NOTE Nolan and Danllowicz: Use of crest nets for sampling presettlement larvae of reef fishes in the Caribbean Sea 
215 
meters are equipped with a high-resolution rotor for 
low-speed flow and had a minimum threshold of approxi- 
mately 6 cm/sec. The mean nightly measurement of flow 
was used to calculate the total volume of water filtered 
by each net. Underwater temperature loggers (Hobo 
Pendant Temperature Logger, Onset Computer Corp., 
Bourne, MA) provided a fine-scale record of the tem- 
perature of water being sampled (temperature data were 
not available for 2005). Wind speed and wind direction 
data were obtained from an automated weather station 
at Belize City International airport (17°53'N, 88°30'W). 
These wind reports provided a reasonable record of 
prevailing conditions at Turneffe Atoll because of the 
proximity and lack of geographic obstacles between the 
two points. The mean nightly wind direction was given a 
positive value for an on-shore wind and a negative value 
for an off-shore wind. Finally, a variable incorporating 
both the nocturnal illumination and tidal periodicity of 
the lunar cycle was calculated (see DAlessandro et al., 
2007). The hours of nocturnal flood tides were calculated 
for each sampling night with tide prediction software 
(JTide, vers. 5.1, P. Lutus, freeware software available 
online) and this number was multiplied by the percent- 
age of the moon that was visible (full moon=100%). 
Statistical analyses 
Species-environment ordinations (CANOCO, vers. 4.5, 
Microcomputer Power, Ithaca, NY ) were used to estab- 
lish the relative importance of individual environmental 
factors (sampling season, wind, water temperature, and 
nocturnal flood tides) in explaining the overall vari- 
ance in larval abundance and species richness in the 
catch. The species and environmental data were found 
to be linear and were examined by redundancy analysis 
(RDA). An RDA plot shows the best fit of multivariate 
data in a two-dimensional ordination. 
The temporal supply of fish larvae was investigated 
by using correlation plots and circular statistics (Ray- 
leigh z; Zar, 1984). Cross-correlation plots were used 
to compare the timing of the capture of larvae in the 
two different environments, namely behind the reef 
crest where crest nets were used and the mangrove 
channels where channel nets were used. Once both net 
types were shown to collect larvae synchronously (see 
Results), the data for both nets were combined into a 
single time series. Auto-correlations were then plotted 
to examine the temporal periodicity of the catch. To 
achieve this, all three sampling periods were concat- 
enated into a continuous time series to ensure that 
more than 2.5 continuous lunar cycles were included 
(the minimum necessary for auto-correlation analysis 
for an examination of lunar periodicity). Each day was 
assigned a number corresponding to its point in the 
lunar cycle (lunar days 1-29, l=new moon). To ensure 
that the cycles were continuous, any overlapping lunar 
days between the sampling periods were deleted (from 
the middle period, spring 2006). The final time-series 
had 164 days, from which 14 overlapping days were 
deleted. 
Figure 2 
Redundancy analysis plot of flow-corrected 
data. The angle between two variables repre- 
sents the correlation between them (0°=positive 
correlation, 90°=no correlation, 180°=nega- 
tive correlation), and the length of the arrow 
represents the magnitude, i.e., the longer the 
arrow, the greater the correlation coefficient. 
Environmental variables are shown as labeled 
arrows. Water = water temperature, Dark = 
hours of moonless, nocturnal flood tide, and 
Wind = average speed (km/h) of onshore wind. 
Categorical variables are shown as triangles. 
Species are shown as unlabelled grey arrows; 
individual species names have been omitted. 
Abundance and species richness are shown as 
black arrows labeled as Abund. and Species, 
respectively. 
Results 
A total of 53,579 larval reef fishes were caught that 
represented 33 families and 59 identified species (Table 
1). On an average night, a crest net trapped 166.3 
larvae (standard deviation [SD] = 407.4) and 8.5 species 
(SD = 5.8), whereas a channel net trapped 4.1 larvae 
(SD=12.2) and 0.9 species (SD=1.5). See Table 1 for list 
of families and species sampled by both net types. 
Ordinations 
There was a strong distinction between the species 
assemblages caught in the two net types (Fig. 2). Only 
data for 2006 sampling periods are presented in Figure 
2, as no water temperatures were available for 2005 
