Presentation is loading. Please wait.

Presentation is loading. Please wait.

Static Acoustic Monitoring and Noise

Similar presentations


Presentation on theme: "Static Acoustic Monitoring and Noise"— Presentation transcript:

1 Static Acoustic Monitoring and Noise
Nick Tregenza Chelonia Limited Exeter University

2 Noise 2 ways noise can affect logger performance:
cause false positives increase false negatives = reduced sensitivity False Positives can destroy your data Reduced sensitivity is less serious, and must occur.

3 Some types of noise/interference
Sand movement - ‘sediment transport’ shrimp clicks surface noise WUTS (weak unknown train sources) boat sonars

4 Sand movement - the 'blue grass'
Often has a distinct pitch that depends on particle size. Occurs at times in all sandy places. Missing from all textbooks on marine acoustics … except … see chapter on PODs: Fine sand click frequency spectrum

5 Sand movement Frequency of clicks C-POD data Amplitude:
shows presence of porpoises

6 Shrimp clicks Clicks/m over 4.5 days Loud, very short clicks.
Raise mean noise level. Sound like frying. May resemble dolphin clicks. Diel patterns? Alphaeus sp Britain Clicks/m over 4.5 days

7 False positives from noise
Sand noise can be more ‘porpoise like’ than a real porpoise. Values to the left of the red line are largely false positives from the PAMGuard basic classifier used in a noisy environment - dolphins, shrimps, moving sand. Ouch! So there's no easy answer…

8 What can we do to evaluate noise effects?
As noise increases … any or all of these happen: Click detection impaired Train detection impaired Porpoises move away / stop clicking ?? Dolphins move away / stop clicking ??

9 Some evaluation approaches for
assessment of noise effects in C-POD data: Click detection impairment: look at the ‘MMM’ statistic = mean multipath minima, a measure of how loud the weakest detectable clicks are. Train detection impairment: test by 'injecting' trains into noisy data. Armando Jaramillo has done this and found good train detection performance in noise, but significant changes in classification. Porpoises move away: do a visual analysis of loud clicks in noise. The detection of loud clicks will be resistant to noise and the visual analysis will avoid the problem above. Dolphins move away / stop clicking: look at data from an extensive array during noise events. This is a much harder task, because these clicks are so much less distinctive.

10 Where are we now? PAMGuard data: Basic porpoise detector gives many false positives in noise. Avoid raw click counts cannot be used, because of multipath. Inspection of all detections is essential, but robust guidelines do not yet exist. Calibrate your analysts, but how? C-POD data: most sets show fewer detections when it is noisy but we don’t know how far this is due to detection processes and how much is due to changes in animal behaviour. In most places significant noise levels are not common, and can be excluded or safely treated as likely to be similar between years. … or SAM can be limited to quieter periods. More evaluation of this topic is needed and will produce valuable insights into the acoustic ecology of cetaceans as well as detector performance in noise.

11 the end


Download ppt "Static Acoustic Monitoring and Noise"

Similar presentations


Ads by Google