See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/223738444

Remote sensing data and longline catches of yellowfin tuna (Thunnus

albacares) in the equatorial Atlantic

Article  in  Remote Sensing of Environment · October 2004

DOI: 10.1016/j.rse.2004.07.015

CITATIONS READS

104 405

3 authors, including:

Joao A. Lorenzzetti José Stech

National Institute for Space Research, Brazil National Institute for Space Research, Brazil
200 PUBLICATIONS   1,816 CITATIONS    164 PUBLICATIONS   1,598 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Physical Limnology View project

FITOSAT – Monitoramento de Ocorrências Ambientais na Bacia de Campos com Sensores Remotos e Dados de Campo View project

All content following this page was uploaded by José Stech on 10 June 2019.

The user has requested enhancement of the downloaded file.

 Remote Sensing of Environment 93 (2004) 267 – 281
www.elsevier.com/locate/rse

Remote sensing data and longline catches of yellowfin tuna

(Thunnus albacares) in the equatorial Atlantic
Cláudia R. Zagaglia*, João A. Lorenzzetti*, José L. Stech
Remote Sensing Division, National Institute for Space Research, Av. Astronautas, 1758, Jardim da Granja, CEP: 12227-010, São José dos Campos-SP, Brazil

Received 18 June 2003; received in revised form 19 July 2004; accepted 24 July 2004

Abstract

This study investigates the relationship between yellowfin tuna (Thunnus albacares) caught in the tropical Atlantic between 158S and
158N and 108W and 558W by the northeast Brazilian longline fleet, and environmental variables obtained from remote sensors such as sea
surface temperature (AVHRR/NOAA), chlorophyll-a concentration (SeaWiFS/SeaStar), sea surface height anomaly (TOPEX/Poseidon), and
wind velocity (Scatterometer/ERS-1 and -2). The observed relationships between longline catch data, expressed as catch per unit effort
(CPUE), and the satellite data were analysed using the statistical method of generalized additive models (GAMs). CPUE was calculated as
the number of fish caught by 100 hooks in 18 latitude by 18 longitude square grid and integrated into a month of fishing activity for the period
of 1995–2000. Results obtained showed evidences of non-linear relationships between catch yields and environmental data. The largest
CPUE of yellowfin tuna in the region seems to be strongly associated with the Intertropical Convergence Zone (ITCZ) position and its
temporal variability. In the model, this was confirmed by the highest correlation of CPUE with the meridional wind component. This result
also indicates 58S as the southern limit for a larger concentration of this species during the most productive fishing period between March and
May. In decreasing order of importance, sea surface temperature, chlorophyll-a concentration and sea surface height anomaly, albeit
statistically significant, seemed to be of secondary importance in controlling this species abundance in the region.
D 2004 Elsevier Inc. All rights reserved.

Keywords: Generalized additive models; Remote sensing; Yellowfin tuna; Thunnus albacares; Tropical Atlantic; CPUE; Longline

1. Introduction track the time and spatial evolution of oceanographic and

Pelagic species of commercial value are generally
distributed over very large oceanic areas of fishing potential.
This means the search for these resources by industrial
fishery fleet is a high cost activity. Consequently, the
availability of biological and oceanographic information
that would help fishing operations and, simultaneously,
would lead to a better management of these resources, could
make this activity more efficient, profitable and sustainable.
In particular, we refer to the possibility of access synoptic
information at meso- as well as large scales, essential to
* Corresponding authors. Tel.: +55 12 3945 6485; fax: +55 12 3945
6488.
E-mail addresses: zagaglia@agricultura.gov.br (C.R. Zagaglia)8
loren@dsr.inpe.br (J.A. Lorenzzetti).
0034-4257/$ - see front matter D 2004 Elsevier Inc. All rights reserved.
doi:10.1016/j.rse.2004.07.015
meteorological parameters that could affect the population
biology and the distribution of caught species of high
economic value.
Remote sensing techniques show great potential to
support global fisheries management and the exploitation
of pelagic species. Both short- and long-term environmental
variations are often reflected in satellite-derived patterns of
oceanographic variables such as ocean temperature and
primary productivity. Joint analysis of satellite and bio-
logical/catch data time series can be used to identify habitat
changes and their impact on migration, size or recruitment
of a particular fish stock, helping to regulate maximum
catches and to preserve fishing resources. Satellite-derived
information has been used by several countries for a
relatively long time for this purpose (Santos, 2000;
Yamanaka et al., 1988). With such support, fisheries
268 C.R. Zagaglia et al. / Remote Sensing of Environment 93 (2004) 267–281
management decisions can be made to protect target species
from possible overfishing.
The tuna fishery in the equatorial Atlantic and partic-
ularly along the Brazilian northeast coast is a growing
business activity, both at the industrial and artisanal levels.
Due to the proximity of seaports to the areas of concen-
tration of tunas in the equatorial Atlantic, this is the region
with best potential for the development of longline oceanic
fishing in Brazil (Travassos, 1999). We take this region as
our study area, which is located between the parallels 158N
and 158S and meridians 108W and 558W. In addition to the
Brazilian North and Northeast Exclusive Economic Zone, it
includes part of the international domain waters of the
equatorial Atlantic (Fig. 1).
1.1. Biology and fishery of Thunnus albacares
According to Arocha et al. (2001) most studies about the
reproduction of T. albacares in the Atlantic indicate its
eastern portion, represented basically by the Gulf of Guinea
and adjacent waters, as the main spawning and concen-
tration areas of this species. These authors also suggest two
other reproduction areas located in the Gulf of Mexico and
southeastern Caribbean, where spawning occurs between
May and August and July and November, respectively.
Analysing 12 years of commercial fishing data in the
Bermudas region (328N, 658W), Luckhurst et al. (2001)
found significantly higher values of fish landings of this
species in the second and third quarters of the year.
Information obtained from a tagging program analysed by
these authors additionally suggested the presence of a
resident population in association with a flux of individuals
associated with a seasonal migratory movement.
The likelihood of a unique stock of T. albacares in the
Atlantic (Fonteneau, 1994; Fonteneau & Soubrier, 1996) is
supported by three facts derived from recapture analysis of
Fig. 1. Study area showing the Brazilian N and NE Exclusive Economic Zone. SPSP—Saint Peter and Saint Paul Archipelago; FN—Fernando de Noronha
Archipelago; AR—Rocas Atoll.
tagged individuals and from commercial fishing: (a) main
spawning area localized in eastern Atlantic, (b) east to west
transatlantic migrations and (c) a high proportion of pre-
adults in the western Atlantic leading a west to east
migration.
The presently accepted stock structure of T. albacares in
the Atlantic (Bard & Hervé, 1994; Fonteneau & Soubrier,
1996) suggests that juvenile individuals move along African
coastal waters beginning at the main spawning area in the
Gulf of Guinea. When reaching the pre-adult stage (between
60 and 80 cm and ages between 1.5 and 2 years), the large
part of the individuals swim towards the American coast.
The return to the eastern Atlantic occurs when the
individuals are about 110 cm long, corresponding to,
approximately, 3 years of age.
Tuna movements can be either advective or diffusive.
The advective type refers to massive and oriented motions
having some biological objective, such as spawning or
feeding. The diffusive type occurs inside of the favourable
zone for the distribution and dispersion of the species, being
generally similar to the seasonal displacements of the
environmental parameters of that zone (Fonteneau &
Soubrier, 1996).
In the equatorial Atlantic, the main spawning and feeding
areas of the T. albacares are geographically separated, with
the feeding grounds in the western Atlantic and spawning
grounds in the east. This separation makes both migratory
displacements from east to west (for feeding), and from west
to east (for spawing), advective. So, the largest part of the
fishing effort of the Brazilian longline fleet is mostly located
in areas associated with the migratory route of the T.
albacares and not where, in general, the stock is temporarily
present, such as the spawning or feeding areas.
Hazin (1993) proposed a migratory cycle of the T.
albacares in the tropical Atlantic based upon a data set of
capture and length distribution of individuals. According to
 C.R. Zagaglia et al. / Remote Sensing of Environment 93 (2004) 267–281
this model, the displacement of tuna schools would happen
in four stages: (a) concentration in the eastern Atlantic (Gulf
of Guinea and adjacent waters) between January and March
for spawning; (b) transatlantic migration from east to west
between March and May; (c) concentration in the west (Gulf
of Mexico and southeast Caribbean) between June and
September for spawning and feeding; and (d) transatlantic
migration from west to east between October and December.
The migratory movement of this species seems also to
result in a high abundance of individuals during the first
quarter of the year in the proximities of the Saint Peter and
Saint Paul archipelago (see Fig. 1). This pattern appears to
be related to the reproductive behaviour of the flying fish
(Cypselurus cyanopterus) in this region and period, the
main source of food for T. albacares schools during their
migration (Hazin, 1993). Also, according to Holland et al.
(1999), submarine banks and islands, besides representing
important concentration areas of prey individuals, can act as
orientation points during the large scale movements of the
tuna schools. It seems that these animals are able to sense
the interferences caused by those topographic features in the
Earth’s magnetic field.
The predominant tuna fishing method used in the region
is the surface longline. This is a fishing gear developed in
Japan especially for tuna capture. It is comprised of a main
line that is supported by surface buoys distributed at regular
intervals. Several secondary lines with hooks are attached to
the main line (Fig. 2). For T. albacares, four to seven hooks
are used between each pair of floats, reaching depths
between 50 and 150 m. This is the longline type
recommended for this species, considering their preferred
surface distribution (Travassos, 1999).
1.2. Oceanographic and atmospheric setup
According to Hazin (1993), the strong vertical temper-
ature gradient present in the region is one of the main
causes inhibiting vertical motion and hindering the
replenishment of nutrient salts in the euphotic layer. In
addition, the western tropical Atlantic is characterized by a
deep thermocline, which limits the access of deeper waters
to the euphotic zone. However, in terms of chlorophyll-a
Fig. 2. Schematic of a surface longline.
269
concentration, some regions in the tropical Atlantic, which
is predominantly oligotrophic, show high biological
primary productivity. This happens mostly due to the
following regional phenomena: African coastal upwellings,
equatorial divergence and associated upwelling, Amazon
river outflow and nutrient discharge and the North Brazil
Current retroflection, which can produce anticyclonic
eddies and upwellings around 68N and 88N (Didden &
Schott, 1993).
In addition to these factors, upwelling of nutrients is
observed in the region caused by the interaction of local
currents with the submarine relief (Travassos et al., 1999).
The bottom topography of the region is characterized
basically by abyssal plains, but near the Brazilian coast
there are also islands and rocks, such as Rocas Atoll,
Fernando de Noronha and Saint Peter and Saint Paul
Archipelagos (see Fig. 1). As part of the North Brazil and
Fernando de Noronha Chain, several seamounts at depths
ranging from 20 to 250 m are also present in the region.
These vertical transports result in local increases of primary
productivity. Probably as the result of such phenomenon, the
areas subject to these topographic features represent one of
the main fishing regions for commercially exploited pelagic
species in the northeast coast of Brazil (Hazin, 1993).
One of the most important environmental forcing
functions in the region is the Intertropical Convergence
Zone (ITCZ), the boundary between the southeast and
northeast trade winds. Atmospheric surface low pressures,
low winds, strong cloud cover and heavy precipitation rates
are normally associated with the ITCZ. To a great extent
associated with the seasonal meridional migration of the
ITCZ, a strong seasonal cycle is observed in the atmosphere
and ocean over the tropical Atlantic. The most northern
displacement of the ITCZ normally occurs between August
and September when it reaches 88N and 148N at the western
and eastern portions of the Atlantic, respectively. Its most
southern position is in April, when the north and northeast
Brazil regions are subject to the enhanced precipitation
produced by this low pressure belt (Moura & Shukla, 1981;
Uvo & Nobre, 1989). The position, strength and the timing
of the ITCZ have a profound influence over a series of
meteorological variables such as, winds, rainfall, shortwave
solar radiation, and air temperature. Changes in these fields
are related in a feedback manner to changes in oceanic
variables such as, depth of mixed layer, sea surface
temperature, currents, chlorophyll concentration, etc. Obvi-
ously, all these changes impact on the pelagic living
resources of the region.
The intent of this paper is to examine the dependence of
T. albacares catches in the Equatorial Atlantic, as reported
by the Brazilian northeast-based longline fishing fleet, on
some environmental variables obtained by satellite sensors.
This analysis is carried out using a generalized additive
model (GAM). Data from the following satellite sensors
were used: AVHRR/NOAA, SeaWiFS/SeaStar, TOPEX/
Poseidon and ERS-1 and -2 scatterometer.
270 C.R. Zagaglia et al. / Remote Sensing of Environment 93 (2004) 267–281
2. Materials and methods
2.1. Longline data
The fishing data used in this study were collected and
recorded in logbooks by the longline fleet operating in the
region. The logbooks were then made available to the
Fishing and Aquaculture Department of the Brazilian
government. Data resolution is 18 latitude18 longitude
per fishing day. Information consisted of position of
beginning of longline set, date, number of hooks used in
each longline set and the number of individuals captured in
each fishing day. The data set encompasses the period from
January 1995 to December 2000.
From this data set, the Catch Per Unit of Effort (CPUE)
was derived and used as the relative abundance index. The
CPUE was calculated as the number of individuals captured
by 100 hooks in a 18 latitude18 longitude grid, integrated
in to 1 month of fishing activity. Fig. 3 shows the spatial
distribution of the CPUE for the period 1995–2000. As
noted by Bigelow et al. (1999), this index, which represents
the dsuccessT of fishing (Ricker, 1975), is primarily used to
Fig. 3. (a) April and (b) September CPUE distributions for the period 1995–
2000. Units are number of individuals captured by 100 hooks in a 18
latitude18 longitude cell.
investigate trends in resources abundance. However, in
addition of being related to the abundance, it is also affected
by factors such as the degree of dispersion of resources and
fishing strategy (He et al., 1997; Lange, 1991). Therefore,
this index can be considered a measure of the relative
apparent abundance (Marr, 1951), or a relative quantity of
the number of fish of a population affected by availability
that is captured per unit of effort.
According to Maury et al. (2001), a non-linear relation-
ship is observed between the commercial CPUE and the
abundance of marine species, and this seems to be related
mostly to the spatial non-homogeneity of the aquatic
resources and to the fishing tactics used by fishermen.
Therefore, if the CPUE is to be used as an index of local
relative abundance, it is necessary that key factors related to
the ability of capturing tuna should be included. Due to the
highly migratory and seasonal behaviour of the tuna
fisheries, a space–time factor was systematically included
in the final model. This space–time factor was considered by
fitting a three-dimensional smoother to the product of the
three variables: latitude, longitude and month. This factor
was taken as a variable in the analysis.
2.2. Remote sensing data
2.2.1. Sea surface temperature (SST)
SST data were generated from the information collected
by the Advanced Very High Resolution Radiometer
(AVHRR) sensor on board the National Oceanic and
Atmospheric Administration (NOAA) satellites. This data
set is produced and distributed by the Physical Ocean-
ography Distributed Active Archive Center (PODAAC) of
the Jet Propulsion Laboratory (JPL)/National Aeronautics
and Space Administration (NASA) in the Hierarchical Data
Format (HDF). The monthly average SST data were
provided in a regular grid of 99 km including the period
from January 1985 to December 2000.
From the 15-year monthly mean SST time series, a
bclimatologicalQ mean SST value was calculated for each
month of the year. These values were then used to calculate
the SST monthly average anomalies (SSTA).
2.2.2. Chlorophyll-a concentration (Chl-a)
The oceanic total primary production resulting from the
photosynthetic process can be defined as the amount of
organic matter produced in a given period of time. It is
proportional to the Chl-a values in the surface layer of the
ocean, defined as a layer from the surface down to one
attenuation depth. The attenuation depth is the level where
downwelling irradiance is reduced to 1/e=0.36788 of its
value just below the water surface (Stewart, 1985).
According to Gordon and McCluney (1975), the depth at
the base of the surface layer is K d 1, where K d (in inverse
meters) is the diffuse attenuation coefficient and 90% of the
remotely sensed ocean colour radiance is backscattered from
this layer. It is possible to use the Chl-a as a proxy of the
 C.R. Zagaglia et al. / Remote Sensing of Environment 93 (2004) 267–281
oceanic productivity for type I waters, defined as water
bodies whose optical properties are basically determined by
phytoplankton (Stewart, 1985).
The Chl-a data generated from the SeaWiFS sensor are
produced and distributed by Distributed Active Archive
Center (DAAC) of the Goddard Space Flight Center
(GSFC)/NASA in HDF format. Only data from September
1997 to December 2000 were available due to the launching
of SeaStar satellite on August 1997. Among the several
SeaWiFS data products, available in different processing
levels, the level 3 data set was used, which corresponds to a
regular grid with spatial resolution of 99 km and monthly
mean Chl-a.
2.2.3. Sea surface height anomaly (SSHA)
SSHA data generated from information collected by the
TOPEX and Poseidon altimeters are also produced and
distributed by the PODAAC/JPL in a binary format. For each
10-day cycle of the satellite, starting in 1992, data files are
available with different products. These files contain among
others, the following parameters: time of data collection,
latitude, longitude, and sea surface anomaly with respect to
the geoid. Starting from a 10-day and 7-km resolution along
track anomaly data set, an interpolation was computed for a
18 latitude18 longitude grid. Monthly means were then
derived for the period January 1995 to December 2000.
2.2.4. Surface winds
Zonal and meridional surface wind components were
generated from the information collected by the scatterom-
eters onboard the European Remote Sensing Satellites ERS-
1 and -2, which are processed and distributed by the Centre
ERS d’Archivage et de Traitement (CERSAT) of the Institut
Français de Recherche pour L’exploitation de la Mer
(IFREMER). Global fields of surface wind vector (at a
nominal height of 10 m) over the ocean have been collected
since July 1991, with the launching of ERS-1 satellite. After
March 1996, these fields have been derived from observa-
tions made by the ERS-2, launched in 1995. Starting from
the 3-day and 25-km time–space resolution of this data set,
an interpolation was carried out for a 18 latitude18
longitude grid and monthly means were computed for the
period January 1995 to December 2000.
2.3. Generalized additive models (GAMs)
The methodology of generalized additive models was
used in this investigation to assess the relative influence of
the environmental factors described above (SST, Chl-a,
SSHA and surface winds) on the performance of the fishing
fleet, defined as the local commercial CPUE of T. albacares.
Despite the advantages of the linear regression techni-
ques in determining model parameters and their interpreta-
tion, the linear assumption is considered of little flexibility
in the sense that it restricts its range of application (Chong &
Wang, 1997). Generalized linear models have also been
271
used to study T. albacares CPUE variability in eastern
Pacific Ocean (Punsly, 1987). Using this technique, CPUE
data is normalised to incorporate non-environmental factors
which affect fishing efficiency, such as speed of fishing
ship, season, fishing area, and association with another
species or floating objects.
The interest in using GAMs is normally justified when
the effects of multiple independent variables need to be
modelled non-parametrically. This means that there is no
need for a previous assumption of a linear relationship or a
pre-determined probability distribution. The only assump-
tion when using GAMs is that the dependent variable is
influenced in some way by each independent variable
considered in the model. The functional form of this effect
will be indicated by the data itself (Hastie & Tibshirani,
1990; MathSoft, 1999).
The relationship between the dependent and each inde-
pendent variable is estimated separately via a univariate
smoothing operator, or one that shows low dimension when it
is multivariate. Therefore, these models allow the analysis of
several non-parametric relationships simultaneously, when
taking the additive nature of the effects of the independent
variables. This has the effect of facilitating the identification
of the form to which the dependent variable is being modified
by each one of the explaining variables. Besides the
advantage of allowing this type of analysis, the additive
generalized models proposed by Hastie and Tibshirani (1990)
also extend the use of the additive models to data sets that
present non-Gaussian distributions, such as the binomial,
Poisson and gamma (the exponential family of distributions).
Among the several possible smoothing operators used in
the estimation of the unknown (non-parametric) functions f j
(X j ) which describe the relationship between the trans-
formed expectancy of the dependent variable and the jth
independent variable, the bLoessQ, Localized Weighted
Regression was used in this investigation. This procedure
was used since it can extract an underlying low frequency
data pattern in extremely noisy data. Also, it is robust in the
presence of outliers, i.e., it is not much influenced by
outliers (SAS Institute, 2001).
A partial residual can be calculated that shows the relative
effect of the X j variable over the interest dependent variable,
considering that the remaining variables have already been
considered in the model. According to Neter et al. (1989), the
plot of the partial residuals tends to show the nature of the
relationship between the independent variables and the
Table 1
Analysis of variance for significant parameters included in the GAM for
yellowfin tuna
Parameter F* statistic p-value
Latitudelongitudemonth 7.94 i0
Meridional wind velocity 8.96 0.00005
Sea surface temperature 3.14 0.01301
Chlorophyll-a concentration 3.75 0.01936
Sea surface height anomaly 2.31 0.04757
272 C.R. Zagaglia et al. / Remote Sensing of Environment 93 (2004) 267–281
Fig. 4. Effect of variables (a) latitude, (b) longitude, and (c) month over
transformed local relative abundance index dependent variable [ln(CPUE+1)]
for T. albacares as depicted by the standardized partial residuals. Tick marks
at abscissa axis represent the observed data points. Full line is the explaining
function. Dashed-dotted lines indicate the 95% confidence interval,
equivalent to approximately to two standard deviations (F2 S.D.).
dependent variable. In a graphical form, the partial residuals
are plotted in the ordinate axis and the abscissa axis contains
the observed values of the X j independent variable.
In this paper, the interpretation of the observed relation-
ships between the studied variables is done based on partial
residuals, which were determined for each significant
variable. The relative influence of each effect was judged
on the basis of the values normalized with respect to the
standard deviation of the partial residuals. The partial
residual plots also contain the 95% confidence intervals,
as well as tick marks in abscissa showing the location of
data points.
The set of variables included in the models was
determined in a similar way to the forward stepwise
procedure of Neter et al. (1989). In essence, this procedure
is based upon some of the possible models that could be
formed out of the independent variables, and it identifies the
best subset of variables. The criteria used to include/exclude
a variable in the model is based upon the value of the
probability associated to the observed F* statistics. This
statistical test determines the probability ( p-value) of the
contribution of an X p variable in explaining the variance has
happened by chance. Thus, if this probability is sufficiently
low, in other words, lower than the assumed significance
level (a=5%, p-valuethreshold=0.05), the X p variable is
considered significant for explaining the variance of the
dependent variable.
After the initial test of all models including only one
explanatory variable, the p-value associated with this term
was analysed for each model. The initial model chosen was
the one corresponding to the variable that presented the
smallest value of the stated criterion. Next, all remaining
variables and interactions were individually tested in this
initial model to assess their possible inclusion. The
significance level of 5% was used in all statistical analysis.
Besides the evaluation of p-value associated with each
explaining term of the model, the final adjustment of the
model was made by the analysis of the pseudo determi-
nation coefficient (pseudo-R 2), defined as the fraction of
total variance explained by the model, and by the residual
variance. In this way, this procedure continued until the
Fig. 5. Effect of meridional wind component over transformed local relative
abundance index dependent variable [ln(CPUE+1)] for T. albacares as
depicted by the standardized partial residuals. Tick marks at abscissa axis
represent the observed data points. Full line is the explaining function.
Dash-dotted lines indicate the 95% confidence interval, equivalent to
approximately to two standard deviations (F2 S.D.).
 C.R. Zagaglia et al. / Remote Sensing of Environment 93 (2004) 267–281 273

inclusion of new significant variables did not contribute to
either, an increasing of the pseudo coefficient or a reduction
of the associated variance.
Despite of the fact that the GAMs allow distributions
other than Gaussian for the dependent variable, we use a
logarithmic transformation of the CPUE to normalize its
asymmetrical frequency distribution. As described in Maury
et al. (2001), we also assumed a normal distribution for the
natural logarithm of [CPUE+1], from now on simply
referred to as brelative abundanceQ. After this procedure, it
is possible to show that the appropriate link function is
identity, that is, the dependent variable is not subject to
transformations (SAS Institute, 2001).
The predictive skill of the final model was evaluated
through linear regression between randomly chosen
observed values of the relative abundance and those
generated by the model using the included independent
variables as input. The 100 samples randomly chosen from
the total fishing data set for this purpose were not included
in the process of model generation. Simple analysis of
variance was done to determine the significance of the
regression.

Fig. 6. ERS-2 scatterometer monthly mean wind magnitude (m/s): (a) April/2000 and (b) September/2000. The black ellipses indicate ITCZ position.
274 C.R. Zagaglia et al. / Remote Sensing of Environment 93 (2004) 267–281
3. Results and discussion
The method of fitting the generalized additive model
indicated five significant parameters ( p-valueb0.05) (Table
1). A set of 2373 samples was used, representing the
observed CPUE in 18 latitude18 longitude grid and
integrated in to 1 month of fishing activity.
The space–time factor, defined as the interaction between
the geographic position (latitude and longitude) and the
month when the data were collected, explains the largest
portion of the data variance, being the most significant
factor amongst all the independent variables included in the
model ( p-valuei0).
Fig. 4, with the individual plots of partial residuals
associated with each variable composing the space–time
factor, shows that CPUE values tend to increase to the north
between 158S and 58S, stabilizing approximately between
58S and 58N. For the region between 58N and 158N, the
95% confidence interval is, however, relatively large, not
allowing a conclusive assessment of an increasing tendency.
The uncertainty in the determination of a trend in this last
latitudinal band seems to be related with the lower data
density there (see tick marks in the abscissa axis). In spite of
this, based on the analysed data, we can suggest that the T.
albacares shows a predominantly equatorial distribution,
with the largest observed values of CPUE between 58S and
58N.
Regarding the longitudinal distribution of data, it is
possible to see in Fig. 4 that most of the fishing effort of the
Brazilian-based longline tuna fishing fleet was concentrated
in the 258W and 388W zonal band (see tick marks on
abscissa). If we consider the highly migratory behaviour of
this species, the higher CPUE values found towards more
negative longitudes seems to indicate that this species is
closer to the NE Brazilian coast in a particular period of the
year. This result agrees with other studies (Arocha et al.,
2001; Brill et al., 1998), which suggest that this species
comes near the NE Brazilian coast as part of their migratory
behaviour.
The explaining function associated with the variable
month (Fig. 4), shows the period from February to May as
that with the highest abundance for the T. albacares. The
CPUE shows its highest values in April and the smallest
captures from August to October with a minimum in
September. These results seem to confirm the migratory
behaviour of the T. albacares described earlier, whereby the
largest CPUEs observed close to the Brazilian coast in April
can be a result of the passage of individuals in that area in
their transatlantic displacements from the eastern to the
western Atlantic (Hazin, 1993; Travassos, 1999). Con-
versely, the smallest CPUEs, which are observed in
September, could be caused by the concentration of
individuals in this time of the year in the Gulf of Mexico
and Caribbean for spawning.
The explaining function for the relation between the
meridional wind component, the most significant environ-
mental variable included in the model ( p-valuei0.00005),
and the relative abundance index for T. albacares increases
until a value of wind speed component of, approximately, 2
m s 1 (Fig. 5). From this point on, an inflection of the curve
occurs and the function decreases as the meridional wind
component becomes more positive. This relationship
indicates that best catches are mostly observed close, and
to the south of the ITCZ, where wind speeds are relatively
weak and the meridional wind component is positive
(Southeast Trade winds).
As expected from climatology, Fig. 6a shows that the
ITCZ (low wind region) was in April 2000 at its most
southerly position. This coincides with the best fishing
period indicated by the present work, between March and
May. In September 2000 (a period of low catches) Fig. 6b
shows that the ITCZ had been displaced to its most
northerly position. In addition, this period is characterized
by stronger winds, a condition unfavourable for the fishing
activities. As seen in Fig. 6, weak winds blowing from south
(characteristic of the Southeast Trade winds close and to the
south of the ITCZ) are expected north of 58S. Therefore, this
latitude could be considered the southern limit of the
favourable area for a larger concentration of T. albacares.
High wind speed is reported to affect fish vulnerability to
capture by modifying fish behaviour and availability to a
specific depth of the fishing gear (Bigelow et al., 1999). For
example, Carey and Robinson (1981) show that swordfish
swim to greater depths with strong winds, reducing their
vulnerability to capture. It is, therefore, possible that the
decrease of T. albacares in the equatorial western Atlantic in
higher winds could be affected by these factors, favouring
higher catches in the low wind zone of the ITCZ.
The other remaining environmental factors, SST, Chl-a
and SSHA, despite being statistically significant, showed
associated p-values well above those found for the space/
time factor and meridional wind (Table 1). In ascending
Fig. 7. Effect of sea surface temperature over transformed local relative
abundance index dependent variable [ln(CPUE+1)] for T. albacares as
depicted by the standardized partial residuals. Tick marks at abscissa axis
represent the observed data points. Full line is the explaining function.
Dash-dotted lines indicate the 95% confidence interval, equivalent to
approximately to two standard deviations (F2 S.D.).
 C.R. Zagaglia et al. / Remote Sensing of Environment 93 (2004) 267–281 275

order, their p-value begin to approximate the chosen
significance level p=0.05. So, we tend to consider that their
influence over the CPUE for the T. albacares in the tropical
western Atlantic, although not negligible, should not be
dominant.
These results show that SST influences but it does not
seem to be the main determinant of the area distribution and
the relative abundance of this species in the equatorial
Atlantic. In fact, Pereira (1995) states that SST is frequently
used in describing the tuna habitats mostly because it is a
relatively easy parameter to obtain. On the other hand, the
SST analysis may become particularly useful in the presence
of strong surface thermal gradients, where the fishing
grounds are normally located in the limit areas of
distribution of these species (Laurs & Lynn, 1977; Laurs
et al., 1984).
According to Stretta (1991), the local relative abundance
of the T. albacares increases near the equator, indicating a
preference of this species to warmer waters, with temper-
ature limits between 18 and 31 8C. Additionally, it is
reported that in the tropical Atlantic most of the catches of
this species occurs in waters with temperatures in a narrower
temperature range, between 22 and 29 8C, and preferentially
above 25 8C. The data distribution of T. albacares catches in
our database agrees with this (see tick marks in abscissa in
Fig. 7), but no distinct temperature preference is suggested

Fig. 8. AVHRR/NOAA sea surface temperature (8C) monthly mean fields for (a) April/2000 and (b) September/2000.
276 C.R. Zagaglia et al. / Remote Sensing of Environment 93 (2004) 267–281
Fig. 9. Effect of chlorophyll-a concentration over transformed local relative
abundance index dependent variable [ln(CPUE+1)] for T. albacares as
depicted by the standardized partial residuals. Tick marks at abscissa axis
represent the observed data points. Full line is the explaining function.
Dash-dotted lines indicate the 95% confidence interval, equivalent to
approximately to two standard deviations (F2 S.D.).
by the almost flat relationship between SST and CPUE in
the temperature range of 26–28.5 8C. Above 28.5 8C, there
is some indication of preference to higher temperatures, but
this should be taken with care since the number of data
points in this high temperature range decreases and the
confidence interval starts to broaden.
Some factors can be suggested to explain the lower
correlation between transformed CPUE for T. albacares
and SST. The western equatorial Atlantic, where Brazilian-
based longline fishing fleet (see Fig. 3) concentrates, is
generally characterized by low SST variability and the
absence of strong oceanic fronts. Stronger SST horizontal
gradients can be found in the tropical Atlantic more to the
east along the equator due to equatorial divergence and
Fig. 10. SeaWiFS monthly mean chlorophyll-a concentration (mg m 3) for April/2000. White pixels indicate cloud-covered areas.
along the African coast, south of equator and part of the
coast of the Gulf of Guinea north of equator (Servain et al.,
1987). The high water vapour content and increased
atmospheric absorption in the equatorial atmosphere act
to reduce the accuracy of the SST obtained from orbital
sensors, tending to produce a lower signal/noise ratio in
this region (Emery et al., 1995; Minnett, 1995;). Power and
May (1991) also report no apparent relationship between T.
albacares abundance and SST in the Gulf of Mexico.
Uncertainty in longline position and T. albacares migratory
behaviour were indicated as possible causes for the lack of
correlation between SST and CPUE. It should also be noted
that the Thunnus genus possess a blood circulatory system
that regulates their body temperature very efficiently and
allows the tunas to make large horizontal and vertical
excursions as a way of enlarging their predation areas. This
high mobility, and particularly the vertical excursions, is
likely to reduce the correlation of tuna capture with the SST
field.
According to Maury et al. (2001), the annual reproduc-
tive transatlantic displacements of the adult population are
probably driven by temperature to warm locations that are
favorable for juveniles. However, on smaller scales, the
relationship of catches with temperature at 150 m depth is
statistically more significant. Block et al. (1997) and Brill et
al. (1999) found that more important than the SST was the
depth of the mixed layer. Adult T. albacares were found
most of the time inside the mixed layer or immediately
below it. It was also observed that maximum swimming
depths of adult T. albacares were limited to water temper-
atures 8 8C cooler than the surface layer.
Fig. 8 shows the SST distributions for April and
September 2000, periods of the year of highest and lowest
CPUE indices of T. albacares, respectively. It is observed
 C.R. Zagaglia et al. / Remote Sensing of Environment 93 (2004) 267–281 277

that waters with temperatures between 29 and 30 8C showed
a seasonal and meridional displacement that encompasses
practically all the northeast Brazilian coast in April reaching
higher northern latitudes in September.
According to these observations, the regions for both
hemispheres with SST values above 28 8C seem to form a
pathway of favourable thermal conditions to the migratory
movements of T. albacares towards the Venezuelan north-
ern coastal region. The meridional displacement of the
surface thermal field could bsteerQ the tuna schools to this
area, which in turn would later reach the Brazilian coast
between the austral summer and autumn seasons.
Hazin (1993) links the southern most position of the 27
8C isotherm with the closest position of the T. albacares
to the Brazilian coast. In addition, it was observed that the
largest captures of this species in the equatorial Atlantic
are distributed in warm waters, normally with temperatures
above 27 8C, and that the migratory movements of T.
albacares seem to be coupled with the seasonal displace-
ment of sea surface isotherms and, possibly, with the
oceanic currents. As the position of the ITCZ is tied to the
meridional migration of the maximum SST in the western
tropical Atlantic (Hastenrath & Lamb, 1977), T. albacares
distribution might be indirectly related to the space–time

Fig. 11. TOPEX/Poseidon monthly mean sea surface height anomaly (mm) fields for (a) April/2000 and (b) September/2000.
278 C.R. Zagaglia et al. / Remote Sensing of Environment 93 (2004) 267–281
Fig. 12. Effect of sea surface height anomaly over transformed local relative
abundance index dependent variable [ln(CPUE+1)] for T. albacares as
depicted by the standardized partial residuals. Tick marks at abscissa axis
represent the observed data points. Full line is the explaining function.
Dash-dotted lines indicate the 95% confidence interval, equivalent to
approximately to two standard deviations (F2 S.D.).
changes of this environmental variable (SST) in the
region.
In addition to being less significant than SST (see Table
1), primary productivity, as depicted by Chl-a concentra-
tion, was inversely related to changes in CPUE (Fig. 9). The
inverse correlation could be due to the fact that tuna are
visual predators and need clear waters (Laurs et al., 1984). A
deviation from this pattern is, however, indicated by the
presence of a small peak near 0.12 mg m 3. Few regions of
higher primary productivity characterize April, considered
the central month of the period of closest proximity of this
species to the Brazilian coast. Fig. 10 shows that Chl-a
values near 0.12 mg m 3 occur, predominantly, north of
58S. This agrees with the observed southern limit of the
largest abundances of T. albacares in the region (Hazin,
1993; Travassos, 1999).
Although some Chl-a frontal regions can be places of
high concentration of tuna (Brill & Lutcavage, 2001; Fiedler
& Bernard, 1987), such well-developed colour fronts are not
common in our study region. As indicated before, such
features can be found in the western Atlantic mostly in the
Amazon River mouth plume and at the North Brazil Current
retroflection zone (Didden & Schott, 1993; Müller-Karger et
al., 1988).
A high concentration of flying fish (C. cyanopterus),
considered the main food source to the T. albacares
individuals in their migratory route, is observed near the
Saint Peter and Saint Paul Archipelago (see Fig. 1) during
January, February and March (Hazin, 1993). So, rather than
being driven primarily by water temperature and chlorophyll
concentration, tuna abundance in the region could be more
dependent on prey concentration, indicating a typically
opportunistic feeding strategy. As indicated before, local
increases of productivity produced by interaction of currents
with topography (islands and submarine banks) could
support flying fish population in the area. However, these
are local phenomenon which could not be resolved by our
coarse resolution 18 latitude18 longitude grid. A study of
the relation between flying fish abundance and SST and
Chl-a concentration, or their gradients is needed to further
examine this hypothesis.
The sea surface height anomaly (SSHA) field is coupled
to the dynamics (currents) and thermodynamics (heat
balance) of the upper ocean. Convergences and divergences
of the mass transport in the surface layer of the ocean result
in positive and negative sea level anomalies, respectively.
Variations in water density, which are dominantly controlled
by changes in temperature or in the heat storage (changes in
mixed layer depth or its temperature), also give rise to sea
level anomalies (Polito et al., 2000). It is natural, therefore,
to expect that changes in SSHA can be related to variations
of the CPUE.
Fig. 11 shows the spatial distribution of SSHA for the
months of April and September of 2000. A seesaw type of
SSHA pattern is present during the year, with negative
anomalies dominating during April and positive anomalies
in September. The inclinations of the SSHA anomalies and
their positions are suggestively related with the ITCZ
position, which, as discussed before, is also linked to the
SST field (see Figs. 6 and 8). Despite the contrasting
configuration of SSHA between the high and low tuna catch
periods and its association to the ITCZ, the relation between
this variable to the CPUE is surprisingly weak as indicated
by its high associated p-value=0.04757 (Table 1) and the
almost flat explaining function observed between the relative
abundance and the SSHA (Fig. 12). SSHA is in fact the least
significant of the included environmental variables. This low
correlation could be linked to the fact that, as indicated
before, SSHA is the result of a complex combination of
dynamical and thermodynamical factors, which could affect
in opposite ways the concentration of the fishing resources.
Fig. 13 shows the scatter plot of the randomly selected
values of relative abundance from capture information of the
longline fishing fleet and the forecasted values generated by
the GAM using observed explaining variables as input. Out
of the 100 randomly chosen samples, 4 had to be excluded
from the analysis due to the absence of valid values in one
Fig. 13. Scatter plot between the observed and GAM model predicted values
of ln(CPUE+1) for T. albacares. Solid line is the best linear fit to the data.
 C.R. Zagaglia et al. / Remote Sensing of Environment 93 (2004) 267–281
Table 2
Analysis of variance for the relation between the observed ln(CPUE+1) values and those predicted by GAM model for the yellowfin tuna
Degrees of freedom Sum of squares
Regression 1 5.1945
Residual 94 5.0459
or more environmental variables. This occurred to, e.g., the
persistence of clouds in some of our 18 latitude18
longitude sampled quadrants and the absence of valid SST
and Chl-a satellite observations.
Despite the spread observed in Fig. 13, Table 2 shows
that the simple linear regression adjusted line is statistically
significant ( p-value well below to the significance level of
5%). Therefore, although not as good as expected, we
consider that the CPUEs generated from the generalized
additive model can represent in a statistically sound way the
observed values as registered in logbooks of T. albacares
catches in the region and contain a valid predictive skill.
4. Conclusions
The GAM partial residual plots confirmed the non-linear
nature of the relationship between the statistically significant
analysed variables and the T. albacares catches. This
reinforces the necessity of using non-linear models in the
investigation of the response of the fishing resources to
changes in the environmental factors. The use of linear
regression models should show very low predictive skills
and should be, therefore, avoided.
As far as spatial distribution is concerned, the relative
abundance index of the T. albacares for the study region
shows a growing tendency towards the northern latitudes
and western longitudes. It is, however, difficult to separate
to what degree the higher catches in the west are dominated
by a larger concentration of individuals in that region from
the effect that the area of action of the Brazilian-based tuna
fleet tends to concentrate towards the west due to their
intrinsic range and endurance limits. The temporal behav-
iour of the stock is characterized by a clear period of higher
local abundance from March to May, while the smallest
captures are observed between August and September.
The largest abundance indices of T. albacares seem to be
predominantly associated with the presence and the varia-
bility of the Intertropical Convergence Zone. The meridional
wind component was the environmental variable with
highest correlation with CPUE. In the tropical Atlantic,
the ITCZ position is the main mechanism controlling the
seasonal variability of this variable. The largest T. albacares
catches were associated with weak winds to the south of
ITCZ position, that is, part of the SE trades. The fact that in
April the ITCZ shows its most southerly position (Fig. 6a)
indicates the 58S as the southern limit favourable to a larger
concentration of this species in the fishing period suggested
by the present work, between March and May.
279
Mean square F* statistic p-value
5.1945 96.768 i0
0.0536
The other environmental variables analysed in this paper
(SST, Chl-a and SSHA), although statistically significant,
seem not to be key main factors controlling the distribution
and the relative abundance of T. albacares in the study
region. It is true, however, that they are dynamically linked
to the position of the ITCZ, which is related to them via its
modulating effect over the surface winds and the surface
heat budget. The surface heat budget is strongly controlled
by the incident solar short wave radiation input which
depends heavily on cloud cover. A strong cloud band also
characterizes the ITCZ. The next important factor in the
surface heat budget is the latent heat flux, which depends
among other variables, on the wind magnitude, and it is
therefore also dependent on the ITCZ position. Considering
that the surface heat budget is one of the key factors
controlling the temperature of the surface layer of the ocean,
and that the SSHA is influenced by this temperature, it is
expected that these variables are connected to the ITCZ
position.
The observed low correlation of these environmental
variables to CPUE in the region could also be related to the
fact that this species already shows a pre-determined
migratory behaviour. T. albacares individuals are available
for capture in the studied region mostly during their
transatlantic east–west displacements. It should therefore
be expected that the migration patterns of tuna schools in the
region should take advantage of current systems. We
consider that the speed and direction of surface ocean
currents should be experimented as new environmental
forcing variables. It is difficult, however, to measure these
variables at the required spatial and temporal resolutions. It
is possible that the results of three-dimensional ocean
models for the tropical Atlantic could be used as a substitute
for the in situ measurements. The tropical Atlantic current
system shows a high space and time variability, which is
strongly tied to the surface wind field. It is plausible to
suppose that the surface wind intensity and direction should
be variables influencing these long distance migratory
displacements.
A cautionary consideration must be made. The use of
commercial CPUE as the relative abundance index is
subject to limitations considering the assumptions associ-
ated with three situations rarely found in fishing data time
series: (a) equal vulnerability of individuals to the fishing
gear; (b) random distribution of the fishing effort; and (c)
random distribution of individuals. As a consequence,
variations in CPUE are not solely related to the abundance,
being also dependent on the availability of the resources or
changes in their vulnerability to the used fishing gear,
280 C.R. Zagaglia et al. / Remote Sensing of Environment 93 (2004) 267–281
which could be caused by environmental changes and fish
response to these variations. For example, the vertical
positioning of the longline should ideally be conditioned
by the vertical thermal structure. The changes in the
vertical temperature profile, such as, for example, the rise
of the thermocline can result in a situation where the
fishing gear is out of reach of the target species which are
preferentially distributed above the thermocline, as it is the
case for T. albacares.
As it becomes available, additional fishing data for the
region should be added in the analysis. This inclusion is
expected to make the GAM model more robust with reduced
confidence intervals and with better predictive skill.
Acknowledgements
The authors would like to thank the National Institute for
Space Research (INPE) and its Graduate Program Service
and researchers Olga Sato, Paulo Polito and Paulo Eduardo
U. de Souza, the SeaWiFS Project and the Distributed
Active Archive Center (DAAC) at the Goddard Space Flight
Center (GSFC), the Fishing and Aquaculture Department of
the Brazilian government (DPA), Fábio Hazin, Paulo
Travassos and Humberto Hazin at Federal Rural University
of Pernambuco (UFRPE), the Centre ERS d’Archivage et de
Traitement (CERSAT) at the Institut Français de Recherche
pour L’exploitation de la Mer (IFREMER) and the Physical
Oceanography Distributed Active Archive Center
(PODAAC) at Jet Propulsion Laboratory (JPL). This work
was partly supported by the Coordination for Improvement
of University Level Personnel (CAPES) of Brazil.
References
Arocha, F., Lee, D. W., Marcano, L. A., & Marcano, J. S. (2001). Update
information on the spawning of yellowfin tuna, Thunnus albacares, in
the Western Central Atlantic. Collective Volume of Scientific Papers, 52,
167 – 176.
Bard, F. X., & Hervé, A. (1994). Structure de stock de l’albacore (Thunnus
albacares) Atlantique d’apres les marquages compares aux lieux de
ponte. Collective Volume of Scientific Papers, 42(2), 204 – 208.
Bigelow, K. A., Boggs, C. H., & He, X. (1999). Environmental effects on
swordfish and blue shark catch rates in the US North Pacific longline
fishery. Fishery and Oceanography, 8(3), 178 – 198.
Block, B. A., Keen, J. E., Castillo, B., Dewar, H., Freund, E. V., Marcinek,
D. J., et al. (1997). Environmental preferences of yellowfin tuna
(Thunnus albacares) at the northern extent of its range. Marine Biology,
130(1), 119 – 132.
Brill, R. W., Block, B. A., Boggs, C. H., Bigelow, K. A., Freund, E. V., &
Marcinek, D. J. (1999). Horizontal movements and depth distribution of
large adult yellowfin tuna (Thunnus albacares) near the Hawaiian
Islands, recorded using ultrasonic telemetry: Implications for the
physiological ecology of pelagic fishes. Marine Biology, 133(3),
395 – 408.
Brill, R. W., Lowe, T. E., & Cousins, K. L. (1998, July 26–20). How water
temperature really limits the vertical movements of tuna and bill-
fishes—it’s the heart stupid. Abstract from the American Fisheries
Society, International Congress on Biology of Fish. Baltimore7 Towson
University.
Brill, R. W., & Lutcavage, M. E. (2001). Understanding environmental
influences on movements and depth distributions of tunas and billfishes
can significantly improve population assessments. American Fisheries
Society Symposium, 25, 179 – 198.
Carey, F. G., & Robinson, B. H. (1981). Daily patterns in the activities of
swordfish, Xiphias glaudius, observed by acoustic telemetry. Fisheries
Bulletin of United States, 79, 277 – 292.
Chong, Y. S., & Wang, J. L. (1997). Statistical modelling via dimension
reduction methods. Nonlinear Analysis, Theory, Methods and Applica-
tions, 30(6), 3561 – 3568.
Didden, N., & Schott, F. (1993). Eddies in the North Brazil Current
retroflection region observed by Geosat altimetry. Journal of Geo-
physical Research, 98(C11), 20121 – 20131.
Emery, W. J., Wick, G. A., & Schluessel, P. (1995). Skin and bulk sea surface
temperatures: Satellite measurement and corrections. In M. Ikeda, &
F. W. Dobson (Eds.), Oceanographic Applications of Remote Sensing
(pp. 146–164). Boca Raton, FL7 CRC Press.
Fiedler, P. C., & Bernard, H. J. (1987). Tuna aggregation and feeding near
fronts observed in satellite imagery. Continental Shelf Research, 7(8),
871 – 881.
Fonteneau, A. (1994). Structure de la population d’albacore de l’Atlantique:
quelques considérations sur les migrations et la modélisation. Collective
Volume of Scientific Papers, 42(2), 215 – 218.
Fonteneau, A., & Soubrier, P. P. (1996). Interactions between tuna fisheries:
A global review with specific examples from the Atlantic Ocean. Status
of Interactions of Pacific Tunas Fisheries in 1995. Proceeding of the
Second FAO Expert Consultation on Interactions of Pacific Tuna
Fisheries, January 23–31, Shimizu, Japan, 612 pp.
Gordon, H. R., & MacCluney, W. R. (1975). Estimation of the depth of
sunlight penetration in the sea for remote sensing. Applied Optics, 14,
413 – 416.
Hastenrath, S., & Lamb, P. J. (1977). Climatic Atlas of the tropical Atlantic
and eastern Pacific oceans. Madison, WI7 University of Wisconsin Press
112 pp.
Hastie, T., & Tibshirani, R. (1990). Generalized additive models. London,
UK7 Chapman & Hall Press.
Hazin, F. H. V. (1993). Fisheries-oceanographical study on tunas, billfishes
and sharks in the Southwestern Equatorial Atlantic Ocean. Doctor’s
thesis, 286 pp., Tokyo University of Fisheries.
He, X., Bigelow, K. A., & Boggs, C. H. (1997). Cluster analysis of longline
sets and fishing strategies within the Hawaii-based fishery. Fisheries
Research, 31, 147 – 158.
Holland, K. N., Kleiber, P., & Kajiura, S. M. (1999). Different residence
times of yellowfin tuna, Thunnus albacares, and bigeye tuna, T. obesus,
found in mixed aggregations over a seamount. Fishery Bulletin, 97(2),
392 – 395.
Lange, A. M. (1991). Alternative survey indices for predicting availability
of longfin squid to seasonal northwest Atlantic fisheries. North
American Journal of Fisheries Management, 11, 443 – 450.
Laurs, R. M., Fiedler, P. C., & Montgomery, D. R. (1984). Albacore tuna
catch distributions relative to environmental features observed from
satellites. Deep-Sea Research, 31(9A), 1085 – 1099.
Laurs, R. M., & Lynn, R. J. (1977). Seasonal migration of North Pacific
albacore, Thunnus alalunga, into North American coastal waters:
Distribution, relative abundance, and association with Transition Zone
waters. Fisheries Bulletin, 75, 795 – 822.
Luckhurst, B. E., Trott, T., & Manuel, S. (2001). Landings, seasonality,
catch per unit effort and tag-recapture results of yellowfin tuna and
blackfin tuna at Bermuda. American Fisheries Society Symposium, 25,
225 – 234.
Marr, J. C. (1951). On the use of terms abundance, availability and apparent
abundance in fishery biology. Copeia, 2, 163 – 169.
MathSoft (1999). S-Plus guide to statistics. Seattle, USA7 MathSoft
Engineering & Education, Inc.
Maury, O., Gascuel, D., Marsac, F., Fonteneau, A., & Rosa, A. L. (2001).
 C.R. Zagaglia et al. / Remote Sensing of Environment 93 (2004) 267–281
Hierarchical interpretation of nonlinear relationships linking yellowfin
tuna (Thunnus albacares) distribution to the environment in the Atlantic
Ocean. Canadian Journal of Aquatic Sciences, 58, 458 – 469.
Minnett, P. J. (1995). Sea surface temperatures from the Along-Track
Scanning Radiometer. In M. Ikeda, & F. W. Dobson (Eds.), Oceano-
graphic Applications of Remote Sensing (pp. 131–142). Boca Raton,
FL7 CRC Press.
Moura, A. D., & Shukla, J. (1981). On the dynamics of droughts in
Northeast Brazil: Observations, theory and numerical experiments with
a general circulation model. Journal of the Atmospheric Sciences,
38(12), 2653 – 2675.
Mqller-Karger, F. E., McClain, C. R., & Richardson, P. L. (1988). The
dispersal of the Amazon’s water. Nature, 333(6168), 56 – 59.
Neter, J., Wasserman, W., & Kutnr, M. H. (1989). Applied linear regression
models. USA7 IRWIN, 667 pp.
Pereira, J. A. G. (1995). A pesca do atum nos Açores e o atum patudo
(Thunnus obesus, Lowe 1839) do Atlântico (bThe tuna fishery in
Azores and the Atlantic bigeye tuna (Thunnus obesus, Lowe 1839)Q.
PhD thesis, University of Açores, Horta, 330 pp.
Polito, P. S., Sato, O. T., & Liu, W. T. (2000). Characterization and
validation of the heat storage variability from TOPEX/Poseidon at four
oceanographic sites. Journal of Geophysical Research, 105(C7),
16911 – 16921.
Power, J. H. & May Jr., L. N. (1991). Satellite observed sea surface
temperatures and yellowfin tuna catch and effort in the Gulf of Mexico.
Fisheries Bulletin of United States, 89, 429 – 439.
Punsly, R. (1987). Estimation of the relative abundance of yellowfin
tuna, Thunnus albacares, in the eastern Pacific Ocean during 1970–
1985. International American Tropical Communication Bulletin, 19,
265 – 306.
View publication stats
281
Ricker, W. E. (1975). Computation and interpretation of biological statistics
of fish populations. Bulletin-Fisheries Research Board of Canada, 191
(382 pp.).
Santos, A. M. P. (2000). Fisheries oceanography using satellite and airborne
remote sensing methods: A review. Fisheries Research, 49, 1 – 20.
SAS Institute (2001). The GAM procedure. SAS/STAT software: changes
and enhancements, Release 8.2 (pp. 23 – 74). Cary, NC, USA7 SAS
Institute.
Servain, J., Seva, M., Lukas, S., & Rougier, G. (1987). Climatic atlas of the
tropical Atlantic wind stress and sea surface temperature: 1980–1984.
Ocean-Air Interactions, 1, 109 – 182.
Stewart, R. H. (1985). Methods of satellite oceanography. Scripps
institution of oceanography. San Diego, CA7 EUA 360 pp.
Stretta, J. M. (1991). Forecasting models for tuna fishery with
aeroespatial remote sensing. International Journal of Remote Sensing,
12(4), 771 – 779.
Travassos, P. E. P. (1999). L’étude des relations thons-environnement dans
l’océan Atlantique intertropical ouest: Cas de l’albacore (Thunnus
albacares), du germon (Thunnus alalunga) et du thon obèse (Thunnus
obesus). PhD thesis, University of Paris, Paris, 253 pp.
Travassos, P. E. P., Hazin, H. V. V., Zagaglia, J. R., Advı́ncula, R., &
Schober, J. (1999). Thermohaline structure around seamounts and
islands off North-Eastern Brazil. Archive of Fishery and Marine
Research, 47(2/3), 221 – 222.
Uvo, C. R. M., & Nobre, C. A. (1989). The Intertropical Convergence Zone
(ITCZ) and the precipitation on northeast Brazil: Part I. The ITCZ
position over the Equatorial Atlantic. Climanálise, 4(7), 34 – 40.
Yamanaka, I., Ito, S., Niwa, K., Tanabe, R., Yabuta, Y., & Chikuni, S.
(1988). The fisheries forecasting system in Japan for coastal pelagic
fish. FAO Fisheries Technical Paper, 301 (72 pp.).