Warding off winter woes: Grapevine cultivation in cool climates faces significant challenges from extreme winter temperatures causing severe frost damage. Traditional monitoring methods like differential thermal analysis (DTA) provide limited, infrequent data, hindering effective vineyard management. Given these challenges, there is an urgent need for a robust, real-time system to predict grapevine responses to fluctuating winter conditions.
Researchers from Cornell University, in collaboration with several institutions across North America, have developed a pioneering automated machine learning (Auto-ML) model, NYUS.2, which was published (DOI: 10.1093/hr/uhad286) in the Horticulture Research journal on December 29, 2023. The study focuses on the large-scale, real-time simulation of grapevine freezing tolerance, a pivotal tool for viticultural regions adapting to climate variability.
The NYUS.2 model integrates extensive grapevine freezing tolerance data collected from various regions in North America, encompassing different climate conditions and grape cultivars. The dataset includes over 10,000 measurements from hybrid and Vitis vinifera cultivars, recorded between 2002 and 2023.
The model leverages AutoGluon, an Auto-ML platform, to process these data, generating a robust prediction model with high temporal and spatial resolution. The model's performance was rigorously evaluated, showing a root-mean-square error (RMSE) of 1.36℃, outperforming previous models like WAUS.2 and NYUS.1 in multiple test regions. Key features influencing the model's predictions include chilling accumulation, exponential weighted moving average (EWMA) temperatures, and cultivar-specific characteristics.
These features help the model account for the complex biological processes underlying grapevine cold acclimation and deacclimation. The deployment of NYUS.2 during the 2022–23 dormancy season demonstrated its practical applicability, providing daily updates on grapevine freezing tolerance for 16 cultivars across 2035 weather stations in the United States. This real-time monitoring capability enables growers to make informed decisions, optimizing vineyard management practices to mitigate frost damage effectively.
"The deployment of NYUS.2 represents a quantum leap in our ability to manage grapevine cold hardiness," says Dr. Hongrui Wang, the lead researcher of the study. "By integrating real-time weather data with our advanced ML algorithms, we can now provide farmers with precise, actionable insights to protect their crops from the ravages of winter, ultimately bolstering the resilience of the grape and wine industries in the face of climate change."
NYUS.2 significantly impacts cool-climate viticulture by providing real-time freezing tolerance predictions, aiding in timely frost protection and boosting productivity. Its adaptability to various climates helps future-proof against climate change. These methodologies can also be applied to other perennial crops, extending their benefits beyond grapes. Newswise/SP