Schobernd et al.: A comparison of numbers of fish larvae extruded from plankton nets of different mesh size 
243 
article. The actual number of specimens measured de¬ 
pended on the taxonomic group and level of identifica¬ 
tion. To increase the number of observations available 
for analysis, up to 50 randomly chosen specimens in 
the 5 families targeted for analysis (Engraulidae, Clu- 
peidae, Scombridae, Sciaenidae, and Lutjanidae) and 
unidentified and Percoidei larvae were measured at 
the NMFS laboratory in Pascagoula, Mississippi. These 
families were examined because they contain either 
ecologically or economically important species, many of 
which are federally managed. Additionally, larvae with¬ 
in these families represent the 2 body shapes of larval 
fish, clupeiform (slender) and perciform (robust) that 
have been shown to influence susceptibility to extru¬ 
sion (Smith and Richardson, 1977). Unidentified larvae 
(with mixed body shapes) and those identifiable only 
to the suborder Percoidei (perciform) were measured 
because larvae in these 2 categories were among the 
smallest specimens in the samples and were, therefore, 
most likely to be extruded from the coarser mesh net. 
Although larval size can shrink as much as 22% to 33% 
because of tissue damage during capture and preserva¬ 
tion (Miller and Sumida, 1974; Theilacker, 1980), this 
potential damage was not accounted for in our length 
measurements. This factor may explain why nominal 
lengths of the smallest larvae in the samples collected 
with the nets of 2 different mesh sizes were smaller 
than reported larval sizes at hatching. Larval shrink¬ 
age rates, however, were not expected to differ between 
samples from the nets with fine and coarse mesh sizes. 
Species-level identification based on published lar¬ 
val descriptions for the Gulf of Mexico region requires 
the morphological presence of characters not gener¬ 
ally present in larvae <3 mm BL (Richards, 2006). As 
such, many small specimens in early stages of devel¬ 
opment from the samples taken with the 2 nets were 
identifiable only to family. To use data over all sizes 
represented in the study collections while maintaining 
taxonomic groups of distinct body shapes, specimens 
of the 5 targeted families identified to genus or spe¬ 
cies were combined with specimens at the family level 
for subsequent analysis (i.e., analysis occurred at the 
family level). 
Statistical analyses 
Paired Wilcoxon signed-rank tests were used to de¬ 
termine significant differences in means between 
samples from nets with the 2 mesh sizes, 0.202 mm 
and 0.333 mm. Means were examined for the follow¬ 
ing values: volume filtered, total sample displacement 
volume, total fish eggs (raw counts), total fish larvae 
(raw counts), and standardized larval abundance (the 
number of larvae under 10 m 2 sea surface, and number 
of larvae per 1000 m 3 ). To reduce the chance of type- 
I errors, a values were adjusted by using a sequen¬ 
tial Bonferroni adjustment (Rice, 1989). Analyses were 
not stratified by time of day because the samples from 
each plankton net attached to the bongo frame were 
taken at the same time (paired tows) so that any diel 
influence would be the same for both samples. Plots of 
volume filtered versus tow depth, by mesh size, were 
examined to determine whether clogging between the 
meshes of the 2 nets over the entire range of sampling 
depths had occurred in our study. Plots of tow duration 
(related to depth) by larval abundance for each mesh 
size were also examined to determine whether shorter 
tows at shallower, inshore stations collected abundanc¬ 
es similar to those of longer tows at deeper, offshore 
stations. Paired t-tests were used to test for signifi¬ 
cant differences in mean larval abundances between 
the 2 mesh sizes for the groups of interest: unidenti¬ 
fied larvae and larvae of Percoidei, Engraulidae, Clu- 
peidae, Scombridae, Sciaenidae, and Lutjanidae. The 
Kolmogorov-Smirnov test (K-S test) was used to de¬ 
termine whether length-frequency distributions varied 
significantly for larvae under 10 mm BL for samples 
from the nets with 0.202-mm and 0.333-mm meshes. 
Functional relationships were constructed by com¬ 
paring the ratio of the numbers of larvae collected 
with the 0.202 mesh net to the numbers of larvae col¬ 
lected with the standard 0.333 mesh net to assess the 
numbers of larvae extruded through the coarser mesh. 
Models were constructed for unidentified larvae, per¬ 
coidei larvae, and larvae from the 5 targeted families. 
Ratios of mean standardized abundance (number under 
10 m 2 sea surface) from the nets with 0.202-mm mesh 
to the mean standardized abundance from the nets 
with 0.333-mm meshes were calculated for each taxon 
by 0.1-mm size classes. Only size classes where both 
nets had positive catches of the target taxa were used 
as data for fitting the models. All functions were fitted 
with maximum likelihood estimation and log-normally 
distributed error structures by using the ‘bbmle’ pack¬ 
age in the software R, vers. 3.3.1 (Bolker, 2008; R Core 
Team, 2016). Power (Eq. 1) and exponential (Eq. 2) 
models were used to describe the relationship between 
the larval abundance ratios: 
P r = aL h and (1) 
P r = de~ u , (2) 
where P T - the predicted ratio of abundances in sam¬ 
ples collected with nets of the 2 mesh sizes 
(0.202-mm:0.333-mm); and 
L = the size class (in millimeters). 
Parameters a, b, d, and f are constants estimat¬ 
ed during the fitting process. 
Akaike information criterion (AIC) was used to deter¬ 
mine which model was the best fit for a given taxon 
(Burnham and Anderson, 2002). In this study, AAIC 
scores are presented as the relative difference between 
the AIC score of each model from that of the best fit¬ 
ting model within a taxonomic group. 
Results 
Fish eggs and larvae were collected in all 162 samples 
from 81 tows of paired bongo nets. Samples included 
