Assessing uncertainty in genomic offset forecasts from landscape genomic models (and implications for restoration and assisted migration)

Abstract

IntroductionEcological genomic models are increasingly used to guide climate-conscious restoration and conservation practices in the light of accelerating environmental change. Genomic offsets that quantify the disruption of existing genotype–environment associations under environmental change are a promising model-based tool to inform such measures. With recent advances, potential applications of genomic offset predictions include but are not restricted to: (1) assessing in situ climate risks, (2) mapping future habitat suitability while accounting for local adaptations, or (3) selecting donor populations and recipient areas that maximize genomic diversity and minimize maladaptation to future environments in assisted migration planning. As for any model-based approach, it is crucial to understand how arbitrary decisions made during the modeling process affect predictions and induce uncertainty.MethodsHere, we present a sensitivity analysis of how various modeling components influence forecasts of genomic offset-based metrics, using red spruce (Picea rubens), a cool-temperate tree species endemic to eastern North America, as a case study. We assess the effects of genomic marker set, climatic predictor set, climate change scenario, and “not-to-exceed” offset threshold and evaluate how uncertainty in predictions varies across space.ResultsClimate change scenario induced by far the largest uncertainty to our forecasts; however, the choice of predictor set was also important in regions of the Southern and Central Appalachians that are of high relevance for conservation and restoration efforts. While much effort is often expended in identifying candidate loci, we found that genomic marker set was of minor importance. The choice of a maximum offset threshold to limit transfers between potential donor and recipient locations in assisted migration programs has mostly affected the magnitude rather than geographic variation in our predictions.DiscussionOverall, our model forecasts suggest high climate change risks across the entire distributional range of red spruce and strongly underscore the potential for assisted migration to help ameliorate these risks. In that regard, populations in the Southern and Central Appalachians as well as along the US and Canadian east coast seem the best candidates for both in situ conservation and relocation.

The limits of predicting maladaptation to future environments with genomic data

“…Because of the assumption that locally adapted populations will be necessary for satisfactory model performance, initial implementations of genomic offset models focussed on putatively adaptive markers where this signal may be strongest (Keller & Fitzpatrick, 2015; Rellstab et al, 2016). More recently, investigators have varied the set of markers used to train models but have found little influence on performance (Fitzpatrick et al, 2021; Lachmuth, Capblancq, Keller, et al, 2023; Láruson et al, 2022; Lind et al, 2024). Our results are similar to previous investigations, finding that the adaptive marker sets provide minimal advantage over all or neutral marker sets, but not universally or by great margins.…”

“…One hypothesis put forth as to why adaptive marker sets perform similarly to all markers is that genome-wide data captures sufficient signatures of IBE (Lachmuth, Capblancq, Keller, et al, 2023; Lind et al, 2024). Our analysis found weak positive relationships between performance and levels of IBE within marker sets.…”

“…We observed performance declines below the analogous z E =3.13 Climate Novelty scenario, indicating offset predictions will likely be inaccurate in many real-world climate change predictions. These issues are also germane to measures derived from offset values (Gougherty et al, 2021; Lachmuth, Capblancq, Keller, et al, 2023; Lachmuth, Capblancq, Prakash, et al, 2023), which currently do not consider the degree of climate novelty in the prediction (but see DeSaix et al, 2022).…”

“…Before genomic offsets can be incorporated into management plans, considerable thought must be put into the sensitivity of model outcomes from input data (Lind et al, 2024), the uncertainty inherent in environmental or climate forecasts (Lachmuth, Capblancq, Keller, et al, 2023), as well as the degree of novelty of future climates (DeSaix et al, 2022, this study). While accurate predictions are limited for novel climates of the future, these offset methods could still be used to guide management in the intervening time in a stepwise manner where experiments can be used to validate model performance in practice.…”

“…If the marker set has little impact on offset performance, this could indicate that offset methods are giving weight to genome-wide signals present in the data. Previously, some (e.g., Lachmuth, Capblancq, Keller, et al, 2023; Lind et al, 2024) have postulated that this signal may be related to isolation by environment ((IBE, i.e., when genetic and environmental distances are positively correlated, independent of geographic distance; Wang & Bradburd, 2014).…”