Table 5.

Segmentations of series of mean temperatures during the HA period using piecewise constant models (2 or 1 segment when the 2-segment model was irrelevant): recording period, change-point instant and amplitude, global standard deviation, optimal segmentation posterior probability, model posterior probability, mBIC model, correlation coefficients between flowering date (BBCH 61 stage for all locations except Conthey–BBCH 65 stage) and mean temperature during the HA period (^**significant at 1% level; ^*significant at 5% level; n.s., non-significant).

Location	Recording period	Change point		Standard deviation	Posterior probability		mBIC model	Correlation coefficient
		Instant	Amplitude		Segmentation	Model		Golden D.	Gala	Fuji
Angers	1963–2013	1987	1.35	0.94	0.63	0.16	3	−0.77**
Nîmes	1974–2013	1988	1.31	0.88	0.57	0.37	3	−0.7**	−0.72**	−0.76**
	1966–2013	1988	1.26	0.9	0.56	0.45*
Forli	1970–2013	1988	0.91	0.92	0.26	0.42	1	−0.8**
Trento	1983–2013	1988	1.72	0.89	0.54	0.81*		−0.78**
Gembloux	1984–2013	1988	2.84	1.19	0.48	0.46*		−0.81**
†	1964–2013	1989	1.49	1.16	0.29	0.57*
Bonn	1959–2013	1989	1.34	1.26	0.27	0.59*		−0.84**
Conthey	1970–2013	1988	1.31	1.06	0.31	0.7*		−0.78**	−0.74**
Ain Taoujdate	1984–2013	1993	1.32	0.86	0.33	0.75*		−0.47**
	1973–2013	1993	1.21	1.08	0.14	0.52*
Sao Joaquim	1972–2013	-	-	0.76	1	0.91*		−0.34*	−0.27 n.s.	−0.12 n.s.
	1955–2013	-	-	0.79	1	0.97*
Caçador	1982–2013	-	-	0.69	1	0.74*		−0.23 n.s.	−0.2 n.s.	−0.05 n.s.
	1961–2013	-	-	0.8	1	0.76*

In the model posterior probability column, an asterisk indicates that the model is the one given by mBIC. If this is not the case, the model given by mBIC is indicated in the next column (mBIC model). We analyzed the mean temperatures during the HA period over the range of years corresponding to the flowering dates and in certain cases (Nîmes, Gembloux, Ain Taoujdate, Sao Joaquim, Caçador) over an extended range of years.

^†Gembloux (1964–2013): Posterior probability of 0.29 for 1989 but of 0.27 for 1988.