FISHERIES OCEANOGRAPHY

Fish. Oceanogr. 19:5, 382396, 2010

Habitat characteristics of skipjack tuna (Katsuwonus pelamis)

in the western North Pacific: a remote sensing perspective

ROBINSON MUGO,1,2,* SEI-ICHI SAITOH,1

AKIRA NIHIRA3 AND TADAAKI
KUROYAMA3
1

Laboratory of Marine Environment and Resource Sensing,

Graduate School of Fisheries Sciences, Hokkaido University,
3-1-1 Minato-cho, Hakodate, 041-8611, Hokkaido, Japan
2
Kenya Marine and Fisheries Research Institute,
P.O. Box 81651, Mombasa, Kenya
3
Ibaraki Prefecture Fisheries Research Station, Hitachinaka,
Ibaraki, Japan

ABSTRACT
Skipjack tuna habitat in the western North Pacific was
studied from satellite remotely sensed environment
and catch data, using generalized additive models and
geographic information systems. Weekly resolved remotely sensed sea surface temperature, surface chlorophyll, sea surface height anomalies and eddy kinetic
energy data were used for the year 2004. Fifteen generalized additive models were constructed with skipjack catch per unit effort as a response variable, and
sea surface temperature, sea surface height anomalies
and eddy kinetic energy as model covariates to assess
the effect of environment on catch per unit effort
(skipjack tuna abundance). Model selection was based
on significance of model terms, reduction in Akaikes
Information Criterion, and increase in cumulative
deviance explained. The model selected was used to
predict skipjack tuna catch per unit effort using
monthly resolved environmental data for assessing
model performance and to visualize the basin scale
distribution of skipjack tuna habitat. Predicted values
were validated using a linear model. Based on the fourparameter model, skipjack tuna habitat selection was
significantly (P < 0.01) influenced by sea surface
temperatures ranging from 20.5 to 26C, relatively
oligotrophic waters (surface chlorophyll 0.080.18,
0.220.27 and 0.30.37 mg m)3), zero to positive

*Correspondence. e-mail: robin_mugo@salmon.fish.hokudai.

ac.jp robinmugo@gmail.com
Received 16 May 2009
Revised version accepted 18 May 2010
382

anomalies (surface height anomalies 050 cm), and

low to moderate eddy kinetic energy (0200 and 700
2500 cm2 s2). Predicted catch per unit effort showed
a trend consistent with the northsouth migration of
skipjack tuna. Validation of predicted catch per unit
effort with that observed, pooled monthly, was significant (P < 0.01, r2 = 0.64). Sea surface temperature
explained the highest deviance in generalized additive
models and was therefore considered the best habitat
predictor.
Key words: generalized additive models, geographic
information systems, habitat characterization, remote
sensing, skipjack tuna, western North Pacific

INTRODUCTION
Skipjack tuna (Katsuwonus pelamis) is a highly migratory pelagic species inhabiting all tropical and subtropical waters of the worlds oceans (Matsumoto,
1975; Arai et al., 2005). The species is commercially
important, ranked among the first 10 species that have
contributed highly to global catches in previous years
(FAO, 2009). A significant portion of these catches
are from the Pacific Ocean, which has one of the most
productive fisheries in the world, particularly the
western North Pacific. Skipjack tuna catches from
the Pacific Ocean have increased consistently since
the 1980s (Miyake et al., 2004). Despite the high
catches and exploitation rates, the western Pacific
stock is said to be capable of sustaining even larger
catches (Lehodey et al., 1998). Catches are highest
from May to August off Japan (Wild and Hampton,
1993). Katsuwonus pelamis are caught almost entirely
by surface gears such as pole and line (Langley et al.,
2005) and purse seines, although other miscellaneous
gears are also used (Miyake et al., 2004).
Distribution in sub-tropical waters is confined to
the 15C surface temperature isotherm (Wild and
Hampton, 1993). In the western Pacific, skipjack tuna
have been captured as far as 44N off Japan (Wild and
Hampton, 1993; Langley et al., 2005). Migration patterns in the western North Pacific follow a north
south seasonal cycle where the poleward movement
occurs in the fallsummer season (Kawai and Sasaki,

doi:10.1111/j.1365-2419.2010.00552.x

 2010 Blackwell Publishing Ltd.

Skipjack tuna habitat from RS & GIS in western NP

383

Figure 1. Schematic illustration of the northern migration of skipjack tuna in the western North Pacific, off the southeast and
east coasts of Japan. The northern migration routes of different groups of skipjack tuna are shown with green lines (14). The
typical path of the Kuroshio Current is indicated by the continuous red line; the Oyashio Current is shown in blue. Figure
modified from Nihira (1996).

1962; Matsumoto, 1975; Watanabe et al., 1995;

Ogura, 2003). This migration is also influenced by
ocean currents and the fish move along prevailing
currents, utilizing them as foraging habitats (Uda and
Ishino, 1958; Uda, 1973). The westernmost groups
comprise one originating from the Philippine islands
and a second group from the MariannaMarshall islands. These groups migrate northwards along the
Japanese coastal waters (Fig. 1). The third group
originates east of the Marshall Islands and moves
northwest into Japanese offshore waters (Matsumoto,
1975). Part of this group could move farther downstream of the Kuroshio to the east of Midway Island. In
late summer and early autumn, the fish begin their
southward migration. Skipjack tuna are known to
associate with fronts, warm-water streamers and eddies
during their northward migration from sub-tropical to
temperate waters (Tameishi and Shinomiya, 1989;
Sugimoto and Tameishi, 1992). We hypothesized that
skipjack tuna were utilizing oceanographic features
recognizable from satellite remotely sensed data, and
that such information could be used in a multivariate
model to derive habitat signatures for the species in
the western North Pacific.
Skipjack tuna physiology and morphology play a
major role in determination of habitat and, by
extension, distribution (Wild and Hampton, 1993).
They lack a swim bladder, which allows for rapid
vertical movements within the near surface habitat.

They also have high oxygen demands (33.5 mL L1)

(Barkley et al., 1978) due to high metabolic rates.
Oxygen concentration usually approaches saturation
(4.5 mL L1) in surface waters, but is often less than
the minimum requirement for skipjack (2.45 mL L)1)
in waters below the thermocline (Wild and Hampton,
1993), restricting skipjack tuna mainly to the mixed
layer above the thermocline. This makes oceanographic satellite observations ideal for habitat studies
for such a species. Nihira (1996) explained a size
screening mechanism for migration of skipjack tunas
across the Kuroshio Front, a phenomenon where only
fish above 45 cm in fork length are able to move from
the Transitional Zone, across the front, and into the
southern area of the Sub-tropical Counter Current
during the southward migration. This was due to their
ability to raise their body temperature while in the
transitional area. Those smaller than 45 cm were
incapable of raising their body temperature and thus
remained in the Transitional Zone.
Global demand for fish is exerting more pressure on
fish stocks, in addition to climate change-induced
impacts (changes in species distributions and disruption of marine ecosystems) (Cheung et al., 2009).
Developing robust tools for near-real time habitat
assessments will facilitate prediction of stocks responses to externalities such as climate change and
fishing pressure. In many studies (Saitoh et al., 1986;
Sugimoto and Tameishi, 1992; Nihira, 1996; Andrade

 2010 Blackwell Publishing Ltd, Fish. Oceanogr., 19:5, 382396.

384

R. Mugo et al.

and Garcia, 1999; Andrade, 2003) temperature has

been the main environmental variable used to explain
skipjack tuna occurrence and abundance. Although
temperature is important, other factors such as chlorophyll concentration and ocean mesoscale variability
could have direct or indirect effects on forage distribution and hence on the distribution of apex predators. Chlorophyll concentration is good indicator of
albacore tuna habitats (Laurs et al., 1984; Zainuddin
et al., 2008), and mesoscale variability is known to
influence catch per unit efforts (CPUEs) of albacore
tuna (Domokos et al., 2007) and foraging habitat for
seabirds (Nel et al., 2001). Some of these variables
may synergistically form suitable habitats for pelagic
species.
Generalized additive models (GAMs) were used to
model skipjack tuna habitats from catch data and
satellite remotely sensed oceanographic data. A GAM
(Hastie and Tibshirani, 1990) is a semi-parametric
extension of a generalized linear model, with an
assumption that the functions are additive and that
the components are smooth (Guisan et al., 2002). It
uses a link function to establish a relationship between
the mean of the response variable and a smoothed
function of the explanatory variable(s). The strength
of GAMs lies in their ability to deal with highly nonlinear and non-monotonic relationships between the
response and the set of explanatory variables. This
makes them ideal for expressing underlying relationships in ecological systems. Statistical models and GIS
(Valavanis et al., 2008) are tools with the potential to
enhance species habitat research. The objective of this
work was to study skipjack tuna habitat from multisensor satellite remotely sensed environment and
fishery data, using GAMs and GIS.

ders twice after leaving the coast of Hokkaido, generating the first and second intrusions (Kawai, 1972).
The meanders are separated by a warm core ring
(WCR) originating from the northward movement of
the ring produced by the Kuroshio (Yasuda et al.,
1992). The southern limit of sub-polar waters is often
referred to as the Oyashio Front (Talley et al., 1995).
The Oyashio ecosystem is an important fishing ground
for several sub-arctic species and sub-tropical migrants
(Saitoh et al., 1986). The Kuroshio originates from the
sub-tropical gyre and is distinguished by low density,
nutrient-poor, warm and high salinity surface waters
(Kawai, 1972; Talley et al., 1995). The Kuroshio
Extension is an eastward-flowing inertial jet characterized by large-amplitude meanders and energetic
pinched-off eddies, with high eddy kinetic energies
(Qiu, 2002). Confluence of the two currents results in
a mixed region, the KuroshioOyashio Transition
Zone (Yasuda, 2003). The behavior of the Kuroshio
Extension, warm streamers and WCRs in the Transition Zone is important to the fishing industry (e.g.,
Saitoh et al., 1986; Sugimoto and Tameishi, 1992).
Fishery data
Skipjack tuna daily catch data obtained from the
Ibaraki Prefecture Fisheries Research Station, for the
period March to November (2004) were digitized from
fishing logs of a pole and line fishery (173 vessels) and
compiled into a database. These data comprised daily
geo-referenced fishing positions (latitude and longitude), catch in tonnes and effort, from which catch per
unit effort (CPUE) was determined in tonnes per boat
day. The data were mapped using ARCGIS 9.2 ( ESRI,
Redlands, CA, USA) and further compiled into
weekly and monthly resolved datasets.
Remotely sensed environmental data

MATERIALS AND METHODS

Study area
This study was conducted in the western North Pacific
(1850N and 125180E) (Fig. 1), an area where
Japanese skipjack tuna fishing vessels operate off the
east coast of Japan. It is a productive ecosystem
influenced mainly by the Tsugaru Warm Current, the
Oyashio Current and the Kuroshio Current (Talley
et al., 1995). The Tsugaru Warm Current originates
from the Tsushima Current and flows with warm and
saline water from the Sea of Japan (Talley et al.,
1995). The Oyashio waters, formed from the Okhotsk
Sea and the Western Sub-arctic Gyre (Yasuda, 2003),
flow southward, transporting low temperature, low
salinity and nutrient-rich waters to the sub-tropical
gyre (Sakurai, 2007). The Oyashio commonly mean-

Weekly and monthly environment databases were

compiled for sea surface temperature (SST), sea surface chlorophyll (SSC), sea surface height anomaly
(SSHA) and eddy kinetic energy (EKE). Daily SST
and SSC, Moderate Resolution Imaging Spectroradiometer (MODIS) Aqua standard mapped images
(SMI) for the year 2004, with a spatial resolution
of approximately 4.63 km were downloaded from
the PO.DAAC (http://podaac.jpl.nasa.gov) and the
Ocean Color (http://oceancolor.gsfc.nasa.gov) sites
respectively, and composited into 7-day images using
SEADAS version 5.3 (NASA, Greenbelt, MD, USA).
Given that it is extremely challenging to match daily
fishery data to daily chlorophyll images (which usually
have very sparse data due to clouds), we used 7-day
composite images. The 7-day MODIS SST and SSC

 2010 Blackwell Publishing Ltd, Fish. Oceanogr., 19:5, 382396.

Skipjack tuna habitat from RS & GIS in western NP

images also matched the temporal scale for SSHA and

geostrophic velocities from AVISO (Archiving, Validation and Interpretation of Satellite Oceanographic
data) which have a weekly temporal resolution. Ninemonthly (MarchNovember) SST and SSC SMI were
also downloaded from the PO.DAAC and Ocean
Color sites, respectively. Weekly mean sea level
anomaly data were downloaded from the AVISO
(http://www.aviso.oceanobs.com), into ARCGIS 9.2
using the Marine Geo-spatial Ecology Tool (MGET)
(Roberts et al., in press), a custom made add-in program for marine-GIS applications. We downloaded
the delayed time, updated and merged product of
mean sea level anomaly. The data are global images
with a 1 3 resolution. They were re-sampled to the
SST and SSC resolution and subset to the study area,
using ARCGIS 9.2. The weekly SSHA images were
averaged to monthly images in SEADAS 5.3. Weekly
global geostrophic current velocity images, 1 3
(u and v components) were downloaded as ARCGIS
rasters from AVISO using the MGET. The u and v
weekly rasters were used to calculate EKE with the
Raster Calculator function in Spatial Analyst extension (ARCGIS 9.2) using Eqn 1 (Robinson, 2004). The
calculated EKE rasters were subset to the study area.
The weekly images were averaged to monthly EKE in
SEADAS 5.3.
EKE 1=2u2 v2

Matching fishery data to remotely sensed environment data

Weekly resolved skipjack tuna data were matched to
corresponding images for SST, SSC, SSHA and EKE
using a C-shell script. The ship-track function in
SEADAS 5.3 was used to extract values corresponding
to latitude and longitude positions from the fishery
dataset. The result was a full matrix of CPUEs and the
respective environmental variables (Valavanis et al.,
2008). This matrix was used to fit GAMs.
Generalized additive models
GAMs were constructed in R (version 2.7.2) software,
using the gam function of the mgcv package (Wood,
2006), with CPUE as the response variable and SST,
SSC, SSHA and EKE as predictor variables. A model
of the form shown in Eqn 2 was applied
gui a0 s1 x1i s2 x2i s3 x3i :::sn xni ;

where g is the link function, ui is the expected value of

the dependent variable (CPUE), a0 is the model constant, and sn is a smoothing function for each of the
model covariates xn (Wood, 2006). The CPUE follows
a continuous distribution; therefore, we chose the

385

Gaussian family which is associated with the identity

link function. We used a logarithmic transformation on
CPUEs to normalize the asymmetrical distribution
(Zainuddin et al., 2008). A factor of 0.1 was added
before log-transformation to account for zero CPUEs
(Howell and Kobayashi, 2006). Models were constructed from the simplest form using one independent
variable, e.g., SST only, with subsequent addition of
predictor variables. Model selection was based on significance of predictor terms, deviance explained, and
reduction in Akaike Information Criterion (AIC)
value (Johnson and Omland, 2004). Constructed
GAMs can be used to predict skipjack tuna CPUEs
using the predict.gam function in mgcv package, given a
set of covariates similar to those used to build the
model. Such an approach was employed by Howell and
Kobayashi (2006) and Zagaglia et al. (2004). We made
predictions from the best model selected from a set of
15 models. Due to sparseness of data on SST and SSC
weekly images, we made predictions from monthly
composites of the four environmental predictors.
Spatial mapping and validation of predicted CPUEs
Predicted CPUEs were mapped using GENERIC MAPPING
TOOLS (Wessel and Smith, 1998) (GMT 4.4.0) and
subsequently overlain with observed CPUEs. The
mapped grids were sampled using the latitude and
longitude positions of the observed CPUEs, thus creating a matrix of observed versus predicted CPUEs.
We further compared observed and predicted CPUEs,
pooled monthly, using a linear model.
RESULTS
Temporal variability of CPUE and environmental
variables
The monthly spatial distribution of fishing sets from
March to November 2004 relative to oceanographic
variables is shown in Fig. 2ad. Figure 3 exemplifies
the weekly fishing ground-oceanographic environment
relationship in September (week 38) where the association with SST and SSC gradients and eddies is
more apparent. The fishing fleet moved north from
March to August, and south from September to
November. The latitudinal displacement is shown in
Fig. 4a. Figure 4bf presents weekly time series plot of
mean CPUE, SST, SSC, SSHA and EKE. The mean
CPUE (Fig. 4b) increased gradually until week 27,
after which there was a decline. The lowest mean
CPUE values were recorded in the last weeks of the
fishing season (OctoberNovember).
Mean SST rose to 27.8C (week 29) (Fig. 4c) and
thereafter declined to 17.5C in the last week. The

 2010 Blackwell Publishing Ltd, Fish. Oceanogr., 19:5, 382396.

386

R. Mugo et al.

Figure 2. Spatial distribution of skipjack tuna fishing locations overlaid on 9-monthly images for each of the four environmental
variables (a) SST, (b) SSC, (c) SSHA and (d) EKE). Fishing locations are shown as red dots.
(a)

(b)

mean SSC concentration was lowest in the 12th week

(0.09), after which it increased to about 0.4 mg m3
(week 17), with sharp rises in November (Fig. 4d).
Mean SSHA (Fig. 4e) shows a pattern where anoma-

(c)

(d)

lies were positive (week 1133). During this period

(MarchJune), the fishing fleet was at or south of the
Kuroshio Front. Anomalies from week 34 to 44 were
all negative, a period when the fleet had sailed beyond

 2010 Blackwell Publishing Ltd, Fish. Oceanogr., 19:5, 382396.

Skipjack tuna habitat from RS & GIS in western NP

387

Figure 3. Fishing positions (gray dots)

in one of the weeks in September 2004
(week 38) overlaid on weekly averaged
SST, SSC, SSHA and EKE. The 20C
SST and a 0.3 mg m3 SSC contour (red
line) are plotted on the respective images to emphasize SST and SSC gradients.
Eddies are observable on the SSHA and
EKE images.

the Kuroshio area. The mean EKE (Fig. 4f) ranged

from 110 to 2315.7 cm2 s2. All mean EKE values
were below 1000 cm2 s2 except for weeks 13, 14, 17,
26, 30, 31 and 41.
Distribution of CPUE and habitat variables from skipjack
tuna fishing sets
Distributions of CPUE and the four environmental
variables utilized by skipjack tuna in the western
North Pacific in 2004 are shown in Fig. 5. The
distribution of CPUE was asymmetrical (Fig. 5a). A
log-transformation of CPUEs indicated that our
assumption to transform the data was appropriate
(Fig 5b). Skipjack tuna were caught between 16 and
30C SST, with the highest frequency of fishing sets
occurring at 20 and 25C (Fig. 5c). Chlorophyll-a

concentration range for fishing sets was 02 mg m3,

with 0.10.4 mg m3 being the preferred concentration (Fig. 5d). The range for SSHA was )30 to 80 cm,
with )20 to 30 cm showing the highest frequency of
fishing sets, with a peak at zero (Fig. 5e). Distribution
of EKE was asymmetrical, with the highest frequency
of fishing sets between zero and 200 cm2 s2 (Fig. 5f).
GAM-derived habitat characteristics for skipjack tuna
Results for each of the 15 models (model, predictor
variables used to construct it, the respective degrees of
freedom, AIC, P-value and deviance explained) are
shown in Table 1. The table presents single parameter
models constructed from one predictor variable, twoparameter models, made from a combination of any
two predictor variables, three-parameter models from

 2010 Blackwell Publishing Ltd, Fish. Oceanogr., 19:5, 382396.

388

R. Mugo et al.

Figure 4. Time-series plots of weekly averaged (a) latitudinal displacement of fishing positions, (b) CPUE, (c) SST, (d) SSC,
(e) SSHA and (f) EKE.
(b)

(a)

(c)

(d)

(e)

(f)

three predictor variables and four-parameter models

from all variables. In all the models, the predictor
variables were highly significant (P < 0.01). Single
parameter models had the lowest deviance explained,
especially for SSHA and EKE. SST showed the highest
deviance explained among the single parameter
models, then chlorophyll-a. Addition of predictor
variables at different levels resulted in slight increase
or decrease in deviance explained, while the AIC
value varied. For instance a model constructed from

SSHA and EKE shows the two may not account for
much variability in skipjack tuna CPUE. SST and
EKE, or SST and SSC account for relatively higher
variability in CPUE, according to AIC and deviance
explained. Among the three-parameter models, combination of SST, SSC and EKE showed that all predictor variables were highly significant, had the lowest
AIC value and the highest deviance explained
(12.8%). Overall, the four-parameter model had the
lowest AIC value and the highest deviance explained.

 2010 Blackwell Publishing Ltd, Fish. Oceanogr., 19:5, 382396.

Skipjack tuna habitat from RS & GIS in western NP

389

Figure 5. Histograms of CPUE and environmental variables showing (a) distribution of CPUEs; (b) distribution of log-transformed CPUE, which helped to normalize the asymmetrical distribution observed in (a); and (c) utilization of SST, (d) SSC, (e)
SSHA and (f) EKE by skipjack tuna in the western North Pacific in 2004.
(a)

(b)

(c)

(d)

(e)

(f)

 2010 Blackwell Publishing Ltd, Fish. Oceanogr., 19:5, 382396.

390

R. Mugo et al.

Model

Variable

EDF

AIC

P-value

SST
SSC
SSHA
EKE
SST+SSC

SST
SSC
SSHA
EKE
SST
SSC
SST
SSHA
SST
EKE
SSC
SSHA
SSC
EKE
SSHA
EKE
SST
SSC
SSHA
SST
SSC
EKE
SST
SSHA
EKE
SSC
SSHA
EKE
SST
SSC
SSHA
EKE

8.3
8.7
8.2
8.1
7.8
8.6
8.2
7.7
8.2
8.2
8.7
8.2
8.7
8.2
6.4
8.2
7.8
8.7
8
7.7
8.6
8.2
8.1
2.8
8.4
8.7
5
8.3
7.6
8.7
2.5
8.3

4593.50
4736.30
5052.40
5003.40
4527.90

< 2.00
< 2.00
< 2.00
< 2.00
< 2.00
8.09
< 2.00
1.84
< 2.00
< 2.00
< 2.00
1.33
< 2.00
< 2.00
< 2.00
< 2.00
< 2.00
1.19
2.24
< 2.00
< 6.27
< 2.00
< 2.00
8.68
< 2.00
< 2.00
1.09
< 2.00
< 2.00
9.93
2.59
< 2.00

SST+SSHA
SST+EKE
SSC+SSHA
SSC+EKE
SSHA+EKE
SST+SSC+SSHA

SST+SSC+EKE

SST+SSHA+EKE

SSC+SSHA+EKE

SST+SSC+SSHA+EKE

4548.80
4479.30
4659.00
4544.40
4829.30
4488.10

4402.40

4431.10

4502.90

4371.10

A normal quantilequantile (QQ) plot fitted to

examine sample versus theoretical quantiles shows a
nearly straight 1 : 1 line (Fig. 6), implying that the
application of a Gaussian distribution was ideal.
A QQ plot is useful for assessing the relationship of
sample to theoretical quantiles (Wood, 2006).
GAM plots (Fig. 7) can be interpreted as the
individual effect of each predictor variable on CPUE.
Rug plots on the horizontal axis represent observed
data points and the fitted function is shown by the
thick line. The gray shade shows the 95% confidence
interval. A negative effect on CPUE was observed
from SST values of 1620.5C. From 20.5 to 21C
there was a sharp positive effect on CPUE. Beyond
26C, the plot shows a decline, because the number of
sets done within 2630C is low. Consequently the
confidence intervals are wider. For SSC, a positive
effect on CPUE is evident from 0.080.18 mg m3,
0.220.27, and 0.30.37 mg m3 (Fig. 7b). A decline

CDE%
1016
1016
1016
1016
1016
1014
1016
1009
1016
1016
1016
1015
1016
1016
1016
1016
1016
1012
1008
1016
1016
1016
1016
1011
1016
1016
1008
1016
1016
1013
1007
1016

9.9
8
3.6
4.3
11
10.7

Table 1. GAMs fitted in the model

selection process (N = 6747). The
model, predictor terms used, the estimated degrees of freedom, Akaike
information criterion (AIC) value and
percent cumulative deviance explained.
The best model was selected based on
significance of predictor terms, reduction
of AIC and increase in cumulative
deviance explained (CDE).

11.6
9.26
10.8
6.9
11.7

12.8

12.3

11.5

13.3

occurs towards higher SSC values over 0.4 mg m3.

The plot on SSHA (Fig. 7c) shows a range from )40
to 80 cm, with fewer data points on the extremes.
A positive effect on CPUE was noted from 0 to 50 cm.
The plot on EKE shows a high density of data points
between 0 and 1200 cm2 s2 (Fig. 7d), with a positive
effect on CPUE on values below 200 cm2 s2, also
with the highest number of fishing sets. In addition,
7002500 cm2 s2 shows a positive effect over the
mean, but these values (especially after 1000 cm2 s2)
represent fewer data.
Prediction and validation of CPUEs
Predicted and mapped CPUEs are shown in Fig. 8.
The predicted CPUE for skipjack tuna show a northward displacement from March (around 25N),
extending to about 41N in September, after which
there is a southward displacement. In October and
November, the areas with relatively high CPUEs are

 2010 Blackwell Publishing Ltd, Fish. Oceanogr., 19:5, 382396.

Skipjack tuna habitat from RS & GIS in western NP

Figure 6. A normal QQ plot for the four-parameter model

showing a plot fitted to examine sample versus theoretical
quantiles. The nearly straight 1 : 1 line of the plotted points
shows that the sample distribution is very similar to a standard normal distribution. The model was constructed using a
Gaussian distribution.

extending southwards. The predicted values lie between zero and 2 tonnes per boat day. Correlation of
observed and predicted CPUEs pooled monthly
showed a significant relationship (P < 0.01; r2 = 0.64)
(Fig. 9).
DISCUSSION
We combined GAMs and GIS to study skipjack tuna
habitat from fishery data and satellite remotely sensed
datasets in the western North Pacific. Fishing data,
though not free from sampling bias based on the
fishermens choice of fishing locations, are low-cost
species distribution data sets available to fishery scientists. Fishing data CPUE are often used as an index
of fish occurrence and abundance (Lehodey et al.,
1998) and therefore high CPUEs can be said to indicate preferred oceanographic conditions for a species.
Here we further examine our results and their inherent
relevance as environmental indicators of skipjack tuna
habitat.
During weeks 1729, the fishery operated between
30 and 35N (Fig. 4a), a period when the mean weekly
SSTs (Fig. 4c) are increasing as a result of rising surface water temperatures due to flow of warm waters
transported by the Kuroshio Current. This is also reflected by the mean SSC, which shows that the waters
were correspondingly oligotrophic (Fig. 4d). However,
after July the fishery advanced further north, where the

391

waters get progressively cooler and eutrophic owing to

the intrusion of Oyashio waters. This may partly explain the decline in mean SSTs towards November.
The northeastward migration of skipjack tuna is
thought to be strongly influenced by temperature
(Uda, 1973; Sugimoto and Tameishi, 1992; Nihira,
1996), which can be observed through rising SSTs in
the north, and the gradual extension of the Kuroshio
Current. According to Uda (1973), when the warm
Kuroshio water spreads over a broader northern area,
more skipjack become available to the Japanese fishery, but a strong Oyashio Current hinders skipjack
tuna migration and leads to a lower catch. Ito et al.
(1998) also showed that temperature affected the
migration pattern of skipjack tuna. Temperature as a
habitat signature may explain part of the observed
spatio-temporal variability in skipjack tuna fishing set
distribution and CPUE; indeed, SST is known for
influencing tuna migration (Sund et al., 1981). Our
work shows that SST significantly influenced skipjack
CPUE (abundance). The results are consistent with
previous findings in the western North Pacific, where
optimum SST range was 20.523C in the Kureshio
region (Mishra et al., 2001) and 2226.5C in the
southwest Atlantic (Andrade and Garcia, 1999).
However, our results also show a broader range of SST,
from 17 to 26.5C. A possible explanation for this
discrepancy could be the northward migration of
skipjack tuna in distinct groups (Nihira, 1996) that
occupy varying SST regimes (Fig. 2a).
Temperature limits horizontal and vertical distribution of skipjack tuna, and this varies by region and
size (Sund et al., 1981). Loukos et al. (2003) suggested
significant large-scale changes of skipjack habitat in
the equatorial Pacific due to increased ocean temperatures caused by global warming. Such a scenario
could expand available habitat for warm water pelagics
to higher latitudes (Cheung et al., 2009). Ogura
(2003) reported that tagged skipjack tuna in the
western North Pacific showed daily vertical movements, where they spent much of the time near the
surface during night time, swimming in deeper waters
with occasional dives during the day. Night time
depths ranged from surface to 30 m, whereas day time
dives often were beyond 100 m. The fish appeared to
avoid waters below 17C. In the equatorial Pacific,
skipjack tuna spent 98.6% of their time above the
thermocline (depth = 44 m) during the night but less
time (37.7%) below the thermocline during the day
(Schaefer and Fuller, 2007). According to the study,
one fish swam in waters where the ambient temperature reached 10.5C, while the peritoneal cavity
temperature reached 15.9C. Such tagging experi-

 2010 Blackwell Publishing Ltd, Fish. Oceanogr., 19:5, 382396.

392

R. Mugo et al.

Figure 7. GAM-derived effect of the four oceanographic variables on CPUE, from the model constructed with: (a) SST, (b)
SSC, (c) SSHA and (d) EKE. Gray-shaded area indicates the 95% confidence intervals; the solid line shows the fitted GAM
function which describes the effect that a predictor variable has on the response variable (CPUE). The relative density of data
points is shown by the rug plot on the x-axis. Values of a predictor variable showing a positive effect on CPUE were read as all
values for which the fitted GAM function was above the zero axis.
(a)

(b)

(c)

(d)

ments are important for understanding skipjack tuna

vertical habitat utilization. Our results (based on
SSTs) indicate that few fishing sets (< 5%) occurred at
temperatures between 16 and 17C (Fig. 5c).
Chlorophyll concentration, the only biological
component available to satellite remote sensing, is an
index of phytoplankton biomass which provides valuable information about trophic interactions in marine
ecosystems (Wilson et al., 2008). Skipjack tuna fishing
sets occurred in waters with relatively low SSC
(Fig. 5d). Previous work shows that skipjack tuna in
the western North Pacific were caught within relatively
low chlorophyll waters which corresponded to warm
water SSTs (Wilson et al., 2008). Preference for relatively low chlorophyll waters, especially on the frontal
edges of warm oligotrophic waters has both physio-

logical and trophic implications. This enables skipjack

tuna to not only locate and forage on the periphery of
highly productive frontal or upwelling zones, but also
stay within tolerable temperatures (Ramos et al.,
1996). Nihira (1996) reported that stomach contents
of skipjack tuna caught near the front of a warm
streamer were two to five times heavier than those
caught at the center of the warm streamer and thus
concluded that the front was the most suitable feeding
place. Fiedler and Bernard (1987) found that skipjack
tuna aggregated on the warm edge of waters near cold
and productive water masses. They concluded that this
enabled skipjack tuna to feed on large numbers of
euphausiids found in upwelled cold waters. A similar
conclusion was reached by Andrade (2003), indicating
that areas with thermal or color gradients provide both

 2010 Blackwell Publishing Ltd, Fish. Oceanogr., 19:5, 382396.

Skipjack tuna habitat from RS & GIS in western NP

393

Figure 8. Predicted skipjack tuna CPUEs overlaid with observed fishing locations from March to November, 2004. Predictions
were done on monthly averaged images due to lack of data in many of the weekly images, especially for SST and SSC.

Figure 9. A scatter plot of pooled monthly observed against

predicted CPUE values (P < 0.01, r2 = 0.64).

physiological and foraging advantages. While high

tuna abundances occur close to highly productive
frontal and upwelling zones, primary production per se
does not aggregate tuna (Lehodey et al., 1998). It is the
downstream development of secondary production
that provides attractive habitat for skipjack tuna
(Roger, 1994; Lehodey et al., 1998).
Oceanic fronts are broadly understood to mark
the boundary between two different water masses,
manifested as regions of strong horizontal gradients in

temperature, salinity, chlorophyll, and concentration

of zooplankton and micronekton (Olson et al., 1994;
Kirby et al., 2000). Our results indicate that the
highest frequencies of skipjack tuna fishing sets were
within 0.10.3 mg m3, with a peak frequency at
0.2 mg m3 (Fig. 5d). However, later in the year
(AugustNovember), the fishing fleet was more closely
associated with the 0.3 mg m3 SSC isopleth (e.g.,
Fig. 3). The 0.2 mg m3 isopleth in the North Pacific,
a proxy for the transition zone chlorophyll front
(TZCF), has been shown to be a major feature
important for migration and foraging of apex predators
such as tunas (Polovina et al., 2001; Polovina and
Howell, 2005). Reasons proposed to account for the
association of tunas with fronts include: (i) confinement to a physiologically optimum temperature range;
(ii) use of frontal gradients for thermoregulation; (iii)
limitation of visual hunting efficiency owing to water
clarity; and (iv) the availability of appropriate food
(Kirby et al., 2000). Frontal systems are important
aggregating mechanisms for plankton and micronekton (Laurs et al., 1984; Lehodey et al., 1998). Tunas
are predominantly visual predators, feeding opportunistically and unselectively on micronekton (Blackburn, 1968). Highly turbid waters are therefore
unsuitable for them (Laurs et al., 1984; Ramos et al.,

 2010 Blackwell Publishing Ltd, Fish. Oceanogr., 19:5, 382396.

394

R. Mugo et al.

1996; Kirby et al., 2000), and extremely oligotrophic

waters would contain little food (Sund et al., 1981).
Prey abundance and water clarity affect the rate of
food encounter (Kirby et al., 2000).
We used SSHA data to address the relationship
between skipjack tuna habitat and mesoscale variability. Among the four variables used to construct
GAMs, SSHA was the least significant (Table 1).
Predominantly positive SSHAs had a greater effect on
CPUE (Fig. 7c), implying preference for areas either
with zero anomalies or closely associated with warm
eddies. Zagaglia et al. (2004) also found SSHA to have
had the least influence on yellofin tuna CPUE, among
two other environmental variables, SST and surface
chlorophyll-a. They attributed this finding to the
possibility that SSHA could have varying effects on
aggregation of fishery resources, owing to its complex
combination of dynamical and thermo-dynamical
factors, thus its influence on distribution and abundance of pelagic resources may be less apparent.
Eddy kinetic energy results suggest that skipjack
tuna fishing sets were made in areas with low to
moderate EKE, indicating that there were instances
when catches were made in areas associated with eddies (e.g., Fig. 3). In the western North Pacific, eddies
have been shown to influence formation of skipjack
tuna fishing grounds (Sugimoto and Tameishi, 1992).
In the Gulf of Mexico, Bluefin tuna significantly preferred areas with moderate EKEs (251355 cm2 s2)
(Teo et al., 2007), while in the American Samoa
albacore, tuna catches were higher at eddy edges
(Domokos et al., 2007). Mechanisms leading to
aggregation of tuna along eddy edges include nutrient
injection or entrainment to the euphotic zone (Olson,
1991) and development of phytoplankton blooms
which trigger secondary production (Bakun, 2006).
This attracts nekton, with a net effect of aggregation
of apex predators to forage on the lower trophic level
organisms around the eddy edge (Ramos et al., 1996;
Fonteneau et al., 2009).
Predicted CPUEs show an appreciable concurrence
with actual fishing locations for the months May to
August (Fig. 8). This is also the period that showed
consistent increases in mean CPUE over the study
period, suggesting that the model performed fairly well
in predicting areas that showed increasing abundances.
However, given that the range of observed CPUEs was
>2 tonnes per boat day (Fig. 5a), the model appears to
have difficulties in predicting higher catch rates. The
relationship between observed and predicted CPUEs
was significant. Zagaglia et al. (2004) also reported a
significant relationship between observed CPUEs and
those predicted from a GAM (r = 0.5073) for yel-

lowfin tuna in the equatorial Atlantic Ocean. Howell

and Kobayashi (2006), working on GAMs for longline
big eye tuna fishery in the Palmyra Atoll, found a
significant relationship between observed and predicted catch rates (r = 0.35) on a set by set basis, and
on monthly averaged catch rates (r = 0.78), but reported difficulties predicting low and high catches in
their model. They suggested that additional data with
enhanced biological characteristics were likely to improve such models. Our model explained 13.3% of
variability in skipjack tuna abundance and was based
on environmental variables only, which are important
habitat predictors but probably not the only factors
influencing fishing locations for skipjack tuna. In
addition, data that include more years are likely to
generate a more robust model with greater predictive
power. The models were constructed from weekly resolved data sets due to absence of valid daily SST and
SSC data. We suggest that better habitat signatures
could be derived from satellite data with improved
temporal resolution and coverage.
CONCLUSIONS
We conclude that:
SST was the most important habitat predictor for
skipjack tuna migration in the western North
Pacific, followed by SSC.
The oligotrophic side of the Kuroshio Front and the
Kuroshio Extension were important skipjack tuna
habitat features.
Meso-scale features such as eddies played a role
in formation of skipjack tuna habitat, although
that role may not be as profound as that of SST and
SSC.
Our work was based on 1-yr fishery and remotely
sensed environmental data. Future work based on
longer term fishery data sets, with environment data at
higher temporal and spatial resolutions could further
improve skipjack tuna habitat models.
ACKNOWLEDGEMENTS
We are greatly indebted to one anonymous reviewer
for providing useful comments on earlier versions of
this manuscript. We also appreciate Mr. Fumihiro
Takahashis advice on some aspects of satellite image
analyses. We also appreciate the use of AQUA-MODIS
SST and chlorophyll-a data sets, downloaded from the
Physical Oceanography Distributed Active Archive
Center (PODAAC) at the Jet Propulsion Laboratory
(JPL) (http://podaac.jpl.nasa.gov) and the ocean color
portal (http://oceancolor.gsfc.nasa.gov), respectively.

 2010 Blackwell Publishing Ltd, Fish. Oceanogr., 19:5, 382396.

Skipjack tuna habitat from RS & GIS in western NP

Use of SSHA and geostrophic velocities data distributed by AVISO is also acknowledged.
REFERENCES
Andrade, H.A. (2003) The relationship between the skipjack
tuna (Katsuwonus pelamis) fishery and seasonal temperature
variability in the south-western Atlantic. Fish. Oceanogr.
12:1018.
Andrade, H.A. and Garcia, C.A.E. (1999) Skipjack tuna fishery
in relation to sea surface temperature off the southern Brazilian coast. Fish. Oceanogr. 8:245254.
Arai, T., Kotake, A., Kayama, S., Ogura, M. and Watanabe, Y.
(2005) Movements and life history patterns of the skipjack
tuna Katsuwonus pelamis in the western Pacific, as revealed
by otolith Sr:Ca ratios. J. Mar. Biol. Assoc. UK 85:1211
1216.
Bakun, A. (2006) Fronts and eddies as key structures in the
habitat of marine fish larvae: opportunity, adaptive response and competitive advantage. Sci. Mar. 70(S2):105
122.
Barkley, R.A., Neill, W.H. and Gooding, R.M. (1978) Skipjack
tuna, Katsuwonus pelamis, habitat based on temperature and
oxygen requirements. Fish. Bull. 76:653661.
Blackburn, M. (1968) Micronekton of the eastern tropical
Pacific Ocean: family composition, distribution, abundance
and relation to tuna. Fish. Bull. US 67:71115.
Cheung, W.W.L., Lam, V.W.Y., Sarmiento, J.L., Kearney, K.,
Watson, R. and Pauly, D. (2009) Projecting global marine
biodiversity impacts under climate change scenarios. Fish
Fish. 10:235251.
Domokos, R., Seki, M.P., Polovina, J.J. and Hawn, D.R. (2007)
Oceanographic investigation of the American Samoa albacore (Thunnus alalunga) habitat and longline fishing grounds.
Fish. Oceanogr. 16:555572.
FAO. (2009) The State of World Fisheries and Aquaculture 2008.
Rome: FAO.
Fiedler, P.C. and Bernard, H.J. (1987) Tuna aggregation and
feeding near fronts observed in satellite imagery. Cont. Shelf
Res. 7:871881.
Fonteneau, A., Lucas, V., Tewkai, E., Delgado, A. and Demarcq,
H. (2009) Mesoscale exploitation of a major tuna concentration in the Indian Ocean. Aquat. Living Resour. 21:109121.
Guisan, A., Edwards, C.T. and Hastie, T. (2002) Generalized
linear and generalized additive models in studies of species
distributions: setting the scene. Ecol. Modell. 157:89100.
Hastie, T. and Tibshirani, R. (1990) Generalized Additive Models
New York: Chapman and Hall, 356pp.
Howell, E.A. and Kobayashi, D.R. (2006) El Nino effects in the
Palmyra Atoll region: oceanographic changes and bigeye
tuna (Thunnus obesus) catch rate variability. Fish. Oceanogr.
15:6. 477489.
Ito, S., Ogura, M., Tanabe, T., Takeuchi, K. and Nonaka, M.
(1998) Estimation of the skipjack migration pattern by a
skipjack general migration model. Tohoku Natl. Fish. Res.
Inst. 60:4148.
Johnson, J.B. and Omland, K.S. (2004) Model selection in
ecology and evolution. Trends Ecol. Evol. 19:101108.
Kawai, H. (1972): Hydrography of the Kuroshio Extension.
p. 235352. In: Kuroshio, Its Physical Aspects. H. Stommel
& K. Yoshida (eds) Tokyo: University of Tokyo Press,
pp. 517.

395

Kawai, H. and Sasaki, M. (1962) On the hydrographic condition

accelerating the skipjack northward movement across the
Kuroshio Front [in Japanese, English abstract]. Bull. Tohoku
Reg. Fis. Res. Lab, 20:127.
Kirby, D.S., Fiksen, O. and Hart, P.J.B. (2000) A dynamic
optimization model for the behaviour of tunas at ocean
fronts. Fish. Oceanogr. 9:4. 328342.
Langley, A., Hampton, J. and Ogura, M. (2005) Stock assessment of skipjack tuna in the western and central Pacific
Ocean. WCPFCSC1 SA WP4.
Laurs, R.M., Fiedler, P.C. and Montgomery, D.R. (1984)
Albacore tuna catch distributions relative to environmental
features observed from satellites. Deep-Sea Res. 31:1085
1099.
Lehodey, P., Adre, J.M., Bertignac, M. et al. (1998) Predicting
skipjack tuna forage distributions in the equatorial Pacific
using a coupled dynamical bio-geochemical model. Fish.
Oceanogr. 7:317325.
Loukos, H., Monfray, P., Bopp, L. and Lehodey, P. (2003) Potential changes in skipjack tuna (Katsuwonus pelamis) habitat
from a global warming scenario: modeling approach and
preliminary results. Fish Oceanogr. 12:474482.
Matsumoto, W.M. (1975) Distribution, relative abundance
and movement of skipjack tuna, Katsuwonus pelamis, in the
Pacific Ocean based on Japanese tuna longline catches,
1964-67. NOAA Technical Report NMFS SSRF-695:
30pp.
Mishra, P., Tameishi, H. and Sugimoto, T. (2001) Delineation of
Meso-Scale Features in the Kuroshio-Oyashio Transition Region
and Fish Migration Routes Using Satellite Data Off JAPAN.
Singapore: National University of Singapore, 5pp.
Miyake, M.P., Miyabe, N. and Nakano, H. (2004) Historical
Trends of Tuna Catches in the World. FAO Fisheries Tech.
Paper 467: Rome: FAO, 74pp.
Nel, D.C., Lutjeharms, J.R.E., Pakhomov, E.A., Ansorge, I.J.,
Ryan, P.G. and Klages, N.T.W. (2001) Exploitation of
mesoscale oceanographic features by gray-headed albatross
Thalassarche chrysostoma in the southern Indian Ocean. Mar.
Ecol. Prog. Ser. 217:1526.
Nihira, A. (1996) Studies on the behavioral ecology and physiology of migratory fish schools of skipjack tuna (Katsuwonus
pelamis) in the oceanic frontal area. Bull. Tohoku Natl. Fish.
Res. Inst. 58:137233.
Ogura, M. (2003) Swimming behavior of skipjack, Katsuwonus
pelamis, observed by the data storage tag at the Northwestern
Pacific, off northern Japan, in summer of 2001 and 2002.
SCTB16 Working Paper.
Olson, D.B. (1991) Rings in the Ocean. Annu. Rev. Earth Planet. Sci. 19:283311.
Olson, D.B., Hitchcock, G.L., Mariano, A.J. et al. (1994)
Life on the edge: marine life and fronts. Oceanography 7:52
60.
Polovina, J.J. and Howell, E. (2005) Ecosystem indicators from
remotely sensed oceanographic data. ICES J. Mar. Sci.
62:319327.
Polovina, J.J., Howell, E., Kobayashi, D.R. and Seki, M.P.
(2001) The transition zone chlorophyll front, a dynamic
global feature defining migration and forage habitat for
marine resources. Prog. Oceanogr. 49:469483.
Qiu, B. (2002) The Kuroshio Extension system: its large-scale
variability and role in the midlatitude oceanatmosphere
interaction. J. Oceanogr. 58:5775.

 2010 Blackwell Publishing Ltd, Fish. Oceanogr., 19:5, 382396.

396

R. Mugo et al.

Ramos, A.G., Santiago, J., Sangra, P. and Canton, M. (1996)

An application of satellite derived sea surface temperature
data to the skipjack (Katsuwonus pelamis Linnaeus, 1758)
and Albacore tuna (Thunnus alalunga Bonaterre, 1788)
fisheries in the north-east Atlantic. Int. J. Remote Sens
17:749759.
Roberts, J.J., Best, B.D., Dunn, D.C., Treml, E.A. and Halpin,
P.N. (in press) Marine Geospatial Ecology Tools: an integrated framework for ecological geoprocessing with ARCGIS,
Python, R, MATLAB, and C++. Environm. Modell. Software, http://mgel.env.duke.edu/tools.
Robinson, I.S. (2004) Measuring Oceans From Space: The Principles and Methods of Satellite Oceanography. Chichester:
Praxis Publishing, 669pp.
Roger, C. (1994) The plankton of the tropical western Indian
ocean as a biomass indirectly supporting surface tunas (yellowfin, Thunnus albacares and skipjack, Katsuwonus pelamis).
Environm. Biol. Fishes 39:161172.
Saitoh, S., Kosaka, S. and Iisaka, J. (1986) Satellite infrared
observations of Kuroshio warm-core rings and their application to study of Pacific saury migration. Deep Sea Res.
33:16011615.
Sakurai, Y. (2007) An overview of the Oyashio Ecosystem. Deep
Sea Res. II 54:25262542.
Schaefer, K.M. and Fuller, D.W. (2007) Vertical movement
patterns of skipjack tuna (Katsuwonus pelamis) in the eastern
equatorial Pacific Ocean, as revealed with archival tags. Fish.
Bull. 105:379389.
Sugimoto, T. and Tameishi, H. (1992) Warm-core rings,
streamers, and their role on the fishing ground formation
around Japan. Deep-Sea Res. 39:S183S201.
Sund, P.N., Blackburn, M. and William, F. (1981) Tunas and
their environment in the Pacific Ocean: a review. Oceanogr.
Mar. Biol. Ann. Rev. 19:443512.
Talley, L.D., Nagata, Y., Fujimura, M. et al. (1995) North
Pacific intermediate water in the Kuroshio Oyashio mixed
water region. J. Phys.Oceanogr. 25:475501.
Tameishi, H. and Shinomiya, H. (1989) Formation of southbound skipjack fishing grounds and its discriminant
prediction off Tohoku sea area. Nippon Suisan Gakkai shi,
55:619625.
Teo, S.L.H., Boustany, A.M. and Block, B.A. (2007) Oceanographic preferences of Atlantic bluefin tuna, Thunnus thynnus, on their Gulf of Mexico breeding grounds. Mar. Biol.
152:11051119.

Uda, M. (1973) Pulsative fluctuation of oceanic fronts in association with the tuna fishing grounds and fisheries. J. Fac.
Mar. Sci. Technol. 7:245264.
Uda, M. and Ishino, M. (1958) Enrichment pattern resulting
from eddy systems in relation to fishing grounds. J. Tokyo
Univ. Fisheries 44:1 2, 105118.
Valavanis, D.V., Pierce, G.J., Zuur, A.F. et al. (2008) Modelling
of essential fish habitat based on remote sensing, spatial
analysis and GIS. Hydrobiologia 612:520.
Watanabe, Y., Ogura, M. and Tanabe, T. (1995) Migration of
skipjack tuna, Katsuwonus pelamis, in the western Pacific
Ocean, as estimated from tagging data. Bull. Tohoku Natl.
Fish. Res. Inst. 57:3160.
Wessel, P. and Smith, W.H.F. (1998) New, improved version of
Generic Mapping Tools released. EOS Trans. Am. Geophys.
Union 79:579.
Wild, A. and Hampton, J. (1993) A review of the biology and
fisheries for skipjack tuna, Katsuwonus pelamis, in the Pacific
Ocean. In: Interactions of Pacific tuna fisheries Proceedings of
the first FAO Expert Consultation on Interactions of Pacific
Tuna Fisheries, Nourmea, New Caledonia. R.S. Shomura, J.
Majkowski & S. Langi (eds) FAO Fisheries Tech. Paper, 336
Rome: FAO, pp. 151.
Wilson, C., Morales, J., Nayak, S., Asanuma, I. and Feldman, G.
(2008) Ocean-color radiometry and fisheries. In: Why Ocean
Colour? The Societal Benefits of Ocean-Color Technology. T.
Platt, N. Hoepffner, V. Stuart & C. Brown (eds), Reports of
the International Ocean-Color Coordinating Group, No. 7,
Dartmouth, Canada: IOCCG, pp. 4758.
Wood, S.M. (2006) Generalized Additive Models, An Introduction
with R. London: Chapman and Hall, 392pp.
Yasuda, I. (2003) Hydrographic structure and variability in
the Kuroshio-Oyashio transition area. J. Oceanogr. 59:389
402.
Yasuda, I., Okuda, K. and Hirai, M. (1992) Evolution of a Kuroshio Warm-Core Ring variability of the hydrographic
structure. Deep-Sea Res. 39(Suppl.):S131S161.
Zagaglia, C.R., Lorenzzetti, J.A. and Stech, J.L. (2004) Remote
sensing data and longline catches of yellowfin tuna (Thunnus
albacares) in the equatorial Atlantic. Remote Sensing Environm. 93:267281.
Zainuddin, M., Saitoh, K. and Saitoh, S. (2008) Albacore
(Thunnus alalunga) fishing ground in relation to oceanographic conditions in the western North Pacific Ocean using
remotely sensed satellite data. Fish. Oceanogr. 17:6163.

 2010 Blackwell Publishing Ltd, Fish. Oceanogr., 19:5, 382396.