Using kernels and ecological niche modeling to delineate conservation areas in an endangered patch‐breeding phenotype

Efficient delineation of conservation areas is a great challenge in maintaining biodiversity. Kernel density estimators (KDEs) are a powerful tool in this perspective, but they have not been applied at the population level on patch‐distributed organisms. This would be particularly worthy for species that need broad habitats beyond those where they can be sampled; such as terrestrial lands for pond‐breeding amphibians. The aim of this study was to compare different approaches for the identification of suitable areas for conservation: KDE, ecological niche modelling, and a combination of KDE and niche models. Paedomorphosis was chosen as a model system because this is an important form of intraspecific variation that is present in numerous taxa, but geographically localized within species and globally endangered. 277 ponds were sampled in one of the hotspots of paedomorphosis to determine the abundance and distribution of paedomorphs (i.e., individuals retaining gills at the adult stage) of the palmate newt (Lissotriton helveticus), with emphasis on the connections between the most valuable populations. KDEs gave insights into the surface areas required to balance the maintenance of certain number of connected ponds and the respective number of disjoint areas in which the whole population is divided. The inclusion of barriers in the models helped in accurately designing the limits of the areas to protect. Alone, habitat models were not able to successfully delineate the area to protect, but the integration between terrestrial suitable areas or barriers and KDE allowed an objective identification of areas required for conservation. Overall, the best performance was observed by the KDE integrating ecological barriers, and by the combination between KDE and niche modelling. In a broader perspective, KDEs are thus a pertinent tool in providing quantitative spatial measurements to delineate conservation areas based on patch‐abundance data with a specific focus to connectivity.