A double-threshold temperature sum model for predicting the flowering duration and relative intensity of Betula pendula and B. pubescens

The accurate prediction of the flowering period and the amount of released pollen of wind-pollinated trees is important for predicting the aerobiological pollen amounts and for estimating the intensity of background pollination of orchard seed production. Current phenological models, such as the commonly used temperature sum (thermal time) model, are efficient for predicting the timing of point events, e.g. the onset of flowering or leaf unfolding. However, the flowering period of boreal trees, including Betula pendula and B. pubescens, may last up to 5 weeks with widely varying flowering intensity. Therefore, the ability to predict the duration of the whole flowering period and the intensity of pollination is very important.In this paper, we present a new phenological model that predicts the whole flowering period and daily (normalized) variation of flowering intensity. The model is based on similar principles as the temperature sum model. The model was calibrated against pollen data recorded with pollen traps placed in several locations in Southern and Central Finland. Our new model predicted accurately the variation in the intensity throughout the flowering period. Moreover, it was able to predict the start and end of the pollination season with accuracy comparable to that achieved with an ordinary temperature sum model for point events. Due to limited amount of test data, the data-dependency of the model had to be tested with a bootstrapping approach. With this method, the model fit and parameter values showed to be independent of the parameter fitting data. The model can be used to predict the whole flowering period of wind-pollinated boreal trees for many different purposes, such as aerobiological forecasts of allergenic pollen, or generation of input data for models of long-range pollen transport.

Research highlights▶ A new model describing the variation of pollen release through the flowering period. ▶ Gives daily predictions of pollen release based on temperature sum accumulation. ▶ Can predict onset and end of flowering with similar accuracy as point-event models. ▶ Can be utilized for aerobiological forecasts or with models of long-range transport.