Kuykendall et al.: A management strategy evaluation for Spisula solidissima 
307 
Table 1 
Varying characteristics of model simulations used in the management strategy evaluation for the Atlantic surfclam ( Spisula 
solidissima ) in the Mid-Atlantic Bight. The structure of each set of cases is shown in bold. Incidental mortality applies to 
clams that are not retained by the dredge. Nine individual simulations (in bold) represent one set of cases, an ennead (see 
also Fig. 4). An ennead was run for each level of incidental mortality, each management option, each closure duration, and 
each definition of a small clam for a total of 72 (2x3x3x4) enneads. Market-size clams are >120 mm shell length (SL). LPUE 
is measured as number of bushels fished per hour. Three initial stock distributions ranging from dense to sparse levels of 
patchiness were specified to cover a range in stock patchiness. Patchiness was established by assigning new recruits to each 
10' square by using a negative binomial random distribution that produced distinctive variance in the abundance of clams 
in each 10' square relative to the mean abundance for all 10' squares. 
Model configurations Ennead complement 
Performance 
metrics 
Levels of 
incidental 
mortality 
Management 
options 
Closure 
duration 
(yr) 
Definitions 
of a small 
clam 
Patchiness 
Captain 
type 
Stock density 
(number of clams 
>75 mm SL/m 2 ) 
0% 
Present-day — no closures 
3 
104 mm SL 
High 
Standard 
Confident 
Survey 
Number of clams 
per bushel 
20% 
Rule 1 — close the 10' square 
with the highest ratio of the 
number of small clams to the 
number of market-sized clams 
5 
93 mm SL 
Medium 
Standard 
Confident 
Survey 
LPUE (bushels/h) 
Rule 2 — close the 10' square 
with the highest number of 
small clams/m 2 
7 
80 mm SL 
Low 
Standard 
Confident 
Survey 
Number of 10' squares 64 mm SL 
fished 
Distance traveled 
during fishing (km) 
10' squares is used to identify groups of clams with 
the same maximal elapsed time to market size for our 
presentation of simulation results. These averages are 
104, 93, 80, and 64 mm SL for 2, 3, 4, and 5 growth 
years, respectively, to reach 120 mm SL. 
Statistical evaluation of alternative management strategies 
Each comparison between present-day and alternative 
management is based on an ennead of simulations with 
varying captain behavior and patchiness of recruitment 
(Table 1) for each defined size of a small clam and each 
specified closure duration. This results in 4 enneads per 
closure duration (1 set of 9 for each definition of a small 
clam) and 12 enneads per closure location rule (Fig. 4). 
The structure of the base case is composed of the same 
states of nature and captain behaviors as the alternative 
management cases, excluding area closures. Each enne- 
ad is designed to evaluate the interaction of patchiness 
of distribution with captain behavior in order to evalu- 
ate the sensitivity of outcomes to this key interaction. 
Management strategies were compared by using 
the nonparametric Wilcoxon signed-rank test (Conover, 
1980). This test uses the difference in metric values be- 
tween 2 outcomes — in this case the difference between 
the base case and the otherwise equivalent simulation 
under alternative management (e.g., closure location 
rules 1 and 2, closure durations, and small clam defini- 
tions (i.e., years to market size, Fig. 5). Each compari- 
son was based on 76 years of simulated time, with 1 
difference calculated for each of the 76 years; therefore, 
a single Wilcoxon test was based on n = 76 (Fig. 6). Be- 
cause each year was different from each succeeding or 
preceding year because a 10' square was opened and 
closed each year and because each vessel and captain 
operated independently in each year with their own be- 
haviors and differing memories, each year represented 
a unique pairwise comparison of the area management 
option and the otherwise equivalent base case. There- 
fore, 9 Wilcoxon tests were conducted for each ennead. 
The likelihood of the number of significant outcomes 
from these 9 tests exceeding chance was evaluated by 
an exact binomial test (Conover, 1980). Any comparison 
yielding more than 1 significant difference between the 
base case and area management option out of the 9 
simulations performed exceeded the number expected 
by chance at a=0.05. 
Performance metrics were evaluated by the propor- 
tion of simulations that resulted in an increased per- 
formance metric in comparison with the base case with 
the same composition and the amount of increase seen 
in those significant simulations. Management strate- 
