Chen et at: A modeling approach to identify optimal habitat and suitable fishing grounds for Ommastrephes bartramii 
11 
sensing, and they are more important than SSS in 
forecasting O. bartramii habitat. Although the chi a and 
SST data obtained from remote sensing have limita- 
tions (Arrigo et al., 1998; Santos, 2000) and errors may 
occur in areas where the cloudy weather is frequent, 
these environmental data are commonly used in marine 
fisheries (Santos, 2000; Wang et al., 2003; Zagaglia et 
al., 2004; Zainuddin et al., 2006; Chen et al., 2008b). 
Moreover, because O. bartramii undertake vertical diel 
movements, inhabiting water depths of 0-40 m during 
night and 150-350 m during the day (Wang and Chen, 
2005), other environmental variables, such as water 
temperature at the different depths and the vertical 
structure of water temperature, need to be considered 
in future analyses. 
Fishing effort is a good indicator in estimating SI 
values when commercial fishery data are used. In the 
squid jigging fishery of North Pacific Ocean, Chinese 
mainland fisherman first determine the fishing area 
with the help of near real-time SST and SSHA data 
from remote sensing and then locate the fishing posi- 
tion by using an echo-sounder (Chen, 2004; Wang and 
Chen, 2005). Therefore, fishing effort can be considered 
an index of squid occurrence. Although the nominal 
CPUE is affected by fishing boats, fishing technology, 
light power, and other environmental factors in the 
commercial fishery, CPUE is not suitable for estimating 
SI values. Tian et al. (2009) also reported that a fish- 
ing effort-based HSI model performed better than the 
CPUE-based HSI model in defining optimal habitats for 
neon flying squid, whereas the CPUE-based HSI model 
tended to overestimate the range of optimal habitats 
and underestimate monthly variations in the spatial 
distribution of optimal habitats. 
Clearly, we found that O. bartramii are not randomly 
distributed in relation to environmental conditions. 
Outputs produced from HSI modeling can indicate the 
spatiotemporal variation of squid habitat conditions. 
Many fish habitat models have been developed by us- 
ing combined empirical and GIS-based spatial model- 
ing techniques (Rubec et al., 1999; Brown et al., 2000; 
Feng et al., 2007; Chen et al., 2008b). These approaches 
differ in their assumptions, inputs, and outputs. This 
study indicates that the HSI modeling approach, which 
is relatively simple and straightforward, may be an 
appropriate method for pinpointing optimal habitats 
and potential fishing grounds of squid. Because nearly 
every commercially important marine species is sensi- 
tive to SST and has a seasonal optimum SST range, 
and because the near real-time SST can be obtained 
from remote sensing, SST is usually considered a basic 
input variable in developing an HSI model (Eastwood 
et al., 2001; Le Pape et al., 2003; Zagaglia et al., 2004; 
Zainuddin et al., 2006). 
Different HSI models with one to four variables tend- 
ed to yield varying results. For the same set of envi- 
ronmental variables, the AMM model performed better 
than the GMM model because its AIC value was smaller 
(Table 3). The AMM model with three variables, SST, 
SSHA, and chi a, had the lowest AIC value for fishing 
60 
50 
40 
£ 30 
20 
10 
60 
J August 
[] September 
| October 
0-0.2 0.2-0. 4 0. 4-0.6 0.6-0.8 0.3-1 .0 
HSI 
B 
Figure 9 
The relationship between the habitat suitability index 
(HSI) values estimated from the arithmetic mean model 
with three environmental variables (sea surface tem- 
perature, sea surface height anomaly, and chlorophyll-a 
concentrations), and (A) actual percentage of the total 
catch, (B) percentage of fishing effort, and (C) catch 
per unit of effort (CPUE) for Ommastrephes bartramii 
from August to October 2005. 
effort. To further evaluate the performance of the AMM 
models, we compared their outputs with corresponding 
abundance density (CPUE). We found that the average 
