Are we overestimating the niche? Removing marginal localities helps ecological niche models detect environmental barriers.

Correlative ecological niche models (ENMs) estimate species niches using occurrence records and environmental data. These tools are valuable to the field of biogeography, where they are commonly used to infer potential connectivity among populations. However, a recent study showed that when locally relevant environmental data are not available, records from patches of suitable habitat protruding into otherwise unsuitable regions (e.g., gallery forests within dry areas) can lead to overestimations of species niches and their potential distributions. Here, we test whether this issue obfuscates detection of an obvious environmental barrier existing in northern Venezuela - that of the hot and xeric lowlands separating the Península de Paraguaná from mainland South America. These conditions most likely promote isolation between mainland and peninsular populations of three rodent lineages occurring in mesic habitat in this region. For each lineage, we calibrated optimally parameterized ENMs using mainland records only, and leveraged existing habitat descriptions to assess whether those assigned low suitability values corresponded to instances where the species was collected within locally mesic conditions amidst otherwise hot dry areas. When this was the case, we built an additional model excluding these records. We projected both models onto the peninsula and assessed whether they differed in their ability to detect the environmental barrier. For the two lineages in which we detected such problematic records, only the models built excluding them detected the barrier, while providing additional insights regarding peninsular populations. Overall, the study reveals how a simple procedure like the one applied here can deal with records problematic for ENMs, leading to better predictions regarding the potential effects of the environment on lineage divergence.