It can be difficult to monitor and manage landscape scale ecologically significant habitats and nationally threatened or endangered species that rely on those habitats. Species Distribution Models (SDMs) are useful for identifying areas for conservation using expert opinion about the likelihood (or probability) of the presence of a species given the combined output of a set of spatially defined habitat parameters at a given location. For the case of an SDM for rare amphibians that move between wetlands and dry grasslands, the biophysical characteristics and critical habitat features can be very complex. Predictor variables were derived from existing mapping, digital elevation model data and Sentinel-2 13-band multispectral (10-20 m resolution) imagery. We applied Random Forest (RF), a popular SDM modeling algorithm, collected and public occurrence data, and a suite of remotely sensed predictor variables to map the primary suitable habitat elements for Northern Leopard Frog (Lithobates pipiens - NLF), and Western Tiger Salamander (Ambystoma movatorium - WTS) in a 40,395 ha area of Aspen Parkland Ecoregion / Grassland Transition Ecoclimatic Region in south central Canada.The RF models of the two species distributions validated well based on performance metrics, Spearman correlation between area adjusted observations, as well as prediction bins from k-fold cross validation. The average AUC value (for the model from k-fold cross validation) of above 0.90 for both models is considered good to excellent, and the Kappa (0.82 NLF and 0.78 WTS) values are considered useful. Derived Sentinel-2 spectral indices including Tasseled Cap Brightness, Tasseled Cap Greenness and Enhanced Water Index were all strong predictors. SDMs indicate species are typically distributed closer to wetlands, warmer temperature ranges and areas with greenness and brightness values associated with vegetation cover correlated with amphibian presence. However, SDM for WTS indicated higher ranked habitat in several patches where wetlands are near grassland areas. This may indicate the association of this species with grasslands and small mammal burrows. In addition, NFL distribution seemed to be in wetlands corridors potentially associated with amphibian refuge and overwintering sites. Future sampling efforts should on focus areas identified as having moderate to high suitability, breeding occupancy and also wetlands with small mammal excavation mounds within 250 m of the wetland edge to help indicate important refuge habitat. The SDMs created can be further refined with additional occurrence data but already it has given managers great insight for where to prioritize conservation efforts.
