When and how can we predict adaptive responses to climate change?

Abstract Predicting if, when, and how populations can adapt to climate change constitutes one of the greatest challenges in science today. Here, we build from contributions to the special issue on evolutionary adaptation to climate change, a survey of its authors, and recent literature to explore the limits and opportunities for predicting adaptive responses to climate change. We outline what might be predictable now, in the future, and perhaps never even with our best efforts. More accurate predictions are expected for traits characterized by a well-understood mapping between genotypes and phenotypes and traits experiencing strong, direct selection due to climate change. A meta-analysis revealed an overall moderate trait heritability and evolvability in studies performed under future climate conditions but indicated no significant change between current and future climate conditions, suggesting neither more nor less genetic variation for adapting to future climates. Predicting population persistence and evolutionary rescue remains uncertain, especially for the many species without sufficient ecological data. Still, when polled, authors contributing to this special issue were relatively optimistic about our ability to predict future evolutionary responses to climate change. Predictions will improve as we expand efforts to understand diverse organisms, their ecology, and their adaptive potential. Advancements in functional genomic resources, especially their extension to non-model species and the union of evolutionary experiments and “omics,” should also enhance predictions. Although predicting evolutionary responses to climate change remains challenging, even small advances will reduce the substantial uncertainties surrounding future evolutionary responses to climate change.


Investigating changes in heritability expected under climate change
We aimed to assess whether conditions expected under climate change would affect evolutionary potential.For that purpose, we searched on Web of Science on 26/04/2022 for articles using keywords "heritability", "additive genetic variance", or "evolvability" associated with either "climate change", "global warming", or "extreme climatic event." This search led to 535 references, the list was reduced to 97 studies based on reading of the abstract, retaining publications where heritabilities estimates were given under at least two conditions (so this excludes estimates of heritability of tolerance), and including effects expected under climate change such as increased temperature but also changes in precipitation/humidity; or increased CO2 and acidity for aquatic environments.For each study, one reviewer noted heritability, evolvability and additive genetic variance when available, the type of stress (heat, drought, acidity, wetness, hypoxia, salinity, carbon dioxide), the type of heritability (broad vs narrow sense) and the trait type (life history, morphology, physiology, behavior).When reported, 95% confidence intervals were transformed into standard errors using the formula, . When compiling the data, we removed studies investigating the heritability of tolerance, providing a single heritability estimate by pooling together data from multiple environments, or not clearly indicating the reference point with future environments.Evolvability was measured as additive genetic variance divided by trait mean (Houle, 1992).
The analyses were run using MCMCglmm (Hadfield, 2010) with a Gaussian distribution.We analyzed the raw values and tested for a difference between normal and future climates.All models included organism class and study number as random effects.Although the model including only the intercept never had the best support, we found no significant effects of stress from climate change on heritability or evolvability estimates.Note -Including only studies where standard errors are reported reduced the data set to 536 estimates from 26 studies across 23 species.In this model, we used the mev function in MCMCglmm to account for sampling variance around the estimates.We used a non-informative prior with V = 1 and nu = 0.02.

Table S1 :
Studies included in the heritability and evolvability meta-analysis.38

Table S2 :
Model selection for evolvability analysis

Table S3 :
Summary of best model (M10) for evolvability analysisNote -Morphological traits have lower evolvability in drought experiments, but not in relation to stress itself.

Table S4 :
Model selection for heritability analysis including all estimates.Note -Many studies did not report errors around heritability estimates, so we first looked at the effect of stress and other factors on heritability.This analysis relied on 677 estimates from 37 studies across 35 species.The best model is the one with lowest DIC.

Table S5 :
Output from the best heritability model (M10).Life history traits had larger heritability under drought (aridity experiments in 60 plants + pond drying in amphibians).61

Table S6 :
Model selection for heritability analysis including only studies where standard errors were reported.

Table S7 :
Summary of the best model (M2) for heritability analysis including only studies where standard errors were reported.

Table S8 :
Model selection for heritability analysis including only studies where standard errors were reported AND restricting to the most common factors: heat and drought.Note -This analysis was based on 429 observations for 20 studies across 18 species.

Table S9 :
Summary of best model (M4) for heritability analysis including only studies where standard errors were reported AND restricting to the most common factors: heat and drought.