Projected Climate Response

The predicted climate response fluctuated annually (Fig. 3) due to the variability of temperature projections, although projected conditions all trended towards warming (Appendix S6). Earlier bolting was predicted in CEUR (both cohorts) and SMER, contrasting with the delay in bolting predicted for the fall cohorts in SCAN and NMER. Interestingly, the magnitude of this delay is lower under RCP8.5 than RCP2.6 in SCAN (mean change in DTB for 2090-2099 under RCP8.5ΔDTBRCP8.5 = +2 days, under RCP2.6ΔDTBRCP2.6 = +22 days), whereas the opposite was predicted for NMER (ΔDTBRCP8.5 = +18 days,ΔDTBRCP2.6 = +1 day).
As for DTB, the predicted change in SP differed between fall and non-fall cohorts, although this should be interpreted cautiously because SP was predicted much more poorly. Spring and summer cohorts (CEUR, SMER) were predicted to greatly decline in fecundity under both climate change scenario, whereas the mean change in SP in 2090-2099 issmaller for fall cohorts in SCAN (mean SP in 2006SP2006 = 10264.25,ΔSPRCP8.5 = +1172.466,ΔRFRCP2.6 = -2731.709) and NMER (SP2006 = 9413.171,ΔSPRCP8.5 = +127.441,ΔSPRCP2.6 = -2860.370). Interestingly for SCAN and NMER, fecundity was predicted to increase under the more severe RCP8.5 and decrease under the milder RCP2.6.
We next compared the predicted change in days-to-bolting and seed proxy for different landscapes of A. thaliana genetic variation but strictly comparing 2006 and 2099. Predicted change in days-to-bolting is described in Appendix S7. Here, we focused on predicted change in seed proxy despite the lower prediction accuracy of SP because its consequence on plant fitness is straightforward to interpret. Under the baseline scenario using the krieged genetic landscape, which assumes the current distribution of genetic variation remains constant, we predicted a decrease in fecundity throughout most of A. thaliana ’s European range (Fig. 4a). We considered a hypothetical revegetation scenario where Eden-2 or Ll-2 genotypes were introduced throughout Europe. Despite identical environmental conditions, the fecundity response differed dramatically between genotypes (Fig. 4b-c). Eden-2 had higher predicted fecundity than local genotypes in the north, particularly in its native Scandinavian range, but had lower fecundity than local genotypes in the south. In contrast, Ll-2 had higher predicted fecundity than local genotypes in both its native Mediterranean range and parts of Scandinavia, although it was still less fecund than Eden-2 in the north (Fig. 4d).
Discussion
The interaction between genotype and the environment greatly differed across regions and was shown to be important for predicting plant response to climate change. This can be modelled by including environmental (here, temperature) predictors in a quantitative trait model, which improves prediction accuracy and allows forecasting the climate response for different genetic and climate scenarios. For the different scenarios tested in European A. thaliana , we predicted heterogeneous climate response depending on both the distribution of genetic variation and spatiotemporal patterns of temperature variation. Our predictions highlighted the complexity of the response to climate change and the breakdown of local adaptation over time that should be accounted for when envisioning ecological restoration.