Abstract

Morning vocal activity data for birds were collected using automatic sound recorders at 165 sites in fragmented forests at Tumut in south-eastern Australia. A model was developed to describe the statistical properties of the vocal activity data and study the relationship between the number of 'elements' (or notes) broadcast by birds and the number of birds recorded by human observers. We discuss the practical issues of finding a model where variance heterogeneity and skewness is a feature of the data, while the context of the problem required that multiplicative relationships be preserved. A satisfactory fit to our vocal activity data was obtained by using a Gamma distribution.

For most groups of birds, the observed relationships between vocal activity and the number of birds recorded by human point counts were weak but statistically significant.

Our analysis suggests that these data provide limited useful information about vocal activity per bird. Automatic sound-recording data may be informative for drawing inferences about temporal patterns in vocal activity but do not seem useful as a method for estimating the abundance of birds. A significant relationship was identified between the number of species vocalising, as measured by sound recorders, and the number of species observed by point count method. However, the relationship was far from isometric and there were many more species detected by the sound recorders than by human observers.

Results from analyses of vocal activity data gathered at Tumut were confirmed by additional data obtained in a subsequent, small calibration study in the Namadgi National Park in the Australian Capital Territory.