A model framework for tree leaf colouring in Europe

Ecosystem productivity is influenced by the start and end of the growing season, and ecosystem models that simulate productivity need reliable representations of the phenology. For the seasonal development, autumn events are less understood than spring events, with comparatively fewer modelling attempts have been made for leaf senescence than for budburst. The few existing models for autumn phenology represent the influence of temperature and photoperiod. In this study, the aim was to evaluate which type of temperature response, photoperiod requirement and interaction between temperature and photoperiod captured the variation in leaf colouring more accurately. We tested existing models on a large dataset and developed new models by combining seven model components: linear or sigmoid temperature response above or below a base temperature, with or without modification by photoperiod, and photoperiod requirements (starting day of temperature response). Potential photoperiod requirement for leaf senescence induction was assessed by using a calibrated starting day or day of budburst that instead of a requirement represent the start of the ageing processes. Day of leaf colouring was simulated using 37 models for birch, beech and oak in Austria, Germany and the United Kingdom, in total 111 model runs that was compared to average day of leaf colouring. In 109 out of 111 simulations, average day of leaf colouring provided a better estimate. Some of the better performing models resembled average day of leaf colouring by counting number of days. Overall, the results indicate that the models estimated response to temperature and photoperiod do not support the use of a fixed degree-day requirement, especially across large regions. No photoperiod requirement could be inferred, and photoperiod in combination with temperature response provided little or no improvement on model performance.