Our case studies leverage sensor technologies, infrastructure, and on-line, automated analysis techniques to investigate the relationship between acoustics and ecological indicators. Our preliminary studies are promising and show that there is a positive correlation between acoustics and ecological indices.

Shown in Figure 1a is a plot of the soundscape quality index (SQI) for the month of May 2006. To produce this plot, the acoustics at a pond site in Okemos, Michigan were recorded every 30 minutes for a duration of 30 seconds for every day in may. The SQI is computed as the ratio of the sound intensity found in the frequencies where biological sounds (biophony) are most prevalent (2-8 kilohertz) to the frequencies where mechanical sounds (technophony) are most prevalent (1-2 kilohertz). SQI has values in the range +1 to -1, with +1 indicating that a signal contains only biophony. As shown in the figure, SQI has a larger biophonic component between 2100 and 0730 hours.

One method that we have implemented for automated species identification is to compute the correlation coefficient between a species signature and the recorded acoustic clips collected by in-field sensor platforms. To construct a signature for an organism, we extract a sub-image from a spectrogram containing the species vocalization that is to be identified. Figure 1b depicts a plot of the signature match correlation for identifying the spring peeper (Pseudacris crucifer crucifer). As shown, the spring peeper signaled most often between 2030 and 0600 hours throughout May.

Although these examples illustrate the potential for using acoustics for ecosystem assessment, there is still much more work to be done. At REAL, we are exploring machine learning and other techniques to further enable soundscape interpretation and facilitate automated ecosystem assessment. Moreover, many of these techniques enable better data stewardship.