Poussard et al.: Discriminating between high- and low-quality field depletion experiments 
E4 
K N1 
4 
T4 | _S4 
$2.54 P4 
nu La! E1 
MV} _D4 10 
abt R1 S1 
3.81 
R2 Al 
P4 
Dimension 3 
Dimension 2 
Figure 7 
Correspondence analysis for dimensions 2 and 3 for the data set from deple- 
tion experiments conducted during 1997-2011 to examine efficiency of 
hydraulic dredges for capturing clam species and to estimate stock density for 
populations of ocean quahogs (Arctica islandica) and Atlantic surfclams (Spi- 
sula solidissima) off the mid-Atlantic coast of the United States. Error terms 
Err1 (R1), Err2 (R2), Err3 (R38), and Err4 (R4) are supplementary variables. 
Estimates of the following characteristics are entered as quartiles: dredge 
efficiency (E); clam density (D); the k parameter (K), which is the negative 
binomial dispersion parameter; coefficient of variation (CV) of efficiency (C); 
CV of density (N); CV of the k parameter (P); effective area swept (S); overlap 
score (T); latitude (L); and depth (Z); only quartiles 1 and 4 are shown in the 
plot. Other variables that describe the experiments include species, the ocean 
quahog (O) and Atlantic surfclam (S); region, Long Island (LI) in New York, 
New Jersey (NJ), and the Delmarva Peninsula (DMV) of Delaware, Maryland, 
and Virginia; and dredge width. Dredge widths are 2.54, 3.30, and 3.81 m. 
Error estimates are entered as 1 (below the 80th percentile) or 2 (at or above 
the 80th percentile). The inner box demarcates the area with loading factors 
from —0.5 to 0.5 on both axes. 
289 
similarity in efficiency estimates for the 
2 species, which essentially are sepa- 
rated solely by depth. 
In the correspondence analysis, Err2, 
Err’, and Err4 are associated with 
experiments characterized by smaller 
dredges, higher CV, values, and higher 
CVp values than those of other experi- 
ments. The co-occurrence of these char- 
acteristics instills suspicion about the 
quality of the results obtained from a 
subset of the depletion experiments. 
Essentially, experiments with estimates 
of Err2, Err3, and Err4 that fall at or 
above the 80th percentile produced low 
efficiency estimates. Experiments most 
similar to simulations with high average 
error in efficiency estimates were flagged 
by Err3 and Err4, indicating that these 
experiments could have high uncer- 
tainty in efficiency estimates and that 
these experiments should be removed 
from further evaluation of the inherent 
efficiency of hydraulic clam dredges. 
Estimation of density 
Interestingly, experiments with high CV) 
are grouped with the experiments with 
low efficiency identified by Err2, Err3, 
and Err4 in the correspondence analysis, 
indicating that experiments with more 
uncertain estimates of clam density also 
produced low efficiency estimates and 
were flagged by the error terms. Pous- 
sard et al. (2021) clearly showed that 
the accuracy of efficiency estimates and 
the density of clams in the area are not 
correlated in simulated depletion exper- 
iments, save for instances where low 
clam density combines with an irregular 
distribution of clams in the benthos to 
bias efficiency to a low value. Efficiency 
estimates not being influenced heavily 
falls near the origin in all 3 dimensions, indicating that 
the experiments identified by this error estimate are 
more or less randomly distributed throughout the field 
depletion data set. A tendency for large dredges to be 
associated with improved experimental performance is 
shown in Figure 6; however, the influence of dredge size is 
complex, as the various dredge sizes do not fall in order of 
size on dimensions 1 or 2. Dredge size to some extent is 
likely conflated with other variables, such as species, 
year, and depth, and was determined more by boat avail- 
ability and increased familiarity of the crew and scientific 
staff with depletion experiment methods over time than 
by experiment performance; the clear exception is 
the largest dredge size. The fact that species as a variable 
falls near the origin on dimensions 1 and 3 indicates the 
by clam density is a logical outcome on the basis of an 
expectation that hydraulic dredges should be equally effi- 
cient whether used in low-density or high-density regions. 
The apposition of high CV for the density estimate and 
low efficiency is likely a product of high uncertainty in the 
density estimate co-occurring with high uncertainty in 
the efficiency estimate. Uncertainties in the efficiency and 
density estimates could be indicative of an experimental 
design failing or of environmental parameters not being 
conducive to accurate and precise estimation of efficiency 
and density. 
The accuracy of the density estimate from the Patch 
model was evaluated thoroughly in Hennen et al. (2012). 
The k parameter was not evaluated for accuracy in that 
study because a negative binomial distribution was not 
