How useful is genomic data for predicting maladaptation to future climate?

Abstract

Methods using genomic information to forecast population maladaptation to climate change are becoming increasingly common, yet the lack of model validation poses serious hurdles toward their incorporation into management and policy. Here, we compare the validation of two methods - Gradient Forests (GF) and the Risk Of Non-Adaptedness - using exome capture pool-seq data from 35 to 39 populations across three conifer taxa: two Douglas-fir varieties and jack pine. We evaluate sensitivity of these algorithms to the number of input populations as well as the source and number of input loci (markers selected from genotype-environment associations [GEA] or those selected at random). We validate these methods against two-year and 52-year growth and mortality measured in independent transplant experiments. Overall, we find that both genetic offset methods often better predict transplant performance than climatic or geographic distances. We also find that while GF models are surprisingly not improved using GEA outliers, they are sensitive to the populations included in analysis. Even with promising validation results, ambiguity of model projections to future climates makes it difficult to identify the most maladapted populations using either method. Our work advances understanding of the sensitivity and applicability of these approaches, and we discuss recommendations for their future use.

Novel genomic offset metrics integrate local adaptation into habitat suitability forecasts and inform assisted migration

“…In two of our common gardens, scaled genomic offsets better predicted growth performance than scaled climate transfer distances (Figure 3). This is in accordance with results on the performance of populations of lodgepole pine ( Pinus contorta ) and balsam poplar ( Populus balsamifera ) in common gardens (Capblancq & Forester, 2021; Fitzpatrick et al, 2021; Lind et al, 2023). These findings are likely to be attributable to the fact that GF turnover functions involve re‐weighting climate predictor variables based on their ability to explain variation in climate‐associated genomic composition (Fitzpatrick & Keller, 2015).…”

“…In two of our common gardens, scaled genomic offsets better predicted growth performance than scaled climate transfer distances (Figure 3). This is in accordance with results on the performance of populations of lodgepole pine (Pinus contorta) and balsam poplar (Populus balsamifera) in common gardens (Capblancq & Forester, 2021;Fitzpatrick et al, 2021;Lind et al, 2023).…”

“…Additionally, caution should be applied, as the same concerns about the intermingling of adaptive and neutral signals that apply to the use of genomic offsets more broadly also apply to the use of scaled offsets. Last, the need for rigorous experimental validation assessing the fitness response to predicted offsets using common gardens or similar approaches is essential before offsets can gain confidence for use in conservation management planning (see also Lind et al, 2023). Nevertheless, these challenges are rather general cautions related to the complexity of determining true adaptive signals, the use of correlative models for forecasting, and the extrapolation of such models beyond the training data, but do not undermine the broader utility of the innovations proposed in this manuscript.…”

“…To be most informative and to avoid confounding effects of nonadaptive genetic differentiation, genomic offsets ideally should be based on loci that have been identified as potentially adaptive via association genetic methods linking genomic variation to locally adaptive traits (GWAS) or to environmental gradients themselves (genotype–environment association; GEA). In addition, genomic offsets should be validated by analyzing the fitness effects of the estimated G–E disruption, ideally in controlled experiments (Capblancq, Fitzpatrick, et al, 2020; Fitzpatrick et al, 2021; Lind et al, 2023; Rellstab et al, 2021). Further, genomic offsets have great untapped potential for advancing predictive provenancing through a genomically informed approach to assisted migration.…”

Spatial autocorrelation of the environment influences the patterns and genetics of local adaptation

“…Such methods analyse present-day relationships between allele frequency and the environment to predict how species will fare given predicted patterns of environmental change. If the agents of local adaptation are highly spatially autocorrelated, neutral population structure may be partially aligned with gradients of selection, which could explain why the use of “adaptive” genetic markers and randomly chosen markers seem to perform equally well in some offset analyses (Fitzpatrick et al, 2021; Láruson et al, 2022; Lind et al, 2023). Violating the assumption of homogeneous local adaptation in offset analyses, for example, would likely introduce noise into predictions but could potentially lead to spurious results.…”