Soh et al: Role of marine reserves in the management of Sebastes borealis and 5. aieutianus 
171 
Shortraker rockfish 
60 , 
50 - 
40 
30 
20 •* 
03 
§ 
Rougbeye rockfish 
0 2 4 6 8 10 12 
Duration (hour) 
Figure 2 
Distribution of haul CPUE (t/haul) versus tow dura- 
tion for shortraker and rougheye rockfish in the Gulf 
of Alaska (domestic observer data for trawlers during 
1987-96). 
raker and rougheye production. Precautions are needed 
because these stocks are very slow growing, long-lived, 
and patchy in their distribution. Once “hot spot” stocks are 
depleted, several decades may be required for them to re- 
cover (Francis, 1988). Because this study aimed to develop 
a method for an objective design of refuge networks and 
to evaluate the feasibility of alternative management sce- 
narios, no specific area was suggested for practical usage. 
Size of refugia was determined by using quantiles from 
cumulated catch data. The selection of the size for har- 
vest refugia depends on the goals of refuge management 
in general. Here, three sizes were arbitrarily selected 
and the effects of refuge size on the fish community and 
fishery practices were then evaluated. These refuge sizes 
are referred to as SSR (small-size refugia) MSR (middle- 
size refugia), and LSR (large-size refugia) and include all 
1987-96 catches that exceeded the cutoff points for the 
99.8%, 99.5%, and 99.0% quantiles, respectively (Fig. 3). 
The SSR, MSR, and LSR refugia defined in this manner 
represented 2.4, 5.4, and 9.5% of the total habitat occupied 
by shortraker or rougheye rockfish, or by both species. 
Because the fishery data only provided net retrieval lo- 
cations, the exact catch locations of each tow were un- 
known. This lack of starting-tow locations led us to postu- 
late a circular area around the retrieval point to represent 
the occurrence of the catch. Although the average tow dis- 
tance was about 18 km, the width of the continental slope 
region containing high catches for both species was usual- 
ly less than 15 km, and trawling was generally conducted 
along depth contours. Therefore, the basic cell of the har- 
vest refugium was arbitrarily chosen as a 20-km diameter 
circle, a group of which comprised the refugia for each spe- 
cies. Two species-specific harvest refugia were constructed 
by using GIS. Because shortraker and rougheye rockfish 
are currently managed together, the two independent re- 
serves were united in the final design of the harvest refu- 
gia. Once a GIS database was established, it became pos- 
sible to visualize the impact of different design criteria 
on the spatial distribution of the refuge network almost 
instantaneously. 
Modeling population behavior under refuge 
management 
There were two main goals for the assessment and projec- 
tion of shortraker and rougheye rockfish stock conditions. 
The first was to examine the current stock status, to proj- 
ect future biomass, and to evaluate uncertainties in the 
key parameters under the current management system. 
The other was to examine the effectiveness of harvest 
refugia in the management of these species and compare 
it with the current management system. We employed a 
population dynamics model to address these goals (Soh, 
1998). The model was coded by using AD Model Builder 
(Fournier, 1994). Determination of parameter uncertain- 
ties was based on the Markov Chain Monte Carlo method 
implemented as part of the AD Model Builder software 
libraries. Two separate time schedules were used in the 
modeling process: an assessment section (1961-1996) and 
a projection section (1997-2016). Inputs to the model 
included reconstructed catch histories (1961-1996), priors 
of the parameters, survey biomass estimates, and fixed 
fishing mortality rates. 
Two conditions of fishing mortality were selected for fu- 
ture projection because fishing mortalities in establishing 
ABC and actual fishing mortalities at sea were not equal. 
Fishing mortalities used to calculate recommended ABC 
level were F=0.023 for shortraker and F= 0.025 for rough- 
eye rockfish (which were natural mortalities determined 
by NMFS and referred to as “F for ABC”) during 1991-96. 
These ABC levels were not attained, however, and actual 
average fishing mortalities estimated by the model were 
F=0.G63 for shortraker and F=G.015 for rougheye rockfish 
(referred to as “actual F”) during 1991-96. If no alterna- 
tive actions were applied to the current management pol- 
icy, then projection results under the “actual F” scenario 
would be more realistic than those under “F for ABC”. 
In a refuge management system, no fishing was allowed 
in the refuge areas, and a fixed exploitation rate was ap- 
plied to the harvestable areas. A major difference between 
projections of the current management and those of ref- 
uge management systems is that refuge management in- 
corporates an adjustment for a discard rate of 29.5% (Ta- 
ble 1) into the system on the assumption that no discards 
will be necessary under refuge management. As a result. 
