Experimental analysis of the auditory detection process on avian point counts

We have developed a system for simulating the conditions of avian surveys in which birds are identified by sound. The system uses a laptop computer to control a set of amplified MP3 players placed at known locations around a survey point. The system can realistically simulate a known population of songbirds under a range of factors that affect detection probabilities. The goals of our research are to describe the sources and range of variability affecting point-count estimates and to find applications of sampling theory and methodologies that produce practical improvements in the quality of bird-census data. Initial experiments in an open field showed that, on average, observers tend to undercount birds on unlimited-radius counts, though the proportion of birds counted by individual observers ranged from 81% to 132% of the actual total. In contrast to the unlimited-radius counts, when data were truncated at a 50-m radius around the point, observers overestimated the total population by 17% to 122%. Results also illustrate how detection distances decline and identification errors increase with increasing levels of ambient noise. Overall, the proportion of birds heard by observers decreased by 28 ± 4.7% under breezy conditions, 41 ± 5.2% with the presence of additional background birds, and 42 ± 3.4% with the addition of 10 dB of white noise. These findings illustrate some of the inherent difficulties in interpreting avian abundance estimates based on auditory detections, and why estimates that do not account for variations in detection probability will not withstand critical scrutiny.