Where the toad crosses the road: multi-method and cross-taxa Texas herpetofauna roadkill modeling for conservation planning

Herpetofauna are globally declining but often overlooked in conservation efforts. Since many herpetofauna are susceptible to localized road mortality, we develop the first North American herpetofauna roadkill models to inform conservation planning. The iNaturalist Roadkills of Texas database is employed to develop and compare ensembles of feature-selected MaxEnt species distribution models (SDMs) and kernel density estimate models (KDEMs) of roadkill, including broad-taxa models (btSDMs and btKDEMs) of combined Texas herpetofauna, and single-species models for the state threatened Texas Horned Lizard (Phrynosoma cornutum) and common generalist Gulf Coast Toad (Incilius nebulifer). We also novelly evaluate cross-taxa roadkill models (ctSDMs; ctKDEMs) using both broad-taxa (bctSDMs; bctKDEMs) and single-species (sctSDMs; sctKDEMs) models, assessing how well models for one taxa project sympatric roadkill of individual species, including two rare herpetofauna with insufficient data for modeling: the state threatened Texas Tortoise (Gopherus berlandieri) and federally endangered Houston Toad (Anaxyrus houstonensis). Texas herpetofauna MaxEnt roadkill btSDMs and Texas Horned Lizard roadkill SDMs both outperformed corresponding poor accuracy KDEMs. Gulf Coast Toad MaxEnt roadkill SDMs and KDEMs had poor accuracy, but KDEMs had lower complexity. Texas herpetofauna roadkill bctSDMs had excellent and significantly highest accuracy in projecting cross-taxa roadkill for the Texas Horned Lizard and Texas Tortoise. Texas herpetofauna roadkill bctSDMs also best projected Houston Toad roadkill. Important MaxEnt SDM variables included distance to residential roads, human population density, percent cover developed open space, and live herpetofauna KDE. Texas herpetofauna bctSDMs can guide further roadkill surveys needed to improve models and develop mitigation plans for herpetofauna conservation.