<i>Hyperspectral Imaging-Enabled High-Throughput Screening of Loblolly Pine (Pinus taeda) Seedlings for Freeze Tolerance</i>

Abstract

Abstract. Loblolly pine (Pinus taeda L.) is a commercially important timber species that is planted across a wide temperature gradient in the southeastern United States. Ensuring the planting stock is suitably adapted to the growing environment is critical to achieve high productivity and survival. Mean winter minimum temperature (MMT) is an important climatic variable related to growth and survival, and is used to guide the transfer of improved seed throughout the species distribution. Advanced generation families are assigned to cold hardiness zones based on the MMT of the region from which their founding ancestors were adapted. This method has been successful, but as the number of cycles in the breeding program increases, so too will the number of founding ancestors increase for a given selection. The precise assignment of the correct cold hardiness zone for a new selection will become more challenging, particularly when the founding ancestors originate from varying geographic regions with a range in MMTs. Long term field studies, although considered the most reliable method for identifying cold-tolerant families, are extremely resource intensive and time consuming. The development of a high-throughput screening tool to characterize and quantify freeze tolerance among different genetic entries of seedlings will facilitate the accurate deployment of highly productive and well-adapted loblolly pine across the landscape. This study presents a novel approach to assess freeze damage of loblolly pine seedlings using hyperspectral imaging. A seedling population, comprising 98 families representing a wide range of MMT at selection origin, was raised in the nursery. Using a freeze chamber, a total of 1549 seedlings were subjected to an artificial mid-winter freeze. A custom-assembled hyperspectral image system was used for scanning the seedlings before and after the freeze event periodically. A hyperspectral data processing pipeline was developed to segment and extract spectra from individual pine seedlings. Examination of spectral features of pine seedling suggested reductions of chlorophylls and water concentrations in the freeze-susceptible seedlings. Cost-sensitive linear support vector machine (SVM) was utilized for classifying the visually scored seedlings into stressed and healthy. Results showed that hyperspectral imaging was able to achieve the geometric classification accuracies of 75-78% for the non-symptomatic seedlings before and within 10 days after the freeze event, and of up to 96% for the seedlings at day 41 day after the freeze event. The top portion of seedlings was found to be more indicative of freeze events than middle and bottom portions. The newly developed freeze tolerance evaluation method will provide breeders with a valuable tool that offers improved efficiency and objectivity in characterizing and screening of freeze tolerance and potentially other resistance attributes for loblolly pine.