Background/Question/Methods Ecological corridors ensure the dispersion of individuals among populations that is essential to counterbalance the negative effect of natural habitat destruction and isolation. Landscape connectivity analyses can help face the challenge of locating ecological corridors to define key areas to protect or restore. These frameworks are based on different assumptions including how to maximize the number of species that are likely to use ecological corridors. These differences can lead to inconsistent spatial prioritization. This issue could be overcome offering empirical procedures to evaluate the efficiency of different landscape connectivity models in locating ecological corridors, with easily accessible data. We proposed a new procedure to evaluate the accuracy in the location of ecological corridors, using independent data of species presence. Using Maxent models or expert opinions, we created resistance surfaces, then we ran Circuitscape to compare the location of ecological corridors among three landscape connectivity modelling approaches: (i) a habitat-based approach that relies upon a land cover map used as a proxy of species presence; (ii) an umbrella-species approach based on presence-only data of Sitta europea considered as an umbrella species for woodpeckers and (iii) a multispecies approach based on presence-only data of woodpeckers. We selected the most efficient approach applying our validation procedure to the forest species Sitta europea, Dendrocopos major, Dryocopos martius and Dendrocopos minor, in a rural landscape located in the French Alps. Results/Conclusions We found that ecological corridors were better localized with the multispecies – and then the umbrella species – approaches that are based on the use of species data related to habitat suitability index with Maxent models, and lower for the habitat-based approach that consists in attributing resistance cost to each land cover from expert opinion. Our validation procedure has several advantages: (i) to fill the gap of absence of validation procedures with accessible data; (ii) to go beyond the comparison of connectivity analysis outputs, which are based on different hypotheses, by selecting the most realistic; (iii) to detect a misclassification of matrix resistance values; (iv) to be independent of the modelling approach, such as expert-opinions based, least-cost paths or circuit theory, because it evaluates one final output of connectivity models (location of ecological corridors). Our results have important implications for landscape managers and stakeholders concerned with connectivity conservation and restoration planning.
