Bayesian analysis of temperature sensitivity of plant phenology in Germany

Abstract The temperature sensitivity of 43 phenological phases was analysed in Germany within the period 1951–2006 with the help of a Bayesian approach. First a Bayesian model comparison of monthly temperatures and phenological phases throughout the year was conducted. We analysed the data as constant (mean onset date), as linear (constant trend over time) and as change point model (time varying change). The change point model involves the selection of two linear segments which match at a particular time. The matching point is estimated by an examination of all possible breaks weighted by their respective change point probability. Secondly a Bayesian coherence analysis was applied to investigate the relationship between phenological onset dates and an effective temperature generated as a weighted average of monthly means. Temperature weight coefficients were obtained from an optimization of a coherence factor by simulated annealing. Results reveal that late spring, summer and early autumn temperature months exhibit a clear preference for the change point model (>50%) indicating nonlinear change. The temperature development of April and August shows exceptionally high nonlinearities compared to the other months with change point model probabilities of 78% and 81% over the last five decades. For all phenophases a strong dependence of phenology on temperature is determined. We can classify two main temperature response patterns of the studied phenological phases: on the one hand spring phenophases are particularly sensitive to temperatures in April, exhibiting a prompt response. On the other hand summer phenophases are less influenced by temperature during or right before the month of the onset. They reveal a delayed response to nonlinear temperature changes mainly of April. Especially abrupt changes during the temperature sensitive stage of species cause a pronounced change in plant phenology regardless of the time of onset.