206 
Fishery Bulletin 113(2) 
LLI 
D 
CL 
o 
8 
-I 
Feb 
O 
O 
O O 
o 
o 
o 
o 
Mar Apr May Jun Jul Aug Sep Oct Nov Dec 
Figure 4 
Monthly box-and-whisker plot of CPUE), measured in metric tons (t )/ 
vd (vd=per vessel per day) of Argentine shortfin squid (Illex argen- 
tinus) caught by Taiwanese jiggers from 1986 through 2010 in the 
southwest Atlantic. Open circles indicate outliers detected for specific 
years. The fishing season began in November and was most active 
from January through June. 
lagged after the previous 2 years showed 
that the logfJ had no link with any sub- 
surface seawater temperature or the SOI 
but was significantly affected by the AAO 
in November, December, March, and May. 
The logU and AAO, in the analysis with 
a lag of the previous 2 years, were nega- 
tively correlated in November and Decem- 
ber and positively correlated in March and 
May (Table 1). There were no correlations 
between log U and bottom seawater tem- 
perature with and without time lags. 
Analysis of annual variation with a general- 
ized linear model 
The GLM analysis that used a backward 
stepwise procedure found that 4 vari- 
ables, including the AAO of November 
and March in the previous 2 years and 
the subsurface seawater temperature in 
March of the current and last fishing sea- 
son, explained approximately 83.3% of an- 
nual variations shown by logt/ differences 
(Table 2, Fig. 6). The VIF test showed that 
the 4 variables exhibited no multiple colin- 
earity. Although analysis with the AIC in- 
dicated that a model with 5 variables that 
included the AAO of the previous 2 year’s 
December was better than the model with 
4 variables, the 5-variable model did not 
add 5% to the R 2 and the P-value of the 
AAO lagged from the previous 2 year’s 
December was not significant. Adding the 
previous year’s values of log U did not im- 
prove the percentage of variation in log U 
that was explained (P=0.324) and therefore 
indicated an absence of autocorrelations. 
The progressive analysis of the GLM 
with the data set from 1998 through 2007, 
showed that the AAO variables in Novem- 
ber and March, lagged by 2 years, were 
constantly selected in our model. In addi- 
tion, the seawater temperatures at a depth 
of 5 m during March of the previous had 
significant effects in the model 9 times out 
of 10 in the year-by-year progressive GLM. 
For the southern location, the monthly 
seawater temperature at a 5-m depth in 
March, without a lag time in relation to 
the current fishing season, also was kept 
in the models that included 2 of the years, 
2004 and 2007 (Table 3). 
Discussion 
The results of this study indicate a link be- 
tween fluctuation in the marine environ- 
ment and abundance of Argentine short- 
