group. The PVQEAC will ensure the self-supporting nature of the management plan through the end of the project. The PVQEAC meets annually, the third Wed of March, at the Tidewater Agricultural Research and Extension Center in Suffolk, VA, to discuss the PVQE results and future needs to be addressed by the PVQE research team, and vote for the release when proposed by the breeders.

Progress 05/01/23 to 04/30/24

Outputs
Target Audience:The target audiences that benefit from this project includes breeders, fellow scientists, peanut shellers, processors, and certified seed producers, and agricultural producers. Regularly, these audiences are updated on the achievements of this project. Graduate students working on this project, in particular, will benefit through the opportunity to produce knowledge and grow professionally. Currently, 3 graduate students, 2 Ph.D. and one M.S., have been recently employed and will continue their training and graduate research in relation to this project in the areas of plant physiology, pathology, and remote sensing. Through participation in graduate student competitions at professional and extension meetings, and publications, these students will represent an important venue for information produced by this project to be timely distributed to the stakeholders. Changes/Problems:To address objective 3 of the project and aside from the field trials under this objective, a graduate student in entomology started rearing a colony of SCRW in the greenhouse at the Tidewater AREC. This was part of a subtask of objective 3, meant to the development of models to estimate SCRW injury from images. Unfortunately, this was an unsuccessful attempt even afterthe student went to the University of Georgiato be trained,where success with rearingthis insect was recorded. This student will graduate in May 2024, therefore, we envision this to become a problem difficult to solve in the second and third years of the project, because the new students do not have expertise in entomology. But we will continue to search for a solution. What opportunities for training and professional development has the project provided?Three students were funded from this project and started in 2023 their graduate training, 2 Ph.D. and one M.S. level, in plant physiology, biology/pathology, and biological systems engineering. The two Ph.D. students are with the Virginia Tech, College of Agriculture and Life Scences and the M.S. student is with Clemson University. These students will address objectives 1 and 2 of the project, but currently they are in Blacksburg to complete the course work requirements before moving to Suffolk, in May 2024, to participat at the work. In summer 2023 they were recently hired and, before classes started, they had the opportunity to adequately lern about the goals and expectations of the project. A fourth student, M.S. level, was onboard already and had one year of data collection before the project started. This student, while participating at the project under objective 3, he was not funded through this project. This student will graduate in May 2024. How have the results been disseminated to communities of interest?Even if the work is in inciient phase, just one year of data collection, the graduate students participated at several professional meeting where they presented thier preliminary data collected under this project. Information was also delivered to the growers, at the Peanut-Cotton Field Day at Tidewater AREC, PVQE Field Day at Slade Farm in 2023, and at the Virignia Peanut Growers Conference in Feb 2024. What do you plan to do during the next reporting period to accomplish the goals?As this is a field agronomical study, the experimental trials from 2023 will be repeated in summer of 2024 and 2025, which is an essential requirement for conclusions to be derived and recommendations formulated for breeders and growers.

Impacts
What was accomplished under these goals? ?? Objective 1: Develop and validate models to estimate peanut maturity for implementation in future PVQE testing. Methods and Data Collection: In the 2023 season, two experimental trials were developed for this study. One trial involved five commercial cultivars of Virginia type peanut (Bailey-II, N.C.20, Emery, Sullivan, and Walton) with two treatments: with growth regulator Calcium Prohexadione and without or control plots for each cultivar. Four replicates of each cultivar and treatment were sampled in a randomized complete block design (RCBD) fashion at 5 times starting at approximately 100 days after planting and until harvest. Peanut pod samples were taken for maturity determination after a small unmanned aerial system (SUAS) was used to collect aerial images in visible and near infra-red bands. The second trial utilized PVQE plots planted at the Tidewater AREC research farm in Suffolk VA, at North Carolina State University farm in Rocky Mount NC, and at Clemson University in Blackville, SC. Ten entries from these trials were selected for maturity estimations via direct measurement and aerial imaging. Pod maturity and drone imaging were taken weekly from 100 days after planting to harvest. At both locations, each entry was planted in 2-row plots, 11 m in length and 0.9 cm between rows. Trials were planted in early May and managed according to the Virginia and North Carolina Extension recommendations. Similarly, with trial 1, pod maturity and drone imaging were performed weekly from 100 days after planting to the harvest. A SUASequipped with a multispectral imaging sensor (DJI Phantom 4 multispectral) was flown at an altitude of 20 m above ground, five times in the growing season. TAcquired images were stitched in a photogrammetry and stitching platform (Pix4D mapper) to obtain reflectance orthomosaics of the imaged trials. Next, about 29 vegetation indices (VIs) are currently being derived for each imaging campaign to serve as an indicator of peanut canopy health status. Immediately followed by the imaging campaigns, peanut plant samples were dug from each replicate of both trials. Each sample contained 4-5 plants acquired from 2-3 random locations within a replicated crop row. Approximately 150 pods were removed from each sample and washed with a pressure washer to remove the exocarp layer. After washing, pods from each sample were placed on a maturity assessment board (Williams and Drexler, 1981). The pods were visually assessed for their color among white, yellow, orange, brown, and black. The pods under each color category were counted and the ratio of orange (O), brown (B), and black (B) pods to the total number of pods were calculated for each sample. This ratio represents the peanut maturity index (OBB) which will also be estimated from aerially collected images using aerial imagery data and machine learning techniques. A total of 800 samples were evaluated in the 2023 season. Results: For trial 1, maturity was evaluated for each cultivar. Observations suggested that Emery, Sullivan and Bailey-II were the earliest maturing cultivars while the Walton was the last maturing cultivar. There were also significant differences noted between the maturity for growth regulator treated peanuts and control treated peanuts. Maturity was also evaluated from the pod color on pod samples collected from PVQE 10 entries in VA and NC. In the near future, we will be formulating machine learning models Support Vector Machine (SVM), Random Forest (RF), Partial Least Square Regression (PLSR), K-nearest Neighbor (KNN) to estimate peanut maturity using aerial spectral imagery and results will be reported in the next cycle. Objective 2. Develop and validate models to estimate peanut heat and drought tolerance for implementation in future PVQE testing. Methods and Data Collection: In 2023, two experimental trials were developed to address this objective. The first trial utilized PVQE plots planted at the Tidewater AREC research farm in Suffolk VA, at North Carolina State University farm in Rocky Mount, NC. Images from these plots were collected at the same times as those for Obj. 1, including visual rating of peanut leaf wilting, when wilting was observed. The second trial used plots under the rain shelters to provide controlled rainfall and induce drought and heat conditions. The rain exclusion shelter facility was located at the Tidewater Agricultural Research and Experiment Center (TAREC) in Suffolk, VA. Fifty-six genotypes, inclusive of five commercial cultivars, were sown in three replications following a RCBD. Plants were grown under rainfed conditions with plentiful rain up to 58 days after planting (DAP), after which three rainout shelters were used to impose heat and drought stress. Ground data collection included NDVI using green seeker, canopy temperature (CT), canopy temperature depression (CTD), Relative water content (RWC), and leaf wilting assessment over three consecutive weeks beginning 2 weeks post-stress imposition. Visual scoring of leaf wilting ranged from 0 (no wilting) to 5 (drying of leaves with ground visible). Additionally, UAV data was collected multiple times during pre- and post-rain exclusion shelter deployment. Significant variation was observed among replications (p < 0.001) and genotypes (p < 0.001). Results: Ground data showed significant stress impact, and CT derived from IR thermal imaging showed a significant increase from an average of 30 °C at 2 weeks after stress (WAS) to 39 °C at 4 WAS. Similarly, RWC decreased from 73% to 71%, and ground based NDVI declined from 0.75 to 0.70 over the same period, consistent with visual wilting assessments. From 2 WAS to 4 WAS initiation, the percentage of plants showing no wilting (0) decreased from 51% to 39% whereas plants with mild wilting symptoms (visual score 1) increased from 31% to 51%. Among the commercial cultivars, Walton exhibited the highest susceptibility to stress, with a visual score of 3 (wilting and drooping of all leaves). At the same time, the remaining cultivars also displayed mild wilting symptoms. However, CTD exhibited significant variations across rainout shelters, with shelter or block 1 showing a +3 °C across multiple sampling dates, compared to -1 °C and -3 °C in shelters 2 and 3, respectively. This variation might be attributed to microclimatic differences across blocks or CTD point measurement biases. Three UAV-derived multispectral datasets were utilized to derive seven vegetation indices, the Normalized Difference Vegetation Index (NDVI), Normalized Difference Red-edge Index (NDRE), Green Leaf Index (GLI), Chlorophyll index of green (CIgreen), and red-edge (CIred-edge), chlorophyll vegetation index (CVI), and Ratio Vegetation Index (RVI). The average values across multiple sampling dates for VIs were as follows NDVI (0.85), NDRE (0.3), GLI (0.3), CIgreen (6.4), CIred-edge (0.8), CVI (3.8), and RVI (14). Correlation analysis indicated a negative correlation between ground based NDVI and CTD (-0.42, p < 0.001), as well as CT (-0.45, p < 0.001), while aerial NDVI showed a negative correlation with CTD (-0.38, p < 0.001). All VIs exhibited negative correlations with CTD, emphasizing the importance of this trait (-0.2 to -0.45). Objective 3. Develop and validate a greenhouse/lab assay and models to estimate pod SCRW injury for implementation in future PVQE testing. A graduate student was responsible for this objective. The student collected SCRW injury data from the PVQE field trials at the Tidewater AREC research farm in Suffolk VA, at North Carolina State University farm in Rocky Mount NC, Slade Farm near Williamston, NC, and at Clemson University in Blackville, SC. The student is a MS student and this is the second year of data collection. He is planned to graduate in May 2024, therefore the data from both years are currently being analyzed and thesis written.

Publications

• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Chandel, A., Balota, M., & Rathore, J. (2023). AI-Machine Learning with Aerial Spectral Imagery for Peanut Maturity Prediction. In AI IN AGRICULTURE: INNOVATION AND DISCOVERY TO EQUITABLY MEET PRODUCER NEEDS AND PERCEPTIONS. ORLANDO, FL. Peer Reviewed: (abstract)
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Hoar, E., Rashed, A., & Balota, M. (2023). Alternatives to chlorpyrifos for control of southern corn rootworm in Virginia type peanut. In American Peanut Research and Education Society Annual Meeting, July 11-13, Savannah, GA.
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Chandel, A., & Balota, M. (2023). Feasibility of Aerial Spectral Imagery with AI-Machine Learning for Peanut Maturity Prediction. In ASA, CSSA, SSSA International Annual Meeting 2023. Online. Peer Reviewed: (abstract)
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Beard, K., Balota, M., & Haak, D. (2023). Pollen Viability Responses to Elevated Temperature of Five Virginia-Type Peanut (Arachis hypogaea L.) Cultivars Grown in Southeastern Virginia. In ASA-CSSA-SSSA annual meeting. St. Louis, MS.
• Type: Other Status: Published Year Published: 2023 Citation: Chandel, A., & Balota, M. (2023). Application of small UAS towards management of field cropping systems of Virginia. In USDA S1069: UAV Multistate Annual Meeting 2023. Tidewater AREC, Suffolk.
• Type: Other Status: Published Year Published: 2023 Citation: Biggs, T., Balota, M., Beard, K., & Haak, D. (2023). Evaluation of image-processing techniques as high throughput-phenotyping method for pollen viability assessment under heat stress. Blacksburg, VA.
• Type: Other Status: Published Year Published: 2023 Citation: Hoar, E., Rashed, A., Taylor, S., Balota, M. (2023). Evaluation of Current Virginia Peanut Cultivars for resistance to southern corn rootworm. Virginia Peanut Growers Annual Meeting, Feb 23, 2023, Franklin, VA.
• Type: Other Status: Published Year Published: 2023 Citation: Chandel, A.K.*, Balota M., Jjagwe, P, Chappell, M., 2024. Pod maturity assessments for Virginia type peanuts: 2023 season results. Virginia-Carolinas peanut news. 72(1), p.14.
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Beard, K., Haak, D., & Balota, M. (2023). Variation in Pollen Viability Responses to Elevated Temperature Between Five Commercial Peanut Cultivars Grown in Southeastern Virginia. In American Peanut Research and Education Society Annual Meeting, July 11-13, Savannah, GA.
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: 4. Chandel, A.K.*, Balota, M., Rathore, J.G, 2023. Small UAS based spectral imaging and machine learning for peanut maturity estimation. Poster session presented at the meeting of American Peanut Research and Education Society Annual Meeting, July 11-13, 2023. Savannah, Georgia.