Hettler et a I.: Variability in abundance of larval fishes inside Beaufort Inlet 
491 
Daily ageing of B. tyrannus caught during the 
study revealed that a rapid and distinct shift in lar- 
val populations occurred in mid-March (Fig. 7). In 
early March, the recruiting larvae were primarily 
from a cohort that was spawned in mid-December, 
first reached the inlet in mid-January, and suddenly 
disappeared on 16 March. On 18 March a new co- 
hort, spawned in mid-January, appeared. Its appear- 
ance was coincident with the year’s highest daily 
mean density. This change in age structure coincided 
with a 3-day shift to southwesterly winds and full- 
moon spring tides. Advanced very high resolution 
radiometer sea surface temperature (AVHRR-SST) 
imagery revealed that sea surface temperatures 15 
km off Beaufort Inlet warmed to 15°C on 9 March 
1992, up from 11°C a week earlier (and down to 
12.3°C a week later), possibly bringing in the younger 
B. tyrannus larvae from warmer offshore water to 
the vicinity of Beaufort Inlet (Stegmann 1 ). This 
warming was also detected by our temperature mea- 
surements at the sampling platform, rising to 17.9°C 
on 9 March followed by a decrease to 12°C on 18 
March (Fig. 8), when the large number of younger 
larvae were caught. Until that date, the average es- 
timated age of larvae caught in Beaufort Inlet in- 
creased from about 35 days in late November to about 
80 days in mid-March. About 57% of the menhaden 
captured in the 1991-92 season were spawned in two 
2-week periods (mid-December and late January). 
Although substantial spawning may have occurred 
at other times, the population from which these lar- 
vae were caught contained the survivors of the cross- 
shelf transport. It appears that size at estimated age 
was significantly larger, about 3 mm SL, for ages 
between 40 and 80 days for the 1991-92 collections 
than the larval size reported by Warlen ( 1992) for B. 
tyrannus collected mainly offshore of Beaufort Inlet 
in 1979-80. Because the daily ageing method was 
the same, this observed difference in growth rates of 
the 1991-92 ingressing larvae is attributed to a 
higher growth rate among the larvae that survive to 
reach the inlet. 
Coastal marine environments experience periods 
of diurnal or semidurnal tidal cycles imbedded within 
lunar and semilunar cycles, and these have been 
shown to influence a broad range of organisms and 
processes (Hutchinson and Sklar, 1993). Cyclical 
phenomena impose a particular set of requirements 
for their adequate measurement and for the avoid- 
ance of bias arising from aliasing (Kelly, 1976) be- 
cause the temporal sequence of observations will be 
1 Stegmann, P. 1996. Graduate School of Oceanography, Univ. 
Rhode Island, S. Ferry Road, Narragansett, RI 02882. Per- 
sonal commun. 
autocorrelated. A circannual rhythm is a feature of 
the life history of most vertebrates, and one mani- 
festation of this is the restriction of reproduction of 
a species to a season of several weeks or months. If 
the purpose of a sampling program is to compare 
recruitment of a species of larval fish from year to 
year, then it should be designed so that it describes 
each year’s temporal pattern accurately. In this re- 
spect it differs from the normal random sampling 
situation designed to estimate the mean and vari- 
ance of some statistical population. Because its pur- 
pose is to enable description of a temporal pattern, a 
systematic design is normally chosen to ensure equal 
(or near equal) spacing of samples. Equal spacing of 
samples is required for many approaches for the 
analysis of a time series. However, if there are other, 
shorter cycles within the seasonal pattern, then care 
must be taken to avoid spurious results (aliasing) 
that can arise when the sampling interval is greater 
than one half the wavelength of a significant compo- 
nent cycle. Sampling to determine a seasonal flux of 
larvae should be designed to detect temporal patterns, 
as well as estimate a mean abundance and variance. 
The question of sampling frequency may have a 
lower priority than considerations of sampling costs 
and vessel availability, and thus it is important to 
quantify the effect that sampling frequency has on 
estimates of larval abundance, size, and age. For 
example, one may be interested in estimating the 
flux of larvae across a boundary over some unit of 
time. If one is interested in the strength of a year 
class, the sampling effort must include the entire 
season of larval recruitment of that species. If one is 
interested in evaluating the influence of a meteoro- 
logical event on larval distribution, then the appro- 
priate sampling interval would be measured in hours. 
As we shall see, both sampling designs also require 
due consideration of various physical and biological 
rhythms that have an important bearing on the num- 
ber of larvae collected at a given point in space and 
time (e.g. tidal, circadian, circannual). Prior to es- 
tablishing sampling protocol for future studies, we 
attempted to determine a sampling interval that 
would provide acceptable larval fish abundance es- 
timates. Larval fish surveys have been made in Beau- 
fort Inlet and other North Carolina inlets on differ- 
ent sampling intervals, i.e. weekly (Warlen, 1994), 
bi-weekly (Lewis and Mann, 1971), every new and full 
moon period (Hettler and Chester, 1990) and every new 
moon period (Hettler and Barker, 1993 ). When we com- 
pared the weekly sampling method that has been used 
for monitoring at Pi vers Island for the past 10 years 
(Warlen, 1994) with our daily sampling experiment, a 
difference in estimated abundance ofR. tyrannus was 
detected. Differences in the two types of nets may have 
