You are on page 1of 21

Fisheries Acoustics

Error sources in acoustic surveys

1
1
Error sources in acoustic surveys
• Bias sources
– Non-linearity
– Non-accessibility
– Non-representative sampling
– Incomplete coverage
– etc
• Random error sources
– Under-sampling biologically
– Under-sampling sA
– etc

2
2
Main types of bias
Bias in acoustic surveys :

• Non-linearity (population distribution)


• Instrument error
• Attenuation
• TS - not reflecting population
• Species allocation - interpretation error
• Length compositions sampling - selectivity
• Accessibility : shadow zones
• Accessibility : avoidance
• Stock coverage - incomplete
3
3
Non-linearity
Assumption :

Ibs/Ii ~ Nfish/m3

Ibs/Ii

Random 
Nfish/m3

Ibs/Ii
Regular d

“not proved 
Nfish/m3
in fish”
Ibs/Ii

Dense
Nfish/m3
4
4
Instrument bias
Instrument error may be random error over longer time but
represents a bias in an individual survey. Includes :

•Calibration error
•Instrument drift
•Directivity not conforming to equivalent beam angle
value
•Poor S/N ratio

The risk of instrument errors has reduced considerably with


technological development.

Action : monitor instrument performance 5


5
Target strength (TS) bias
TS values generally taken from
• literature
• measurements detached from echo integration
But TS dependent on
• behaviour
• condition
• maturity
• depth
• etc etc
Actual TS during survey may differ systematically
Action : difficult, it may have to be accepted that stock
estimates are relative estimates 6
6
Species allocation bias

Target identification subject to :

• Experience of biologist

• How characteristic are the targets appear on echo


recordings :
• danger of bias relative to earlier surveys if
behaviour has changed

• Ability to sample targets (avoidance ?)

• Personal bias ? Do you want fish ??


7
7
Length composition sampling
bias
Both:

Gear selectivity and avoidance


are related to length !

Sample layer - representative ?

Sampling sometimes not possible

8
8
Non-accessibility : shadow zones

Draft
Nearfield Surface BLINDZONE 0.5 + 5 m

Bottom resolution
Bottom DEADZONE
0.1 - 2 m depending on:
Pulse length
Back-step range
depth and beam

Steep bottom :
extended bottom shadow zone
9
9
Non-accessibility : avoidance
Horizontal avoidance:
- demonstrated for pelagic fish in
schools
and layers

Vertical avoidance:
- demonstrated for layers of
gadoids and clupeoids
Depends on
• Noise level
• Noise spectrum (frequency spectrum of noise)
• Light (at night)
• Possibly visible clues 10
10
Example of avoidance

• Norwegian spring spawning herring


(Clupea harengus) in the Ofotfjord
• Echogram showing the avoidance using a
stationary echosounder
• Measured noise spectrum using a
hydrophone

11
11
The experiment

Herring

Float

Hydrophone

30 m
Echosounder
Acoustic
lander

Release
Anchor 12
12
The spectrogram

Measured noise from at different frequencies using the


hydrophone 13
13
Avoidance
Pass: e:\goS_nov_Des_2004\lan\Passeringsforsøk 2 - 031204\L0000-D20041203-T225915-EK60.raw.processed.mat

20 Herring
40
Depth [m]

60

80

100

120
Sea surface
140
01:00 01:30
Time

x 10
4 Research vessel
3
2.5
sA [m /n.mi. ]

2
2

1.5
2

0.5
Integrated values
0

-0.5
(NASC)
01:00 01:30
Time

14
14
Bubble Attenuation
Bubbles under the transducer from

•Wave action
•Vessels movements in the sea

Act as an absorber of sound

Can introduce considerable bias up till nearly full loss of


signal

Action :
•Correction based on empirical relationship
•Keel
•Towed body
•Stop integration in extreme weather 15
15
Attenuation by fish
In dense layers and schools, some of the acoustic signal
may be attenuated by the fish itself, and the density in
the deeper parts of the school is underestimated.

•Theory and methods to correct for this are available,


but should be separately studied for the surveyed
population

•Shadowing must be measured and the ratio (e/b)


must be determined.

Action:
•Correction based on empirical relationships
•See Foote et al. 1992, Zhao & Ona (2003)
16
16
Attenuation by herring

Zhao and Ona (2003)

17
17
Stock coverage
If total stock distribution not covered :

Result dependent on actual distribution between


covered an not covered areas

This distribution can not be expected to be constant

Special problems :
• non-accessibility in shallow water
• national boundaries
• cost

18
18
Stock migration
If stock is moving or migrating during the survey, a
systematic bias may occur.

Result dependent on actual migration speed, and


direction relative to track lines. See attempts to correct in
MacLennan & Simmonds (1992).

Action:
• No surveys in strong migration periods.
• Measure school migration speed, relative to vessel by
sonar.

19
19
Main types of
random error

• Biological under-sampling : too few hauls in high


variance zones
• Acoustic under-sampling : too few miles in high
variance zones

Action :
Improve survey design !!

20
20
Reference literature
• Aglen A. 1994 Sources of error in Acoustic estimation of
fish abundance.

21
21

You might also like