Modelling density surfaces from integrated passive acoustic and point count data streams

Point counts employing distance sampling methods can be used to accurately estimate animal detection probabilities and absolute abundances, but are difficult to implement over broad spatial and temporal scales. Passive acoustic monitoring (PAM) via autonomous recording units is an efficient method for monitoring populations of vocalizing animals but on its own can only produce relative, not absolute, abundance estimates. Here, we demonstrate a novel model that can integrate point count and PAM data streams into a single model of the spatial distribution of a species (an integrated Density Surface Model; DSM) via a multiplicative scaling constant to account for data‐specific detectability. We also show how this model can be estimated using standard generalized additive model software. To evaluate the effectiveness of this approach, we estimated an integrated DSM using data from overlapping observer and PAM surveys for two Hawaiian bird species and compared estimated density surfaces with a model with separate density estimates for each data stream. Parameter surfaces had minimal differences in density and variance estimates, indicating parameter estimates were reliable for both distance sampling and PAM data streams. Scaled estimator precision for predicted surfaces was consistent near sampling points. Our integrated model was more precise than data stream‐specific models, accurately captured spatial correlation in densities, and with calibration through the integrated model, PAM‐derived data can be used to monitor absolute abundances.