bs_bs_banner
Austral Entomology (2016) ••, ••–••
An appraisal of sampling methods and effort for investigating moth assemblages
in a Fijian forest
Siteri Tikoca,1 Simon Hodge,2* Marika Tuiwawa,1 Gilianne Brodie,1 Sarah Pene1 and John Clayton3
1
University of the South Pacific, Suva, Fiji.
Lincoln University, Canterbury, 7647, New Zealand.
3
Broughty Ferry, Dundee, UK.
2
Abstract
The moth assemblages in forest ecosystems are often used as indicators of forest quality and to monitor the effects
of habitat degradation or ecological restoration and management. However, to provide meaningful data on nocturnal moth faunas, it is important to evaluate the efficacy of available sampling methods and identify the minimum
number of samples needed to obtain a reliable estimate of moth diversity. This study compared three light-based
collecting methods to sample nocturnal moths in Colo-i-Suva Forest Reserve, a lowland mixed forest 8 km north
of Suva, Fiji Islands. On average, over eight nights collecting, a mercury vapour light (MV) with manual capture
obtained approximately 14 times more individuals and five times more species than a white fluorescent light with
automatic capture and ultraviolet light with automatic capture. Of the 84 moth taxa recorded in total, only two were
not obtained by the MV trap, suggesting the moth assemblages obtained by the fluorescent light and ultraviolet
light methods were subsets of the larger MV collection. Using a bootstrap method to estimate the total species
collected after successive nights sampling, we found that after four nights almost 90% of the predicted total moth
species would be obtained by the MV method. These results identify the MV method as a high-performing
technique to investigate nocturnal moth diversity in Fijian forests, and that a minimum of four nights sampling
with this protocol would produce reliable data for use in habitat evaluation.
Key words
Fiji Islands, Lepidoptera, mercury vapour, species curves, statistical power.
IN TR ODUCTION
Numerous studies have proposed the use of moth assemblages as
indicators of forest quality, habitat disturbance, degradation or
restoration success (e.g. Holloway et al., 1992; Kitching et al.,
2000; Fiedler & Schulze, 2004; Summerville et al., 2004;
Brehm, 2005; Lomov et al., 2006). Light trapping is generally
considered an efficient collection method for adult night-flying
moths, attracting large numbers of individuals and diversity of
species for a comparatively low collecting effort (Fiedler &
Schulze, 2004; Beck & Linsenmair, 2006). However, as with
many insect trapping methods, the composition of moth assemblages captured using light traps can be strongly affected by
the style of trap used and the type and power of the light source
(Taylor & French, 1974; Bowden, 1982; Canaday, 1987;
Muirhead-Thompson, 1991; Intachat & Woiwod, 1999; Fayle
et al., 2007; Jonason et al., 2014).
Common forms of artificial light sources include mercury vapour lamps, ultra violet, standard white fluorescent lights and
‘black lights’ (Robinson, 1975; Brehm, 2005). Not all moth species respond to light types and intensities to the same extent, and
thus light trapping can be considered to sample moths selectively
rather than randomly, and the catches a reflection of activity and
attraction as well as moth abundance (Bowden, 1982; Butler
et al., 1999; Merckx & Slade, 2014). However, although they
*simon.hodge68@gmail.com
© 2016 Australian Entomological Society
might not provide a true measure of relative abundance, light
traps can still legitimately be used to compare and monitor the
moth assemblages at different sites and at different time of year
if protocols are standardized (Wolda, 1992; Beck & Linsenmair,
2006). Furthermore, the constraints of light trapping are shared
with many other insect sampling methods such as baited traps
and ‘sugaring’ (Southwood & Henderson, 2000; Fiedler &
Schulze, 2004).
The mechanism by which moths are captured can also influence which species and how many individuals are obtained by a
trapping event (e.g. Axmacher & Fiedler, 2004). Manual trapping involves the hand collection of moths attracted to the light
source which land on white sheets or gauze towers placed beside
the light. Automatic traps generally include a ‘bucket’ or holding
container beneath the light source and often contain a chemical
agent which sedates the moths and helps restrain them until
eventual collection.
In addition to consideration of sampling method, it is also important to quantify the sampling effort required to obtain relatively complete information on the diversity of insect
communities (Morrison, 2007; Christensen & Ringvall, 2013;
Molloy et al., 2013). Recently, Leather et al. (2014) suggested
that some of the major flaws in insect conservation and biodiversity research may be associated with inadequate sampling protocols (see also Jennions & Moller, 2003). Numerous
extrapolation and rarefaction methods are now available to
examine how species number accumulates with sampling
effort (Colwell et al., 2004). Similarly, a high frequency of
doi: 10.1111/aen.12209
2
S Tikoca et al.
singletons in insect collections can indicate that the community
has been under-sampled (Novotny & Basset, 2000; Coddington
et al., 2009). These mathematical methods can be used to
establish the minimum sampling effort required to record a
certain fraction of the predicted total species present: as the
inventory of species approaches completion, the species
accumulation curve should reach an asymptote and the number
of singletons should tend towards zero as additional samples
are obtained.
Although the statistical considerations of an insect sampling
regime are undoubtedly important, practical and economic
compromises are often equally important when designing an
actual scheme of study. Although it is usually desirable to
include as many replicate samples as is possible, this will be
limited by access to technical resources (e.g. light traps; storage
facilities), labour availability, travel costs and the time available
for sample processing. Consequently, it is important to
‘optimize’ sampling effort in terms of obtaining sufficiently
high-quality estimates of the parameters of interest without
performing excessive sampling that may only slightly improve
the quality of the final data set (Thomas & Thomas, 1994;
Jonason et al., 2014).
The moths of Fiji have received only periodic attention:
Robinson (1975) provided a comprehensive checklist of species
and distribution of moths in Fiji and Rotuma, which has since
been updated by Evenhuis (2013), and an online resource containing more recent Fijian moth records has been made available
by Clayton (2004). Intensive field work on the moths of Fiji’s
forest systems is currently being planned by researchers at The
University of the South Pacific, and it was therefore desirable
to systematically assess the efficacy of different trapping
methods and propose a standardized sampling effort that could
be used in these subsequent studies. Thus, the specific aims of
the current investigation were (i) to assess how the moth
assemblage of a lowland mixed forest near Suva was characterized by three commonly utilized light-trapping methods:
ultraviolet light with automatic-capture, white fluorescent light
with automatic-capture and a mercury vapour light combined
with manual-capture and (ii) to estimate the number of samples
required to obtain a meaningful proportion (i.e. 90%) of the
total number of species present at the site. By taking into account
the variation in daily catches, a simplistic power analysis
was used to determine the minimum detectable differences in
moth faunas (abundance and species richness) given a certain
number of samples.
M A T E R I A L S AN D M E T H O D S
The study was conducted in a forest reserve situated in the National Forest Park at Colo-i-Suva, located 8 km from Suva
( 18.034 and 178.432). This forest consists of a mixture of
healthy lowland rain-forest and plantings of exotic mahogany
which now contain a broad range of native species at various
growth stages in the understorey (Tuiwawa & Keppel, 2013).
During sampling, all light-trapping stations were situated along
paths in the forest and avoided densely covered forested areas.
© 2016 Australian Entomological Society
Moth abundance and species richness were investigated
using three light-trapping methods: two using automatic traps
and one involving a hand collecting protocol. The two automatic
traps consisted of a polypropylene bucket, smooth-surfaced
polypropylene vanes, light tube frames and a rain drain. One of
the automatic traps was equipped with a 15 W Sylvania
UV/blue tube light and the second with a standard Phillips
15 W white fluorescent tube light. Each automatic trap was
powered by a 12 V portable battery. The automatic traps were
emptied at the end of each collecting period.
The light trap used for the manual collecting consisted of a
125 W mercury vapour lamp powered by a Yamaha portable
generator. A 2 × 2 m white sheet was spread out and secured onto
nearby trees or branches, and all moths that settled on the white
sheet were collected directly into jars charged with ethyl acetate.
All three traps were used on eight nights between October
2011 and February 2012 when weather was forecast to be fine
(low wind and no rain). The traps were operated simultaneously
for 4 h between 18:30 (around dusk) and 22:30 h. The three light
traps were set up approximately 200 m apart to prevent betweentrap interference, and the positions of the traps were rotated over
the collecting nights to avoid confounding trap type and location.
Specimens were generally assigned to species level by reference to keys, images and nomenclature provided by Robinson
(1975); Clayton (2004); CSIRO (2011); and Evenhuis (2013),
following the family-wise taxonomy given by Zahiri et al.
(2012). Individuals of the large genus Cleora were not identified
to species level, with ‘Cleora spp.’ treated as a single taxon for
the purposes of this study. Similarly, apart from two conspicuous
species (Locastra ardua Swinhoe and Botiodes asialis Guenée),
the Pyralidae and Crambidae were also treated as a single familylevel taxon. These distinct taxa at various levels of nomenclature
are generally referred to as ‘species’ for ease of language
throughout this paper.
Statistical analysis
All statistical analyses were performed in GENSTAT v15 (VSN
International Ltd, Hemel Hempstead, UK). To compare the total
number of individuals and species caught in each trap, analysis
of variance was performed treating trap-type as a fixed factor
and sampling night as a random ‘blocking’ factor. Trap types
were then compared in a pairwise fashion using Tukey’s test.
The abundance data and species richness data were transformed
as log10(x) prior to analysis to help meet the assumptions of the
analysis of variance model.
To compare how total accumulated species richness was related to the number of sampling nights performed with each trap
type, a bootstrap procedure was performed using 1000 random
permutations (with replacement) of the raw catch data obtained
from the eight trapping nights. These permutations were then
used to calculate the mean predicted species richness obtained
for each trap type for each additional number of trapping nights.
The ‘power’ of statistical tests to detect differences between
groups is positively related to the sample size but negatively related to the variance of the data (Johnson et al., 2008). When
comparing groups of samples statistically, the minimum
Light-trapping methods for Fijian moths
acceptable ‘power’ of the statistical test used (i.e. the probability
of the test detecting a significant result when a difference actually
occurs) is often set at 0.8 (or 80%). Conversely, the probability
of making a Type II statistical error, that is, concluding there is
no statistical difference between groups when one actually exists
is equal to one minus the power of the test (i.e. 1 0.8 = 0.2).
Given that the sample size is fixed by the researcher, and the variance of a sample is known once the data are collected, or can be
estimated from prior studies, the minimum detectable difference
between groups for a test with acceptable power of detection
(e.g. 80%) can then be calculated (Bloom, 1995).
To estimate the minimum detectable differences (MDDs) between two samples of moths, we performed a simplistic power
analysis utilizing the means and standard deviations (SD) for
abundance (N) and species richness (S) we obtained from the
MV data. Calculating the coefficient of variation as CV = SD/
mean × 100%, and assuming this is constant regardless of sample
size, the power analysis investigated how MDDs were related to
sample size based on a two-sample t-test (two-tailed), with statistical power set at 0.8 and significance probability set at P < 0.05.
The test assumed the variances and sample sizes for both samples were equal, and then the sample size was increased in a stepwise manner from 2 to 16. The MDDs between the two samples
were obtained for raw data and log10 transformed data and
expressed as a percentage of the lower mean.
RESULTS
Trapping methods
Over the eight nights of sampling, a total of 1028 moths were
collected, with 10 families and 84 species being recorded
(Table 1; Appendix). On average, the hand sampling with the
MV lamp caught more individuals (F2,14 = 22.8; P < 0.001)
and species (F2,14 = 16.9; P < 0.001) than both the UV and FL
3
Fig. 1. Moth abundance (individuals) and species caught per sampling night (mean ± SE; n = 8) in Colo-i-Suva forest, Fiji, using three
collecting methods for 4 h from dusk. (MV, mercury vapour light
with hand sampling; FL, autosampler with fluorescent lamp; UV,
autosampler with ultraviolet lamp).
automatic traps (Fig. 1; Table 1). There was no statistically
significant difference in the catches obtained by the two types
of automatic traps (Tukey test, P > 0.05).
All 10 moth families were recorded by the MV trapping,
whereas the FL and UV trap only obtained six families each
(Table 1). Of the 84 species collected overall, only two (the
erebids Adetoneura lentiginosa and Simplicia cornicalis) were
not captured in the MV traps (Fig. 2; Appendix) [although both
of these species were captured by the authors using the MV trap
during subsequent studies]. The UV traps obtained a total of 28
species, 27 of which were also recorded by the MV trap. Similarly, the FL traps collected a total of 22 moth species, 20 of
which were also encountered at the manual MV trap (Fig. 2).
An individual based rarefaction procedure using the entire moth
collection obtained by all methods over the whole 8 days
collecting, indicated that 27 species were expected to be obtained
by the FL trap based on a collection with 56 individuals, and 33
species by the UV trap based on a collection of 83 individuals.
These predicted values are both higher than the actual species
collected of 22 and 28, respectively, and suggests that the low
Table 1 Summary of total moth catches in Colo-i-Suva Forest
Reserve obtained by eight nights of sampling using three different
light-trapping methods
Individuals
Species
Family
MV FL
UV
Total MV
Pyralidae + Crambidae
Geometridae
Erebidae
Noctuidae
Nolidae
Thyrididae
Uranidae
Limacodidae
Yponomentidae
Sphingidae
Total
Singletons
Singletons (%)
297
213
172
114
26
23
18
16
8
2
889
–
–
12
17
31
14
2
0
0
7
0
0
83
–
–
324
242
225
131
28
23
18
26
9
2
1028
–
–
15
12
22
3
0
0
0
3
1
0
56
–
–
3
24
28
11
4
3
1
6
1
1
82
6
7.3
FL
UV Total
2
2
7
7
9
10
1
4
0
1
0
0
0
0
2
4
1
0
0
0
22
28
10
10
45.5 35.7
3
24
30
11
4
3
1
6
1
1
84
–
–
MV, mercury vapour with hand sampling; FL, autosampler with fluorescent lamp; UV, autosampler with ultraviolet lamp.
Fig. 2. Venn diagram illustrating distribution and overlap of 84
moth species recorded using different light-trapping methods (MV,
mercury vapour light with hand sampling; FL, autosampler with
fluorescent lamp; UV, autosampler with ultraviolet lamp).
© 2016 Australian Entomological Society
4
S Tikoca et al.
species number obtained by these methods could be more than a
simple reflection of the low numbers of individuals caught.
Overall, the results indicate that the moths obtained by the UV
and FL automatic traps did not offer an alternative assemblage
of species to that obtained using the MV light, but rather they
obtained subsets of the more substantial collection obtained by
the latter method.
catches by the UV and FL traps would yield only 38% and
35% respectively of the total moth fauna. The bootstrappredicted catch using the MV trap after three sampling nights
was 68 species (83% of total species recorded) and by four nights
74 species (90% of total species recorded). Thus, a sampling regime of four sampling nights with the MV light could be
expected to obtain a sufficiently high proportion of moth species
whilst avoiding excessive sampling effort.
Sample size and species richness
The proportions of species represented by singletons in the total
collections made by the FL and UV traps were high, 45.5% and
35.7% respectively, suggesting these methods were grossly
under sampling the moth assemblage present (Table 1). However, the proportion of species represented by singletons in the
collections made by the MV light was only 7.3%, suggesting that
the sampling program using this technique was much more
complete.
From the bootstrap analysis, asymptotic curves for each trapping method were fitted for the predicted species catch obtained
by each extra sampling event, of the form:
S ¼a
b:r samples
where, S = (predicted) species recorded, a = b = asymptote and
r < 1, so that the curve was constrained to the zero intercept
and a represents the maximum number of species that would
be collected given ‘infinite’ sampling. The asymptotes from the
curves obtained suggest that species available to be collected
by the UV trap would be 31, by the FL trap 29 and for the MV
method 81 (Fig. 3). The fitted curve for the MV data was more
affected by constraining to the origin than those for the MV
and FL data; without constraining, the ‘best fit’ curve produced
a slightly higher asymptote of 83 species.
If the actual catch using the MV method of 82 species is taken
as our ‘best estimate’ of the available species in the forest (in this
season) (Table 1), then even after eight samples, the predicted
Fig. 3. Predicted number of moth species collected by three lighttrapping methods estimated using a bootstrap procedure (1000 permutations of raw data). (MV, mercury vapour light with hand
collecting; FL, fluorescent light with automatic collecting; UV, ultraviolet light with automatic collecting). Trend lines are based on
asymtotic regression equations based on values estimated from the
bootstrap analyses.
© 2016 Australian Entomological Society
Sample size and minimum detectable difference
On average, 143.5 individuals and 39.8 species were caught per
night when hand sampling with the MV light was used (Fig. 1).
However, the catch was highly variable between nights; the
number of individuals captured ranged from 53 to 517, and species number ranged from 20 to 62 per night. This high variability
would have a negative effect on the power of any statistical tests
used to detect significant differences between groups of moth
samples, for example, when comparing habitats, sites or times
of year.
If considering raw data for moth abundance, the mean of
143.5 and SD of 154.2 produced a CV of 107.5%. The MDDs
between two groups using this CV value for a given number of
samples tend to be substantial (Fig. 4). Considering the use of
four MV-light samples (as advocated previously for obtaining
90% of available moth species), the estimated MDD between
two groups (using a power of 80% detection) was 255.9%
(Fig. 4). For species richness, the mean of 39.8 and SD of 15.9
produced a CV of 40%, giving a MDD of 95.2% with the sample
size of four. Thus, if using raw moth sampling data, to achieve
the desired power of detection with sample size of four,
then > 3.5 times the number of individuals and ≈ 2 times the
number of species must occur in one group compared with the
other to be likely to produce a statistically significant outcome.
Fig. 4. Minimum detectable differences (MDD; % of smallest
mean) estimated from power analysis using means and standard deviations (SD) of the raw and log(10) transformed data obtained from
eight nights sampling with the MV light. Power analysis was based
on comparison of two hypothetical moth samples using a two-tailed
two-sample t-test, with power set at 80% and significance probability set at P = 0.05. MDDs for the log-transformed data have been
back transformed to be in same units as MDDs for raw data. (N,
abundance; S, species richness).
Light-trapping methods for Fijian moths
For the log-transformed data, the CVs obtained for the abundance and species richness were much smaller than their counterparts calculated from raw data: 15.2% and 12.4%, respectively.
When performing the power analysis using these CV values,
and then back transforming the log-MDD obtained to represent
the MDD on the same scale as the raw data, the effects on the
MDD of N and S are quite different (Fig. 4). Because the CV
of N has decreased considerably (107.5% to 15.2%), the MDDs
required to provide a statistically significant result were generally
much lower than when using the raw data: if the four sample
comparison is utilized the MDD dropped from 255.9% to
130.1% (Fig. 4). However, for the species richness data, the
effect of transforming the data was much less dramatic; the
four-sample MDD actually increasing slightly from 95.2% in
the raw data to 97.2% with the log transformed data (Fig. 4).
Only with a sample size ≥ 5 did transforming the data result in
a reduction in the MDD for species richness.
Although the MDD naturally decreased for abundance (N),
species richness (S), and the log-transformed data, as sample size
was increased, for sample sizes ≥ 6 the MDD for each variable
tended to level off (Fig. 4). Thus, with regard to increasing the
power of a statistical test, there would be a tendency for
diminishing rewards if undertaking any additional sampling
effort beyond that point.
D I S C U S S IO N
Trapping methods
The results of this study indicate that, among the sampling
methods we compared, the most successful method for
collecting nocturnal Fijian forest moths was the MV light with
hand sampling. Furthermore, if this method is to be used for
ecological monitoring, then a sampling regime of four sampling
nights with a 4 h collecting period could be expected to return a
reasonable proportion of the total moth species present. Our results are similar to a number of previous studies where the
strength and type of light source influenced the abundance and
diversity of moth catches. Jonason et al. (2014) reported that a
powerful 250 W MV lamp collected approximately 50% more
individual moths and 10% more species than a 40 W UV lamp
during long-term monitoring of farmland moths in Germany.
Merckx and Slade (2014), using mark-recapture methods, found
that very weak light traps (6 W) had narrow ranges of attraction
for moths (<50 m) and that only a small fraction of released
individuals were retrieved. So, more powerful light sources can
lead to a larger number of species responding to the trap and
attract specimens from a further distance, and these are likely
major factors in the MV method obtaining more individuals
and species in our results (Yela & Holyoak, 1997).
Jonason et al. (2014) and Bates et al. (2013) also reported that
MV lamps attracted more individuals and species when compared with other types of lamps tested, as did Fayle et al.
(2007), although in the latter study, the compositions of the assemblages obtained by the distinct lights were very different.
In our study, almost all (97%) of the species collected by the
5
UV and FL lights were also obtained by the MV sampling,
strongly suggesting that the UV and FL lights were providing
small subsets of the MV catch rather than collections with a distinct faunal composition.
The hand collecting used with the MV light, although more
painstaking and labour intensive, resulted in a higher proportion
of the attracted moth specimens being collected. Axmacher and
Fiedler (2004) also found that hand collecting from a ‘light
tower’ produced 10-times more geometrid specimens and twice
as many species, even when the light sources were the same
(15 W blacklight) in each case. Similar results were reported by
Brehm and Axmacher (2006), but in this case the improved catch
obtained with hand sampling was dependent upon which moth
family was considered and which habitat was being sampled.
With the automatic light traps, moths were frequently observed
to fly towards the light source and land on the plastic vanes of
the trap but then did not actually fall into the collecting bucket
and were thus not recorded as part of the catch (see also Kitching
et al., 2000). We also found that direct hand collecting of moths
produced good runs of undamaged specimens that simplified the
subsequent sorting and identification processes. The hand
collecting method also had an additional benefit in that rare species which can be identified in the field can be recorded without
the need of killing excessive numbers of individuals.
There are other practical considerations in deciding which
trapping method is feasible in any given study. Obviously,
methods involving hand collecting will require collectors to be
present throughout the sampling period and at each location,
meaning that labour costs tend to be higher than when using automatic collecting methods. Also, when very powerful lamps are
used, these often need to be operated using a portable generator
(as opposed to smaller 12 V batteries). These generators tend to
be expensive and can require considerable man power to transport if study sites are in isolated and inaccessible areas. Less
powerful lamps can be powered by portable batteries, which tend
to be cheaper, are smaller and lighter, and thus more convenient
in terms of transportation to and from the study site. Also, because of their reliance on cheaper portable batteries, automatic
traps can be more readily replicated amongst or within study
areas.
We concede that our study compares three distinct sampling
methods and that the effects of important experimental factors
such as light type, light intensity, mechanism of specimen
collecting and labour time/costs have not been examined separately. This confounding of factors meant that the most successful method, the MV-hand collecting combination, also had by far
the most powerful lamp, used the most expensive and labourintensive power source, and required the constant attention of
collectors throughout the sampling period. However, given the
magnitude of the increases in individuals and species obtained
by using this system compared with the automatic samplers,
and that the UV and FL lights did not offer alternative assemblages to that collected by the MV light, this extra effort and
use of resources appears well justified. Also, there is a need to
perform studies such as this one so that moth assemblages obtained by different mechanisms can be compared directly and
that cross-calibration of methods can be performed in order to
© 2016 Australian Entomological Society
6
S Tikoca et al.
help understand discrepancies in moth assemblages that occur
among different studies (Brehm & Axmacher, 2006)
Sampling effort
In terms of obtaining a good representation of species present in
the forest, the bootstrap analysis indicated that undertaking only
four nights sampling using the MV method would obtain around
90% of the estimated total species of moths occurring at this
locality. Extra sampling would be highly likely to result in
additional species being recorded but these new species would
be added to the inventory at a much lower rate. When taking into
account these diminishing rewards, and the associated costs of
extra nights spent in the field and the time required to sort and
process the specimens collected, samples in excess of four
appear to be unwarranted in terms of enhancing the species
count.
The use of moth assemblages as bioindicators or for ecological monitoring often involves groups of samples from different
habitats, or from different stages or methods of ecological restoration, being compared using typical statistical significance tests
such as t-tests or ANOVA. When studies involve numerous
collecting sites, habitats or methods, often combined in a factorial design, the workable sample number ‘per treatment’ can be
limiting. For example, four nights sampling per site was used
by Lomov et al. (2006), six nights per habitat per season was
used by Kitching et al. (2000) and the trap comparison paper
of Fayle et al. (2007) also used six nights sampling per treatment.
However, the power analysis we performed based on moth
samples from the MV-method indicated that, if this number of
samples was used, considerable differences would be needed
between groups before these differences would likely be found
statistically significant.
The power of a statistical test can be increased by proposing a
one-tailed rather than two-tailed hypothesis if the research situation allows. Similarly, log-transformation is known to provide a
number of advantages with the form of count data obtained from
insect sampling studies of this kind (although see O’Hara &
Kotze, 2010). After log-transformation, the data tend to become
less positively skewed and the relationship between the mean
and variance is removed to some extent. These effects improve
the likelihood of the data meeting the conditions of the ANOVA
model regarding normal errors and homogeneity of variance,
both of which can increase the test’s power of detection in some
circumstances. With our data, transforming using log(10) resulted in lowering the CV% of abundance and species richness
values, meaning that smaller differences between groups would
be detected as statistically significant. However, the effect sizes
expressed in the transformed state must be ‘back transformed’
to be readily understandable in natural units. For example, an increase of 17.6% in the log-transformed data would indicate an
increase of 50% in the raw data, and an increase of 30% in the
log-transformed data would indicate an increase of 100% (i.e. a
doubling) in the raw data (Quinn & Keough, 2008). In our data,
although transforming the data had a considerable effect on reducing the size of the MDD for abundance, there was little effect
observed on the MDD for species richness. These examples
© 2016 Australian Entomological Society
reaffirm that power analyses used to determine adequate sample
sizes based on raw data and log-transformed data are not interchangeable, and decisions regarding transformation of the data
should preferably be a priori rather than after the data have been
collected (Chow & Liu, 2008; Quinn & Keough, 2008).
When adopting typical parametric statistics based on a normal data distribution, the results of our study would advocate
the routine log-transformation of, at least, the raw data for abundance. We acknowledge that the use of alternative nonparametric comparison tests would be more appropriate if the
data assumptions of a parametric test are strongly violated,
although non-parametric tests tend to lack power compared with
their parametric equivalents. For example, using a Mann–
Whitney test to compare two groups of four samples requires that
the sample ranks are totally non-overlapping to give a statistically significant result. Simplistic power analyses for nonparametric tests appear to be lacking, but the use of
bootstrapping or jack knife methods can indicate the likelihood of a significant result being found given certain replicate
numbers (see Mumby, 2002). Similarly, the analysis of insect
count data using generalized liner models (GLMs) and generalized linear mixed models (GLMMs), where a more suitable
non-normal data distribution is assumed, can be employed
where parametric tests appear inappropriate. Again, although
these techniques appear to lack readily available power analysis, simulation methods are providing a means of assessing
the ability of different tests to detect a significant result for
given sample sizes, variability of the data and effect sizes
(Walters, 2004).
CO NC LUSI ON S
The results of this assessment indicate that, in the Fijian forest
habitat we examined, moth assemblages are best sampled by
attracting moths using a 125 W mercury vapour light and hand
collecting specimens from a white sheet. Depending on the
needs of subsequent research, four nights sampling using this
MV-method in fine weather should provide a high proportion
of the species present at any one time, whereas a sample size
of six nights should provide adequate statistical power to detect
real differences in abundance and species richness if the raw data
are log-transformed prior to the analyses being performed. Further research is required to isolate the effects of light type, bulb
Wattage and specimen capture method on the moths obtained
in similar and other types of Fijian forests and at different time
of the year.
REFERENCES
Axmacher JC & Fiedler K. 2004. Manual versus automatic moth sampling at
equal light sources–a comparison of catches from Mt. Kilimanjaro. Journal of the Lepidopterists’ Society 58, 196–202.
Bates AJ, Sadler JP, Everett G et al.. 2013. Assessing the value of the
Garden Moth Scheme citizen science dataset: how does light trap
type affect catch? Entomologia Experimentalis et Applicata 146,
386–397.
Light-trapping methods for Fijian moths
Beck J & Linsenmair KE. 2006. Feasibility of light-trapping in community research on moths: attraction radius of light, completeness of samples,
nightly flight times and seasonality of Southeast-Asian hawkmoths
(Lepidoptera: Sphingidae). Journal of Research on the Lepidoptera 39,
18–37.
Bloom HS. 1995. Minimum detectable effects: a simple way to report the statistical power of experimental designs. Evaluation Review 19, 547–556.
Bowden J. 1982. An analysis of factors affecting catches of insects in lighttraps. Bulletin of Entomological Research 72, 535–556.
Brehm G. 2005. Diversity and community structure of geometrid moths of
disturbed habitat in a montane area in the Ecuadorian Andes. Journal
of Research on the Lepidoptera 38, 1–14.
Brehm G & Axmacher JC. 2006. A comparison of manual and automatic
moth sampling methods (Lepidoptera: Arctiidae, Geometridae) in a rain
forest in Costa Rica. Environmental Entomology 35, 757–764.
Butler L, Kondo V, Barrows EM & Townsend EC. 1999. Effects of weather
conditions and trap types on sampling for richness and abundance of forest Macrolepidoptera. Environmental Entomology 28, 795–811.
Canaday CL. 1987. Comparison of insect fauna captured in six different trap
types in a douglas–fir forest. Canadian Entomologist 119, 1101–1108.
Chow S-C & Liu J-P. 2008. Design and Analysis of Bioavailability and Bioequivalence Studies, 3rd edn, pp. 760. Chapman and Hall/CRC, London.
Christensen P & Ringvall AH. 2013. Using statistical power analysis as a tool
when designing a monitoring program: experience from a large-scale
Swedish landscape monitoring program. Environmental Monitoring
and Assessment 185, 7279–7293.
Clayton J. 2004. Moths in Fiji [Online]. The University of the South Pacific.
[Accessed 2 Feb 2012]. Available from URL: http://www.usp.ac.fj/index.php?id=8504
Coddington JA, Agnarsson I, Miller JA, Kuntner MZ & Hormiga G. 2009.
Undersampling bias: the null hypothesis for singleton species in tropical
arthropod surveys. Journal of Animal Ecology 78, 573–584.
Colwell RK, Chang XM & Chang J. 2004. Interpolating, extrapolating, and
comparing incidence-based species accumulation curves. Ecology 98,
2717–2727.
CSIRO. 2011. Australian moths online: a photo gallery. [Accessed 2 Feb
2012]. Available from URL: http://www.csiro.au/outcomes/environment/biodiversity/australian-moths
Evenhuis NL. 2013. Checklist of Fijian Lepidoptera. [Accessed 2 Nov
2015]. Available from URL: hbs.bishopmuseum.org/Fiji/checklists/
lepidoptera.html
Fayle T, Sharp RE & Majerus ME. 2007. The effect of moth trap type on
catch size and composition in British Lepidoptera. British Journal of Entomology and Natural History 20, 221–232.
Fiedler K & Schulze CH. 2004. Forest modification affects diversity (but not
dynamics) of speciose tropical pyraloid moth communities. Biotropica
36, 615–627.
Holloway JD, Kirk-Spriggs AH & Khen CV. 1992. The responce of some
rain forest insect groups to logging and conversion to plantation. Philosophical Transactions Royal Society London B 335, 425–436.
Intachat J & Woiwod I. 1999. Trap design for monitoring moth biodiversity
in tropical rainforests. Bulletin of Entomological Research 89, 153–163.
Jennions MD & Moller AP. 2003. A survey of the statistical power of research in behavioral ecology and animal behavior. Behavioral Ecology
14, 438–445.
Johnson SE, Mudrak EL, Beever EA, Sanders S & Waller DM. 2008. Comparing power among three sampling methods for monitoring forest
vegetation. Canadian Journal of Forest Research 38, 143–156.
Jonason D, Franzen M & Ranius T. 2014. Surveying moths using light traps:
effects of weather and time of year. PLoS ONE 9, e92453.
Kitching RL, Orr AG, Thalib L, Mitchell H, Hopkins MS & Graham AW.
2000. Moth assemblages as indicators of environmental quality in
7
remnants of upland Australian rain forest. Journal of Applied Ecology
37, 284–297.
Leather SR, Basset Y & Didham RK. 2014. How to avoid the top ten
pitfalls in insect conservation and diversity research and minimise
your chances of manuscript rejection. Insect Conservation and
Diversity 7, 1–3.
Lomov B, Keith DA, Britton DR & Hochuli DF. 2006. Are butterflies and
moths useful indicators for restoration monitoring? A pilot study in
Sydney’s Cumberland Plain Woodland. Ecological Management and
Restoration 7, 204–210.
Merckx T & Slade EM. 2014. Macro-moth families differ in their attraction to
light: implications for light-trap monitoring programmes. Insect Conservation and Diversity 7, 453–461.
Molloy PP, Evanson M, Nellas AC et al. 2013. How much sampling does it
take to detect trends in coral-reef habitat using photoquadrat surveys?
Aquatic Conservation: Marine and Freshwater Ecosystems 23,
820–837.
Morrison LW. 2007. Assessing the reliability of ecological monitoring data:
power analysis and alternative approaches. Natural Areas Journal 27,
83–91.
Muirhead-Thompson RC. 1991. Trap Responses of Flying Insects: The
Influence of Trap Design on Capture Efficiency. Academic Press,
London.
Mumby PJ. 2002. Statistical power of non-parametric tests: a quick guide
for designing sampling strategies. Marine Pollution Bulletin 44,
85–87.
Novotny V & Basset Y. 2000. Rare species in communities of tropical insect
herbivores: pondering the mystery of singletons. Oikos 89, 564–572.
O’Hara RB & Kotze DJ. 2010. Do not log-transform count data. Methods in
Ecology and Evolution 1, 118–122.
Quinn GP & Keough MJ. 2008. Experimental Design and Data Analysis for
Biologists, 3rd edn, pp. 537. Cambridge University Press, London.
Robinson GS. 1975. Macrolepidoptera of Fiji and Rotuma. E.W. Classey
Ltd, London.
Southwood TRE & Henderson PA. 2000. Ecological Methods, 3rd edn.
Blackwell Science, Oxford.
Summerville KS, Ritter LM & Crist TO. 2004. Forest moth taxa as indicators
of Lepidopteran richness and habitat disturbance: a preliminary assessment. Biological Conservation 116, 9–18.
Taylor LR & French RA. 1974. Effects of light-trap design and illumination
on samples of moths in an English woodland. Bulletin of Entomological
Research 63, 583–594.
Thomas AW & Thomas GM. 1994. Sampling strategies for estimating moth
species diversity using a light trap in a northeastern softwood forest.
Journal of the Lepidopterists’ Society 48, 85–105.
Tuiwawa SH & Keppel G. 2013. Species diversity, composition and the regeneration potential of native plants at the Wainiveiota Mahogany
Plantation, Viti Levu, Fiji Islands. The South Pacific Journal of Natural
and Applied Sciences 30, 51–57.
Walters SJ. 2004. Sample size and power estimation for studies with health
related quality of life outcomes: a comparison of four methods using
the SF-36. Health and Quality of Life Outcomes 2, 26.
Wolda H. 1992. Trends in abundance of tropical forest insects. Oecologia 89,
47–52.
Yela JL & Holyoak M. 1997. Effects of moonlight and meteorological factors
on light and bait trap catches of noctuid moths (Lepidoptera: Noctuidae).
Environmental Entomology 26, 1283–1290.
Zahiri R, Holloway JD, Kitching IJ, Lafontaine JD, Mutanen M & Wahlberg
N. 2012. Molecular phylogenetics of Erebidae (Lepidoptera,
Noctuoidea). Systematic Entomology 37, 102–124.
Accepted for publication 3 April 2016.
© 2016 Australian Entomological Society
8
S Tikoca et al.
(CONTINUED)
A P PE N D I X I
Moth catches in Colo-i-Suva Forest Reserve obtained by eight
nights of sampling using three different light-trapping methods.
Trapping was carried out between October 2011 and February
2012.
Family
Erebidae
Geometridae
Limacodidae
Species
MV
UV
FL
Total
Achaea robinsoni
Adetoneura lentiginosa
Aedia sericea
Bocana manifestalis
Calliteara fidjiensis
Cosmophila flava
Dysgonia duplicata
Dysgonia prisca
Ericaea inangulata
Ericaea leichardtii
Eudocima fullonia
Hydrillodes surata
Hypena rubrescens
Hypenagonnia barbara
Hypenagonnia catherina
Hypenagonnia diana
Hypenagonnia emma
Hypospila similis
Mecodina variata
Mocis frugalis
Oenistis delia
Oxyodes scrobiculata
Palaeocoleus sypnoides
Polydesma boarmoides
Progonia sp.
Rhesalides curvata
Rusicada vulpina
Serodes mediopallens
Simplicia cornicalis
Thyas miniacea
Agathia pisina
Perixera gloria
Perixera lautokaensis
Perixera lautokaensis
Perixera monetara
Perixera oblivaria
Bulonga philipsi
Chlorochaeta cheromata
Chloroclystis nina
Cleora sp.
Episteira nigrilinearia
Gelasma stuhlmanii
Horisme chlorodesma
Luxiaria sesquilinea
Mnesiloba eupitheciata
Petelia aesyla
Pseuderythrolopus bipunctatus
Pyrrhorachis pyrrhogona
Scardamia eucampta
Thalassodes chloropis
Thalassodes figurata
Thalassodes pilaria
Thalassodes fiona
Thalassodes liquescens
Beggina albafascia
Beggina mediopunctata
Beggina minima
Beggina unicornis
2
0
2
10
19
2
2
13
7
6
2
9
2
3
1
3
5
2
2
12
12
17
5
9
1
2
10
2
0
10
7
3
3
2
12
9
12
6
2
92
1
6
7
7
3
2
2
9
4
9
2
4
4
5
2
1
8
2
0
0
0
6
1
0
0
0
0
0
0
6
0
0
0
0
1
1
0
1
6
0
6
2
0
0
0
0
1
0
2
0
0
0
8
0
0
0
0
2
0
1
1
0
0
2
0
0
0
0
0
0
0
1
0
1
3
0
0
2
0
0
1
0
0
0
0
0
0
2
3
0
0
0
1
0
0
5
2
0
0
1
0
0
0
0
5
0
0
0
0
0
2
0
1
0
0
4
0
0
0
0
0
0
0
0
0
1
1
0
2
1
0
0
2
0
2
2
2
16
21
2
2
13
7
6
2
17
5
3
1
3
7
3
2
18
20
17
11
12
1
2
10
2
6
10
9
3
3
2
22
9
13
6
2
98
1
7
8
7
3
4
2
9
4
10
3
4
6
7
2
2
13
2
Continues
© 2016 Australian Entomological Society
Family
Noctuidae
Nolidae
Pyralidae + Crambidae
Sphingidae
Thyrididae
Uranidae
Yponomentidae
Species
MV
UV
FL
Total
Beggina zena
Beggina sp.
Aegilia vitiscribens
Agrotis ipsilon
Athetis thoraicica
Chrysodesis eriosoma
Condica illecta
Leucania venalba
Sasunaga oenistis
Spodoptera litura
Spodoptera mauritia
Stictoptera stygia
Stictoptera vitiensis
Barasa triangularis
Earias flavida
Maceda mansueta
Maceda savura
Locastra ardua
Botyodes asialis
unknown
Daphnis placida
Banisia anthina
Banisia messoria
Striglina navigatorum
Urapteroides anerces
Atteva aleatrix
1
2
9
2
2
11
14
2
2
9
45
11
7
5
1
4
16
2
42
253
2
7
4
12
18
8
2
1
0
0
0
2
3
0
0
2
7
0
0
2
0
0
0
0
7
5
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
3
0
0
0
0
0
0
0
1
14
0
0
0
0
0
1
4
3
9
2
2
13
17
2
2
11
55
11
7
7
1
4
16
2
50
272
2
7
4
12
18
9