Autumnal migration patterns of hoverflies (Diptera: Syrphidae): interannual variability in timing and sex ratio

Study area and sampling

Our study was carried out at the Červenohorské sedlo mountain pass in Jeseníky Mountains, NE of the Czech Republic, 50.1245331N, 17.1537733E, 1,020 meters above sea level (m a.s.l.) (see Figs. 1A and 1B). Červenohorské sedlo is a deep break that divides the Jeseníky mountains into the western and eastern ridge, generating an ideal path for aerial migration of insects, birds, and bats moving in the north-south direction (Vavřík et al., 2016), see Fig. 1C.

Two Malaise traps were placed at the top of a ski slope (1,020 m a.s.l.), rising from the river dale (850 m a.s.l.), which provides an ideal and shortest fly path for migratory flying animals to cross the mountains in the lowest possible elevation. We used two one-side blocked Malaise traps to distinguish between southward and northward migration, see Fig. 1D. The traps were operated from 2 August to 18 October in 2018, from 4 August to 31 October in 2019, from 2 August to 2 October in 2020, and from 27 September to 28 October in 2021, according to the annual weather conditions. We emptied the traps daily, approximately at 1900 (UTC+1). Data from the pilot project conducted in 2018, when the placement of the trap was adjusted, and data from the year 2021, lacking August and September observations, are displayed; however, they were omitted from the analysis of the sex ratio, species ratio, and weather conditions on migration.

To explore the effect of Malaise trap size and proportion of the Malaise traps, various traps were used from 2018 to 2020 and in 2021: from 2018 to 2020, smaller traps were used (lower side height = 1 m, higher site height = 2 m, wide = 1.3 m), and in 2021, higher traps (lower side height = 2 m, higher site height = 2.5 m, wide = 1.3 m ) were used in order to test the effect of trap diameter and shape on its effectivity. Data from 2019 and 2020 were collected by the same traps and they were used for quantitative analysis of inter-annual differences in hoverfly abundance.

Weather data (temperature (°C), wind speed (m/s), wind direction, precipitation (mm/day), and relative barometric pressure (hPa)) were retrieved from Hadex WH 1080 weather station, placed in the close proximity (<100m) of the traps.

Determination and statistics

All specimens were pinned and are deposited in the private collections of the authors. The species were identified using keys: Hippa, Nielsen & Steenis (2001), van Veen (2010), Láska, Mazánek & Bicík (2013), Bot & van de Meutter (2020). In genus Sphaerophoria, females were not identified, but they were assumed to be Sphaerophoria scripta (L., 1758), as no other species of the genus was recorded within males.

Statistical analysis and data visualization were performed in R software, version 4.1.2 (R Core Team, 2021). We used a generalized linear model (GLM) assuming a quasibinomial distribution of errors, in order to assess sex ratio discrepancy between years and species, where only data from peaks of migration (PM–period, when more than 95% of individuals per day were caught by north-facing trap and was preceded by at least one day without any observations, to avoid the bias from local populations), were used (2019: 13 to 29 October; 2020: 3 September to 2 October). Significance at the 0.05 level was determined by ANOVA. Due to the extremity of the p-values, post-hoc corrections were not necessary. The analysis of species phenology was performed by fitting local polynomial regression (Cleveland, Grosse & Shyu, 1992), α = 0.5. Shannon & Weaver (1949) diversity index was used to compare species diversity between years. To assess influence of weather, we used Canoco software v. 5.0 (Ter Braak & Šmilauer, 2012), to perform partial RDA analysis, with stepwise selection of variables, where year was included as random factor.

Species diversity

Overall, 31 species of hoverflies were recorded flying southward during the peaks of migration (Table 1). Only four of them made up 81.51% of migrants. The most abundant species was Melanostoma mellinum (L., 1758) (40.26%), followed by Episyrphus balteatus (19.64%), Sphaerophoria scripta (14.66%), and Eupeodes corollae (6.95%), see Table 1. According to Shannon-Weaver index, species diversity differed between years, with highest diversity in 2020 (H = 1.947) and lowest in 2021 (H = 0.178), see Table 1.

Timing of migration

The median of PM differed between years (21 October in 2019, 17 September in 2020). Besides the overall timing, even species differed in the phenology of migration. In 2019, the abundance of Episyrphus balteatus, Melanostoma mellinum, and Sphaerophoria scripta peaked on 14 October, whereas Eupeodes corollae was most abundant on 23 October. In 2020, migration of most species started almost a month earlier than in the 2019, and the PM of species differed. Eupeodes corollae and Episyrphus balteatus appeared first (5 September) and were consistently abundant during the migration period. Sphaerophoria scripta and Melanostoma mellinum were observed later, with peak of abundance on 24 September (see Fig. 3).

The timing of migration differs for males and females. The PM was similar for both sexes during both years in Melanostoma mellinum. In Eupeodes corollae, males peaked in 2019 at the end of the migration period, when females were consistently abundant. On the contrary, in 2020, the peak of Eupeodes corollae female migration was observed at the beginning of the migration period, when males were consistently abundant with small peak at the end of the migration period. However, in both years, on average, females migrated earlier than males in E. corollae. The abundance peak in Episyrphus balteatus was similar for both sexes in 2019. In 2020, females of Episyrphus balteatus peaked at the beginning of the migration period, followed by males later on. In Sphaerophoria scripta, the peak of female migration in 2019 was at the beginning of the migration period, where males were consistently abundant. In 2020, both sexes of Sphaerophoria scripta migrated at the same time; see Fig. 4.

Sex ratio of migrants

The sex ratio of species varied between years. Considering the overall sex ratio, males and females did not differ (2019: 51.75% F, 48.25% ; 2020: 54.82% F, 45.18% M; binomial GLM, F-test, p = 0.4771, Df = 1, R²_L = 0.003, nM = 2,156, nF = 2,391). However, the sex ratio of migrants differed between species (binomial GLM, F-test, p = 2.365e−06, Df = 3, R²_L = 0.205) and there was a statistically significant interaction between species and year (binomial GLM, F-test, p = 6.421e−05, Df = 3, R²_L = 0.15); see Table 2.

Female-biased sex ratio was observed in Sphaerophoria scripta during both years; see Table 3. The male-biased sex ratio was observed in Eupeodes corollae in 2019; however, in 2020, the sex ratio was insignificantly female-skewed; see Table 3. In Episyrphus balteatus, the sex ratio was even in 2019; but females prevailed in 2020; see Table 3. The sex ratio of Melanostoma mellinum was shifted toward the females in 2019, and toward the males in 2020; however, in both years, observed trend was insignificant; see Table 3. For the graphical comparison of the sex ratio of species, see Fig. 5.

Weather conditions

Intensity of migration of all species was positively influenced by southerly winds, temperature, and relative barometric pressure, and negatively by precipitation. Wind speed marginally influenced migration of Eupeodes corollae and Sphaerophoria scripta, see Fig. 6A. The partial RDA provided quantification of proportion of variance, explained by weather conditions. Partial RDA axis one (eigenvalue = 0.4892) and axis two (eigenvalue = 0.0112) explained 51.04% of variation in species data. Weather conditions explained 51.8% of variation in data. Monte Carlo permutation test was performed on all constrained axes (pseudo-F = 7, p = 0.0001, number of permutations = 10,000). Highest abundance of migrants was observed during the southern wind, see Fig. 6B; however, the wind on the Červenohorské sedlo was rarely northerly, see Fig. 6C.

Species diversity

From 53 hoverflies reported as migratory by Speight (2020), we recorded 29 species. Consistently with previous studies, we found that Platycheirus albimanus (F., 1758) and Syrphus torvus Osten Sacken, 1875 are migratory (Aubert, Aubert & Goelding, 1976; Jensen, 2001; Gatter et al., 2020; Martens, 2020), even though they were not listed as migratory by Speight (2020). Two species, Cheilosia canicularis (Panzer, 1801) and Myathropa florea (L., 1758), were recorded in a single year, 2021, and due to the lack of any other observations, it will be too early to conclude them as migratory. The composition of the migrants differed between all monitored years. This pattern could be driven by the specificity of each year given e.g., by climate variation and weather conditions which may affect the length of development or survival of particular species as well as the timing and intensity of migration on both local and large-scale spatial level.

Timing of migration

The timing of migration of the four most abundant migrants differed between species and years. Timing of migration is a complex phenomenon, believed to be influenced by the circadian clock (Denlinger et al., 2017; Zhan et al., 2011), weather conditions (Drake, Drake & Gatehouse, 1995; Knoblauch, Thoma & Menz, 2021), or co-migration with predators (Cohen & Satterfield, 2020). However, none of the presented studies explain the discrepancy between years and species. Promising theory, proposed for birds, seemed to be ‘body size theory’ (Ketterson & Nolan, 1976), suggesting that larger individuals are ‘better suited to survive the colder and less predictable climates at higher latitudes’ (Nebel, 2007). This theory is perfectly suitable for endotherms (explaining even a sex ratio discrepancy). However, the observed migration patterns of hoverflies directly contradict body size theory, with smaller species (Melanostoma mellinum, Sphaerophoria scripta) migrating later or simultaneously with larger ones (Episyrphus balteatus, Eupeodes corollae). Moreover, Knoblauch, Thoma & Menz (2021) observed the same pattern in dragonflies, when the median day of passage came later for smaller species. We hypothesize that the timing of migration may be influenced, in addition to weather and overall seasonal conditions, by a (1) dispersion capacity, where larger species have bigger wing area and higher wing aspect ratio, so they arrive earlier, (2) a minimal temperature threshold for species, where smaller, purely ectothermic, species are able to withstand harsh conditions. This study provides a first step towards a better understanding of the phenology of migration among broader taxa; however, confirmations of our finding require further research, where work is currently underway to determine the temperature threshold and flight capability of migrants.

We observed differences in phenology between males and females. In the four most abundant species, the peak of female abundance occurred earlier than the migration peak of males, with the exception of Melanostoma mellinum, where the peak of migration occurred at the same time for both sexes. Earlier migration of females is believed to be influenced by their earlier emergence, in order to search for suitable oviposition habitat (Guo et al., 2019). However, earlier migration of females was observed even in Episyrphus balteatus, which overwinter mainly as adults (Speight, 2020). Furthermore, in Melanostoma mellinum, which overwinter as larvae (Speight, 2020), the females migrated at the same time as the males. We cannot rule out the theory of oviposition in Eupeodes corollae or Sphaerophoria scripta, which overwinter as larvae (Speight, 2020). On the basis of our results, we challenge the theory of oviposition as a general explanation for earlier female migration. It is very likely that the earlier appearance is driven by population dynamics or higher flight capability of females (Cui et al., 2016; Guo et al., 2019).

Sex ratio of migrants

The sex ratio of the four most common migrants differed between years. Previous studies suggested various explanations for the sex ratio bias in migrating swarms: methodological bias, driven by placement of traps or various migration routes (Nielsen, 1968; Svensson & Janzon, 1984; Steinbauer, 2003), differential mortality of sexes (Tabadkani et al., 2013), various flight and dispersion capabilities (Steinbauer, 2003; Guo et al., 2019), or fat reserves allowing to survive harsh conditions of higher latitudes for longer time (Steinbauer, 2003; Brattström et al., 2018; Guo et al., 2019). Given the same placement of traps in our study, our results do not support hypothesis explaining biased sex ratios in the light of spatial heterogeneity in the distribution of migrating insect (Nielsen, 1968; Svensson & Janzon, 1984). The significant variation of the sex ratio between years challenges the theory of higher flight and dispersion capability of one of the sexes (Guo et al., 2019). However, most studies dealing with the sex ratio discrepancy were not based on long-term observations. At this point, we cannot fully resolve this question, and further research is needed. A tentative explanation is that the sex ratio may be altered on population level by various factors, such as interannual variation in weather or parasitism (Davis & Rendón-Salinas, 2010).

Weather conditions

Intensity of migration was greatly affected by weather. Hoverfly migration was positively related to the southerlies (headwind for migrating hoverflies), temperature, and barometric pressure and negatively by rainfall. We assume that highest abundance of migrants in headwind conditions is largely affected by methodology of sampling. While masses of migrating insects keep high above the ground, when flying with tailwinds southward, thus being hardly detectable by Malaise traps, they migrate at the ground-level at so-called flight boundary layer (Taylor, 1974), to reduce energy costs and increase flight efficiency, when facing headwinds (Gao et al., 2020), hence they are easily detectable by Malaise traps. However, southerlies are strongly predominant on the study site, so the testing of possible selection for northern tailwinds is further limited. Moreover, autumnal migrants are strongly limited in the selection of wind direction, as the weather in the north is gradually becoming unbearable for them (Gao et al., 2020). The observed correlation between actual weather conditions and detected migration could be either driven by direct effect of weather on migration behaviour (Hu et al., 2016), or by constrains of methodology used in our study.

Malaise trap as method for migration research

We have demonstrated the functionality of two single-sided Malaise traps for the study of hoverfly migration at ground-level. One-side-blocked Malaise trap represents a cost-effective method, helping to understand the natural history of migrants. The method of two single-sided Malaise allowed us to distinguish between migratory and resident individuals and, in addition, to determine the start of migration according to the proportion of individuals flying southward and northward, contrasting with studies using classical Malaise-like traps such as those used by Aubert, Aubert & Goelding (1976) or Gatter et al. (2020). On the contrary, Malaise traps are known to be selective (Campbell & Hanula, 2007) and placement sensitive (Malaise, 1937). Furthermore, some hoverflies are known to evade the Malaise traps (Burgio & Sommaggio, 2007), as we observed where for example individuals of the Eristalis tenax (L., 1758) dodged or rammed and escaped the traps. Our design could be improved by combining the Malaise trap with window traps or Möricke traps.