Using paired acoustic sampling to enhance population monitoring of New Zealand’s forest birds

Large-scale bird monitoring can provide valuable insights about drivers of population change across different spatial and temporal scales. Yet, challenging terrain and survey costs hinder the collection of data needed to estimate absolute abundance or population densities for New Zealand’s forest birds. Acoustic sampling is being used more frequently to increase efficiency in avian monitoring and paired sampling facilitates robust density estimation from acoustic data. In paired sampling, point counts are conducted simultaneously by human observers and autonomous recording units (ARUs) to allow estimation of statistical offsets that correct biases in ARU data relative to human observers. These offsets can then be used to calibrate count data collected only by ARUs in a larger sampling scheme. However, the effectiveness of paired sampling has not yet been evaluated in New Zealand. We assessed bias in bird counts from ARUs relative to traditional point counts and evaluated whether paired sampling reduced ARU bias, when present, at 280 count stations in six indigenous forest patches on the North Island from January to April 2017. For 13 forest bird species, we estimated δ, a statistical offset that represents the ratio of the effective detection radius (EDR) of the ARU data to human count data and compared bias in density estimates from ARUs relative to human observers between models with and without δ offsets. We found that δ estimates of EDR ratios were near 1.0 and 95% confidence intervals around δ overlapped 1.0 for nine of 13 species. Furthermore, densities produced by ARU counts were unbiased relative to human point counts for nine of 13 species. When models included δ offsets, ARU density estimate bias was removed for all species. Thus, paired acoustic sampling offers a promising strategy for increasing the efficiency, and spatial and temporal coverage of bird population monitoring across New Zealand.