The effect of region of origin and maternal environment on dormancy level
Seeds produced by Italian populations had on average stronger seed dormancy 12 weeks after seed maturation than had seeds produced by Fennoscandian populations both at the Italian field site and in the greenhouse (F1,41 = 69.7, p < 0.001,Fig. 3A and Table 1 ). At the Swedish field site, very few Italian populations managed to produce sufficient amounts of seeds to allow germination tests. However, the few lines tested tended to produce seeds with stronger dormancy compared to the seeds produced by Fennoscandian populations also at that site (Fig. 3A andTable 1 ).
Germination proportions were on average 37% higher among seeds matured at the Italian field site compared to those matured in the greenhouse (F1,331 = 15.4, p < 0.001), and this effect did not vary with region of origin (region of origin × maternal environment interaction, F1,331 = 1.2, p = 0.27;Fig. 3A and Table 1 ). Germination proportions varied among populations within regions (L-ratio6,7 = 107.5, p < 0.001, Fig. 3A and Table 1 ). Although germination proportions 12 wk after seed maturation in each of the three maternal environments were positively correlated (Spearman rank correlation, Italy field vs. Sweden field, rs = 0.63, N = 30, p < 0.001; Italy field vs. Greenhouse, rs = 0.52, N = 43, p < 0.001; Sweden field vs. Greenhouse, rs = 0.88, N = 32, p < 0.001), the effect of population differed between the two maternal environments (significant population × maternal environment interaction, L-ratio7,9 = 158.8, p < 0.001, Fig. 3A and Table 1 ). The interaction reflected the fact that variance in germination proportion among Fennoscandian populations 12 weeks after seed maturation was much larger when seeds had been produced in the greenhouse compared to when matured at the Italian site, whereas for Italian populations the opposite was true (Fig. 3A ). For both regions of origin, the highest variance in population means was observed when the overall mean germination proportion was intermediate (seeds of Fennoscandian populations grown in the greenhouse, and of Italian populations grown at the Italian field site) rather than very high (Fennoscandian populations at the Italian field site) or very low (seeds of Italian populations in the greenhouse; Fig. 3A ). The seeds of all populations except one (Ørnes, the northernmost sampled population from Fennoscandia) had higher germination proportions when matured in Italy compared to in the greenhouse. Removal of this population from the dataset did not affect the significance of the population × maternal environment interaction (not shown).
Germination proportions of seeds produced at the two field sites varied among Fennoscandian populations (F17,113 = 11.3, p < 0.001, Fig. 3Aand Table 2 ), and the variance among populations was markedly larger for seeds produced at the Swedish field site compared to seeds produced at the Italian field site (maternal environment x population interaction, F17,113 = 5.4, p < 0.001,Fig. 3A and Table 2 ).
Among seeds matured in the greenhouse, Fennoscandian populations had on average higher germination proportions than had Italian populations in all tests conducted during one year of after-ripening. The average germination proportion of Fennoscandian populations increased from about 10% one week after seed harvest to 50% eleven weeks later, and reached a maximum of about 75% after one year of after-ripening (Fig. 3B ). By contrast, the average germination proportion of Italian populations was below 17% during the first 12 weeks of after-ripening, and reached 50% only after one year (Fig. 3B ). Still after 54 weeks of after-ripening, germination proportion varied widely among both Fennoscandian and Italian populations (Fig. 3B ).