Poussard et al.: Discriminating between high- and low-quality field depletion experiments 283 
Efficiency 
T 
10 
Depletion experiment 
Efficiency 
experiments. In addition, experiments 
flagged by Err3 and Err4 had CV, and 
EAS values that were substantially dif- 
ferent from the values from the experi- 
ments that were not flagged (Table 8). 
Interestingly, all 9 experiments flagged 
by Err3 were among the 10 flagged by 
Err4, yet experiments flagged by Err3 
produced significantly different CV, and 
CVp estimates and the group of experi- 
ments flagged by Err4 did not. 
Correspondence analysis 
A plot of correspondence analysis shows 
that variance in descriptor characteris- 
tics is primarily explained by the first 
2 axes (Fig. 6). The 4 error terms are 
included as supplementary variables. 
The dispersions of clams (Fig. 2) were 
also added as supplementary variables; 
however, they are not included in the fig- 
ures because each falls near the center 
of the correspondence plot. Dimension 1 
(Figs. 6 and 7) was determined primarily 
on the basis of characteristic values from 
the Patch model, including estimates of 
efficiency, CVz, CVp, width of the dredge, 
and EAS (Table 9). Low EAS (which indi- 
cates more dredge overlap, low efficiency, 
or small experimental area), low effi- 
ciency estimates, high CV, and CVp, and 
small dredge sizes (2.54 m [8.33 ft] and 
3.05 m [10.00 ft] wide), along with exper- 
iments with estimates of Err2, Err3, and 
T T 
T 
10 15 20 
Depletion experiment 
Figure 3 
Estimates of capture efficiency of hydraulic dredges (A) for each of the 19 
depletion experiments that targeted ocean quahogs (Arctica islandica) and 
(B) for each of the 31 depletion experiments that targeted Atlantic surfclams 
(Spisula solidissima) between 1997 and 2011 off the mid-Atlantic coast of 
the United States. The black horizontal line indicates the mean efficiency 
for the group of experiments in each panel. Error bars indicate standard 
deviations. 
identified experiments were distributed randomly within 
the full set of field experiments with respect to the differ- 
ent error terms tested. If the experiments flagged by Err1 
were removed from the analysis, the mean and median 
efficiency estimates of the remaining field depletion exper- 
iments were not significantly different from the mean and 
median estimates for the entire data set. In dramatic con- 
trast, for experiments flagged by Err2, Err3, and Err4, the 
efficiency estimates from the Patch model differed sub- 
stantially from the estimates from the remaining field 
Err4 that fall at or above the 80th per- 
centile, fall on the positive (right) side of 
dimension 1. High efficiency estimates, 
high EAS, the largest dredge size (3.81 m 
[12.50 ft]), and low CV, and CVjp fall on 
the negative (left) side of dimension 1. 
Dimension 2 (Figs. 6 and 7) is catego- 
rized by species (the ocean quahog and 
Atlantic surfclam) and other variables 
that relate to the location of the deple- 
tion experiments for the 2 species, such 
as depth, latitude, and region. The pos- 
itive values are variables that relate 
to depletion experiments that targeted 
ocean quahogs, variables such as higher latitudes and 
deeper depths. Negative values are variables that relate 
to depletion experiments that targeted Atlantic surfclams: 
lower latitudes and shallower depths. Experiments with 
ocean quahogs were typically conducted farther north 
(Fig. 1) than experiments with Atlantic surfclams, and 
the species is generally found at deeper depths than the 
Atlantic surfclam. Dimension 3 (Figs. 7 and 8) is charac- 
terized by the CVx, and high efficiency estimates and high 
EAS fall on the positive side. 
