182 
Fishery Bulletin 111(2) 
20 
-20 
50 
-50 
150 
-150 
75 
A 
Mode 1 SSHA 
1 1 i i i 
Q. 
E 
< 
-75 
100 
E 
6 
r y — \ Mode 2 SSHA 
C ._J, 
1 Mode 1 SST 
D 
1 1 Mode 2 SST 
Mode 1 Chl-a 
-100 
40 
-40 
Mode 2 Chl-a 
Jan98 Jul98 Jan99 Jul99 JanOO JulOO 
Year 
Figure 5 
Amplitude function from the temporal mode of the analysis for 
the study area in the eastern Indian Ocean off Java from Sep- 
tember 1997 to December 2000: (A) first mode of sea-surface- 
height anomaly (SSHA) (interannual signal), (B) second mode of 
SSHA (seasonal signal), (C) first mode of sea-surface tempera- 
ture (SST) (interannual signal), (D) second mode of SST (season- 
al signal), (E) first mode of chlorophyll-a (chl-a) concentrations 
(interannual signal), and (F) second mode of chl-a concentrations 
(annual signal). The x-axis represents the year, and the y-axis 
shows the amplitude function for each mode ( nondimensional ). 
The first EOF mode of SSHA and SST showed 
an inverse relationship with the first EOF mode of 
chlorophyll-a, with a negative SSHA value and rela- 
tively low SST followed by higher chlorophyll-a con- 
centrations along the southern coast of Java. The first 
mode of chlorophyll-a contained 63.39% of the energy 
variance, and the second mode contributed 
15.21% of the energy variance, with notably 
higher chlorophyll-a levels concentrated along 
the southern coast of Java (7-9°S). The ampli- 
tude function of the first mode of chlorophyll- 
a corresponded with interannual variability, 
and the second mode corresponded with the 
annual cycle. Positive values (chlorophyll-a 
concentrations greatly elevated above values 
seen in other periods) occurred during Sep- 
tember-November 1997 (Fig. 5, E-F). 
Generalized additive models 
The results of the GAMs are presented as 
1-parameter, 2-parameter, and 3-parameter 
models (Table 2). All of the variables used were 
statistically highly significant (PcO.OOOl) for 
SSHA, SST, and chlorophyll-a concentrations. 
The addition of predictor variables at different 
levels resulted in an increase in the deviance 
in catch rates explained. In the 1-parameter 
models, SST explained the highest deviance 
(6.48%) and chlorophyll-a concentrations ex- 
plained the lowest deviance (2.04%). The 3-pa- 
rameter combination models explained the 
highest deviance (16.30%) and had the lowest 
AIC values. 
GAM plots can be interpreted as the indi- 
vidual effects of predictor variables associated 
with SSHA, SST, and chlorophyll-a concentra- 
tions on Bigeye Tuna catch (Fig. 6, A-C). High 
probabilities of Bigeye Tuna presence were 
observed for SSHA ranging from -21 to 5 cm, 
for SST ranging from 24° to 27.5°C, and for 
chlorophyll-a levels ranging from 0.04 to 0.16 
mg m~ 3 . Negative effects on Bigeye Tuna were 
observed for SSHA >5 cm, SST values >27.5°C, 
and chlorophyll-a values of 0.01-0.03 mg mr 3 
and >0.16 mg m 3 . 
Spatial predictions for catch distribution of 
Bigeye Tuna were compared with the actual 
monthly fishery data collected during the El 
Nino (September and October 1997) and La Nina 
(March and April 1999) events. The predicted 
catch distribution of Bigeye Tuna in September 
1997 during the El Nino event indicated a po- 
tential area with higher catch probability of ap- 
proximately 70-80% at 10-16°S and 104-122°E, 
and the actual Bigeye Tuna fishing locations 
(with a HR of 0.41) occurred in the area between 
12-16°S and 110-115°E (Fig. 7A). In October 
1997, spatial predictions for catch distribution 
of Bigeye Tuna indicated locations with higher catch 
probability (60-70%) in the west at 7-12°S, 104-108°E 
and 14-16°S, 109-1 14°E, but the actual Bigeye Tuna 
fishing locations were located at 12-15°S, 110-116°E, 
where there was a predicted catch probability of around 
20-40% and actual HR of only 0.20 (Fig. 7B). The 
