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