Forecasting climate change response in an alpine specialist songbird reveals the importance of considering novel climate

Species persistence in the face of climate change depends on both
ecological and evolutionary factors. Here, we integrate ecological and
whole-genome sequencing data to describe how populations of an alpine
specialist, the Brown-capped Rosy-Finch (Leucosticte australis) may be
impacted by climate change. We sampled 116 Brown-capped
Rosy-Finches from 11 sampling locations across the breeding range. Using
429,442 genetic markers from whole-genome sequencing, we described
population genetic structure and identified a subset of 436 genomic
variants associated with environmental data. We modelled future climate
change impacts on habitat suitability using ecological niche models (ENMs)
and impacts on putative local adaptation using gradient forest models (a
genetic-environment association analysis; GEA). We used the metric of
niche margin index (NMI) to determine regions of forecasting uncertainty
due to climate shifts to novel conditions. Population genetic
structure was characterized by weak genetic differentiation, indicating
potential ongoing gene flow among populations. Precipitation as snow had
high importance for both habitat suitability and changes in genetic
variation across the landscape. Comparing ENM and gradient forest models
with future climate predicted suitable habitat contracting at high
elevations and population allele frequencies across the breeding range
needing to shift to keep pace with climate change. NMI revealed large
portions of the breeding range shifting to novel climate
conditions. Our study demonstrates that forecasting climate
vulnerability from ecological and evolutionary factors reveals insights
into population-level vulnerability to climate change that are obfuscated
when either approach is considered independently. For the Brown-capped
Rosy-Finch, our results suggest that persistence may depend on rapid
adaptation to novel climate conditions in a contracted breeding range.
Importantly, we demonstrate the need to characterize novel climate
conditions that influence uncertainty in forecasting methods.