Recipient Organization
AUBURN UNIVERSITY
108 M. WHITE SMITH HALL
AUBURN,AL 36849
Performing Department
(N/A)
Non Technical Summary
The Southeastern U.S. contributes more than 23% of the total $1B of blueberry production and has the privilege of producing the earliest blueberries in the U.S. However, most blueberries in the Southeast are produced in Georgia, Florida, and North Carolina, whereas small- to mid-sized growers in other states only maintain marginal production due to a lack of locally adapted cultivars. To sustain blueberry production in underserved regions of the southeastern U.S., the new small fruit breeding program at Auburn University is collaborating with four established breeding programs in the Southeast to:(i) develop climate-resilient southern highbush blueberries for small- and mid-sized growers in Alabama and nearby regions through collaborative cultivar evaluation,(ii) identify genotype-by-environment interactions for key traits to better allocate future cultivars to appropriate environments, and(iii) enable high-throughput yield phenotyping for more efficient cultivar development.A total of 30 advanced selections and 8 checks from four established breeding programs will be evaluated for 3 years in central and south Alabama, and testing sites in Florida, Mississippi, and North Carolina, respectively.Towards the end of this project, we expect to release new cultivars for the general and specific environments of the southeastern region, to benefit all growers and especially small- and mid-scale stakeholders; train a field-based fruit breeder; and develop a smartphone app to enable high-throughput yield phenotyping. Cultivars developed from this collaboration will set the foundation for a more sustainable and profitable blueberry industry in the Southeast for years to come.
Animal Health Component
80%
Research Effort Categories
Basic
20%
Applied
80%
Developmental
(N/A)
Goals / Objectives
The southeastern U.S. contributed more than 23% of the total $1B of blueberry production in the U.S. in 2021 (USDA NASS). The southeast has the privilege of producing the earliest blueberries in the U.S. and therefore enjoys higher market values. However, most of the southeastern blueberries are produced in Georgia, Florida, and North Carolina, whereas small- to mid-sized growers in other states such as Alabama, South Carolina, and Tennessee only maintain marginal blueberry production despite a similar environment with major producing areas. The long-term goal of this project is to sustain blueberry production in underserved regions of the southeastern U.S. by developing climate-resilient cultivars suitable for small- to mid-sized growers. Specifically, this proposal aims to:Objective I. develop climate-resilient southern highbush blueberry cultivars for small- and mid-sized growers in Alabama and nearby regions through collaborative cultivar evaluation,Objective II. evaluate genotype-by-environment interactions of key traits to better allocate future cultivars to the appropriate environments, andObjective III. enable accurate and high-throughput yield phenotyping through image-based analysis and machine learning.The southeastern U.S. contributed more than 23% of the total $1B of blueberry production in the U.S. in 2021 (USDA NASS). The southeast has the privilege of producing the earliest blueberries in the U.S. and therefore enjoys higher market values. However, most of the southeastern blueberries are produced in Georgia, Florida, and North Carolina, whereas small- to mid-sized growers in other states such as Alabama, South Carolina, and Tennessee only maintain marginal blueberry production despite a similar environment with major producing areas. The long-term goal of this project is to sustain blueberry production in underserved regions of the southeastern U.S. by developing climate-resilient cultivars suitable for small- to mid-sized growers. Specifically, this proposal aims to:Objective I. develop climate-resilient southern highbush blueberry cultivars for small- and mid-sized growers in Alabama and nearby regions through collaborative cultivar evaluation,Objective II. evaluate genotype-by-environment interactions of key traits to better allocate future cultivars to the appropriate environments, andObjective III. enable accurate and high-throughput yield phenotyping through image-based analysis and machine learning.
Project Methods
Objective I.Germplasm and experimental design. Material transfer agreements have been established between Auburn University and the University of Georgia, University of Florida, North Carolina State University, and USDA Thad Cochran Southern Horticultural Laboratory, respectively, to evaluate a total of 30 advanced selections and 8 checks of southern highbush blueberries in central and south Alabama. Four checks including Arcadia, Keecrisp, Legacy, and O'Neal will be evaluated in all five locations, and the other four checks New Hanover, Colossus, Optimus, and Patricia will be evaluated in four locations. Randomized complete block design will be used in each location and environment with two to three replicates for each genotype and three plants per plot. In all locations, soil testing and amendments were applied during site preparation. Pine bark was applied to raised beds (7-15 cm high and 90 cm wide) to further reduce soil pH to an ideal range of 4.5 - 5.5. Beds were covered with heavy-weight U.V. polypropylene fabric or pine bark mulch for weed control. Dripped irrigation or overhead sprinklers were installed before planting. Spacing for plants is 0.9 m within the row and 3 m between rows. Standard management will be applied in all locations.Data collection. Plants at each location will be evaluated for dates of 50% bloom, dates of 50% fruit ripening, the ratio of flower buds damaged by spring frost (in the occasion of frost events), 50-berry weight (g), average berry diameter (mm), berry yield per plant (g), disease score (1-5), Brix (%), titratable acidity (TA, %), and firmness (g/mm) for three years. Each plant will be hand-harvested 3-5 times throughout the harvest season to measure accumulative yield per plant (g). Once harvested, berries will be stored in cold storage under 1 to 4 °C for two weeks before additional phenotyping. After cold storage, the juice of 25 randomly selected but represented berries will be used to measure Brix and TA. Thirty berries will be randomly selected to measure firmness (g/mm) and average berry diameter. Frozen juice of additional berries will be sent to the USDA Thad Cochran Southern Horticultural Laboratory for polyphenolic content (mg/ml) and total sugar content (mg/ml). Weather data such as hourly temperature (°C), relative humidity (%), wind speed (m/s), wind direction, and solar radiation (W/m2) will be obtained from local weather stations for each testing site.Data analysis. Analysis of phenotypic data will take two steps. First, the original data will be grouped into three datasets based on the origin of the materials for analysis of variance components and breeding values. Dataset 1 contains Florida materials and checks; dataset 2 contains North Carolina materials and checks; and dataset 3 contains USDA materials. Materials in each dataset will be tested in 3 locations. Additionally, data from all five locations will be integrated through pedigree and common checks to analyze breeding values for all environments.Objective II. Joint G×E analysis will be conducted on the entire dataset connected through common checks. Additionally, G×E interactions will also be evaluated on the three datasets separately (as described under Obj. I) for comparison. Analysis methods will follow Bakare et al. (2022). A linear mixed model will be used. Environments will be treated as fixed effects, while genotypes, genotypes by environment interactions, blocks, and residuals will be treated as random effects. The significance of G×E interactions will be tested with the likelihood ratio test by comparing the likelihood of a full model, including G×E with a reduced model without G×E. When significant G×E interactions are detected, patterns of G×E interactions will be further analyzed through various models such as Finlay-Wilkinson regression (Finlay & Wilkinson, 1963), the Additive Main effect and Multiplicative interaction (AMMI) model (Gauch, 1988) using the statgeneG×E package (Bart-Jan van, 2021).Objective III. Per-plant yield data and image data will be collected under objective III to improve the accuracy of the existing deep-learning model for yield prediction. Existing deep learning models for berry detection, as developed prior to this project, will be used to count berries captured in the image data. Berry count, yield, and other parameters such as genotype, plant age, and estimated canopy density will be used to train machine learning models for yield prediction, to better account for occlusion effects. The best yield prediction model will be implemented in a smartphone app to allow near real-time berry detection and yield prediction in the field and potentially other production systems.In addition to per-plant yield data collected under Obj. I, we will also collect image data from each plant before harvest and use that data to test and improve the current yield prediction model. Specifically, a 12-megapixel RGB camera on a Google Pixel 5a smartphone (or other cameras with equivalent or higher resolution) will be used for image acquisition. Images will be taken from two sides of a field-grown blueberry plant, one from each side of the row. Four ArUco markers mounted onto two 48-inch rods will be placed on the left- and right side of the plant to mark the vertical boundaries of the plant. Total yield and average berry weight will be measured for each plant as described in Obj. I. Average berry weight will be calculated as 50-berry weight divided by 50. The maturity level of a plant will be calculated as the percentage of fully ripe berries over the total number of berries, based on manually annotated images of each plant.Field images will be manually annotated with the open-source image annotation software COCO Annotator to create a custom blueberry dataset for deep-learning research. Pretrained Mask R-CNN models based on He et al. (2017) and our 2022 data in Detectron2 (Facebook AI Research) will be further improved with the 2023 dataset to detect individual berries and estimate their maturity levels. Improved berry detection models will be applied to process single-plant images collected. Estimated berry count based on deep learning models, will then be used to train yield prediction models on the cultivars and selections in this multi-environment trial. Statistical models (e.g., multiple linear regression) and machine learning models (e.g., support vector machine, random forest, artificial neural network, etc.) will be evaluated to predict the ground truth yield based on estimated berry count, average berry weight, plant age, genotype, and visually estimated canopy density to better account for occlusion effect.A smartphone app will be developed for blueberry yield prediction in the field setting. TensorFlow Lite, the open-source real-time object detection android app from Google, will be modified towards a blueberry yield prediction application. The generic object detection model will be replaced by a blueberry detection and yield prediction model developed from this study. The smartphone app will be developed in 2024 with a user-friendly interface. User-friendliness and robustness will continue to be tested and improved throughout the entire funding period.The results from this project will be measured through surveys conducted after field days and extension workshops. Participants change of knowledge will be measured through pre-and post-workshop/field day surveys. Also, survey questions will be designed to measure participants change of behavior by evaluating their intent to adopt the new technology/cultivars based on the new knowledge gained via the extension activities.