decaf – density estimation for cetaceans from passive ... · decaf – density estimation for...
TRANSCRIPT
DECAF – Density Estimation for Cetaceans from passive
Acoustic Fixed sensors
Presented by: Len Thomas
JIP Program Review Meeting October 28-30, 2008
JIP program review, October 28JIP program review, October 28--30 200830 2008
About the project• 3 year project: May 2007 – 2010• Funded under NOPP by JIP and NOAA.
Leveraging previous work funded by ONR.• Investigators:
– University of St Andrews: Len Thomas, David Borchers, Tiago Marques, Catriona Stephenson
– US Naval Undersea Warfare Center: Dave Moretti, Jessica Ward, Nancy DiMarzio, Ron Morrissey
– US Space and Naval Warfare Systems Center: Steve Martin– University of Oregon: Dave Mellinger, Elizabeth Küsel– Woods Hole Oceanographic Institution: Peter Tyack
• Steering group:– Walter Zimmer, Jay Barlow, Steve Buckland
JIP program review, October 28JIP program review, October 28--30 200830 2008
Goals1. Develop methods for estimating the density of
cetacean species from fixed passive acoustic devices.
2. Demonstrate the utility and generality of the methods by implementing them in a set of key test case studies.
3. Promote adoption of the new methods in the marine mammal research community by
a. publishing results in the peer-reviewed literatureb. archiving data and results in publicly available
electronic storehousesc. holding workshops open to all interested
researchers
JIP program review, October 28JIP program review, October 28--30 200830 2008
Why estimate absolute density?
• Why not use an easier-to-obtain, relative measure such as detections per unit time?
• Relative measures not good enough when:– population size is of intrinsic interest– doing long term monitoring– regulatory framework – need to determine “take limits”
and “takes”– need to determine population effect of potentially
harmful activity (e.g., via BACI)– want priors on real-time monitoring studies
JIP program review, October 28JIP program review, October 28--30 200830 2008
• Minke whales at PMRF• Sperm whales at AUTEC
• Humpback whales at PMRF
Case studies
All use US Navy instrumented testing ranges
( )
• Beaked whales at AUTEC
• Sperm whales from a single sensor at AUTEC
JIP program review, October 28JIP program review, October 28--30 200830 2008
Conceptual framework
How to estimate whale density from fixed passive acoustic monitoring
JIP program review, October 28JIP program review, October 28--30 200830 2008
Focus here on cue-based methods
• We have K hydrophones• operating for T hours• and we detect n cues
• How many whales are there? or, equivalently• What is the whale density?
(numbers per unit area)
Cue=discrete, identifiable vocalization – e.g., click, song unit, etc.
JIP program review, October 28JIP program review, October 28--30 200830 2008
Example: Beaked whales at AUTEC
K=82 hydrophones
T=5,062 minutes
n= 2,943,789 Mesoplodon densirostriswhale clicks detected
What is the average beaked whale density at AUTEC over this 6-day period?
JIP program review, October 28JIP program review, October 28--30 200830 2008
Estimating density (1)• Let w be a distance beyond which no cues
can be detected.• If all cues with distance w
are detected• then density of cues per
unit time is
TwKnDc 2π
=0 2 4 6 8 10
0.0
0.2
0.4
0.6
0.8
1.0
distancep(
dete
ct)
example: w=7
JIP program review, October 28JIP program review, October 28--30 200830 2008
Estimating density (2)• but, realistically, not all cues within w will
be detected.
• Let the average proportion detected be p.
• Then,
TpwKnDc 2π
=0 2 4 6 8 10
0.0
0.2
0.4
0.6
0.8
1.0
distance
p(de
tect
)
example: p=0.12
JIP program review, October 28JIP program review, October 28--30 200830 2008
Estimating density (3)• Also, most detectors have a non-negligible
false positive rate.• Let c be the proportion of false positive
detections. • Then,
TpwKcnDc 2
)1(π−
=
JIP program review, October 28JIP program review, October 28--30 200830 2008
Estimating density (4)• Lastly, to convert density of cues to
density of animals, we need the cue rate –number of calls per unit time.
• Let r be cue rate. • Then,
TprwKcnD 2
)1(π−
=
JIP program review, October 28JIP program review, October 28--30 200830 2008
Canonical formula
So, we need estimates of 3 quantities:• p – the detection probability• r – the cue rate• c – the proportion of false positive
detections
rpTwKcnD
ˆˆ)ˆ1(ˆ
2π−
=
e.g., hand calibrate a sample of data
e.g., tagging study or focal follow
lots of ways…
JIP program review, October 28JIP program review, October 28--30 200830 2008
0 1 2 3 4
1020
3040
5060
distance
n.de
tect
ed
0 1 2 3 4
1020
3040
5060
distance
n.de
tect
ed
0 1 2 3 4
1020
3040
5060
distance
true
n
Estimating p using distance sampling• Need (a) distances to detected cues, and
(b) “randomly” placed hydrophones
12
34
ππ3
π5
π7
0 1 2 3 40.0
0.2
0.4
0.6
0.8
1.0
distance
p(de
tect
)
JIP program review, October 28JIP program review, October 28--30 200830 2008
What if you can’t get distances and/or don’t have random hydrophones?
• Lots of options – see poster on DECAF web site
• (or chat to me anytime!)• Also covers individual
and group-based (as opposed to cue-based) methods – don’t need cue rate (but may need group size)!
JIP program review, October 28JIP program review, October 28--30 200830 2008
Estimating p for beaked whales at AUTEC
• Can’t get distances (narrow beam pattern)• Used data from tagged animals
– Dives geo-referenced– Clicks from diving animals associated with those
received on surrounding range hydrophones
– Preliminary estimate of with w=7km– Need to adjust for different ambient noise conditions
057.0ˆ =p
JIP program review, October 28JIP program review, October 28--30 200830 2008
Estimating cue rate for beaked whales at AUTEC
• Use tagged whales• Do “tag-on” effect – confirmed by 72 hour
dataset from Robin Baird• Mean cue rate 0.41 clicks/second.
JIP program review, October 28JIP program review, October 28--30 200830 2008
Estimating false positive rate forbeaked whales at AUTEC
• Hand checked 30 10-minutes samples from all hydrophones
• False positive rate (for simple FFT-based detector used) 0.44 or 0.49 depending on assumptions made about calls ambiguous even to human operator
JIP program review, October 28JIP program review, October 28--30 200830 2008
Preliminary density estimate for beaked whales at AUTEC
rpTwKcnD
ˆˆ)ˆ1(ˆ
2π−
=
41.00573.0303720782)443.01(2984379
2 ×××××−×
=π
21000kmper whales18=
• Note (1) preliminary estimate; (2) no variance!
JIP program review, October 28JIP program review, October 28--30 200830 2008
Progress on other case studies• Humpback whales
Niihau Kauai
Localizations of manually detected songs, 10 minute period 15 Apr 07
range hydrophones
song locations
• Many are far from the range and/or close together, making localization difficult
• Assumption of known distribution with respect to hydrophones not justified
• Therefore, not the “simple”case study we thought it would be
JIP program review, October 28JIP program review, October 28--30 200830 2008
Progress on other case studies• Minke whales at PMRF
– Found within range hydrophones, so provides a better “simple” case study
– 10 minute test datasets have been manually analyzed and we are developing automated detectors and localizers
– Call rates currently unknown• Sperm whales at AUTEC
– Test datasets identified, and preliminary analysis has been conducted
• Sperm whale single sensor– Postdoc Küsel to start at OSU soon
JIP program review, October 28JIP program review, October 28--30 200830 2008
Open workshops on estimating cetacean density from passive
acoustics – You’re invited!• San Diego, California
– 16th July 2009– Informal, one-day workshop, after
distance sampling training workshop
• Pavia, Italy– 13th September 2009– One day workshop, linked to
Detection, Classificiation and Localization meeting
• More details on DECAF web site
COME TO AN ACOUSTIC DENSITY
ESTIMATION WORKSHOP
JIP program review, October 28JIP program review, October 28--30 200830 2008
Conclusions
• Obtaining robust density estimates from fixed PAM is hard…
• … but it is possible
• Often auxiliary information is required– Tagging studies / Visual-acoustic data
JIP program review, October 28JIP program review, October 28--30 200830 2008
Future priorities“Beyond DECAF”…
• Further case studies– vertical arrays– other biological systems– …
• Transition– software, training workshops…
• Survey design
JIP program review, October 28JIP program review, October 28--30 200830 2008
For more information• DECAF web site
http://www.creem.st-and.ac.uk/decaf/– Background white paper– Marques et al. Poster paper
• Nice PAM overview paper:– Mellinger, D. K., Stafford, K. M., Moore, S. E.,
Dziak, R. P. and Matsumoto, H. 2007. An overview of passive acoustic observation methods for cetaceans. Oceanography 20: 36-45.
• Come to the workshops• Come talk with me!