Specimens
All of the trapped insects underwent species classification, though we paid special attention to macromoths, which are regarded as nuisance insects by visitors.
Experimental periods
Gypsy moth outbreaks in Hokkaido started in 2012, peaked in 2013 (see https://www.youtube.com/watch?v=BKeW-MtlZIs), and declined in 2014. Following the spread of nuclear polyhedrosis virus (NPV), commonly known as baculovirus, among larvae and pupae in 2014, outbreaks ceased from 2015 to 2019, a period referred to as the innocuous or endemic phase [27]. Hence, we were not able to collect sufficient numbers of gypsy moths at the Mt. Usu rest area in 2015 and 2016. In the early spring of 2017, we spotted gypsy moth egg masses on the a of an elevated expressway close to Yubari Interchange, and we conducted field experiments in the summers of 2017 and 2018 (Additional file 1).
Light trap surveys were conducted from mid-July to the end of September in 2014, 2015, 2017, and 2018. In this study, we paid special attention to the relationships between insects and weather parameters in 2014 and 2018, years for which daily counts of trapped insects were available.
Study sites
During an intense outbreak of gypsy moths that occurred in central Hokkaido (Tokachi district) in 2009, E-NEXCO set up a balloon-light trap for one night in the back area of the Ikeda Interchange (42°59′45.0″N 143°26′15.7″E, elevation: 23 m) (Figs. 1a and 2e,f). Since then, East Nippon Expressway Company Limited has conducted preinspections for gypsy moth egg masses and larvae from April to June each year. Accordingly, two sites alongside expressways were selected because they were moth-abundant sites. One was located on the coastline, and the other was located inland. The former was the back area of the Mt. Usu (Usuazan) rest area (Fig. 1a) along the outbound line (42°28′02.3“N 140° 54’31.6”E, elevation: 121 m), located approximately 200 m from commercial lighting (Fig. 1b). The latter was a hill on flat ground (42°54′55.8″N 141°58′02.0″E, elevation: 130 m) located beneath an elevated expressway (Additional file 1) near the Yubari Interchange (42°55′26.3″N 142°02′10.6″, elevation: 121 m), approximately 100 m from a branch of Yubari River (Fig. 1c). The site had no commercial light sources for 300 m in all directions and was surrounded by forest (Fig. 1c).
Configurations of light traps
In 2014 and 2015, we used “light tower” traps [6, 12]. Each trap consisted of a U-shaped white tarpaulin (thickness: 3 mm) with an array of fluorescent lights placed vertically in the middle (Fig. 3a, Additional file 2). Traps were set 2 m apart. Basins for catching insects that dropped from the tarpaulin were made inside and outside the base of the tarpaulin and filled with water (Fig. 3a). To ensure a nonadherent surface, a kitchen detergent was applied to both sides of the tarpaulin. All of the insects in the basins were collected manually from each trap the next morning, and their wet weights were subsequently measured.
To reduce manpower, a flight-interception LED trap, also referred to as an “automatic funnel trap” [6, 13], was used in 2017 and 2018 (Additional file 3). Each LED module was fixed in the center of a corrugated plastic board (thickness: 5 mm) that was positioned vertically against the collection box (Fig. 3b). A funnel was placed on the top of the collection box to prevent trapped insects from escaping (Fig. 3b). Traps were placed 1.5 m apart (Fig. 3c) and the distance between lights was set at 2 m (Fig. 3d). Electricity was provided by the main building or mobile batteries via long extension power cables. In all the experimental periods, the lights were turned on and off with a timer.
We rotated the location of the traps clockwise at each site on successive collection nights to ensure at least one night at each position. Unless otherwise stated, all specimens collected were removed from the traps after each collection night and taken to the laboratory for sorting and subsequent species identification.
The emission power spectra of each light and its illuminance (lx) were measured at 80 cm from the light using a spectroradiometer (CL-500A, Konica Minolta, Japan) and are presented as the average of 10 consecutive measurements. The emission intensity of the UV light range (290–390 nm) was measured by a UV intensity meter (UVK-40 M, JEFCOM, Japan). The detailed configurations of the light traps for each year are as follows.
2014 To evaluate the effects of commercial lights that emit broadband wavelengths on insect attraction, four “old-type” of fluorescent tubes were deployed in the back area of the Mt. Usu rest area (See Additional file 2 for specifications of lights.). The lights were 1) neutral white (Ra: 70) fluorescent lights (φ32.5 mm, 1170 lm × 5 = 5850 lm, FL20SS・N/18, Toshiba Lighting & Technology Corporation); 2) LED lights (φ32.5 mm, T8T-S562F50, Inaba Denki Sangyo Co., Ltd.); 3) UV-free lights (Ra: 90) (φ32.5 mm, 850 lm × 7 = 5950 lm, FL20S・N-SDL・NU, Toshiba Lighting & Technology Corporation), and 4) black lights (φ32.5 mm, 12 tubes, FL20S・BLB, Toshiba Lighting & Technology Corporation). To eliminate UV light, each of the lights was entirely covered with a polycarbonate transparent filter that blocks UV light (PDW-VF20, JEFCOM, CO., LTD., thickness: 2 mm). The intensity of total illumination for each trap was adjusted to approximately 6000 lm except for the black light (Additional file 2). The lights were turned on at 18:00 and turned off at 6:00 the following day.
2015 To evaluate the effects of colors of light, we used the following colored fluorescent tubes with UV-free filters (PDW-VF20, JEFCOM, Japan): 1) blue fluorescent tubes (FLR40S.EB/M. A, Toshiba Lighting & Technology Corporation), 2) turquoise-blue fluorescent tubes (FL20S・BW, Toshiba Lighting & Technology Corporation), 3) green fluorescent tubes (FL20S・G, Toshiba Lighting & Technology Corporation), and 4) neutral white fluorescent tubes (FL20SS・N/18, Toshiba Lighting & Technology Corporation). The tubes were deployed in the back area of the Mt. Usu rest area (See Additional file 2 for light specifications). For each trap, the intensity of illumination was adjusted to approximately 5000 lm (Additional file 2). The lights were turned on at 18:00 and turned off at 6:00 on the following day.
2017 and 2018 To evaluate the effect of UV light on insect attraction, LED modules that emit light from UV-A to visible-light wavelengths were deployed in the experimental site near the Yubari Interchange (see Additional file 3). Each LED module consisted of multiple spherical LED chips emitting light with wavelengths of 365, 375, 405, 430, 450, 470, 490, 505, 525, and/or 535 nm (LED-ON, Japan); the modules were set into LED bulb sockets arrayed on a specialized circuit board (Spectrolight SPL-25-CC, LED-ON, Japan), which was connected via an AC adapter to the outlet (AC100V) of a generator. The intensity of the total illumination for each LED module was adjusted to approximately 1500 lm (Fig. 3d). The light source was covered with transparent UV light-transmitting plastic for rain protection (Fig. 3b). The LED modules were turned on at 18:00 and turned off at 2:00 on the following day.
Measurements of meteorological factors
On each night of collection, ambient temperature, visibility range, and wind speed were automatically recorded every 5 minutes by data logging systems on the expressway near the light traps. During the periods when light traps were used, the overnight temperature ranged from 5.9 °C to 33.8 °C, and the wind speed ranged from 0 m to 22.2 m/s. The moon score was calculated as the summation of the weighted score of the moon size and the meridian passage of the moon (highest possible score: 10 points). The time of 21:00 h was regarded as the postsunset time when moths exhibit the highest degree of flight activity. Five points represented a full moon, and zero points represented a new moon. Another five points were assigned when the meridian passage of the moon occurred at 21:00, and zero points were assigned when the meridian passage of the moon occurred at 9:00.
Insect sorting and identification
In 2014, the insects trapped at the Mt. Usu rest area were mostly Lymantria dispar, and further species identification was therefore omitted. The mass weight of the moths was measured immediately after one collection night. In 2015–2018, the insects trapped in each trap were photographed and lightly dried in an oven (50 °C). We then transferred the specimens to plastic dishes and observed them with the naked eye. We sorted the insects by taxonomic order with guidance from the River Environmental Database supplied by the Ministry of Land, Infrastructure, Transport and Tourism of Japan (http://www.nilim.go.jp/lab/fbg/ksnkankyo/mizukokuweb/system/seibutsuListfile.htm). The gypsy moth Lymantria dispar japonica, which inhabits Honshu and Hokkaido, and the closely related species Lymantria umbrosa, which inhabits only Hokkaido, have a very similar external appearance and are distinguishable only by their genitalia [4]. Due to the physical damage to gypsy moths caused by the traps, we regarded the samples as gypsy moths. For the same reason, closely related species of oak silkmoths, Antheraea yamamai ussuriensis and Saturnia jonasii fallax Jordan, were regarded as oak silkmoths. For small insects for which species identification was difficult (e.g., flies, mosquitoes, and winged ants), the order, family or genus to which they belonged was identified.
Data evaluation
Using the nonparametric Kruskal-Wallis H test, we examined the data to determine if the distribution was normal. Subsequently, multiple comparisons were made by the Steel-Dwass test using add-ins in Excel (Excel statistics ver. 7.0, Esumi, Japan). This examination revealed a 95% reliability level. Eight daily meteorological variables were selected as possible meteorological predictors correlated with trap catches: highest temperature, mean temperature, dusk temperature (at 20:00), lowest temperature, moon score (see above), average wind speed from 18:00–24:00, maximum wind speed, and visibility range from 18:00–24:00. The data were subjected to correlation and least squares linear regression analyses [11]. We regarded R2 values of 0.2–0.4 as indicating weak correlations and R2 values of 0.4–0.7 as indicating moderate correlations.