Analysis of Environmental Acoustics in Sequoia National Park

1
Analysis of Environmental Acoustics in Sequoia
National Park

• Stuart Gage
• Bernie Krause
• Rudy Trubitt
• Jack Hines
• Brian Napoletano
• Wild Sanctuary
• and
• Michigan State
• University

2
Research Questions

• How do we characterize ecological soundscapes?
• What is the relationship between the ecological
  soundscape and landscape?
• Can environmental soundscapes, when correlated to
  landscape analysis, serve as quantifiers of
  ecosystem health and change?

3
Places-Times-Seasons

• Objectives of the Study
• Develop a strategy to quantify the character of
  environmental acoustics in SEKI
• Measure SEKI acoustics in representative
  locations
• Quantify the acoustics to enable comparisons
  within and between places, times and seasons
• Recommend strategies to continue to characterize
  environmental acoustics in SEKI and other
  National Parks

4
Project Design

• Record acoustics in locations which represent
  ecological landscapes significant in SEKI
  (Shepard Saddle Sycamore Spring Buckeye Flats
  Crescent Meadow)
• Record acoustics at times of the day that
  represent the diurnal diversity of environmental
  acoustics (Dawn Midday Dusk Night)
• Record acoustics to quantify the biological
  variability in seasons (Fall Winter Spring
  Summer)

5
Four Places Four TimesFour Seasons

• Locations in Sequoia
• Sycamore Springs (SY-1)
• Shepard's Saddle (SH-2)
• Crescent Meadow (CM-3)
• Buckeye Flats (BF-4)
• Recording Time
• One hour recording 4 times/day at each site
• Sectioned each 1 hour recording into twelve-30
  second subsets
• Processing
• Sound file gt Image gt 11 Hz Slices (1000 Hz)
  intervals gt Analytical framework gt Frequency
  analysis

6
May 18,2002
Landsat Scenes of SEKI
7
Buckeye Flat
8
Crescent Meadow
9
Shepherd's Saddle
10
Sycamore Spring
11
CM
SH
Acoustic Monitoring Locations
BF
Sequoia-Kings Canyon National Parks
SY
12
Environmental Acoustics Terminology

• Soundscape Environmental acoustics that occur
  within a biophysical region.
• Biophony Chorus of ambient sounds produced by a
  regions organisms.
• Geophony Sounds produced by the physical
  components of the environment.
• Anthrophony Sound produced due to human presence
  in the ecosystem, particularly mechanical and
  other human introduced acoustics.

13
Soundscape
14
Geographic Representation of a Sonogram

• Each sonogram is converted to a geographic image
  with known and fixed coordinates (x1000 y500)
• Each sonogram sample (geographic image) is 30
  seconds in length (x) with y representing KHz and
  z representing intensity
• Each image can then be analyzed and compared with
  any other image using geospatial technologies
• We use this technology to conduct our analysis

15
Interpretation of Acoustic Signals

• The Niche Hypothesis1
• Acoustic communities with more vocalizing species
  tend to present greater acoustic diversity
  regarding utilized spectral space i.e.
  bandwidth, amplitudes, and durations.
• Habitat bio-diversity tends to correlate
  positively to the acoustic diversity exhibited
  over time within the habitat.
• Environmental acoustics can provide indices of
  biodiversity and ecosystem health.
• 1Adapted from The Niche Hypothesis proposed by
  Dr. Bernie Krause. See Bioacoustics, Habitat
  Ambience in Ecological Balance published in Whole
  Earth Review, Winter, 1993.

16
Hypotheses

• Vocalizing organisms carefully adapt their
  acoustic signals to fit specific acoustic niches,
  thereby minimizing interference in their
  communication.
• When other acoustic signals interfere with a
  specific species niche, the organisms can only
  adapt to a certain threshold. Beyond this point,
  communication is disrupted and the population
  tends to decline or the organisms may leave the
  landscape.
• 2Also Adapted from The Niche Hypothesis proposed
  by Dr. Bernie Krause. See Bioacoustics, Habitat
  Ambience in Ecological Balance published in Whole
  Earth Review, Winter, 1993.

17
AnalysisObjective

• Analyze the amount and distribution of acoustics
  in each recorded sample, including frequency
  domains within each acoustic sample.
• Do this for data collected from the 4 locations,
  for 4 times per day during 4 season of the year
• Perform statistical analysis to compare
  differences and test hypotheses

18
Bands of Concentration
When mapped in an acoustic sonogram, the values
tend to aggregate within two primary frequency
ranges

• Anthrophonic Bands
• Generally below 1.5 kHz
• Contains most stationary mechanical noise,
  including fans, ventilation systems, and
  generators
• Also tends to contain temporal signals, such as
  passing aircraft, automobiles and trains,
  depending on transducer location

• Biophonic Bands
• Generally between 4 and 7 kHz
• Contains most biological activity, including
  avian communication, amphibian vocalizations, and
  other organism vocalizations
• There are exceptions such as doves, crows, quail,
  ruffed grouse and some mammals

19
Typical Bands of Concentration
Anthrophonic Bands (lt 1.5 kHz) (Generator)
Biophonic Bands (3 7 kHz) (crickets)
20
Taxonomy of the Environmental Acoustic Spectrum
21
Spectrogram Parameter Selection
22
Theory
Analysis
Quantitative Image
Signal Distribution
Dominance
Concentration
Frequency
Amplitude
Time
Species-Specific Analysis
Information Theory Pattern Assessment
Temporal
Spatial
Habitat
Regional
Scales
23
Frequency Distribution

• A useful metric for the interpretation of a
  soundscape is the mean amount of acoustics
  (number of pixels in a sonogram) in each
  frequency domains. This allows analysis and
  interpretation of the average amount of sound
  occurring in each domain. This can be used to
  conduct statistical comparisons within and
  between samples and to test hypotheses.

24
Analysis Process

• Acoustic recordings sub-sampled (wav files)
• Convert wav to bitmap images for analysis of
  sonograms using Spectrogram
• Apply ESA (EnviroSonic Analysis) to create
  frequency domains and analyze sonogram images
• Data from ESA analysis compiled and converted to
  database format for plotting and statistical
  analysis
• Mean for all sound sub-samples computed for
  summary
• Statistical Analysis for within and between
  comparisons
• Plot trends
• Compute other indices (Diversity, Dominance,
  Fragmentation, Richness) for each acoustic dataset

25
Automation of Acoustics Samples

• Acoustic Recordings 4-sites-4 times/day-4
  seasons
• 1 hour digital recordings from each
  site/time/season 4x4x464 hours
• 12 sub-samples per hour 12x4 48 Sound Segments
  during each visit to SEKI
• Each Sound Segment divided into 12 subsets (whole
  11 Frequency domains) 48x12576 data segments
• 2304 (576x4) Digital Images 2304 Numerical
  Analysis for each frequency domain
• There can be 1000x500x256 elements in each
  sonogram
• diversity, dominance, richness, fragmentation)
  2304x511520 statistics

26
Processing the Data
27
(No Transcript)
28
Meta DataThese are the data contained in the
statistics files for each of the locations where
acoustics were recorded.

• Location Location of acoustics recoding
• Season Season of Year (1Fall 2Winter
  Spring3 Summer4)
• TOD Time of Day (Dawn1 Midday2, Dusk3
  Midnight4)
• ST (Start Time of Recording)
• FMI 5 minute interval (1 hour recording divided
  into 12 slices)
• Level Hz level where
• 10-1000 Hz
• 21000-2000 Hz
• 32000-3000 Hz
• 43000-4000 Hz
• 54000-5000 Hz
• 65000-6000 Hz
• 76000-7000 Hz
• 87000-8000 Hz
• 98000-9000 Hz
• 109000-10000 Hz
• 1110000-11500 Hz
• 12Full Range 0-11500 Hz

Analysis Analysis where FreqFrequency
RelRich Relative Richness Frag
Fragmentation DomDominance DivDiversity MN
Mean SDStandard Deviation DFDegrees of Freedom
(total pixels possible in Hz domain
29
SEKI Summary Statistics for Dawn Acoustics Sample
(Entire Sonogram)
30
Buckeye Flats
Dawn-Fall 20.455
Dawn-Winter 9.7382
Dawn-Spring 12.718
Dawn-Summer 11.627
31
Crescent Meadow
Dawn-Fall 0.5244
Dawn-Winter 0.2124
Dawn-Spring 0.6955
Dawn-Summer 0.4893
32
Shepard Saddle
Dawn-Fall 1.8027
Dawn-Winter 0.2760
Dawn-Spring 1.5410
Dawn-Summer 0.2856
33
Sycamore Spring
Dawn-Winter 1.1138
Dawn-Fall 1.7446
Dawn-Summer 2.7910
Dawn-Spring 0.9901
34
3D-Soundscapes

• Buckeye Flats
• Crescent Meadow
• Shepard Saddle
• Sycamore Spring

35
Buckeye Flats 1-1
Buckeye Flats 2-1
Buckeye Flats 4-1
Buckeye Flats 3-1
36
Crescent Meadow 2-1
Crescent Meadow 1-1
Crescent Meadow 4-1
Crescent Meadow 3-1
37
Shepard Saddle 1-1
Shepard Saddle 2-1
Shepard Saddle 4-1
Shepard Saddle 3-1
38
Sycamore Spring 2-1
Sycamore Spring 1-1
Sycamore Spring 4-1
Sycamore Spring 3-1
39
Seasonal Patterns of Acoustics
40
Seasonal Patterns of Acoustics(without Buckeye
Flats)
41
Buckeye FlatsFrequency Domains
Fall
Winter
Spring
Summer
42
Buckeye FlatsFrequency Domains
Fall
Winter
Omit FD 1
Spring
Summer
43
Crescent MeadowFrequency Domains
Fall
Winter
Spring
Summer
44
Crescent MeadowFrequency Domains0-1 kHz omitted
Omit FD 1
Fall
Winter
Spring
Summer
45
Shepard SaddleFrequency Domains
Fall
Winter
Spring
Summer
46
Shepard SaddleFrequency Domains 0-1 kHz omitted
Omit FD 1
Fall
Winter
Spring
Summer
47
Sycamore SpringFrequency Domains
Fall
Winter
Spring
Summer
48
Sycamore SpringFrequency Domains 0-1 kHz
omitted
Omit FD 1
Fall
Winter
Spring
Summer
49
Statistical Analysis of AcousticsTukey Means
Comparison
Full Spectrum Dawn Chorus
Each site appears exhibits a unique statistical
trend
50
(No Transcript)
51
Summary

• This analysis quantifies environmental acoustics
  for interpretation of soundscape meaning.
• The division of each soundscape into frequency
  domains provides one way to separate the
  compartments of a soundscape.
• This methodology enables standardized comparison
  within and between soundscapes.
• The computation of landscape indices to
  soundscapes provides the potential to compare
  landscapes and soundscapes.
• The automation of this process provides a
  powerful tool to analyze and interpret
  soundscapes.

52
Implications

• An overall measure of the acoustic-diversity of
  an acoustic habitat, expressed in terms of
  utilized spectral space, may serve as an index
  of ecological integrity and habitat composition.

53
Bernie Krause
Jack Hines
Stuart Gage
Rudy Trubitt
Brian Napoletano
54
(No Transcript)
55
Thanks to David Graber Annie Esperanza and SEKI
Park Staff!
National Park Service And special thanks to
Bernie Krause Who got me into this