It is well known that forecasting the flowering time of wild vegetation is
useful for various sectors of human activity, particularly for all agricult
ural practices. Therefore, continuing previous work by Cenci et al., we wil
l present here three new phenoclimatic models of the flowering time for a s
et of wild species, based on an original data sample of flowering dates for
more than 500 species, observed at Guidonia (42 degrees N in central Italy
) by Montelucci in the period 1960-1982. However, on applying the bootstrap
technique to each species sample to check its basic statistical parameters
, we found only about 200 to have data samples with an approximately Gaussi
an distribution. Eventually only 57 species (subdivided into eight monthly
subsets from February to September) were used to formulate the models satis
factorily. The flowering date (represented by the z variable), is expressed
in terms of two variables x and y by a nonlinear equation of the form z=al
pha x(beta)+gamma y. The x variable represents either the degree-day sum (i
n model 1), or the daily-maximum-temperature sum (in model 2), or the daily
-global-insolation sum (in model 3), while 4 for all three models correspon
ds to the rainy-day sum. Note that all summations involved in the computati
on of the variables x and y take place over a certain period of time (prece
ding the flowering phase), which is a parameter to be determined by the fit
ting procedure. This parameter, together with the threshold temperature (ne
eded to compute the degree-days in model 1), represents the two implicit pa
rameters of the process, thus the total number of parameters (including the
se last two) becomes respectively, five for model 1, and four for the other
two models. The preliminary results of this work were reported at the XVI
International Botanical Congress (1-7 August 1999, St. Louis, Missouri USA)
.