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.