Dependent double-observer surveys reduce abundance bias and identify priority habitat for cryptic arboreal geckos

Résumé fourni par la source

Abstract Species conservation depends on accurate population data. However, monitoring is challenging when species occur at low densities, display cryptic behaviour, or possess inconspicuous phenotypes that reduce detectability. Squamate reptiles often exhibit these traits, and monitoring is especially difficult in structurally complex habitats with restricted visual access, such as tropical forests. We combine a dependent double-observer survey design with multi-species hierarchical Bayesian inference to estimate detection and abundance for endemic Ailuronyx geckos in remnant native palm forests on the island of Praslin, Seychelles. We also quantified the contribution of the second observer by refitting the model to primary-observer counts and validated both designs against simulated data with known population sizes. Detection rates were low, with combined two-observer probabilities of 16.6% for A. seychellensis, 18.6% for A. trachygaster, and 27.5% for A. tachyscopaeus. Secondary observers contributed roughly one-third of all detections, increasing combined detection of these cryptic arboreal species by 39–50%. Observer identity dominated detection heterogeneity, accounting for 78% of its variance. Simulations results showed the primary-observer design underestimated true abundance in most replicates. Estimated abundances across all study sites were 55.5 ind ha-1 for A. tachyscopaeus, 7.3 ind ha-1 for A. seychellensis, and 3.8 ind ha-1 for A. trachygaster. The abundance of A. trachygasterincreased by 46% per standard deviation of the endemic coco de mer (Lodoicea maldivica; 67.4 ind ha⁻¹) palm. We identify mature coco de mer stands as a key conservation priority for A. trachygaster. Protecting existing stands from fire, clearance, nut poaching and invasive species encroachment, alongside continued replanting and restoration of palm-dominated habitat to improve forest extent and connectivity, will benefit this critically endangered species and likely the wider endemic fauna. The low estimated abundance further emphasises the need for targeted conservation management and long-term population monitoring. Integrating structured observer protocols with hierarchical Bayesian modelling provides credible, detection-corrected abundance baselines for species that are difficult to monitor. This approach is straightforward to implement in the field, scalable to additional species and locations, and applicable to evidence-based conservation of cryptic reptiles globally.