Seasonal changes in carbohydrates and water content predict dynamics of frost hardiness in various temperate tree species.

Predicting tree frost tolerance is critical to select adapted species according to both the current and predicted future climate. The relative change in water to carbohydrate ratio is a relevant trait to predict frost acclimation in branches from many tree species. The objective of this study is to demonstrate the interspecific genericity of this approach across nine tree species. In the studied angiosperm species, frost hardiness dynamics were best correlated to a decrease in water content at the early stage of acclimation (summer and early autumn). Subsequently, frost hardiness dynamics were more tightly correlated to soluble carbohydrate contents until spring growth resumption. Based on different model formalisms, we predicted frost hardiness at different clade levels (angiosperms, family, genus and species) with high to moderate accuracy (1.5-6.0 °C root mean squared error (RMSE)) and robustness (2.8-6.1 °C prediction RMSE). The TOT model, taking all soluble carbohydrate and polyols into account, was more effective and adapted for large scale studies aiming to explore frost hardiness across a wide range of species. The ISC model taking the individual contribution of each soluble carbohydrate molecule into account was more efficient at finer scale such as family or species. The ISC model performance also suggests that the role of solutes cannot be reduced to a 'bulk' osmotic effect as could be computed if all of them were located in a single, common, compartment. This study provides sets of parameters to predict frost hardiness in a wide range of species, and clues for targeting specific carbohydrate molecules to improve frost hardiness.