Migratory insects provide a valuable ecosystem service by transporting large amounts of organic matter across regions where they become temporarily abundant prey. In species performing multigenerational migration, such as the painted lady butterfly Vanessa cardui, successive generations face a wide variety of predator communities and may be subject to different predation risks. Here, we analyze the pattern of wing damage of over 2,000 butterflies to investigate, for the first time, the risk of predation of adult painted ladies across a latitudinal range of ca. 3,500 km extending from the northern Mediterranean through the Maghreb to sub-Saharan West Africa. Large number of butterflies showed substantial wing damage attributable to failed attacks, with birds, mantids and lizards being the most likely predators. The risk of attack increased towards the equator, even after controlling for wing wear. In addition, there was a strong effect of butterfly size on predation risk, with larger butterflies facing a higher risk compared to their smaller counterparts, and clear evidence that females suffered more attacks than males. Although size is a major factor, latitude was a stronger predictor of predation risk across the migratory system, as evidenced by greater wing damage in butterflies at lower latitudes, even though their size notably decreased. These results raise an interesting evolutionary conflict, with a trade-off between size and predation risk, as larger butterflies are likely to be more fecund and efficient in migratory flight but, at the same time, more vulnerable to predation.

Butterfly collection and sampling regions

Between 2015 and 2023, a total of 2,276 painted ladies were collected in three regions (Fig. 1, Supporting information). Sampling was designed to obtain representative collections of the study regions, while avoiding spatial and temporal biases: (1) northern Mediterranean, 891 butterflies from 102 sites in northeast Spain (38-43ºN, 0-4ºE); (2) Maghreb, 1,247 butterflies from 141 sites in Morocco (28-35ºN, 2-11ºW); and (3) sub-Saharan West Africa, 138 butterflies from 22 sites in Senegal (11-15ºN, 11-17ºW). In the northern Mediterranean, butterflies were collected each month from March to October; in Morocco, they were collected in autumn (October-November), spring (April), and winter (January-February); and in Senegal, in October.

Butterflies were sexed, and their wings were removed and stored in glassine envelopes for further measurements. To account for the potential influence of age on wing damage, wing wear (i.e., a surrogate of age) was assessed on a scale from 1 to 5, as described in Stefanescu et al. (2016). Wing size was measured with a calliper (to the nearest tenth of millimeter), from the apex to the insertion on the thorax for either forewing.

Wing damage measurements and predation risk

Wing damage was measured for each of the four wings and quantified as the percentage of wing-area loss using digital images and the open-source platform for biological-image analysis Fiji (Schindelin et al. 2012; Supporting information).

As wing damage can result from both wear and tear associated to routine behavioral activities, or be the consequence of unsuccessful predator attacks (Korkmaz et al. 2022), we examined the distribution of wing area loss through histograms (Supporting information). Taking an average of area loss across the four wings, more than half (60.1±10.3%) of the sample showed no loss or a loss of less than 1% on their wings, while the remaining butterflies showed a gradual decrease from 1-2% wing area loss (10.4 ± 1% of individuals) to 40-50% loss in rare and extreme cases (2.2 ± 2.1% of individuals). This pattern suggests that wing damage in the first interval is simply explained by wear and tear from normal activities as the butterfly ages, whereas values below 99% of the wing area may often be related to unsuccessful predator attacks. Therefore, predation risk was treated as a binary variable, with individuals classified as non-attacked (i.e., 0-1% wing area loss, calculated as the average of the four wings) or attacked (>1% wing area loss). Although this method is likely to overestimate the frequency of attacks, the error is distributed homogeneously across the entire sample and does not affect the hypotheses being tested.

It is important to recognize that the actual predation rate cannot be directly inferred from wing loss data alone (e.g., Bengtson 1981, Burger and Gochfeld 2001). Despite this, when accounting for the degree of wing wear, the incidence of wing loss is likely correlated with the frequency of predator attacks, thereby serving as a proxy for assessing predation risk.