Evidence for a southward autumn migration of nocturnal noctuid moths in central Europe

The suggested evolutionary benefits of energetically costly insect migrations (Rankin and Burchsted, 1992) are manifold and a matter of healthy debate (e.g. Holland et al., 2006; Chapman et al., 2015). The following benefits are usually mentioned in the literature: (1) an increase in the production of offspring via exploitation of ‘green waves’ of vegetation, (2) the avoidance of severe weather conditions during winter time and (3) the avoidance of predators and/or parasites (Chapman et al., 2015). Moth migrations additionally incur a considerable agricultural (Porter et al., 1991; Drake and Gatehouse, 1995) as well as ecological (Green, 2011; Hu et al., 2016) impact. In the case of the migration of the Australian bogong moth (Agrotis infusa), the resulting transportation of biomass from the Australian plains into the Snowy Mountains provides a remarkable inflow of nutrients and energy into this harsh alpine environment, a vital resource for the survival of many animal (and presumably even plant) species (Green, 2011; Warrant et al., 2016). Similar to what is known from birds (e.g. Hein et al., 2011; Richardson, 1990), the migratory behaviour of flying insects is influenced by weather conditions, and this has been documented in a wide variety of species (e.g. Drake, 1994; Brattström et al., 2008).

In European moths, the most convincing evidence for truly migratory behaviour, involving southward autumn migrations and northward spring migrations, comes from the British Isles (Sotthibandhu and Baker, 1979; Baker and Mather, 1982; Chapman et al., 2010). In continental Europe, nocturnal species like the silver Y moth (Autographa gamma) and the dark sword-grass (Agrotis ipsilon) are mentioned as ‘very definite migrants’, with seasonally dependent migratory directions (Williams, 1958). However, apart from this, little is known about the migratory behaviours of continental moth species.

The external compass cues used as orientation references by nocturnally migrating insects have also been little explored, although recent work suggests that the Australian bogong moth relies on the Earth's magnetic field and visual landmarks during migration (Dreyer et al., 2018). In contrast, much more is known about the compass cues used by insects migrating during the day – these migrants rely primarily on celestial cues, such as the sun and the celestial polarization pattern, for orientation (e.g. Reppert et al., 2016; Homberg, 2015).

Since seasonally dependent migratory orientations in the large yellow underwing (LYU) have previously been described by Chapman and colleagues (2010), with a northward orientation in spring and a southward orientation in autumn, we hypothesized that noctuid moths captured during autumn in central Europe would show a similar pattern, with a southbound migration towards the Mediterranean. In this study, we tethered moths within a computer-interfaced flight simulator (Mouritsen and Frost, 2002) to investigate the autumn migratory orientations of two species of noctuid moths: the large yellow underwing [Noctua pronuba (Linnaeus 1758)] and the red underwing [RU; Catocala nupta (Linnaeus 1767)]. We also tested which weather conditions were associated with directional migration behaviour. These studies were performed at a migratory nodal point in central Europe (Fig. 1A): the Lake Neusiedl Biological Station, located within the Austrian National Park Seewinkel near the eastern foothills of the European alps. The wetlands and saline lakes of this area provide an important stopover location for many migratory bird species and also act as an important refuge for waterfowl. Nightly light trapping of insects also revealed a similar richness in migratory moth species.

All moths tested in the present study were caught using light traps at the Biological Field Station Lake Neusiedl during the first half of September 2016. The captured moths were transferred into plastic containers before they were stored in a sheltered place to recover from the stress induced by capture and were fed with 10% honey solution. The moths were usually tested 1–3 days after being captured. Prior to tethering, the moths were immobilized in a refrigerator for up to 5 min. The scales on a moth's pterothorax were removed using a micro-vacuum pump. Afterwards, a vertical tungsten stalk was glued to the dorsal thorax using contact cement while the moth was restrained by a weighted plastic mesh. Moths were then placed individually in clear plastic containers and provided with a clear view of the sky prior to their experimental night. Since insulating scales were removed from the pterothorax, and the glued Tungsten stalk could potentially cool the animal down, we placed the animals in a box warmed by hot water bottles wrapped in towels prior to the experiments. All animals were released one day after they were tested.

Ferromagnetic-free, modified Mouritsen-Frost flight simulators (Mouritsen and Frost, 2002) were used to continuously record the heading directions of moths (see Fig. 1). The circular arena had a diameter of 50 cm and a height of 36 cm; the inner wall of the arena was covered with black cloth.

A moth equipped with a vertical tungsten stalk was connected with a piece of thin rubber tubing (inner diameter slightly smaller than diameter of the stalk for a tight fit; length: 1.5 cm) to a vertical tungsten rod (diameter: 0.5 mm; length: 15.3 cm) that was attached very tightly via a very small grub screw to the centre of an optical encoder (US Digital, Vancouver, WA, USA) located at the centre of the upper opening of the arena (Fig. 1C). This optical encoder continuously recorded the rotations of the tungsten rod assembly (between 0 deg and 360 deg). Thus, the attached moths were able to flap their wings and freely choose any flight direction around the yaw axis while their heading directions were recorded relative to magnetic North with a sampling rate of 5 Hz and a horizontal resolution of 3 deg. Using a projector (BENQ, GP3), a mirror and neutral density filters, a very dim moving image (a bitmap satellite image of the local region from Google Earth) was back-projected onto a tracing paper screen (mounted on a clear Perspex sheet below the arena; a hole with the same diameter of the arena was cut into the table top). Custom-written software controlling the movement direction of the image was coupled to the encoder system (USB4; US Digital, Vancouver, WA) via a feedback loop. The resultant ventral flow-field image, which was always moving 180 deg relative to the moth's heading (i.e. from head to tail), created the visual sensation of a ventral ground moving below the moth during forward flight, irrespective of which direction the moth flew in, because it was continuously updated by the moth's own flight direction. The average luminance of the optic flow was 6.7×10⁻⁴ cd m⁻².

Moths were tethered and allowed to fly for several minutes. In order to allow a meaningful comparison of the recorded tracks, only the first 10 min were used for the analysis (except LYU #13, which flew for 7.3 min). Since the RU moths were caught less frequently (∼1–2 per catch), we were only able to record the orientation choices of 14 individuals. 11 out of 14 individuals flew over the full recording time of 10 min, 1 individual flew 20 min, and 2 individuals stopped flying after 7.7 and 9.5 min, respectively. The 36 LYU moths (usually ∼5–15 were caught per night) flew for at least 10 min within the arena. In the case when a moth continuously spiralled after it was attached to the encoder (putatively due to a skewed stalk), or showed a faltering flight behaviour (characterized by several stops in a row), the experiment on this individual was immediately aborted and the data excluded from the analysis.

While the moths were performing in the flight simulator, they experienced a clear view of the sky (viewing angle ∼120 deg; see Fig. 1C,D). A paddock close to the biological field station was chosen as an experimental site (47°46′05.0″N, 16°46′05.5″E). Great care was taken to position the arena at a sufficient distance from putative landmarks (e.g. trees) to remove the possibility that moths might see them. The simulator arena was placed on a wooden table that was given a different orientation each experimental night. The arena itself was rotated by about 90 deg after each individual experiment. A small aluminium camping table and two plastic chairs were deployed about ∼5 m from the arena. Three sides of the experimental table were covered with black cloth to prevent stray light (reflected by the mirror beneath the table) from influencing the behaviour of the tested moths. One side remained uncovered to allow the projection of optic flow onto the mirror (see above). All artificial light sources were either covered with several layers of duct tape or with red filter film. Headlamps equipped with red LEDs were used while the moths were being handled. The experiments were conducted on seven nights between 1 and 13 September 2016. All experiments were conducted after sunset, under more or less clear and windless conditions, and under a natural magnetic field. The moon's disc was not visible to the moths in any of the experiments.

The first three experimental nights were dry and clear. Owing to slight showers on 4 September, the experiments were put on hold. On the following two nights, the weather changed dramatically with a temperature drop from 29.4°C on the 4th to 21.7°C and 16.3°C (maximum temperatures), respectively, on the 5th and 6th, which was accompanied by moderate winds (17 and 20 km h⁻¹) and rain. Considerable changes in humidity were noticed on the experimental nights that followed the rainy nights, probably due to the resulting sogginess of the surrounding fields (see below).

The experimental nights of the 7, 8, 9 and 12 September were quite foggy and moisture condensed noticeably on our gear. The atmospheric pressure and the air temperature varied between the first three experimental days and the last four, but the most notable difference was related to the relative humidity and the dew-point spread (difference between the air temperature and the dew-point). Changes in the dew-point spread give some indication of whether fog will gather or not; the lower the spread, the higher the chance of fog formation. Fog and high humidity were observed on the last four experimental nights, but not on the first three. All meteorological measurements derive from meteorological stations in Andau and Eisenstadt (Burgenland, Austria, both around 20 km from the experimental site), which are available online (kachelmannwetter.com; wetter.com) and from personal notes taken during the experiments.

Data analysis was conducted using custom-written MATLAB scripts (2010b, 2013a; MathWorks Inc.) and Oriana (4.01; Kovach Computing Services). Since the analysis of the orientation choice of each individual moth gives a mean vector in which the r-value encodes the magnitude of the mean vector (the longer the mean vector the higher is the concentration of the data around the mean direction), and the direction of the vector is the mean direction, we were able to conduct second order statistics, using the non-parametric Moore's modified Rayleigh test (MMR; Moore, 1980; Zar, 1999). This test involves different weightings of the respective mean angles according to their respective r-values. The test assumes the null hypothesis that the means of the tested sample are uniformly distributed around the circle. A low P-value (conventionally P<0.05) indicates that the data are distributed non-uniformly and that the tested sample had a preferred migratory direction. The length of the black arrow encodes the R* value (score of MMR test). The dashed circles in the circular diagrams encode the critical R* values needed for statistical significance: inner circle, P<0.05; middle circle, P<0.01; outer circle, P<0.001. A black arrow crossing a respective dashed circle indicates statistical significance at the corresponding level.

In order to compare the orientations of the respective groups, we used the non-parametric Mardia–Watson–Wheeler Test (MWW; Fisher, 1993; Mardia and Jupp, 2000; Batschelet, 1981) and a Moore's paired test (MPT; Zar, 1999), to test against the null hypothesis of identical distributions. All orientation choices with r<0.2 were not included in the evaluation since we assumed that the moths were performing behaviours other than migratory behaviour (see Nesbit et al., 2009), the stalk was not attached correctly to the moth's thorax, or the stalking process had interfered with flight performance by introducing wing beat asymmetries (due for example to glue spill).

To compare our results with those of Sotthibandhu and Baker (1979), we used the open-source program Stellarium (version 0.14.0) to reproduce the stellar constellations of the night sky over Manchester in 1977 (Sotthibandhu and Baker, 1979) and Illmitz during September 2016 at 22:30 and 01:30 h local time, according to the experimental time mentioned in fig. 9 of Sotthibandhu and Baker (1979). To avoid confusion, we want to stress that this procedure was simply part of a post hoc data analysis: Stellarium was not used during the actual experiments. Unfortunately, Sotthibandhu and Baker (1979) did not provide precise dates and locations for their experiments (on moths tested under moonless starlit conditions), apart from the fact that the experiments took place in July 1977 in Manchester or Bristol. We thus chose Manchester on 15 July 1977 for the reconstruction of the night sky. On a clear night in Manchester, the bright star Arcturus – a possible orientation cue – would have been visible throughout July. The orientation angle of each moth was compared with the time the experiments were conducted using linear regression (see Results), performed using the Data statistics Toolbox in MATLAB (2010b). Because the ‘starting angle’ of the visible celestial bodies at the beginning of each experimental night changed only marginally between the experimental nights (1–13 September 2016), and because their rate of rotation is constant, the slight angular differences between the experimental nights included for the data recorded in Illmitz (2016) are negligible versus the respective time of the night.

The RUs oriented SSE when provided with a clear view of the starry sky and while experiencing the natural magnetic field of the Earth (Fig. 2A). According to the MMR test, the tested sample was statistically significantly directed towards 146 deg (α) relative to magnetic North (95% CI: 119 deg, 169 deg; r=0.49; R*=1.71; P<0.001; n=12). Since most of the RUs (n=8) were tested during four foggy nights, the orientation of this species was evidently not affected by the weather conditions prevalent on these final four experimental nights.

The preferred mean direction of all tested LYUs (Fig. 2B) was significantly towards SSW (α: 201 deg; 95% CI: 141 deg, 257 deg; r=0.173; R*=1.1; P<0.05; n=36). We also analysed the orientation choices of the LYU against the presence or absence of fog. A lower dew-point spread (Fig. 3) is associated with an increased presence of water particles in the air, and an increasing chance of fog. From experimental day 4 onwards, the dew-point spread dropped severely, indicating a much higher likelihood of fog during our experiments (and fog was noticed).

Interestingly, as a group, LYUs were significantly oriented on clear nights (Fig. 2C; α: 205 deg; 95% CI: 167 deg, 260 deg; r=0.275; P<0.01; n=19), but not in a groupwise analysis on foggy nights (Fig. 2D; α: 160 deg; r=0.083; R*=0.381; 0.5<P<0.9; n=17). In addition, the mean directions of the RUs and the LYUs were significantly different (MWW, P=0.007, W=9.826; MPT, P<0.01, R′=1.291; all LYU data used for comparison), suggesting that these two species have different migratory directions (Fig. 2A,B).

In order to test for a time-dependent ‘shift’ of the heading directions of moths over the course of a night, as detected by Sotthibandhu and Baker (1979), we plotted the heading directions of moths as a function of the experimental time (see Fig. 4; slopes for RU: −1,29; LYU in Illmitz: −14,8; LYU in Manchester: 16,95). The implications of this analysis for migratory headings measured in our moths will be explored in the Discussion.

RUs have previously been reported to be ‘occasional vagrants’ (Skinner, 2009), and have been considered to merely expand their distribution by invading other territories during favourable seasons (Fox et al., 2011). In Austria, members of the genus Catocala are thought to regularly invade the country from the Mediterranean region (Malicky et al., 2000). We found that when tested in a flight simulator during September in southeast Austria, RUs demonstrated a highly concentrated SSE orientation. A plausible explanation for this behaviour could be that RUs perform a roughly southbound autumn migration towards the Mediterranean (possibly to as yet unknown wintering grounds), and that we recorded the orientation choices of individuals that were undertaking this migration. According to this hypothesis, one could assume that we caught individuals that had flown into the National Park from locations to the north or northwest of the study site. This assumption is supported by a quantitative sampling of RUs until late August in Lower Bavaria using bait traps (Sturm, 2002) – released RUs were never subsequently re-captured, strongly suggesting a lack of site fidelity. Thus, we conclude that the directedness of our tested RUs does not simply reflect dispersal movements (i.e. expansion of the species distribution) but rather a regular autumn southbound migration from central Europe towards regions of the Mediterranean (e.g. Greece; Kailidis, 1964). Judging from the recorded mean direction of the tested RUs, one might draw the conclusion that their conspecifics in the wild would fly farther into the Balkans.

The LYU is listed as a migratory moth species in Central Europe and migratory movements have been reported in regions spanning from Britain to the Caucasus (Chapman et al., 2010; Poltawski, 1982). The southwesterly orientation of the LYUs recorded in Illmitz seems to be seasonally appropriate and in agreement with the data obtained in southern Britain (see brown arrow in Fig. 2; Chapman et al., 2010). This suggests that LYUs, similarly to RUs, perform an annual roughly southbound autumn migration passing through Illmitz and heading towards their wintering grounds situated to the south. The exact wintering locations are, however, still unknown.

The influence of wind and temperature on the behaviour of insects is described extensively in the literature (e.g. Pedgley et al., 1990; Taylor and Carter, 1961; Gregg et al., 1994; Marchand and McNeil, 2000; Chapman et al., 2016). The ‘minimum temperature threshold’ required for migratory flights in two medium-sized noctuid moth species (wingspan: 30–40 mm) was estimated to be ∼8°C (Taylor and Carter, 1961). We conducted our experiments on practically windless nights, with air temperatures distinctly above 8°C, so if the weather influenced our experiments at all, wind and air temperature are presumably only of secondary importance for the interpretation of our data.

RUs have been reported at bait traps even in bad weather conditions (ongoing rain, strong winds and temperatures less than 15°C; Sturm, 2002). This observation is in line with our results indicating that the orientation of RUs was evidently not influenced by the weather. In contrast, we noticed that a subsample of LYUs tested under foggy conditions was less oriented (in fact random, Fig. 2B) compared with a subsample that was highly significantly oriented under clear conditions. Both species readily flew in the arena, so the motivation to fly was not necessarily reduced because of the weather conditions. However, in the case of the LYUs, the motivation to perform oriented migratory flights in the arena might have been reduced as a result of unfavourable weather conditions (e.g. lack of favourable initiation factors for migration).

The presence of fog might technically explain this phenomenon since the lower visibility could negatively influence the moth's ability to orient during migration by obscuring a clear view of the sky. However, we never observed the arena veiled in thick fog. As mentioned earlier, the most dramatic meteorological factors that accompanied foggy nights are changes in the relative humidity and the dew-point spread. However, in a quantitative long-term study in northern New South Wales (Australia), a positive correlation was observed between the local relative humidity and the numbers of migratory moths caught (Gregg et al., 1994). This implies that these particular Australian moth species preferred higher humidity while in flight. Unfortunately, that study did not provide any data on the presence or absence of fog.

Factors that could explain the reduced directedness of LYUs compared with RUs on foggy nights might include an unfavourable presence of water particles in the air, a slightly reduced outside temperature and/or a smaller body size. If water particles in the air moistened the bodies of LYUs, this could have led to an overall critical decrease in their body temperature and a reduced motivation to migrate.

Moreover, in our experiments, the moths were tethered to a tungsten rod and performed flight behaviour within an arena that was located ∼1 m above the ground. Thus, the conditions under which our tested moths performed were quite different to those experienced by their conspecifics in the wild. Besides the obvious limitations that were introduced by our experimental setup, we also wish to stress a less obvious factor that might have influenced the results: we removed the dorsal scales of the pterothorax in order to glue the tungsten shaft to the animal. The ‘coats of hair’ of moths are an ‘excellent insulator’ (Church, 1960) and in noctuids, an intact layer of scales contributes to maintaining up to 90% of the insect's internal temperature while in flight (Church, 1960). If during our experiments the tungsten shaft collected water particles due to the fog, and was fanned by the air current generated by the beating wings of the moth (with a damaged layer of scales), this could have decreased the overall body temperature due to evaporative chill, thereby reducing the motivation of the moth to migrate. The smaller LYU may have been much more susceptible to such a situation than the considerably larger RU. Another putative explanation would be that a humid evening simply resembles an unfavourable trigger for a migratory flight for the smaller LYU.

During the daytime, insects can rely on obvious and reliable celestial cues such as the position of the sun's disc and the polarization pattern of the sky to navigate (e.g. Wehner and Müller, 2006; Homberg, 2015), whereas compass cues that are utilized for long-distance movements at night are still comparatively little studied. Two obvious directional cues at night are stars (Foster et al., 2018) and/or the Earth's magnetic field (e.g. Baker and Mather, 1982; Guerra and Reppert, 2015; Dreyer et al., 2018). Indeed, birds that migrate at night are capable of using both of these cues for compass orientation (Emlen, 1967; Wiltschko and Wiltschko, 1972). In addition, the light conditions are about 1 million times dimmer at night and this might have an impact on the cue hierarchy used during migration (el Jundi et al., 2015). Taken together, using the sky at night might require a more dynamic compass system that is different to the one in diurnally orienting insects.

A moth fixed in our flight arena would have experienced an apparent movement (due to the rotation of the Earth) of celestial bodies from east to west at a constant rate, depending on their distance to Polaris. In figure 9 and table V of Sotthibandhu and Baker (1979), the authors presented orientation directions of LYUs (see Fig. 4A,B) recorded on a moonless starry night in England (using a different orientation apparatus) and suggested that this species is likely to use celestial cue(s) for compass orientation. Since all tested individuals showed a shift in orientation of about 16 deg per hour from east to west between ∼22:00 h and ∼01:00 h local time, and coupled to the fact that both samples were significantly directed, the authors concluded that all individuals used the azimuth positions of particular stars or star groups to orient (Sotthibandhu and Baker, 1979). The stars or star groups could be (estimated from the shift of 16 deg per hour) located at a position about 95 deg away from the pole star (Sotthibandhu and Baker, 1979). The star constellation Boötes, featuring Arcturus (α boo) as the most radiant star of the northern hemisphere (Kaler, 2002), roughly fulfils these requirements, and should have been visible with the naked eye during the experiments conducted in England in 1977 (see Fig. 4D). During our experiments in 2016, at the more southerly experimental location in Illmitz, one constellation would have been clearly visible as well: Aquila featuring Altair as its brightest element. Regardless of which celestial body (or bodies), the tested individuals in Sotthibandhu and Baker (1979) might have used (if any), they clearly exposed a positive ∼50 deg shift of the group's mean vector. If the hypothesis that stellar orientation drove this shift was correct, then the English moths obviously did not show any form of time compensation (see Fig. 4F), as has been found, for example, in the Monarch butterfly (Mouritsen and Frost, 2002). Without the ability to time compensate for the rotation of celestial bodies, moths would continuously hold a constant angle relative to celestial cues, meaning that their flight trajectories would change direction over time (i.e. become curved: Fig. 4E). If Sotthibandhu's and Baker's hypothesis is correct, and their moths indeed relied on stars for orientation, then they must have used a non-time-compensated celestial compass [as discussed in Nesbit et al. (2009) for the day-active painted lady, Vanessa cardui]. This would imply that these moths would not have had the ability to maintain a constant migratory bearing over the course of a night. This is indeed what their data suggest (black solid line in Fig. 4G).

If the LYUs tested in our flight simulators used a particular celestial cue as a reference point, such as the star constellation Aquila or the Milky Way (or some other constellation of stars), and like the English moths, did not apparently compensate for the celestial rotation, their bearing should systematically change over the course of an evening. However, our data did not find any indication of such a systematic change in the bearing of LYUs (orange solid line in Fig. 4G), nor indeed in the bearing of RUs (red solid line in Fig. 4G). A couple of explanations for this difference in our results from those of Sotthibandhu and Baker might be as follows (not an exhaustive list): (1) the Austrian moths did not use the stars as a compass cue, or (2) they did use the stars as a compass cue but were capable of time compensation. An obvious major difference between our experiments and those described in Sotthibandhu and Baker (1979) is that we did not test our animals twice at different times of the night and this, together with the degree of ‘noise’ in our data, may have masked any systematic change in bearing in the Austrian moths.

Another potential directional cue is the geomagnetic field, and a magnetic compass sense was reported in the large yellow underwing by Baker and colleagues (1982). Additionally, the day-flying Monarch butterfly has been suggested to use a light-driven inclination compass (Guerra et al., 2014, but see Mouritsen and Frost, 2002 and Stalleicken et al., 2005). Moreover, a magnetic compass sense has very recently been implicated in the highly directed migration of the nocturnal Australian Bogong moth (Dreyer et al., 2018). Since we used a ferromagnetic-free experimental apparatus, it is at least plausible that our moths used the undisturbed magnetic field of the Earth for navigation.

In conclusion, we believe that our study provides an ideal baseline for further investigations of compass mechanisms used for nocturnal orientation during long-distance moth migration. The two European moth species we have investigated – the red underwing and the large yellow underwing – are ideal for studying the navigational mechanisms that steer nocturnal insect migrations and for investigating the roles of navigational cues proposed in earlier studies, but using different techniques. Follow-up studies using computerized three-dimensional Helmholtz coil systems (in order to manipulate the experimental magnetic field within the arena) and top projections of the starry sky are already planned.

During the preparation of this manuscript, our dear friend, colleague and co-author, Professor Barrie Frost, passed away after a short illness in his home town of Kingston, Ontario, aged 79. His passing has brought much sadness, but also deep gratitude for the wonderful years we shared together, and for the enthusiasm and insight he brought to the many projects we have worked on together. Vale Barrie Frost, you will be greatly missed. We thank the kind and helpful staff of the Biological Field Station in Illmitz for their technical support.