Download - Playing Sherlock Holmes with parasites
Establishing Harvest Location for Atlantic Cod: Playing Sherlock
Holmes with Parasites
Francisco Montero, Eugenia Ferrer, Diana Perdiguero, Toni Raga
& Juan Antonio Balbuena Cavanilles Institute of Biodiversity and Evolutionary
BiologyUniversity of Valencia, Spain
Establishing harvest location is relevant to fisheries and
consumer policies IUU Fishing (up to 30% of total catches)
Legal disputes between countries
Geographical characterization of seafood
Ensuring safety and enhancing confidence of consumers
http://www.codtrace.ie
The European Commission Community Research
Establishing traceability for cod (Gadus morhua) in European waters: determining
location of spawning and harvest
Marine Research Institute – Iceland
(Project coordinator)University College Dublin
Who Is Behind COTRACE?
Faunistic characterization
Differences between localities (exploratory)
Predictive tool assigning individual cod to its harvest location
Conclusions (preliminary)
This presentation concerns preliminary analyses of 148 cod
from 3 Localities
ICELAND
CELTIC SEA
BALTIC SEA
NORTH SEA
IRISH SEA
BALTIC SEA
N= 60
NORTH SEA
N= 28
IRISH SEA
N=60
Cod harboured rich parasite assemblages
32 forms: 25 helminths, 6 crustaceans, 1 leech
About 19,000 parasite individuals
Only 1 cod parasite free
Diversity of parasite assemblages differred
between localities
Species richness Diversity* Mean SD Range Mean SD
Baltic Sea 1.6 0.8 0-4 0.22 0.29
Irish Sea 9.3 2.2 4-15 1.43 0.35
North Sea 8.0 2.3 4-13 1.35 0.34*Shannon’s
Differences could be visualized with catPCA
Axis 1
210-1-2-3
Axis
28
6
4
2
0
-2
-4
North Sea
Baltic Sea
Irish Sea
Variation was related to certain species
Axis 1
1.0.50.0-.5-1.0
Axis 2
0.9
0.7
0.5
0.3
0.1
-0.1
-0.3
-0.5
LBRANC
CELONG
CCURTUS
CORYNO
SPIROID
ANIS1
PDECIPHYSTE
CONTRAC
ASIMP
CUCUL
ASCARO
SPLEUR
TRYPANORABOTGADI
STEPHAN
OPECH
HEMIUR
GONOC
DVAR BUCEPH1
EGADI
North Sea
Irish Sea
Baltic Sea
E. gadi
D. varicus
H. aduncum
Hemiurus sp.
Corynosoma sp.A. simplex
LBRANC
CELONG
CCURTUS
CORYNO
SPIROID
ANIS1
PDECIPHYSTE
CONTRAC
ASIMP
CUCUL
ASCARO
SPLEUR
TRYPANORABOTGADI
STEPHAN
OPECH
HEMIUR
GONOC
DVAR BUCEPH1
EGADI
North Sea
Irish Sea
Baltic Sea
A. simplex
H. aduncumD. varicus
Contracaecum sp.
Sp 1
Sp 4
Sp 2
Sp 3N
B
I
Hidden LayersInput Layer Output Layer
Neural Networks worklike an artificial Sherlock Holmes
We applied 2 different Neural Networks
Radial Basis Function: equalize training set
Bayesian NN: no validation set
33 Baltic33 Baltic
34 Irish34 Irish
33 North33 North
12 Baltic12 Baltic
12 Irish12 Irish
12 North12 North
24 Baltic24 Baltic
22 Irish22 Irish
18 North18 North
Training Validation Test
20 Baltic20 Baltic
20 Irish20 Irish
20 North20 North
40 Baltic40 Baltic
40 Irish40 Irish
8 North8 North
Training Test
04000200Irish
8001910North
10390020Baltic
NorthIrishBalticNorthIrishBalticPredictedPredictedTestTraining
Act
ual
Bayesian Neural Network
180012002850North
220011101330Irish002400121032Baltic
NorthIrishBalticNorthIrishBalticNorthIrishBalticPredictedPredictedPredictedTestValidationTraining
Act
ual
Radial Basis Function
97%97%94%
98% 99%
...and these are the encouraging results
Conclusions
1. Geographical differences in parasite fauna
2. Parasites show promise for cod traceability
3. A few parasite species may do the job (cost efficiency)
4. Artificial Neural Networks seem appropriate statistical tools for this type of problem
Thank you