Migration of the painted lady butterfly, Vanessa cardui, to north-eastern Spain is aided by African wind currents



    1. Butterfly Monitoring Scheme, Museu de Granollers de Ciències Naturals, Francesc Macià, 51, E-08402 Granollers, Spain;
    Search for more papers by this author

    1. Department of Physics and Nuclear Engineering, Universitat Politècnica de Catalunya, Avda Victor Balaguer s/n, E-08800 Vilanova i la Geltrú, Spain; and
    Search for more papers by this author

    1. Center for Ecological Research and Forestry Applications, Edifici C, Universitat Autònoma de Barcelona, 08193 Bellaterra, Barcelona, Spain
    Search for more papers by this author

Constantí Stefanescu, Butterfly Monitoring Scheme, Museu de Granollers de Ciències Naturals, Francesc Macià, 51, E-08402 Granollers, Spain. E-mail: canliro@teleline.es


  • 1Thousands of records of migratory butterfly species such as Vanessa cardui flying just above ground-level on fixed compass bearings have led to the common belief that these insects migrate within the so-called ‘flight-boundary layer’, where movements are relatively independent of the wind.
  • 2Given the selective advantages of windborne migration and the existence of a number of observations of flights of V. cardui from the upper levels of the atmosphere, we tested the hypothesis that migration from North Africa to southern Europe in this species is influenced by synoptic-scale wind currents.
  • 3Even with modern technology, it is extremely difficult to observe high-altitude flights directly, so we rely on an indirect approach that examines whether or not arrival peaks in north-eastern Spain are associated with winds blowing from Africa.
  • 4Arrivals of V. cardui were determined for the spring period (1 March–27 June, 1997–2006) at 79 sites in the Catalan Butterfly Monitoring Scheme. Wind patterns were described on the basis of synoptic-scale maps, transport models and back-trajectories calculated for each day of the spring period.
  • 5We found a strong association between migration and winds from North Africa, both for the whole data set (1997–2006; χ2 = 4·7, P = 0·03) and for a restricted data set that excludes years in which the species was very scarce (χ2 = 7·26, P = 0·007).
  • 6Episodes of massive northward migration within the species’ flight-boundary layer also coincided with spells of winds from North Africa, suggesting a connection between low-altitude (observational) and high-altitude flights (inferred from wind patterns).
  • 7Finally, on the assumption that migration in V. cardui is windborne, a source–receptor transport model applied to spring abundance data in north-eastern Spain enables us to identify the most probable population source areas in North Africa.


Knowledge of insect migration has increased greatly over recent decades as a result of intensive theoretical and applied research (Drake & Gatehouse 1995; Dingle 1996; Dingle & Drake 2007). For instance, the seminal contributions by Rainey (1951, 1963) and Johnson (1969) have shown that long-distance movements of many insect species are influenced by the high-altitude winds that blow several hundred metres above ground level. The development of new techniques, particularly the use of radar, has contributed greatly to a deeper understanding of this phenomenon, albeit to date almost exclusively in the field of insect pests (e.g. Reynolds & Riley 1997).

Although there are many examples of insects, including moths, the migratory movements of which depend on the external force of the wind (Johnson 1969; Drake 1985; Reynolds & Riley 1997; Chapman et al. 2002), similar evidence is surprisingly rare for butterflies. Since the pioneering work by Williams (1930, 1958), it has been widely accepted that butterflies migrate mostly within the so-called ‘flight-boundary layer’– the space within a few metres of the ground where the air speed of the flying insect is greater than the wind speed (Walker 1980; Drake & Farrow 1988; Pedgley, Reynolds & Tatchell 1995). This conclusion is based on thousands of records of butterfly species migrating at low altitudes on fixed compass bearings, their movements being relatively independent of winds (Nielsen 1961; Baker 1978; Walker & Riordan 1981; Walker 1991; Srygley, Oliveira & Dudley 1996; Srygley 2001).

According to Johnson (1969), however, the range of data available for butterflies is insufficient for drawing firm conclusions regarding the nature of the migratory process. Johnson (1969) argued against the tendency to use records for relatively short distances to extrapolate for very long distances. Such a practice would be misleading if migration occurs for the most part in the upper air, as reported for both the monarch, Danaus plexippus (L.), and the red admiral butterfly, Vanessa atalanta (L.). Monarch butterflies often ride the rising layers of warm air that precede the arrival of cold air masses, enabling them to take advantage of the northerly winds that blow in association with such fronts (Calvert 2001). They rise on thermals up to altitudes of over 1000 m and then glide while migrating to their wintering areas (Gibo & Pallett 1979; Gibo 1981). The red admiral also uses northerly winds when migrating from the north of Europe to its breeding areas in the Mediterranean (Mikkola 2003). Although butterflies have only ever been directly observed flying at heights of 20–100 m, radar data indicate that flights may occur as high as 1000–3000 m (Mikkola 2003). It may well be the case, therefore, that the paucity of records of butterflies showing upper-level windborne migration is a consequence of the lack of appropriate methods for detecting this phenomenon.

Here we return to the subject for one of the most well known of all migratory butterflies, the painted lady, Vanessa cardui (L.) (Nymphalidae), the low-altitude directional flights of which have been known all over the world for centuries (see Williams 1930 for a comprehensive review of early data). However, previous investigations have been strictly circumscribed to ground-level observations, and flights at higher levels have not been studied as yet.

The life history of V. cardui is shaped by long-distance migration (Wiltshire 1941; Larsen 1976). In the Palaearctic, each spring the species arrives in Europe from northern Africa (Williams 1970; Pollard et al. 1998) after journeys of several hundred kilometres. In the Nearctic, such flights occur between desert areas in Mexico and Baja California, and latitudes between 35–60° N in the USA and southern Canada (Abbot 1951; Tilden 1962; Williams 1970; Myres 1985; Giuliani & Shields 1995). Taking into account the selective advantages of windborne migration (such as allowing flight to be more fuel-efficient and more rapid), it would be most surprising if V. cardui did not make use of tailwinds to complete its migrations. Although a few observations of V. cardui being carried by high-altitude winds do exist (Giuliani & Shields 1997; Mikkola 2003), it is not known if these flights are a common strategy used on migratory displacements to take advantage of wind currents.

Here we tested the hypothesis that V. cardui migration from North Africa to southern Europe occurs in the form of flights in the upper levels in the atmosphere that benefit from synoptic winds. We examine this idea by testing whether the peaks of individual arrival in north-eastern Spain are associated with winds blowing from Africa. To describe wind patterns, we use a variety of meteorological tools widely employed for describing the transport of atmospheric chemicals (Stohl 1998).

Nevertheless, ground-level observations have also reported flight displacements of V. cardui on fixed compass bearings, indicating that migration at surface level also takes place. Given that a number of episodes of this kind have been observed, we have also examined the specific meteorological conditions during these migratory episodes to check whether or not this phenomenon contradicts our working hypothesis.

Previous work on atmospheric chemistry in eastern Spain has shown that winds from North Africa, loaded with Saharan dust, are frequent in spring (Àvila, Queralt & Alarcón 1997; Rodríguez et al. 2002; Àvila et al. 2007) and that back-trajectories, satellite images and transport models can be used to describe the provenance of the dust. Therefore a further task was to use these meteorological tools to identify the departure areas of V. cardui in North Africa with a simple source–receptor transport model, as there are few published data on potential source areas.

Materials and methods

migratory phenology of v. cardui in the study area

In Catalonia, north-eastern Spain, spring migration of V. cardui is recorded mainly in April, May and early June (Stefanescu 1997; Pollard et al. 1998). African migrants breed throughout the region on a diversity of plants [mainly mallow, Malva sylvestris (L.), and several thistle species] that are in full growth in this period (Stefanescu 1997). At the start of the summer, newly emerged butterflies leave the area quickly as they encounter the hot, dry conditions typical of the Mediterranean summer and migrate northward or towards higher mountains in search of both nectar and egg-laying resources (Pollard et al. 1998). During the period of summer drought, the species is almost absent from lowland areas in Catalonia, although in the second fortnight of August and in September, moderate numbers are recorded again as part of autumn migration. Small-scale breeding also occurs then, and a new local generation is usually detected in October. Although some of these late-summer butterflies may exceptionally survive the winter in the region, virtually all the locally produced population flies southward to Africa, as shown by the sudden appearances of huge numbers of V. cardui in the Maghreb, the Canary Islands and the northern edge of the Sahel in September–October (C.S., unpublished data).

migration peak characterization

For reasons not as yet fully understood, in Catalonia and elsewhere in Europe, the autumn migration of V. cardui remains relatively unnoticed in comparison with the spring migration (Williams 1951; Pollard et al. 1998; but cf. Hansen 1997), thus it is dificult to identify any clear migratory waves during the return flight. Therefore we have restricted our study to spring migration, that is, northward migration from Africa.

We used data from 79 monitoring transects of the Catalan Butterfly Monitoring Scheme (CBMS; Stefanescu 2000), including three recording stations in the Balearic Islands, for the period 1997–2006 (Fig. 1). Overall, the data set comprised 385 yearly transect counts. Counts (standardized to individuals per 100 m of the recording route) were made on a weekly basis from 1 March to 26 September, thus comprising the whole period of northward migrations. Vanessa cardui was recorded at virtually every monitored site, and its yearly frequency of appearance ranges from 0·96 to 1 (0·99 ± 0·02, mean ± SD). The number of transects where it was observed varied each season in line with the number of sites included in the scheme, from a minimum of 24 in 1998 to a maximum of 51 in 2004 (38·80 ± 10·86).

Figure 1.

Location of the 79 recording transects in the Catalan Butterfly Monitoring Scheme used for monitoring Vanessa cardui populations in Catalonia (north-eastern Spain) and the Balearic Islands (two transects in Menorca and one transect in Ibiza, not shown) in 1997–2006. The size of each circle is proportional to the number of recording years.

To identify spring migratory waves, data from the first 17 weeks of the CBMS calendar (1 March–27 June) were used, the latter date being considered the latest when northward migrations can still be detected, as indicated by exceptional observations of poleward flights by African butterflies on 21 and 27 June 1996 (Stefanescu 1997). Mean abundance of the species was calculated for all the available transects for each of these 17 weeks, each year being analysed separately. Abundance estimates for every pair of consecutive weeks were carried out on the assumption that any significant numerical increase would be the result of the arrival of a migratory wave. One-tailed t-tests for paired comparisons were used for comparing abundance between consecutive weeks. Data from each transect for 1 week were paired with data from the same transect the following week, to avoid biases in the butterfly abundance estimates stemming from factors such as missing counts, and spatial heterogeneity in nectar and oviposition resources along transects.

Two different data sets were used: the whole data set (10 migratory periods monitored between 1997 and 2006), and a subset excluding years with low abundance of the species. Population levels of V. cardui were extremely low in 1997, 1998, 1999 and 2005 (Fig. 2), which meant that chances of detecting increases in abundance were greatly reduced (1997 and 2005), and made estimates of migration less reliable (in 1998 and 1999, when two out of three increases in abundance in very late June might have been caused by local emergence).

Figure 2.

Average spring abundance (1 March–27 June) of Vanessa cardui in the study sites from 1994 to 2006. Population levels were estimated using the methodology of Moss & Pollard (1993).

meteorological data and air-mass trajectories

We investigated the possible association between the arrival of waves of migratory butterflies and the occurrence of high-altitude African winds originating from synoptic-scale systems, that is, from a succession of low- and high-pressure cells approaching the region from the west or south. A positive association would be expected if spring migration were mainly windborne at high altitude.

The possible source region of an air mass was determined with backward-in-time trajectories. A single 96-h back-trajectory arriving in Catalonia at 00 UTC (Coordinated Universal Time) was calculated for each day between 1 March and 27 June during the 10-year period 1997–2006 (a total of 1190 back-trajectories). Trajectories were run at 1500 m asl using the hybrid single-particle Lagrangian integrated trajectory (HYSPLIT-4) model of the National Oceanic and Atmospheric Administration (NOAA) (http://www.arl.noaa.gov/ready/hysplit4.html; Draxler & Rolph 2003). The HYSPLIT model calculates isentropic trajectories from the gridded meteorological fields of the FNL archive data. A height of 1500 m, corresponding to 850 hPa standard pressure level, was selected as most representative for transport in the lower troposphere. This layer is typically sensitive to cyclonic wave features, and is the approximate boundary between the surface wind regime and the free troposphere (Artz, Pielke & Galloway 1985). Moreover, a relationship between the 850 hPa wind direction and the prevailing weather patterns associated with the passage of cyclonic waves is well established (Dayan & Lamb 2003). For each recording week, an incursion of African winds was considered to have taken place when at least one of the seven daily back-trajectories crossed the shaded area in Fig. 4b.

Figure 4.

Case study: peak migration on 10–16 May 2004. (a) Increase in abundance detected in the Catalan Butterfly Monitoring Scheme. (b) Back-trajectories at 1500 m asl for the 7 days of the period (arrival date in Catalonia labelled next to each trajectory); the shaded area corresponds to the region over which a trajectory was considered to be of African origin. (c) Displacement of a dust cloud (at 3000 m asl) originating in central Algeria on 11 May 2004, from the DREAM model; and (d) vertical profile of the dust concentration (µg/m3) over Barcelona on 12 May 2004, from the Dust Regional Atmospheric Model (DREAM).

We also investigated which meteorological scenarios were responsible for the synoptic-scale wind systems associated with the arrival of migrants. Weather charts from the NOAA (http://www.arl.noaa.gov/ready.html), showing mean pressure at sea level and the geopotential height at 850 and 700 hPa, were used. The classification by Escudero et al. (2005) of dust mobilization and transport from North Africa to the Iberian Peninsula was applied with slight modifications, and the following three transport scenarios were defined. (1) Atlantic Depression (AD), characterized by a deep low pressure situated west or south-west of Portugal and often associated with a ridge over the central Mediterranean: source areas under these conditions were mostly from the western Sahara. (2) Depression over North Africa or the Iberian Peninsula (PD): this situation favours the transport of African air masses across the Mediterranean to the Iberian Peninsula, and in these cases dust originates mostly from Algeria and Tunisia. (3) A high-pressure system at higher levels (NAH-H), appearing in summer when intense heating over the Sahara induces the development of the North African thermal low and the uplift of dust to around 5000 m. Then dust is transported to the Iberian Peninsula by the western branch of the high situated at upper levels.

low-altitude migratory flights

Several episodes of low-altitude migratory flights were recorded in the CBMS during the study period. We investigated a possible connection between low-altitude (observational) and high-altitude flights (inferred from the wind pattern) for a selected set of six episodes, during which intensive northward flights were reported during butterfly counts. We also included observations from April–May 2003 and 2004 by bird-ringers on L’Illa de l’Aire (a small islet off the south-east coast of Menorca in the Balearic Islands). All six episodes coincided with a strong increase in V. cardui abundance in the region (CBMS data). Details on the nature of the wind regime prevailing during these major migratory events were obtained from back-trajectories at 1500 m asl and from the Dust Regional Atmospheric Model (DREAM; http://www.bsc.es/projects/earthscience/DREAM), which provides information on dust load, wet and dry dust deposition, dust concentration, and the vertical dust concentration profile over the city of Barcelona.

source–receptor models

A final analysis was carried out to identify the most likely source regions of butterflies, under the assumption that migration was aided by winds from Africa. Source–receptor models were applied to estimate the relationship between a receptor point and the probable source areas connecting receptor-recorded data with the regions crossed by the trajectory pathways. For more clarity in the results, only the restricted data set was used. 714 (17 weeks × 7 days × 6 years) daily trajectories at 1500 m were computed using HYSPLIT-4, the segment end-points of which correspond to 60-min time steps (a total of 68 544 end points). Each trajectory was associated with a corresponding value of butterfly abundance, which was the mean abundance averaged for each 7-day period. A grid with 2601 cells of 1 × 1° latitude and longitude was then superimposed on the integration region of the trajectories. The concentration field function (CFF; Seibert et al. 1994) was used to identify the source regions. This model associates an abundance value to each cell based on the residence time of the trajectories in the cells. If Cl is the abundance associated with trajectory l, and nijl the number of end points of the trajectory l in the ij cell, then:

image(eqn 1)

where Cij is the computed abundance in the ij cell. The abundance field map obtained in this manner reflects the contribution of each cell to the abundance at the receptor point. To minimize the uncertainty of the trajectories, smoothing was applied and the value of each cell was replaced by the average between the cell and the eight neighbouring cells. A final filter excluded cells with fewer than five time-steps.


phenology of the spring migration

Between 1997 and 2006, 21 episodes of significant increase in the mean abundance of V. cardui between consecutive pairs of recording weeks were identified, attributable to the arrival of migratory waves (Table 1; Fig. 3). This number increased to 32 if significantly marginal results (0·05 < P < 0·1) were also considered. April (weeks 5–9 of the CBMS calendar) was the month with the most significant increases (47·6%), followed by May (weeks 10–13, 28·6%), June (weeks 14–17, 19·0%) and March (weeks 1–4, 4·8%). Only two significant increases (9·5%) occurred in the second fortnight of June. The number of recording transects was uncorrelated with the number of yearly migrations (P = 0·48), indicating that the method used for detecting migratory waves was not biased against this variable.

Table 1.  Weeks with significant increases in Vanessa cardui in the Catalan Butterfly Monitoring Scheme between 1997 and 2006
YearWeekAfrican windsMeteorological scenario*
  • Italics, significant at P < 0·1; otherwise P < 0·05.

  • +/–, Presence or absence of winds of African origin at the same dates.

  • *

    AD, Atlantic Depression; PD, Peninsular Depression; NAH-H, North African High located at Upper levels.

199710–16 May+PD
199812–18 AprilAD
199814–20 JuneNAH-H
199926 April–2 May+PD
199924–30 May+AD
19997–13 JunePD
199914–20 JuneNAH-H
200015–21 MarchPD
200019–25 April+AD
20007–13 June+AD
200126 April–2 May+PD
200110–16 May+AD
200131 May–6 JuneNAH-H
200114–20 June+AD
200226 April–2 MayAD
20023–9 May+PD
200217–23 May+AD
200231 May–6 June+NAH-H
200312–18 April+PD
200326 April–2 May+AD
200331 May–6 June+AD
200412–18 April+PD
200426 April–2 May+PD
200410–16 May+PD
200421–27 JuneNAH-H
20065–11 AprilAD
200612–18 April+NAH-H
200619–25 April+PD
200626 April–2 May+AD
20063–9 May+PD
200610–16 May+NAH-H
20067–13 June+NAH-H
Figure 3.

Significant increases in Vanessa cardui abundance from consecutive weeks in spring (1 March–27 June) from 1997 to 2006.

back-trajectories associated with migratory waves

There was an association between migratory waves and incursions of winds from North Africa at 1500 m asl in 16 out of 21 cases (76·2%) (Table 2). While African winds also occurred in 76 of the 149 weeks (51·0%) in which no migration was detected, contingency table tests indicate a significant association between migration and African winds for the complete data set (χ2 = 4·70, df = 1, P = 0·03). When the analysis was repeated with the restricted data set, the significance of the association increased (χ2 = 7·26, df = 1, P = 0·007), with 14 out of 16 migrations (87·5%) coinciding with African winds (Table 2). Very similar results were obtained when the significance level for accepting an increase in the abundance of V. cardui was lowered to P < 0·1 (Table 2; complete data set, χ2 = 5·01, df = 1, P = 0·025; restricted data set, χ2 = 7·23, df = 1, P = 0·007). These additional analyses were made because northward low-altitude migratory flights were observed in some of the weeks that showed an increase in abundance at P < 0·1 but not at P < 0·05 (thus suggesting that a type II error was being committed for a 5% significance level).

Table 2.  Contingency tables for testing the independence of Vanessa cardui migrations and the incursions of African winds at 1500 m asl
ParameterP < 0·05P < 0·1
Weeks with migrationWeeks without migrationWeeks with migrationWeeks without migration
  1. Figures for both complete data set and a restricted data set excluding years in which the species was very scarce, and for migrations identified at both P < 0·05 and at P < 0·1 (see text for details).

Complete data set
African trajectories16762369
Non-African trajectories 573 969
Restricted data set
African trajectories14442038
Non-African trajectories 242 539

Of the different meteorological scenarios responsible for an increase in the abundance of V. cardui, the PD and the AD were equally common (37·5%), while the NAH-H accounted for only 25% of cases. Source areas under PD conditions are mostly from Algeria and Tunisia, whereas under AD conditions they are from the western Sahara (Mauritania, West Sahara, West Morocco).

low-altitude migratory flights and winds from africa

All six episodes recording intensive low-altitude northward flights, and giving rise to dramatic increases in V. cardui abundance in the CBMS, coincided with an incursion of winds from Africa, either on the same day or between 1 and 4 days before the migration was observed (Table 3). The origin of winds was confirmed by the DREAM, which forecasted abnormally high levels of African dust over Barcelona during the same periods. Thus, despite the limited number of episodes reporting massive low-altitude northward flights, the evidence provided here indicates a strong link between such phenomena and spells of winds from Africa.

Table 3.  Dates for the six episodes with intensive northward migration within the butterfly's flight-boundary layer, as reported by CBMS recorders and bird-ringers (see text for details)
EpisodeAfrican winds (no. days)Dust concentration (mg m−3)*Observations
  • *

    Dust concentration from the Dust Regional Atmospheric Model (DREAM).

  • CBMS, Catalan Butterfly Monitoring Scheme.

19–25 April 20001 (22 April)No availableNorthward flights on 22 and 24 April, CBMS
17–23 May 20021 (17 May)265 at 4500 m (17 May)Northward flights on 18 and 19 April, CBMS
26 April–2 May 20033 (28–29 April, 2 May)60 at 4000 m (29 April)Very strong northward passage in Illa de l’ Aire on 26–30 April (butterflies resting on sea objects between Menorca and Mallorca)
31 May–6 June 20033 (3–5 June)100 at 3500 (5 June)Very strong northward passage on 5 June, CBMS
10–16 May 20041 (11 May)340 at 4500 m (12 May)Very strong northward passage on 13–16 May, CBMS and Illa de l’Aire
10–16 May 20063 (13, 15–16 May)100 at 1800 m (15 May)Very strong northward passage on 14–15 May, CBMS
  550 at 4000 m (16 May) 

As an example, Fig. 4(a) shows the situation for the migratory wave arriving in the region in mid-May 2004 that was especially conspicuous in week 11 of the CBMS (10–16 May). At least four different recorders reported massive low-altitude northward flights of V. cardui on 14 and 15 May, while other independent observations were made during the same period at places as distant as the Pyrenees (at 2000 m, on 16 May) and L’Illa de l’Aire (Menorca, 11–14 May).

As deduced from back-trajectories at 1500 m, African winds originating around Algeria reached the Balearic Islands and approached the coast of north-eastern Spain on 11 May (Fig. 4b). The most accurate representation provided by the Navy Aerosol Analysis and Prediction System (NAAPS) model (http://www.nrlmry.navy.mil/aerosol/) shows how a cloud of African dust was displaced at high atmospheric levels from central Algeria to the Balearic islands and the north-eastern Spanish coast on that day (Fig. 4c). This dust intrusion was recorded over the city of Barcelona on 12 May, with a maximum level of 340 µg m−3 at 4500 m, and abnormally high levels also evident at heights of 100–2000 m (Fig. 4d). These southerly African winds were provoked by the passage of a PD over northern Algeria on 12 May (Table 1). As soon as the low-pressure cell had moved eastwards, a fierce northerly wind began to blow over Menorca and L’Illa de l’Aire on 13 May and especially on 14 and 15 May (when this wind – the tramuntana – reached a force of 4–5 on the Beaufort scale). This can be seen in Fig. 4(b), which shows how the direction of the back-trajectories changed from south to north between 11and 12 May (11 May back-trajectory in black).

Although some migrating butterflies were recorded on Menorca on 11 May, it was only on 13–15 May that numbers increased dramatically (transect counts on 11, 12, 13, 14 and 15 May = 1, 3, 16, 113 and 81 individuals, respectively), probably because butterflies were forced to fly closer to the ground in the face of unfavourable head winds.

source of migrants

The CFF source–receptor model applied to the restricted data set enabled several sources for spring migrants to be identified, most of which are located along a vast area reaching between 30 and 35°N and 15°W to 20°E within the bioclimatic limit of the Mediterranean region in North Africa (Fig. 5). The three main sources were all located in the Maghreb around the Moroccan Anti-Atlas (and to a lesser extent in the equivalent area south of the Saharan Atlas in Algeria), in the low plain of north-eastern Algeria stretching away across central Tunisia and up to the coast, and in the coastal area of Tripolitania in Libya. A fourth smaller, but apparently still important, source area was located much further south, at the northern limit of the Sahel, between Western Sahara and Mauritania (≈ 21–22°N and 13–15°W). Moreover, although incorrectly situated too far north, the model also indicated the presence of a secondary source area in the Canary Islands.

Figure 5.

Abundance of Vanessa cardui (individuals per 100 m transect) computed with a source–receptor model (concentration field function; Seibert et al. 1994) applied to spring counts for the period March–June 2000–04 and 2006. See text for details.

It should be noted that the other areas in Fig. 5 that seem to be associated with a high abundance of V. cardui do not correspond to true migrant sources. For instance, the area located to the north-east of the British Isles represents an artefact of the model, caused by the fact that, quite commonly, African winds associated with migrants are immediately succeeded by northerly winds originating in Iceland and to the north of the British Isles following the eastward passage of a low-pressure cell. The migrants are presumably forced down to low levels when they encounter a switch to northerly winds, and so will become visible at ground level, as previously shown for the episode of massive migration on 10–16 May 2004. Likewise, the uniform area of high abundances recorded over the south-western Mediterranean represents the main route followed by migrants from their source areas in Africa.


Seasonal poleward movements allow a great variety of organisms to exploit resources that become available in spring and summer in the temperate zone, and to escape from the increasingly inhospitable environments of southern regions as the season advances. The large number of species showing such springtime movements suggests that there is strong selection pressure for migratory behaviour (Johnson 1995) and, not surprisingly, many of the organisms involved in this phenomenon have evolved a set of specific attributes aimed at facilitating migratory journeys (Dingle 1996). One such adaptation is the use of winds associated with the passage of fronts, which has been reported for birds (Gauthreaux 1991) and many insects migrating in the temperate zone (see reviews by Pedgley 1982; Drake & Farrow 1988; Johnson 1995; Pedgley, Reynolds & Tatchell 1995).

However, most migratory butterflies do not appear to conform to this pattern, as they reportedly stay within their flight-boundary layer and exert some control over their direction of displacement (Walker 1980; Walker & Riordan 1981). One particularly good example is provided by V. cardui, a highly mobile species that would be expected to have evolved behavioural adaptations such as windborne migration. Despite this, numerous observations of poleward migration flights occurring near the ground, surprisingly often against northerly winds (Williams 1930; Abbot 1951; Tilden 1962; Stefanescu 1997), seem to contradict this assertion. One possible explanation for this apparent maladaptive behaviour in this and other butterfly species was advanced by Johnson (1969), who indicated that most observations simply recorded individual tracks over relatively short distances that are difficult to interpret. He even suggested that records of low-altitude directional movements may correspond to situations of relatively unfavourable weather for flying, representing only part of a process that also involves flight at much higher altitude. He eventually concluded that tracking the whole population displacement in relation to large-scale weather systems would provide the key to the question.

Although most of these criticisms are fully applicable to the bulk of existing data on V. cardui migration, recent systematic butterfly recording (for example in the form of the Butterfly Monitoring Scheme; Pollard & Yates 1993) has provided a tool for exploring further the nature of migrations of this and other butterfly species. In our study, we have used more than 20 000 individual records obtained from weekly CBMS counts in north-eastern Spain to identify reliable peaks of V. cardui migration. The possibility that spring migrants use synoptic wind systems was tested by backtracking air trajectories at 1500 m asl on days with and without associated migration, and comparing their frequencies in relation to the occurrence of winds of African origin. We consistently found a significant association between spring migrations and African winds, strongly suggesting that migrants were carried on the moving air masses above their flight-boundary layer. These air masses were related to meteorological scenarios involving the passage of depressions approaching the Mediterranean region from the west or south.

Interestingly, a close association between migrants and winds from North Africa was also found during the six episodes in which massive low-altitude northward flights were recorded in the region. In some cases, directional flights occurred with time lags of 1–4 days with respect to the incursions of African winds. A possible explanation is that butterflies undertake high-altitude windborne migration for most of the distance travelled from source areas, but descend to ground levels and then continue migrating within their flight-boundary layer on fixed compass bearings for a period of time, perhaps even for days. As noted by Gatehouse & Zhang (1995), this change in behaviour would presumably commence with the transition from an initial phase of truly migratory behaviour to a second phase, in which the migrants regain responsiveness to stimuli associated with favourable habitats. A similar explanation was provided by Mikkola (1986), who suggested that the migration of several butterflies (including V. cardui) and moths into Finland consisted of two phases, the first involving windborne flights over long distances, the second much shorter displacements near the ground. Physiological studies coupled with behavioural laboratory experiments (such as flight tests using tethered butterflies) may help to elucidate these aspects.

Some reports of high-level flights in V. cardui provide additional clues that windborne migration could indeed be a common phenomenon in this species. Giuliani & Shields (1997) recorded two large spring migrations in California extending up to at least 100–300 m above ground level. Most butterflies appeared to be drifting, allowing themselves to be carried by the wind; in one instance migrants were seen on top of a 2315-m ridge, hence the displacement was actually taking place at ≈ 2500 m asl. Mikkola (2003) also reported a few individuals of V. cardui on autumn migration at a Finnish site, flying south on tailwinds at ≈ 20–100 m above ground level among large numbers of V. atalanta, and radar observations at a nearby site indicated that many unidentified insects were moving south at heights of up to 3000 m. Likewise, unusual records of V. cardui at sea, far from any continental source, and data on the daily rate of advance, have been interpreted as indirect evidence of windborne migration (Bowden & Johnson 1976; Mikkola 1986). All in all, the evidence gathered so far leads us to believe that the scarcity of records for V. cardui of high-altitude flights mainly responds to the practical difficulties of observing butterflies far above the ground. The use of alternative techniques, such as the vertical-looking radar described by Chapman, Reynolds & Smith (2003), could add crucial data for understanding the mechanistic basis of this migratory system.

Although it is widely accepted that V. cardui migrating into Europe are of African origin (e.g. Pollard et al. 1998), there are very few observations of mass emergence of adults of this species in Africa (Skertchly 1879; Egli 1950 cited by Johnson 1969, p. 42). Our study is the first to use testable predictions of the location of these source areas. According to our source–receptor model, migrants arriving in north-east Spain come mainly from three different areas in the Maghreb, roughly 900, 1300 and 1600 km away (Fig. 5). If we assume that migration is mainly windborne, and that it takes place mostly at altitudes of 500–1500 m asl (where air masses move at a mean speed of 30 km h−1 and usually have temperatures in the range 10–20 °C in the spring), the journey from the Maghreb may take 30–53 h with a tailwind. If we also include in the estimate the speed of the insect's flight (15 km h−1; cf. Abbot 1951), these figures drop to 20–36 h. This suggests that part of the migration is done at night, which appears to be supported by fairly frequent records of V. cardui being attracted to electric lights during favourable migration periods (Ryrholm & Källander 1986).

Source areas in Africa could be confirmed by direct observations of massive concentrations of larvae or adult emergences at predicted times or, alternatively, by the application of techniques such as the analysis of isotopes in migrants (Wassenaar & Hobson 1998). In addition, Butterfly Monitoring Scheme data on population levels in spring could be used for testing forecasted breeding success in the hypothetical source areas, according to particular environmental conditions (such as rainfall levels or a combination of several climatic variables; Myres 1985; Vandenbosch 2003). Of further interest would be the use of data from different Butterfly Monitoring Scheme networks to obtain a comprehensive picture of the progress of V. cardui across Europe each season, as well as the identification of possible source areas, which may vary from one European country to another.


Thanks are due to all the recorders who contribute data to the Catalan Butterfly Monitoring Scheme. A special mention is also due to Lluc Julià and Guillem Alfocea for their careful observations of V. cardui migrations while ringing birds on L’Illa de l’Aire in Menorca. Technical assistance was provided by Pedro Arnau, Jordi Jubany and Ferran Páramo, and statistical advice by Javier Retana. We are very grateful to the insights and comments made by Ernie Pollard and Daniel Sol on earlier drafts of the manuscript. Useful comments were also provided by Oriol Battestini, Jason Chapman, Carolyn Jewell, Mike Lockwood, Don Reynolds, Ferran Rodà, David Roy, Nils Ryrholm and Roger Vila. The Butterfly Monitoring Scheme in Catalonia is funded by the Departament de Medi Ambient i Habitatge de la Generalitat de Catalunya. The Diputació de Barcelona, Patronat Metropolità Parc de Collserola, Fundació Territori i Paisatge have also given financial support to this project. Funding from the Spanish Ministerio de Educación y Ciencia through project CGL 2005-07543 is fully acknowledged.