Variance partitioning and identification of important climate
variables
The RDA including all loci with climate and population structure as
predictors explained 71.4% (adjusted R2 = 0.714) of
the genetic variance among populations (F = 4.06, p >
0.001) (Figure 1c). Using partial redundancy analysis on the full data
set, we found an effect of population structure after controlling for
the effects of climate (F = 3.02, p = 0.01), explaining 38.2% of
genetic variance. However, we found no effect of climate alone after
controlling for the effects of population structure (F = 1.34, p
> 0.2). Each climate variable independently explained only
a small amount of the total genetic variance (Table 3).
In the subset of 485 outlier loci that were detected by the global RDA
analysis, the full RDA model explained 51.1% of the genetic variance in
these outliers (F = 3.73, P = 0.03). Climate accounted for 46.7% of the
total amount of genetic variation, and the joint effects of climate and
population structure was reduced to 9.8% (Figure S2).
Table 3. Amount of genetic variance explained by each climate variable
after removing the effect of the other variables including population
structure. Results are based on a partial redundancy analysis (pRDA) of
climate variables after controlling for confounding population structure
for all 21,892 SNPs, or 485 outlier SNPs identified by the RDA analysis.