Estimating the number of vocal bullfrogs using a single passive acoustic sensor.

American bullfrogs pose a significant threat to local biodiversity in regions where they are non-native, as they spread diseases and prey on smaller species. Therefore, being able to monitor their population changes in local habitats is important for effective policymaking and conservation efforts. This study investigates the use of a single passive acoustic sensor, AudioMoth, to estimate the number of vocal bullfrogs in natural habitats. Several noise removal methods are applied to the spectrograms of audio recordings, after which spectral features of bullfrog calls are extracted and analyzed using a spectral clustering method. Calls are grouped into clusters, with the number of clusters corresponding to the estimated number of vocal bullfrogs in the dataset. The proposed method demonstrated close agreement with field observations, achieving a mean absolute error of 0.85 across 13 field observation cases at seven ponds in 2023. In a complementary 2024 dataset, results were consistent with estimates derived from an acoustic localization approach using an array of Global Positioning System-synchronized AudioMoth sensors. These results indicate the feasibility of estimating numbers of vocal bullfrogs using passive acoustic data from a single sensor. The proposed method can be extended to other species, provided the spectral structure of individual calls exhibits differences.