Santa Margarita Species Acoustic Monitoring via HPWREN

The microphone is one of the few sensors that can quantify species specific occurrence in the ecosystem. The goal is to establish long-term acoustic monitoring sites in unique ecological landscapes to characterize its acoustic signature. This research is testing the idea that the soundscape is a valuable ecological attribute and that long term acoustic monitoring can provide important information about ecosystem dynamics.

The Santa Margarita Ecological Reserve is an ideal site in California to examine the acoustic signature. Collaborators are interested in testing the idea that acoustics can provide a unique signature of ecological significance. The participants in the project include Stuart Gage (Michigan State University); John Helly (San Diego Supercomputer Center); Sedra Shapiro (SDSU-SMER), Pablo Bryant (SMER) and Mark VanScoy (SMER).

The site has been active since January 2003. Gage and Helly visited the site in April 2003 and photographed the recording site and surroundings.

Gage visit to SMER in April 2003 to check operation and microphone location

The objective is to test an automated acoustic monitoring instrumentation developed by Gage at Michigan State University. This system involves unattended operation; auto transmission of acoustic samples to a remote server at MSU/CEVL (http://www.cevl.msu.edu). Sound is sampled at a frequency 22,050 Hz monaural every 1/2 hour for 30 second duration. Files are location and date stamped and stored in wav file format on local hard drive. Files are then are transmitted via auto-ftp through HPWREN to a remote server at MSU. Files are stored in digital archive; converted to a sonogram image and each image is quantified to extract the acoustic intensity in each of 11 1-kHz frequency bands. The acoustic intensity values from each band are stored in a relational database. A web system links sound, sonogram and acoustic intensity values.

Weather Station view representing the acoustic soundscape monitored at SMER

Sounds of the SMER environment are recorded 48 times per day. Sounds were drawn from the digital database archive to represent difference in season. Sounds sampled 15 March and 15 May, 2004 at 0700 hrs illustrate the difference in seasonal soundscape. There is more biological information in the May 0700 recording compared to the March 0700 recording based on the greater amount of acoustics in frequencies in the 4-6 kHz frequency range due to increased birdsong in May. The image contains considerable background sound (geophony) due to the elevated location of the microphone, wind and echo from the canyon.

Sonograms above are derived from raw wav file digital recordings made at 0700 hrs on 15 March (left, sound file in .wav format) and 15 May (right, sound file in .wav format) at the SMER site

The 3D sonograms for March and May 2004 at 0700 hrs shown below provide an enhanced comprehension of the relationship between time, frequency and loudness.

Sonograms above derived from background filtered wav file digital recordings made at 0700 hrs on 15 March (left) and 15 May (right) at the SMER site

Animated GIF versions of the unfiltered data for 0700 hrs in March 2004 and 0700 hrs in May 2004 show the dynamic chage in acoustics over the two months.

Analytical software, developed at the Computational Ecology and Visualization Laboratory, conducts image analysis of sonograms and produces a statistical summary of the acoustic intensity in each 1 kHz interval between 0-11 kHz. These are coded as levels 1-11. The full spectrum is coded as level 12. The following tables are created from wav files that filter out the majority of background sounds to enable comparison of biophony between the two time periods. The acoustic intensity (MN) and associated standard deviation (SD) and number of pixels in each level are shown in the table below for the March 0700 hrs recording. Note that levels 4 and 5 contain acoustic intensity from biological sounds.

Table 1.Mean acoustic intensity for 15 March 2004 700 hrs analysis of the sonogram for each of 11 - 1 kHz frequency levels plus the mean acoustic intensity values for the entire spectrum (level 12)

Values of acoustic intensity for biological levels (4 and 5) are 4.9 and 2.1, respectively for the15 March 0700 recording. The acoustic intensity values for the same levels from the recording made on 15 May 0700 are 13.1 and 16.5 (see below).

To compute the relative biological acoustic intensity for bands with biological information, values in levels 4 and 5 are summed and divided by the acoustic intensity for the full sonogram (level 12). The resulting biological intensity values are 5.84 (7.0/1.2) for the March recording and 8.43 (29.6/3.5) for the May recording illustrating that the May recording has a higher biological content than the March recording.

Table 2. Mean acoustic intensity for 15 May 2004 700 hrs analysis of the sonogram for each of 11 - 1 kHz frequency levels plus the mean acoustic intensity values for the entire spectrum (level 12)

Three D images of the sonograms with noise filtering for 15 March and15 May at 0700 below show the patterns and intensity of biophony.

Three D images of sonograms from noise filtered sounds recorded on 15 March 2004 and 15 May 2004 at 0700 at SMER

Research is progressing to enable analysis of sounds in the digital library using the above methods. This will enable us to define ecosystem characteristics and change over time using acoustics as a measure of complexity and diversity.

-- Prof. Stuart H Gage, Department of Entomology, Michigan State University