Cowen et al.: Evaluation of the In Situ Ichthyoplankton Imaging System and comparison with the bongo-net sampler 
9 
A. 
20 - 
15- 
E 
E 
03 
c 
.2 
• • 
• i 
c f) 
o ■ 
i i , 
0 - 
— j 
Bongo 1 
, - — 1 - I ~ 
Bongo 2 ISIIS 1 ISIIS 2 
Sample gear 
Clupeidae - 
1 
Phycidae - 
-ID- • 
Merlucciidae . 
-on — • 
Gadidae - 
i 
03 
Ophidiidae . 
-a- 
o 
Gobiidae - 
to 
o 
Scopthalmidae - 
-D- • 
Paralichthyidae . 
--D— - 
Unknown . 
Clupeidae . 
— i n- 
Phycidae . 
HE — 
Merlucciidae . 
1 1 1 
Gadidae . 
i i( 
Ophidiidae . 
-m — • 
CO 
Gobiidae . 
- i- • 
CO 
Scopthalmidae . 
-H3— 
Paralichthyidae . 
— {D — 
Unknown . 
— 1 ! 1 1 1 — 
0 5 10 15 20 
Standard length (mm) 
Figure 6 
(A) Summary statistics (box plot) of larval size distribu- 
tion by sampling gear and transect (1 and 2). The vertical 
bars of the box plot represent the range, the box repre- 
sents the 1 st (lower) and 3 rd (upper) quartile, and the cen- 
tral (horizontal) line is the median of the distribution of 
observations. Perceived outliers are denoted as separate 
points beyond the range. Sampling was conducted with 
the In Situ Ichthyoplankton Imaging System (ISIIS) and 
a bongo sampler south of Woods Hole, Massachusetts, in 
October 2008. (B) Taxon-specific comparison of fish lengths 
between bongo sampler (top) and ISIIS (bottom). Note: the 
box plots are rotated 90°relative to A; however, basic fea- 
tures are the same as in A. 
cessfully has estimated larval fish concentration, and, 
in an environment of relatively low diversity, as in this 
study, resolved the taxonomic composition of the larval 
ichthyofauna. Under such conditions, the rapid sam- 
pling speed of ISIIS could be used to increase spatial 
and temporal resolution of ichthyoplankton patchiness, 
without the need for additional ship days. For example, 
rapid undulation of ISIIS resulted in 24 vertical forays 
through the water column being repeated every 1.7 km 
along the 41.4-km transect in just 4.6 h. In comparison, 
6 bongo tows were completed along the same transect 
in ~6 h for a spatial resolution of 6.9 km. Therefore, 
ISIIS can provide 3-4 times the spatial resolution of 
a bongo sampler over a comparable (or shorter) time 
frame. Other benefits of ISIIS include its ability to re- 
solve very fine-scale patchiness because its sampling 
rate is both continuous and rapid. Consequently, de- 
pending on how it is towed, ISIIS can be used to assess 
detailed vertical distributional data, a feat that is not 
possible with a bongo sampler, or even with opening 
and closing net systems, without very extensive (and 
expensive) sampling efforts. Further, simultaneous 
sampling by other environmental sensors provides de- 
tailed concurrent image and physical data. Information 
about nearest-neighbor scaling and fish larval distri- 
bution in relation to their predators and prey, as well 
as environmental conditions, would be possible because 
of the fine-scale, in situ information available in the 
ISIIS imagery. Such sampling with ISIIS would allow 
targeted, process-oriented studies, even while general 
survey designs are being employed. 
Still, the results of this study indicate several specif- 
ic functional aspects that need to be considered or ad- 
dressed for ISIIS to be a highly effective sampling tool 
for survey and process-oriented studies. First, ISIIS 
detected fewer smaller larvae than did the bongo sam- 
pler. Further, the small larvae detected with ISIIS were 
largely classified as unknown. These results indicate 
that the image resolution of ISIIS should be improved 
to increase the detectability and identification of small 
larvae, although preflexion larvae will likely always be 
problematic because of their limited morphological dis- 
tinctiveness. An increase in detectability will require 
an increase in the depth of field such that particles 
that pass between the viewing ports are all in focus, 
thereby eliminating regions of out-of-focus particles 
that potentially can obscure the remaining image. The 
current version of ISIIS (ISIIS-2) has been successful 
at extending the depth of field from -30 cm to the full 
50-cm space between viewing ports, adding to the vol- 
ume sampled and the overall clarity of imagery (Cowen 
and Guigand, unpubl. data). 
The second issue is the need for rapid, accurate im- 
age processing. The large number of images produced 
makes computer-aided image analysis a requirement 
for large-scale application of this instrument. We were 
able to use manual assessment of the images taken in 
the current study (by focusing only on fish larvae), but 
further analysis of these data or more extensive surveys 
