A lower developmental threshold of 10 °C was used to estimate growing degree days (GDD) (Pruess 1983). Weather data were measured at a weather station located at the OARDC about 8 kilometers south of the trapping site. The method used to calculate GDD was a modified sign wave method (Allen 1976; Pruess 1983) as outlined at http://www.oardc.ohio-state.edu/gdd/glossary.htm (viewed Jan 2009) and see also (Cardina et al. 2007). We recognize that many of the moths may have developmental thresholds different from 10 °C, but for consistency, we retain the base temperature of 10 °C even for those few species where sufficient research exists to justify a different base. The calculation for GDD in the OARDC site was based on English units, which were converted to metric using GDD in °C = -0.00013+0.555639* GDD in °F (F=57017608; df 1,363; P<0.0001). We used the single triangulation method in cases where we needed to recalculate GDD (Lindsey and Newman 1956), and note that there tends to be close agreement between the various sine and triangulation methods (Roltsch et al. 1999). The use of a fixed threshold temperature for different species has been used to model voltinism changes in Finnish moth species (Poyry et al. 2011).

We used the various approaches to estimating species richness implemented in EstimateS (Colwell 2013) set to run 1000 randomizations without replacement. We calculated species richness using both individual sampling dates and yearly pooled samples. However the difference between the estimates was less than the estimated standard deviation for either method. Therefore we only present results using individual sampling dates.

We used the proportion of species represented by a single capture as an indication of the effectiveness of the sampling protocol (Carlton et al. 2004; Williams et al. 2007). This approach assumes that no viable moth population can be represented by a lone individual, so the capturing of only a single individual indicates that the method missed some individuals. Although some singletons are indicative of an ineffective sampling methodology, e.g., moth species that do not come readily to light, some singletons should be expected since they could come from migrating individuals that have little interest in the trap or its environment.

The study site had bats, birds, and wasps that preyed on moths attracted to the light. There may also have been other vertebrate and invertebrate predators. Moths were released in different locations in the yard to reduce such predation. However, we could not quantify the level of predation or the effectiveness of any effort at reducing predation. Sometimes moths were too worn to be properly identified, and these individuals were ignored.

We expect that three traps run six times per year for one year (Summerville and Crist 2003) would have less impact on the local ecosystem than would one trap run at the same location 115 to 215 times per year for nine years (this study). Long term intensive sampling has shown the potential to negatively impact moth populations (Vaisanen and Hublin 1983). Consequently moths were released after identification. This methodology may inflate abundance estimates, though it would not affect the number of observed species. So we address the issue of multiple captures internally using the frequency of doubletons, and externally by comparing with the published literature.

Quantitative assessment of the effect of multiple captures was made by examining the number of moth species captured once per year versus the number represented by two captures per year. A methodology that increased the probability of recapturing moths should have a disproportionate number of rare species captured twice. The average doubleton÷singleton ratio for each year was 0.574 (standard deviation [SD] of 0.211). We also look at this ratio for each sampling date because in this case doubletons cannot be recaptures of the same individual. The doubleton÷singleton ratio for each night where there were both singletons and doubletons was 0.382 (SD 0.336). The yearly average was not significantly greater than the daily average (F=2.91; df=1, 980; P=0.09). Obviously, a failure to detect a significant difference is not the same as proving that there was no effect. Summerville et al. (2003) reported a ratio of 0.568 (SD of 0.054), whereas Summerville et al. (2004) found a ratio of 0.472 (SD of 0.092) (these numbers based on data provided by Dr Summerville from research published in cited literature). Lower values have been observed in other studies, 0.311 [Vancouver, Canada] (deWaard et al. 2009), 0.552 (SD of 0.247) [Rothamsted insect survey site 336, United Kingdom] (Harrington and Woiwod 2007), as well as higher values 0.932 [Blue Mts, Oregon] (Grimble et al. 1992), 0.618 [Birch Mts., Alberta, Canada] (Macaulay and Pohl 2005). A collection from Inverness Ridge in California had a value of 0.362 (data provided courtesy of Jerry Powell). A collection from Annville, Pennsylvania had a value of 0.222 (data provided courtesy the Pennsylvania Natural Heritage Program) (Ferster et al. 2008). The most comparable study would be the 31 years of data from Rothamsted because the data were yearly counts over multiple years from one locality. We conclude that our result of 0.574 is not unusual compared to these studies, and therefore the possibility of capturing the same individual twice doesn’t seem to result in an excessive bias in this study. However, we don’t know if we got lucky, or if this is a typical result.

The raw data are included as supplemental data. The data file is in Excel format. We recommend that users read the “Introduction”, which is the first page (left-to-right) in the file. The next page to the right in the file includes the weather data. Farther to the right are nine pages with yearly capture data. These pages include the number of growing degree days accumulated by each collection date. Cells are blank if no individuals of a given species were captured on a specific date. Next is a page “Condensed List” that contains total number of each species, and the number of years each species was collected. This page contains the species as they were identified and the equivalent under the system by Lafontaine and Schmidt (2010). It also lists the range in capture date, and range in growing degree days. Then follows total captures per year, and a list of known host plants. Next there is a list of the 13 new county records and their collection date. Next is a page with a list of univoltine and bivoltine species selected based on abundance and environmental fidelity. Next is a listing of the 20 pest species and their yearly abundance. This was extracted from the main list to facilitate access. Lastly is a page with the moon phases. We did not find this of any use, but it may prove useful to someone else.

In 1290 sampling dates from 1 January 2001 through 31 December 2009, a total of 77,581 moths were captured and identified. This averages to 60 moths/night. However this number has little value because it includes early and late season samples that have few moths. In 2001 the range was from 1 to 496 moths per night with an average of 96. Within this nine year sampling effort were 501 species, of which 122 were found in all nine years.

The numbers of species within a family that were represented by a single capture has been used as a metric for evaluating the effectiveness of a sampling methodology. Carlton et al. (2004) reported that singletons accounted for 38 to 43% of their sample, and this was considered indicative of sufficient sampling effort. Using this criterion, the average singleton rate per year for the Thyatiridae, Drepanidae, Mimallonidae, Apatelodidae, and Saturniidae all indicate that the sampling strategy might be ineffective. Either the moths do not respond well to black light traps, or we may be sampling transients. The Noctuidae and Geometridae accounted for the majority of singletons (Table 2). However, relative to the number of species in these families, the number of singletons in these families was low, thereby indicating sufficient sampling effort. The Lasiocampidae, Sphingidae, Erebidae, Euteliidae, and Nolidae also had singleton percentages that were with acceptable limits. The estimated total number of species was between 533 for the Bootstrap method and 599 for the Jacknife 2 method (Table 3).

The number of species only present in a single year was greatest in 2001 (Table 4). At the other extreme, 2005 and 2006 had an unusually small number of species that were only captured in that year. For the first six years, the number of species never before captured declined (Table 4). Eventually it should converge to the speciation rate plus the immigration rate of new species. However, 2007 and 2009 were unusual years in that more new species were added than one would expect from the initial pattern. These two years therefore have a large influence on the estimated total number of species. There was no obvious pattern in the new species for 2007 and 2009. None of the species were pests. In 2007, four of the 15 species fed on oak, maple, or walnut, whereas six of the 15 species from 2009 had these hosts. Seven of the 15 species in 2007 were captured again in either 2008 or 2009. In 2003 there was an F2 tornado that went through the city. Another F2 was within a few miles of the city in 2009. Smaller tornados occur in Wayne County nearly every year (http://www.tornadohistoryproject.com/tornado/Ohio/2003/map). Another environmental disturbance happens twice per year as the city applies insecticide for mosquito control using a truck mounted fogger. None of these events seem related to patterns in our survey.

Given that we documented 501 species at this one location, one might suggest that this urban environment had greater macrolepidopteran diversity than 15 of the 19 North American sites in Table 1. However, this comparison is problematic. Sampling effort both within season and the number of seasons affect the number of species collected (Gotelli and Colwell 2001). The other studies used multiple traps in a variety of habitats, but usually did so over a shorter time span both in terms of the number of years sampled and in terms of the number of trapping nights per year. The type of black light trap may also have an influence (Schweitzer et al. 2011; Truxa and Fiedler 2012). Furthermore, many of the sampled habitats in the other studies might be more homogeneous than an urban landscape with corresponding influence on biodiversity (Fox et al. 1997; Krauss et al. 2003; McKinney 2002; Southwood et al. 1979), especially considering that small plantings of native vegetation augment botanical diversity and concomitant increases in moth biodiversity (Burghardt et al. 2008). Considering the large number of species found in this survey, we would agree with the idea that it should be possible for urban planning committees to design urban landscapes that support an abundant and diverse macrolepidopteran fauna (Ockinger et al. 2009; Valtonen et al. 2007), which might also improve the habitat for birds and other wildlife.

There have been three moth surveys near this survey. The Wilderness Center had the fewest number of shared species with our study (Table 5). The greatest similarity was in the Geometridae where 15% of the combined species were in common. The Atwood Lake Park survey and the Funk Bottoms survey were much more like our survey with a 28% or better overlap in species lists (average overlap 52%). If urbanization at the Wooster site had elevated diversity due to habitat fragmentation and colonization by a diverse assemblage of generalists, then one would expect that most of the species unique to the Wooster survey would be in the Noctuidae. However, this was the case only for comparisons with Atwood Lake Park where 46% of the Noctuid moths captured at Wooster were unique to Wooster, while only 29% of the Geometrids were. This is in contrast to Funk Bottoms where 58% of the Noctuids were unique to Wooster while 62% of the Geometrids were unique, and the Wilderness Center had 35% and 38% respectively. This outcome is inconsistent with the hypothesis that urbanization at this location has increased diversity by attracting more generalists at the expense of naturally occurring moths.

An alternative strategy to assess the value of this urban moth assemblage is to examine specific genera within the Noctuoidea that are not associated with typical urban vegetation. Larvae from moths in the Noctuid genus Schinia are mostly associated with plants in the Asteraceae and Fabaceae. Species in the genus Catocala are specialists on plants in the Fabaceae, Fagaceae, Rosaceae, Juglandaceae, Myricaceae, and Salicaceae. The genus Acronicta larvae feed on woody shrubs and trees, some are specialists. Larvae of moths in the genus Papaipema are borers in stems of plants in the Asteraceae and other weedy species. The tribe Psaphidini primarily feed on members of the Juglandaceae and Fagaceae with the exception of Copivaleria grotei (Morrison), which feeds on ash. The subfamily Herminiinae is a member of the family Erebidae, with larvae that primarily feed on senescent plant material (http://www.eeb.uconn.edu/people/wagner/USDA Noctuid Guide Most Current.doc). Table 6 compares the number of species within these groups collected at the various locations. There doesn’t appear to be any pattern. For example, in the Papaipema there were six species not recovered in this survey but present in one or more of the other surveys: P. lysamachiae Bird, P. rigida (Grote), P. rutila (Guenée), P. unimoda (Smith), P. marginidens (Guenée), P. nelita (Strecker), and P. birdi (Dyar). In the case of Turkey Run (E in Table 6) and Coopers Rock (F in Table 6), half of the species in each case were in common with this survey, and only two species were in common between the two cited works. No species of Papaipema was common to all locations, though P. inquaesita (Grote & Robinson) was only missing from the Funk Bottoms survey (D in Table 6). Finally, the Papaipema in this study are moderately abundant with records of 2 to 90 specimens in total over the nine year survey. The agricultural pest P. nebris (Guenée) ranked fifth in abundance within this group. The point is that there does not seem to be a pattern that would indicate that the Wooster moth fauna are lacking in species associated with non-urban environments. It is possible that the abundance of these species is lower in Wooster than in more natural settings. However, the other local surveys did not record abundance data.

A Whittaker plot showed no obvious difference in ranked abundance between any of the years (Fig. 1). A Whittaker plot for each month showed that September and October had the most even distribution, and evenness decreased on either side (Fig. 1). Alpha diversity was greatest in early August (Fig. 2). Abundance of all macrolepidoptera had three peaks (Fig. 3). The first and last peaks were caused by the emergence of abundant bivoltine moth species, whereas the central peak was caused by abundant univoltine species. The first peak was between 250 and 583 GDD, the second between 750 and 1111 GDD, and the third between 1334 and 1611 GDD. This corresponds to late June, early August, and late September.

Figure 1. 

Whittaker plots for both year (where month is ignored), and month (where year is ignored). The number in parentheses is the total number of sample days and total number of captures in that month from 2001 through 2009.

Figure 2. 

Seasonal and yearly change in a diversity. Bars show the 95% confidence interval. The top date for season was the average date for the midpoint, while the bottom dates give the range in month/day format.

Figure 3. 

Seasonal abundance by growing degree days and Julian Date.

We calculated the number of species that went missing from one year to the next expressed as a proportion of the number of species originally present (e.g., 100* number of species in 2001 not collected in 2002 divided by the total number of species collected in 2001). We also calculated the number of immigrants expressed as a proportion of the number of species present in the year of collection (e.g., 100* number of species in 2002 not collected in 2001 divided by the total number of species in 2002). The missing rate averaged 24.0% (standard deviation 0.0701) while the immigration rate averaged 23.9% (standard deviation 0.0632) from 2001 through 2009. This would suggest that the biodiversity in this area was relatively stable over this nine year period.

The Noctuidae had three peak abundances in the year, with the first peak ending at about 722 GDD (late June), a second peak from 722 to 1056 (early August), and the third peak from 1056 GDD onwards (Fig. 4). The first and third peaks were the most substantial, with the third peak containing about 1/3rd more individuals than the first peak. In contrast, Geometrid abundance gradually increased through July, and declined thereafter. The Geometridae lacked the sharp peaks seen in the Noctuidae (Fig. 4).

Figure 4. 

Seasonal abundance based on growing degree days for Noctuidae, and Geometridae.

The presence of seasonal patterns has been documented previously, though the specific pattern may be unique to a specific location (Szabo et al. 2007). Because diversity in both Noctuidae and Geometridae had a seasonal component, there was also a seasonal component to the ratio of these two groups (Fig. 5). Although the yearly average was nearly constant, the seasonal ratio ranged from 0.09 to 0.91. Thus a small mismatch in season could result in finding significant differences that are an artifact of seasonality interacting with an experimental design. Seasonal changes in biodiversity could introduce a potentially large inadvertent bias into biodiversity research that was based on a sampling a few dates each year (Summerville 2008). Our data provided a concrete example of this bias. We noted the peak in alpha diversity was in early August (Fig. 2). Yearly alpha diversity (Table 4) was less than peak seasonal alpha diversity (Fig. 2), but yearly alpha diversity would overestimate seasonal diversity through most of a season. For this reason one needs to know where sampling has taken place relative to the seasonal shift in alpha diversity if one is to make valid comparisons with similar studies. Otherwise one does not know if differences between studies represent ecological differences or a mismatch in seasonality. We note in Fig. 2, that if one is sampling at peak alpha diversity, then a shift of only two weeks could result in a significant difference. What might be required to cause such a shift? The current total GDD achieved by August 4 could be achieved by July 25 if every hourly observation was raised by 1.53 °C, or if both minimum and maximum daily temperatures were increased by 1.50 °C. Roughly, this is the equivalent of changing elevation by 153 to 416 m based on an environmental lapse rate of 3.6 to 9.8 °C/1000 m (Sheridan et al. 2010; Varmghani 2012). Alternatively, one could drive about 217 km closer to the equator, assuming a change of 6.9 °C/1000 km (Colwell et al. 2008; Jump et al. 2009). In the current context, this also means that we cannot determine how much of the difference between our results and the three previous surveys was due to the collection dates versus biological differences. Furthermore, the simple approach of first selecting only sampling dates that our study has in common with the other studies will not work to overcome this problem, although it would solve the problem of unequal sampling effort.

Figure 5. 

Seasonal and yearly fluctuation in the geometrid/noctuid ratio. Bars are one standard deviation from the mean.

A long term trapping effort is managed as the Hungarian Plant Protection and Forestry Light Trap Network (Szentkiralyi 2002; Szontagh 1975). Results from 55+ traps per year sampled from 1962 through 2006 were recently published (Gimesi et al. 2012). They showed three peaks in the number of captured individuals, although in their case the central peak was much larger than the other two. The Hungarian data had a broad peak in Fisher’s alpha corresponding to warmer summer months, and there was a strong relationship between mean daily temperature and biodiversity. This pattern was present in our data (Fig. 2), but in our results alpha had a distinct peak in early August. Seasonal shifts, multiple traps over a broad geographic range, and averaging results over longer time spans would tend to smooth out seasonal trends into a much broader peak in Fisher’s alpha. The Hungarian data showed distinct losses in both abundance and diversity over this period, but one could find nine year spans in their data where abundance and diversity increased (Szentkiralyi 2002). Furthermore, the Hungarian data showed that seasons have gotten earlier by about 2 to 3 weeks over a 44 year span (Gimesi et al. 2012). Based on these results, our inability to detect a similar trend in our data could be due to having only nine years of data.

We were interested in the difference between using a growing degree day model versus a calendar date. We selected 37 individuals with 350 or more captures in the nine year study, and calculated the average day of capture. For each species we divided the mean by the standard deviation, and used a paired t-test for a significant difference between using Julian day versus GDD (df 36; t=7.12; Pr>|t|<0.001). On average there was a 57% reduction in this ratio for GDD relative to using Julian Day (95% CI: 51.7 to 61.3%). Therefore, the GDD approach should significantly increase the statistical power of tests for treatment differences relative to using calendar date.

Looking at the number of Catocola, Acronicta, and species in the Herminiinae that we collected relative to surveys from less disturbed environments, we would conclude that our sample from an urban environment was not inflated by a large number of generalists attracted to the mix of exotics in the urban landscape. Therefore we would concur with others that urban landscapes can be planned to increase biodiversity relevant to more natural ecosystems (Ockinger et al. 2009; Valtonen et al. 2007). Our survey showed that there were three peaks in moth abundance, whereas biodiversity had a single peak late in the year. We also showed that moth biodiversity was relatively stable with nearly equal missing and immigrating species from year to year. There was also a regular progression of species throughout the year. The sequential gain and loss of species each month resulted in seasonal shifts in the Geometrid:Noctuid ratio such that it is unlikely to be generally useful as a single number describing habitat disturbance. Furthermore, the seasonality demonstrated in these data would suggest that any species ratio would need careful validation prior to use. In describing our data, we made use of a growing degree day (GDD) model. This has the effect of rescaling the data (see Fig. 3). It also corrects for temperature dependent shifts in moth biology. Consequently, some of the variability in the data was removed and this should improve the power of statistical tests involving survey data.

There is no end point to general surveys. No matter how many years of sampling, there will always be an additional species that can be added to the list if sufficient effort is expended. One reason for making such lists is that they provide quantifiable justification for maintaining a natural area to preserve biodiversity. In some cases a threatened local population is being preserved, and those individuals may be locally abundant. More commonly we are preserving rare species associated with a specific habitat. In this case, there is no end to the survey because it is not possible to identify all the species present at an instant in time nor is it possible to identify all the potential species that could live in that habitat. Partly this is a function of forces like climate change, but there are also changes in the spatial distribution of all plant communities. An end point might be reached if the survey goal is to identify those species visitors to the park are likely to encounter and ask “what is this?” In this survey there were 122 species encountered every year. Five species stop flying in May. Five only start flying in August, and three start flying in September. So one could ask how many years it would take to get all 122 species by sampling once per month from May through October. Sampling the first day of these six months will result in recovering an average of 63.2 of these 122 species in any one year. This sampling protocol will only recover 117 of these species in the nine years sampling took place. How does this answer change if we took two or three samples each month? What if we shifted the sampling dates by a few days? Another simple option is to choose the date with the most number of species for the year. In this study that date would fall between 24 July and September 1. The maximum number of species recovered on a single night averaged 51.1. Thus a simple sampling design has difficulty recovering species that we know are present every year. The required sampling effort increases greatly if one desires to go beyond a species list to an understanding of the underlying relationships between these ecologically important organisms.