Phenology is an important adaptive trait since it determines the duration and timing of the growing season as well as the period of reproduction (Chabot & Hicks 1982; Lechowicz 1984; Kikuzawa 1989; Hänninen 1990; Reich et al. 1992). In the context of the climate global warming, many studies have attempted to predict the consequences of increasing temperatures on the phenology of temperate zone trees to determine whether species would break bud or flower later or earlier, and thus experience increased or decreased risks of frost damage (Cannell & Smith 1986; Prentice et al. 1991; Hänninen et al. 1993; Kramer 1994a; Murray et al. 1994; Kramer 1995; Hänninen 1996; Kramer & Mohren 1996; Kramer et al. 1996). In the last 30 years, many studies have developed and tested predictive models of tree phenology using climatic variables such as temperature and photoperiod (Cannell & Smith 1983; Cannell 1989; Murray et al. 1989; Hunter & Lechowicz 1992; Kramer 1994b; Chuine et al. 1998, 1999). Using such models over the entire range of a species requires, however, that model estimates are valid across this range, i.e that there is no significant genetic variation of phenology between the different populations. Such variation could be due to either local adaptation or drift, but both causes require that the trait considered be under genetic control and show genetic variability.
For many plant species, phenology is known to be a variable character with a high degree of heritability (Billington & Pelham 1991; El-Kassaby & Park 1993; Farmer 1993), and in these cases it can therefore be modified by natural selection. Given the large size of most tree populations, genetic variation (if it exists) is unlikely to be explained by drift, but rather by local adaptation. Local adaptation results from a balance between natural selection and gene flow and will occur if selection is stronger than gene flow. Although, most tree species of temperate and boreal zones show high potential gene flow due to pollen (dispersion up to hundreds of kilometres, see Faegri & Iversen 1989 for a review) and seeds (migration distance over hundreds of metres or even kilometres, reviewed by Huntley & Birks 1983 and Delcourt & Delcourt 1991), gene flow can be low because of the non-overlapping reproductive periods of populations that live at different elevations (Phillips & Brown 1977; El-Kassaby et al. 1984). Climatic variables such as temperature may, however, constitute strong selective forces on phenology, although the plasticity of this trait in response to climate may reduce its selective action. In addition, since temperature and frost are highly variable from one year to another, different genotypes are probably favoured by selection at different times. In conclusion, we would therefore expect that the balance between the factors contributing to local adaptation is unlikely to lead to significant genetic variation in phenology between populations of most lowland tree species.
We analysed genetic variation in the timing of flowering between populations of six European angiosperm tree species using phenological models and flowering dates in different locations. First, using data for one of this species and a further three clonal species at a number of locations, we verified that model estimates for an individual genotype do not vary with environment. Second, by fitting model parameters to the dates of flowering of different populations of the six species, we analysed the pattern of variation between populations. Third, using the fitted models, we studied the response of each population to hypothetical transfers. We address the following questions: (i) do model parameters adequately account for genotype and genotype × environment effects; (ii) do the populations studied show significant genetic variation in their timing of flowering; and (iii) what are the implications of the patterns of genetic variation observed for predicting phenology under a global climate warming scenario?