Back

Warming winters, changing springs: cold hardiness dynamics predict budbreak and associated low temperature threats on an intercontinental scale

Campos-Arguedas, F.; Kirchhof, E.; North, M. G.; Londo, J. P.; Bates, T.; van Leeuwen, C.; Destrac-Irvine, A.; Bois, B.; Kovaleski, A. P.

2025-01-14 plant biology
10.1101/2025.01.09.632158 bioRxiv
Show abstract

Temperate woody perennial plants form buds that develop into leaves or flowers that emerge in the following growth season. To survive winter, dormant buds must attain cold hardiness, and timely lose it in spring to break bud while avoiding damage from low temperatures and late frosts. Here, we use a process-based model to predict bud cold hardiness of three grapevine varieties (V. vinifera Cabernet-Sauvignon and Riesling, and V. hybrid Concord) from historical temperature records of eight different locations in North America and Europe (n=329). Based on those predictions, and thresholds of cold hardiness at budbreak from literature, timing of budbreak was extracted. Despite the model being untrained to budbreak data, the cold hardiness model resulted in good predictions (RMSE=7.3d) that were further improved based on expected delays from cold damage (RMSE=7.2d). Both increasing and decreasing trends in freeze damage risk were predicted with increasing temperature, depending on the range of mean dormant season temperature (MDST; 1 Nov-30 Apr) in each location. Predictions in timing of spring phenology in relation to MDST also showed warming to advance (MDST<10{degrees}C) or delay (MDST>10{degrees}C) budbreak. Cold hardiness dynamics represents an advancement in phenological modeling that provides information for the entirety of the dormant season, as well as budbreak.

Published in New Phytologist (predicted rank #2) · training set

Matching journals

The top 3 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.