Alternative modelling approaches significantly differ in simulating summer crops phenology in Mediterranean Europe

Variations in temperature trends are considerably impacting plants' phenology. Most predictive models share the concept of Growing Degree Days (GDDs). Among available formulations, the ones not considering the effects of high temperatures on plants' development seem no longer adequate, due to the increasing frequency of heat waves, leading to misinterpretation of climate effects. The aim of the present work is to compare six different degree-days models, in order to assess which of them could give the best results in terms of GDDs calculation for summer crops in Mediterranean Europe. Specifically, average method, single triangle method (with also three different cut-off techniques: horizontal, vertical, intermediate) and beta-distribution function method were tested. For this purpose 22 years of phenological data were used, comparing "standard" and "warm" years (with average temperature during June - August below and above, or equal to, the median value of the 22-years period, respectively). Models were compared via Mean Absolute Error (MAE), Mean Error (ME), Root Mean Square Error (RMSE) and Diebold-Mariano test, to assess differences in their predictive performance. Results showed that the use of models considering the negative effects of high temperatures in the ripening period significantly boosted predictive accuracy. Among these approaches, the physiologically based beta-distribution function provided the best results. However, simpler methods, which could facilitate the acquisition of modelling novelty in operational contexts, having the advantage of being easy-to-use also proved to be significantly improving, such as intermediate cut-off technique, which among geometrical models can be considered the best approximation of crops physiological response.