Simulation modelling demonstrates differential performance of connectivity methods in their ability to predict genetic diversity in complex landscapes.

• Given the wide application of connectivity modelling tools in applied conservation planning, it is crucial to broadly evaluate how these models perform across resistance, movement, and population structure conditions in predicting genetic diversity patterns. • Our goal was to investigate how the predictions of three connectivity models were related to spatial patterns of genetic diversity complex landscapes, considering factors such as population structure, resistance, genetic drift, genetic disequilibrium, and organism movement abilities. • We used CDPOP, an individual-based, spatially explicit population and genetic simulation model, to simulate genetic diversity across these resistance surfaces. We then simulated landscape connectivity with several popular connectivity methods, including resistant kernels, Circuitscape, and Pathwalker, and evaluated how well they were able to predict spatial patterns of genetic diversity. • Resistant kernel outperformed other connectivity methods in predicting landscape patterns of genetic diversity. The strongest relationships occurred when the population process has created spatial structure but has not yet led to significant genetic diversity loss due to drift. • Our results confirm the strong relationship between genetic diversity and population connectivity, and suggest that computationally efficient incidence function algorithms, such as resistant kernel methods, are well suited to predicting functional connectivity. There have been few evaluations of how well different connectivity modelling methods are able to predict the spatial genetic structure and genetic diversity of populations residing in complex landscapes. Given the wide application of connectivity modelling tools in applied conservation planning, it is crucial to broadly evaluate how these models perform across resistance, movement, and population structure conditions in predicting genetic diversity patterns. Such evaluations are critical to provide rigorous, biologically based guidance for conservation and management applications. Our goal was to investigate how the predictions of three connectivity models were related to spatial patterns of genetic diversity complex landscapes, considering factors such as population structure, resistance, genetic drift, genetic disequilibrium, and organism movement abilities. We evaluated the performance of several connectivity methods across seven a priori landscape resistance surfaces to provide a broad assessment of their performance. We used CDPOP, an individual-based, spatially explicit population and genetic simulation model, to simulate genetic diversity across these resistance surfaces. This provided a pool of genetic diversity patterns that were the response factor in our simulation experiment. We then simulated landscape connectivity with several popular connectivity methods, including resistant kernels, Circuitscape, and Pathwalker, and evaluated how well they were able to predict spatial patterns of genetic diversity. Resistant kernel outperformed other connectivity methods in predicting landscape patterns of genetic diversity. The strongest relationships occurred when the population process has created spatial structure but has not yet led to significant genetic diversity loss due to drift. The time lag disequilibrium was relatively short. Long simulation times resulted in severe reduction in prediction ability due to drift. Resistant kernel predictions were much more strongly related to spatial patterns of genetic diversity than were predictions produced by Circuitscape and Pathwalker, across a large combination of population structures. Strong relationships exist between functional connectivity and genetic diversity, with clearer and stronger associations seen in spatial patterns of allelic richness compared to heterozygosity or spatial effective population size. Our results confirm the strong relationship between genetic diversity and population connectivity, and suggest that computationally efficient incidence function algorithms, such as resistant kernel methods, are well suited to predicting functional connectivity. [Display omitted] [ABSTRACT FROM AUTHOR]