234 
Fishery Bulletin 1 13(3) 
ting a signal. At the Kaneohe nursery, the fish were 
split equally into 2 batches and released at the surface 
over each of the 2 receiver moorings. A snorkeler ob- 
served that all fish swam down (to a depth of ~30 m) 
and disappeared from sight within a minute. Tags had 
an expected life of 95 days and emitted a signal with a 
pseudorandom delay at an average of 100 s. Cultured 
fish were released on 23 June 2006 and monitored for 
58 days; wild fish were released on 10 July 2007 and 
monitored for 76 days. 
Data analysis 
We assumed a detection radius of 400 m for the re- 
ceivers that was based on the results from the manu- 
facturer’s range calculator (Website) when we used 
the inputs of the tag power (151 db) and using the 
average windward Oahu wind speed (15-20 kt). Wind 
speed affects sea-surface conditions and introduces 
background noise that impacts the travel of trans- 
mitter signals underwater. Simultaneous detections 
were rare and only occurred between the 2 receivers 
deployed on the nursery site; they were the receivers 
closest to each other and monitored approximately 
half the area of the nursery habitat. The rest of the 
receivers allowed us to monitor habitat on the slope 
and were spaced far enough apart to avoid overlap- 
ping detections. It was necessary to discern signals 
from a tag on a resident, live fish from signals from 
a tag that was still transmitting but lost on the bot- 
tom for various reasons (e.g., tag expulsion or death of 
fish). For the purposes of this study, signals were as- 
sumed to be emitted from dead fish if 1) they had an 
inordinately high number of detections (e.g., >15,000) 
consistent with the detection of the continuous trans- 
mitting of a tag lost on the bottom for the full dura- 
tion of the surveillance period and 2) they were de- 
tected by only 1 receiver. 
Because prior tracking (Moffitt and Parrish, 1996) 
indicated that movement patterns of fish in the nurs- 
ery differed between day and night, the detections 
of the fish in our study were binned in 6-h intervals 
for analysis (2400-0559, 0600-1159, 1200-1759, and 
1800-2359). Initially, the density of tagged fish (and 
the risk of signal collisions) was high; therefore, we re- 
quired 2 or more successive signals detected within 1 
h to provide greater temporal resolution for the first 
3 days. We defined a signal as “successive” if it was 
a repeat signal detected within 5 min of the previous 
signal — a time period adequate enough for 2 detections 
given the cycle of the tag’s delay between transmis- 
sions. After 3 days, when most of the fish had departed, 
the risk of signal collisions that create false detections 
was reduced; therefore, multiple (>2) isolated detec- 
tions (spaced more than 5 min apart) were accepted as 
long as they fell within the 6-h time interval. Combin- 
ing successive detections binned by time intervals in- 
creased confidence that the fish were actually present 
(94.5% confidence intervals binned by hour, 97.9% by 6 
h) and rendered insignificant the effect on our analysis 
of erroneous detections from signal collisions of mul- 
tiple tags (see Pincock 4 ). 
We used the data from the 2 nursery receivers to 
look at patterns of time spent in the nursery, of in- 
fluences of temperature, and of fish body length. Data 
recorded by the 4 slope receivers were used to exam- 
ine patterns of habitat use by wild fish as they trav- 
elled away from the nursery. The sample size of fish 
was suitable for detecting large effect sizes at a power 
of 0.80 with an alpha of 0.10 (Cohen, 1988). All sta- 
tistical analyses were performed in IBM SPSS, vers. 
22 (IBM, Armonk, NY). Normalized data fitted to a 
negative exponential distribution showed how close 
the decline in fish detections was to a constant rate. 
In other comparisons, analysis of variance (ANOVA), 
correlation, and standard nonparametric tests (e.g., 
Mann-Whitney [MW] and Kruskal-Wallis [KW] ) were 
used to analyze patterns in fish movements over time 
and in relation to body size and habitat (Siegel and 
Castellan, 1988). 
Results 
Tag detections in the nursery 
Movements documented through tag detections indi- 
cated that 18 cultured fish and 28 wild fish were alive 
and suitable for inclusion in the analysis (Table 1). 
Seven other fish were excluded, including 2 wild fish 
that went undetected, 2 wild fish that were present 
only briefly (< 10 signals), and 2 cultured and 1 wild 
fish whose tags emitted continuous signals recorded by 
1 receiver (an indication that the tags were immobile 
and lying on the bottom because of tag expulsion or 
the death of a fish). Detections of tags lost on the bot- 
tom were unaffected by daily tidal changes in the tem- 
perature cycle, indicating that there was little effect 
from the thermocline on the reception of the receivers. 
Almost all (97%) of the tagged fish (cultured and wild) 
used in the analysis moved back and forth between 
the 2 nursery receivers, indicating that the fish were 
active. 
The signals of both cultured and wild fish declined 
exponentially to a low and variable level of presence 
by the end of the surveillance period in each year. 
Within 3 days of release, the portion of released cul- 
tured fish that were detected in the nursery area 
dropped to less than 20% in a pattern that closely fit- 
ted an inverse exponential curve (adjusted coefficient 
of multiple determination, i? 2 =0.821). In contrast, 
75% of wild fish persisted in the nursery for 3 days, 
reducing the fit to the curve (adjusted i? 2 =0.205) and, 
therefore, indicating that the decline was not as con- 
tinuous as the drop observed for cultured fish (Fig. 2). 
4 Pincock, D. G. 2012. False detections: what they are and 
how to remove them from detection data. Application note. 
AMIRIX Systems Inc. DOC-004691-03, 11 p. [Available at 
Website.] 
