From habitat prediction to cost-effective intervention: A spatial modelling and connectivity framework for firefly conservation planning

Abstract

Firefly habitats are increasingly threatened by rapid urbanisation, while encounters with wild fireflies have become ever rarer, fuelling growing interest in firefly watching and increasing pressure on vulnerable populations. Existing conservation research has largely focused on broad-scale habitat assessment, leaving a disconnect between habitat modelling and practical conservation action. To bridge this gap, we developed an integrated conservation framework using the terrestrial firefly Emeia pseudosauteri as the focal species. We compared random forest (RF), boosted regression trees (BRT), and generalized additive models (GAM), and found that a two-stage spatial hurdle model showed the best out-of-sample performance among the frameworks compared, particularly for conditional abundance under spatially independent validation. By modelling occurrence and conditional abundance separately while accounting for spatial dependence, the model accommodated zero-inflated, spatially clustered firefly counts. We integrated Unmanned Aerial Vehicle (UAV)-derived life stage based seasonal canopy information, anthropogenic variables, circuit-theoretic connectivity analysis, and grid-level cost-effectiveness evaluation to prioritise conservation interventions. Summer canopy cover and water-edge cost distance emerged as the principal determinants of distribution, while fine-scale sub-canopy analysis indicated that local plant associations explained observed aggregations better than structural complexity alone. The framework identified cost-effective priorities for improving habitat connectivity while informing visitor management to reduce ecological disturbance. By linking spatial habitat modelling, connectivity analysis, and cost-effectiveness optimisation, this study provides a transferable planning framework for prioritising habitat management under increasing human disturbance.