SUBSAMPLING ACCURACY IN BEAM TRAWL CATCHES
Samuel Van de Walle 1, 2, Jochen Depestele1,2 Hans Polet1, Kris Van Craeynest1 & Magda Vincx2
1Institute
for Agricultural and Fisheries Research, Animal Sciences – Fisheries
ILVO, Biological Environmental Research, Ankerstraat 1, 8400 Oostende, Belgium
2Ghent University, Biology Research Group, Marine Biology Section
Krijgslaan 281 (Campus Sterre – S8), B-9000 Gent, Belgium
N° 58
Introduction
57
In commercial beam trawling data on fish catches is routinely
available, while composition of invertebrate discards is ignored.
However, from an ecosystem perspective such data is equally
important yet is highly labor intensive to acquire. Subsampling the
discards, and accepting the error this inherently imposes, is the only
option. Here we investigate these error rates, which will allow for
more accurate data or estimates in discard composition.
56
55
54
53
52
51
N° 50
-4
W°
-2
Figure 1: Catch Sorting in Progress
A
S
(4)
=
Rare
Common
40
Abundant
Dominant
30
20
10
0
p= proportion of
catch analyzed
a= Species abundance
in subsample
T= Species abundance
in haul
W= Total weight of haul
20
15
N1 (122,0kg)
N3 (19,0kg)
N5 (53,0kg)
10
(1)
Categories of Abundance Index (n)
0
%
Asterias rubens
24068
47,19
Echinidea sp.
21324
41,81
Buccinum undatum
901
1,77
Pagurus bernhardus
675
1,32
Liocarcinus
depurator
636
1,25
Liocarcinus holsatus
608
1,19
Aphrodita aculeata
416
0,82
Palaemoninae sp.
318
0,62
Necora puber
247
0,48
236
0,46
…
…
…
Totaal
51001
Table 1: Top Abundant Species
Discussion
Trawl N1
400
20
40
60
Sampling Error (S) (%)
Rare
Common
Rare 95% CI
Common 95% CI
300
200
150
100
50
0
0
20
40
60
80
100
Proportion of Catch Analysed (%)
Sampling error (S) (%)
40
35
Abundant
Dominant
30
25
Abundant 95% CI Dominant 95% CI
20
15
10
5
0
0
10
20
30
40
50
60
70
80
100
25
20
15
F1 (48,6kg)
F3 (57,7kg)
F5 (28,8kg)
10
80
90
100
Proportion of Catch Analysed (%)
Corresponding author:
 +32(0)472 31 94 88  samuel.vandewalle@outlook.com
F2 (86,3kg)
F4 (28,0kg)
A1 (110,0kg)
5
0
20
40
60
80
100
30
25
20
15
N1 (122,0kg)
95% CI
5
0
250
N2 (154,0kg)
N4 (65,0kg)
N6 (125,0kg)
5
Individuals
Ophiura ophiura
Mean Sampling
error for 3/8 of
haul analyzed
350
10
E°
25
10
C
8
30
Species
50
S
(3)
S
6
B
Number of Species Recorded
Number of Cases (Species x Trawl)
(2)
S
4
30
60
S
2
Figure 2: Trawling Location
Materials & Methods
Twelve trawls were performed, in February 2009(F1-F3), February 2010(F4&F5), April
(A1) and November2011 (N1-N6) at the location in Fig. 2. Every species x trawl
combination was assigned to an abundance category(1) based on an abundance
index (n). Hauls were subdivided into 10L buckets(2) and individuals were identified.
Different numbers of buckets(from 1 to all) from a trawl were used to simulate
different subsample sizes. For a large number of random recombination's of buckets, a
sampling error (S) was calculated(3) and from all these a mean error and a confidence
interval was derived(4). A similar approach was then used to calculate the mean
number of species in a
subsample of a certain
size.
S
0
20
40
60
80
100
(A) Species Abundances
Proportion of Catch Analysed (%)
All hauls were highly dominated by
starfish and sea urchins. A majority of
species was found to be rare.
(B) Species Richness
An average of 23 species per trawl was found. Despite large differences in catch
weight there was comparatively little variation in number of species. Most trawls
follow a similarly shaped curve regardless of their weight or number of species. For
N1 the confidence interval showed a constant range over al subsample sizes.
(C) Sampling Error
Results for N1 are given as example, other trawls revealed similar patterns. Average
error decreased with increasing subsample size. Rare species showed the highest
error rates while common species had intermediate error rates. For abundant &
dominant species errors were up to 10 times lower and surprisingly similar to each
other. This indicates that after a certain abundance threshold is reached(around
5ind./10kg), error rate is unlikely to drop much further, even at highest abundances.
Future Prospects
Currently most results are represent all trawls individually. In the next step, we aim
to produce overall results for all trawls combined. Interpolation and/or a general
linear mixed model approach should allow for this, while also accounting for the
complex interdependence of all data points.