Robustness of resistance surface optimisations: sampling schemes and genetic distance metrics affect inferences in landscape genetics

• Published in: Landscape ecology Vol. 38; no. 11; pp. 2861 - 2883
• Main Authors: Schleimer, Anna, Luttringer, Amanda, Wittische, Julian, Drygala, Frank, Proess, Roland, Cantú-Salazar, Lisette, Frantz, Alain C.
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 01.11.2023; Springer Nature B.V
• Subjects: Biomedical and Life Sciences; Clustering; Conservation; Convergence; Ecology; Environmental Management; Errors; Gene mapping; Genetic distance; Genetics; Genotypes; Habitats; Landscape Ecology; Landscape/Regional and Urban Planning; Life Sciences; Management decisions; Nature Conservation; Optimization; Performance evaluation; Planning; Ponds; Population genetics; Research Article; Robustness; Sampling; Sampling designs; Sustainable Development; Connectivity; Genetic distances; Landscape genetics; Resistance surfaces; ResistanceGA
• ISSN: 0921-2973; 1572-9761
• DOI: 10.1007/s10980-023-01752-5

Context Landscape genetics provides powerful tools to quantify the effects of landscape features on population connectivity, but robust results are imperative to inform conservation planning. Objectives The robustness of landscape genetic inferences was assessed using the case of the northern crested newt ( Triturus cristatus ) in Luxembourg. Specifically, the effect of different study designs and genetic distance metrics was tested in terms of model convergence and misspecification rates (Type I error). Methods The optimisation of resistance surfaces was performed in ResistanceGA, using individual- and population-based sampling designs and 16 genetic distance metrics inferred from 897 multilocus genotypes from 85 locations. Empirical results were complemented with simulations to assess Type I error rates and correlation between ‘true’ and optimised resistance surfaces. Results Individual-based optimisations seemed prone to overfitting, with little convergence among empirical resistance surfaces from different sets of individuals. Simulations showed significant differences in performance among population genetic distance metrics. Linear topographical features exhibited higher Type I error rates (83.3%) than continuous features (44.9%), suggesting potential underestimation of road-induced fragmentation effects. Jost’s D , F ST , and PCA axes 1–45 were the top three genetic distance metrics for recovering true resistance features. Topographic roughness consistently drove spatial genetic clustering of T. cristatus, but variability existed among conductivity maps derived from optimised resistance surfaces. Conclusions These findings underscore the importance of carefully selecting genetic distance metrics and addressing potential sources of uncertainty in resistance surface optimisation. By doing so, we can enhance the effectiveness of conservation planning efforts for T. cristatus and species with similar ecological considerations.