Modelling the first flowering time of Xanthoceras sorbifolium Bunge in north China based on temperature

Xanthoceras sorbifolium Bunge is a potential biofuel plant, owing to its high seed oil content. But planted forest experiments and practice indicate that field management before flowering played a vital role in increasing the female to male ration and further raise seed yield. Thus, knowing the onset of flowering in advance is crucial for taking management measures to regulate floral development. To predict the first flowering date, we developed a predictive model with linear regression, support vector machine, and random forest models. Mean monthly temperatures with chilling and forcing accumulation (November–March) and preseason temperatures one month prior to the onset of flowering were taken as independent variables. Chilling accumulation was calculated using chilling portion (CP) with the Dynamic Model and chilling hours (CH) for the Utah Model, while forcing accumulation was derived from the Growing Degree Hours (GDH) Model. Internal leave-one-out cross validation and external validation were used to examine model performance. Generally, multivariate models had similar errors and performed better than the simple linear regression models. The lowest external and internal error values (3.8 and 2.59 days, respectively) were obtained using GDH with the multivariate linear model for March and CP for November. Regarding practicality, we recommend the use of the linear regression model for CP data from January (external error = 5.7 days, internal error = 5.64 days). The non-significant correlation between one month pre-season temperatures and first flowering date indicates that flowering time is primarily determined by earlier temperature accumulation. The significant influence of chilling accumulation in January on the onset of flowering are in coincidencewith that X. sorbifolium is an early flowering species. The stable external errors related to the use of GDH for March data and CP for November data implies that the combination of forcing and chilling accumulation plays a critical role in initiating flowering onset. Our study provides a basis for forecasting flowering onset and regulating the floral development of X. sorbifolium.