Data to E ti t S ti l Eff tf Estimate Spatial Effort for Unobserved Vessels

Using
U
i
V
Vessell Monitoring
M it i
System (VMS) Data to
E ti
Estimate
t Spatial
S ti l Effort
Eff t ffor
Unobserved Vessels in the
B i
Bering
Sea
S
Pollock
P ll k Fi
Fishery
h
Alan Haynie, NOAA Fisheries
Alaska Fisheries Science Center
P ti kS
Patrick
Sullivan,
lli
C
Cornell
ll U
University
i
it
Overview
• VMS and Observer coverage in the Bering Sea pollock
fi h
fishery
• Methodology
• Results
R
l
• Conclusions
2
The Bering Sea
pollock fishery is
conducted by
y
United States
vessels in the
B i
Bering
Sea
S
3
4
Pollock Fishery
y
•1 – 1.5 million MT/ year
•Pollock make up approximately
30 percent of the fish and shellfish
landed in the US
•75%
5% o
of Alaska
as a g
groundfish
ou d s fishery
s ey
in recent years
5
VMS Coverage in Bering Sea Fisheries
•
All vessels that conduct directed fishing for pollock, Pacific cod,
and Atka mackerel in the Bering Sea and Aleutian Islands are
required to have VMS (beginning in 2003)
•
VMS system transmits the following information 2-3 times/hour
throughout the season
•
Data used here are data for 2004 (approximately 100,000
observations used in this analysis)
— Vessel identifier
— Location
— Bearing
— Speed
6
Observer Program
g
Data
• On
On-board
board observers required for 100% of days at sea
for vessels >124 feet (~ 38 meters);
• On-board observers required for 30% of days at sea
for vessels 60-124 feet (18 – 38 meters)
• More than 80% of catch is observed
7
Salmon Bycatch Spatial Variation
Source: Sea State, Inc.
8
VMS Research Question
Can we use th
C
the VMS d
data
t ffrom
unobserved trips
p to better estimate
spatial effort and bycatch in the
pollock fishery?
9
Generalized Additive Model
(Logit Link)
⎧ π ⎫
l ⎨
log
⎬ = α + s ( St ) + s ( St −1 ) + s (ΔB{t − 2,t + 2} )
⎩1 − π ⎭
where
π = Probability of fishing
α = Intercept
s ( St ) = Smooth function of speed at time t
s ( St −1 ) = Smooth
h function
f
i off speedd at time
i t −1
s (ΔB{t − 2,t + 2} ) = Smooth function of change in bearing
over times t − 2 to t + 2
10
0.2
0.4
Proba
ability of fishing
0.4
0.2
0.0
0
0.0
0
Proba
ability of fishing
0.6
0.6
0.8
Probability of fishing relative to
speed
0
2
4
6
8
Speed at time t
10
12
14
0
2
4
6
8
10
12
14
Speed at time t-1
11
0.4
0.3
0
0.2
0.1
0..0
Probab
bility of fishing
0.5
0.6
Probability of fishing relative to
mean change in bearing
0
50
100
Mean change in bearing
150
12
Receiver-Operator characteristic (ROC)
Curve
0.2
0.4
0.6
•Specificity
Prob(Pred=0|Obs=0)
Prob(Pred=0|Obs=0).
CV Vessels
( 0.94 )
0.0
senssitivity
0.8
1.0
•Sensitivity
Prob(Pred=1|Obs=1)) relative to
the specificity of the prediction
0.0
0.1
0.2
0.3
0.4
1 - specificity
13
Map
p of Statistical Areas
14
Preliminary Results for Full coverage
vessels: VMS vs. Observed % effort by
area and squared difference
Area
VMS100
Observed100
DifSq100
1
16.9
17.9
1.0
2
10 8
10.8
10 1
10.1
05
0.5
3
2.8
2.6
0.0
4
5.9
5.5
0.2
5
4.0
4.5
0.3
6
4.9
5.0
0.0
7
7.3
8.4
1.3
8
5.5
6.1
0.4
9
13
1.3
15
1.5
00
0.0
10
3.5
3.6
0.0
15
Preliminary Results for 30% Coverage
Vessels: VMS vs. Observed % effort by
area and squared difference
Area
VMS30
Observed30
DifSq30
1
18.0
21.1
9.7
2
14.3
20.8
42.6
3
7.7
10.4
7.5
4
5.1
5.3
0.0
5
3.9
3.2
0.5
6
3.7
4.2
0.3
7
3.7
4.1
0.2
8
3.4
2.7
0.6
9
2.5
1.6
0.9
10
2.5
1.8
0.6
11
2.4
2.5
0.0
12
2.2
0.0
4.7
13
22
2.2
19
1.9
01
0.1
14
1.6
0.8
0.7
15
1.6
0.0
2.4
16
Conclusions & Next Steps
p
• Current algorithm allows accurate prediction of Areas
fished for fully observed vessels
vessels.
• Preliminary analysis indicates that there is a
significant
i ifi
diff
difference iin spatial
i l effort
ff ffor observed
b
d and
d
unobserved fishing for partially observed vessels.
• Current work will estimate the spatial variation in
effort between seasons and years; cross-validation is
being conducted
conducted.
• Methods can be used with VMS data in other
locations to accurately estimate spatial effort
17
Acknowledgements
•
Funding support for Dr.
Sullivan from NOAA
Fisheries Science and
Technology and previous
support from NOAA
Fisheries AK Region
•
VMS data provided by
NOAA Fisheries
Enforcement
•
Brilliant symposium
organizers for choosing to
have this meeting in Rio
• Muito obrigado!
18