Phenological models for evaluating the effects of climatic trends on apple cv. “Golden Delicious” flowering dates

In a global warming context, the trend of frost risk is a general topic of concern in fruit farming. A warmer climate is supposed to anticipate flowering dates to some extent, exposing plants in a frost-prone phase to temperatures which can turn out to be higher or lower than in the past. This work aims to the development of a high-performing flowering model for apple, allowing a frost risk assessment for future years in an important fruit-growing area.

Several models based on temperature records have been tested and validated with phenological series from two locations at different altitudes (210 and 650 m a.s.l.). For these sites both meteorological and phenological observations are available for the period 1983 - 2003. The best performing models have been determined by an evaluation of several statistical indices on the differences between simulated and recorded flowering dates: RMSE, median of absolute errors and correlation. The influences of day length and site-depending parameters on flowering dates have been evaluated. Also, various hypothesis on the relationship between the efficiency of growing degree hours (GDH) and the total GDH’s sum requirement have been tested.

The model giving the best estimate, called MISTO, is essentially an adaptation of the “Utah model” (Richardson et al., 1974; Valentini et al., 2001) to the use of hourly temperature data with some modifications. First, the cold requirement is evaluated as Chilling Units (CU) sum to reach a fixed value; then the growing GDH are accumulated to build a heat sum. CU are cumulated as in “Utah model” while GDH efficiency at one day is a function of the fraction of accumulated heat till that day, with reference to the total required value. Values of CU’s and GDH’s thresholds and heat requirements have been varied in order to obtain the best estimate. The model yielded a mean absolute error lower than 2 days, and it showed a very good skill even for years with a pronounced advance or delay of flowering. The need for one model encompassing different climatic conditions (namely valley bottom and hillside) suggested to take into account flowering data from more locations, available and adapted from other sources (Chuine, 1998). Also in this case, the model showed a satisfactory skill in reproducing flowering dates in sites with various altitudes and different morphology and climate characteristics.

Finally, the model has been applied with temperature series derived from several scenarios inferred from IPCC’s SRES (Special Report on Emission Scenarios, 2001), to estimate effects on flowering dates. The results show a general trend in anticipating the bloom date, which partially confirms the well detectable trend in the record of the last 20 years.