332 
Fishery Bulletin 1 14(3) 
Hokkaid6 
Sanriku 
Joban 
byashi(?.te§io|2. 
Kui;oshio/egion^ ^ 
140'’E 142°E 144°E 146°E US^E ISO^E 152'’E 154‘’E 
46“N 
44°N 
42‘’N 
40°N - 
38°N - 
SG'N - 
34''N 
"v- - > ’ ^ '• '"*^^***‘ - 
's^ A'. ’ ^ ‘ ' 
i s.' vv V s ^ > ' » 
V* ■ It*' ' % 
H;- tiV' y V ' 
, ' SuU^dlic-^ptrppicar 3/- 
Trgnsitipnzone' 
\ V- - ■ ' 
.. V -c 
V--* , ‘ 1 V 
I Depth (m) 
-8000 
-6000 
-4000 
-2000 
0 
Figure 1 
Map of the study area in the western North Pacific with the hydro- 
graphic and topographic features of the ocean basin. The line with 
dashes and dots represents the boundary of the EEZ of Japan. The 
lines with dashes correspond to the 3 major oceanic fronts — the Po- 
lar Front (PF), Subartic Front (SAF), and Kuroshio Extension Front 
(KEF). The subarctic-subtropical transition zone is also shown. Re- 
drawn after Murase et al. (2014). 
Detection of fishing vessel 
We examined the histograms of DNs in our analyses 
of OLS images for each month in order to identify the 
fishing areas. Several peaks in DNs were recorded 
over the examined 5-month periods (Fig. 2). To extract 
the areas with fishing-vessel lights, DN thresholds for 
identifying Pacific saury fishing vessels were calculated 
for each month because of the monthly differences in 
DN frequency distribution. A 2-level slicing method 
was used to extract the bright areas thought to be 
caused by the fishing fleet. This method is used to find 
a statistical optimum threshold from the DN frequency 
distribution (Takagi and Shimoda, 1991). 
The thresholds, k, were determined through the 
use of the following method proposed by Kiyofuji and 
Saitoh (2004), and the variance, cf(k), was calculated 
with the equations proposed by Takagi and Shimoda 
(1991); 
where 
N 
P. 
(Oq 
Ao 
(7Hk)=o)o(iJ,-I^Ty^ + ct>i (1) 
the number of pixels at i levels; 
the total number of pixels; 
nJN; 
Y!l=xPi and cOi = 
T,Li^P^ and /i-i = T,Lk+i^P^ 
liT= T!i=iiPi- 
With these methods, 5 thresholds were identified (Table 
2). Class 1, 2, 3, and 4 thresholds indicate ocean water 
or cloud coverage, and the class 5 threshold indicates 
bright areas resulting from fishing vessel lights. There- 
fore, class 5 threshold values were applied to extract 
the bright areas that represented fishing vessel lights. 
Lights from vessels that fish for Pacific saury and 
those that fish for squid are contained in OLS images. 
These lights are difficult to distinguish from each other; 
therefore, it is necessary to generate OLS images with 
less contamination from the lights of vessels fishing for 
