The identification of changes in observational data relating to the climate change hypothesis remains a topic of paramount importance. In particular, scientifically sound and rigorous methods for detecting changes are urgently needed. In this paper, we develop a Bayesian approach to nonparametric function estimation. The method is applied to blossom time series of Prunus avium L., Galanthus nivalis L. and Tilia platyphyllos SCOP. The functional behavior of these series is represented by three different models: the constant model, the linear model and the one change point model. The one change point model turns out to be the preferred one in all three data sets with considerable discrimination of the other alternatives. In addition to the functional behavior, rates of change in terms of days per year were also calculated. We obtain also uncertainty margins for both function estimates and rates of change. Our results provide a quantitative representation of what was previously inferred from the same data by less involved methods.