Journal of Insect Conservation 4: 253–261, 2000. © 2001 Kluwer Academic Publishers. Printed in the Netherlands. How useful is fluctuating asymmetry in conservation biology: Asymmetry in rare and abundant Coenonympha butterflies Jack J. Windig1, Pekka T. Rintamaki¨ 2,∗, Anna Cassel3 &Soren¨ Nylin4 1Institute of Zoology, Stockholm University, 106 91 Stockholm, Sweden and ID-Lelystad Institute of Animal Science and Health, P.O. Box 65 8200 AB Lelystad, The Netherlands 2Department of Population Biology, Evolutionary Biology Centre, Uppsala University, Norbyvagen¨ 18 D, 752 36 Uppsala, Sweden 3Department of Conservation Biology and Genetics, Evolutionary Biology Centre, Uppsala University, Norbyvagen¨ 18 D, 752 36 Uppsala, Sweden 4Institute of Zoology, Stockholm University, 106 91 Stockholm, Sweden ∗Author for correspondence: Pekka T. Rintamaki,¨ Department of Population Biology, Evolutionary Biology Centre, Uppsala University, Norbyv¨agen 18 D, 752 36 Uppsala, Sweden (tel.:+46 736746165; fax:+46 18 4716424; e-mail: [email protected]) Received 14 December 2000; accepted 8 January 2001 Key words: Coenonympha, population size, population quality, fluctuating asymmetry, conservation Abstract It has been suggested that minor, fluctuating differences in size of bilateral traits could validly indicate individual differences in developmental stability. One plausible reason for instability to occur could be lowered population size, which has been suggested to increase fluctuating asymmetry due to inbreeding, for example. Wemeasured seven wing asymmetries of three Coenonympha butterfly species in central Sweden. One species is abundant (C. pamphilus), one rather common (C. arcania), and one rare (C. hero). We expected that if fluctuating asymmetry is a reliable indicator of population quality and thus a useful tool for conservation purposes, the most abundant species should show lowest asymmetry and the most endangered, the highest. Contrary to our expectations, the highest wing asymmetry was found in the relatively common species C. arcania and the most abundant and rare species did not show significant differences in levels of wing asymmetry. Our results obtained from three Coenonympha species hence suggest that the use of fluctuating asymmetry as an indicator of population conservation status may be misleading. Possible increase in asymmetry of small and/or isolated populations of butterflies may be masked by local differences in environmental conditions that could have high impact on bilateral development as well. Introduction asymmetry might be a very useful tool in conserva- tion biology (Clarke 1995; Møller & Swaddle 1997), Fluctuating asymmetry (FA), ‘random departures from in reflecting various forms of environmental change. anticipated bilateral symmetry’ (Van Valen 1962; There are, however, conflicting reports on the rela- Markow 1995), is thought to reflect quality of organ- tionship between quality and asymmetry (reviewed isms (Møller & Swaddle 1997, but see for example amongst others by Palmer & Strobeck 1986; Parsons Clarke 1998; Simmons et al. 1999). The idea behind 1990; Møller & Swaddle 1997; Clarke 1998; Palmer this concept is that developmentally unstable genotypes 1999; Simmons et al. 1999). Two recent examples of a cannot control their development precisely, and conse- relationship between FA and quality are: elevated levels quently more often develop different phenotypes on the of FA in five tropical bird species living in fragmented left and right side. Inbreeding and high levels of stress, habitats, compared to large undisturbed area of rain for example, may induce asymmetry. Thus fluctuating forest (Lens et al. 1999) and increased levels of FA in 254 J.J. Windig et al. threatened butterfly species compared to species with The number of populations is decreasing because of stable populations in Denmark (Poulsen 1996). Two reforestation of suitable habitat. A reduction in fitness recent examples of no relationship between FA and related traits has also been documented in local small quality are cactophilic fruitflies, Drosophila pachea, and isolated populations (Cassel et al. 2001). One of us that did not have elevated levels of FA under food stress (AC) sampled one of the remaining populations in cen- (Hurtado et al. 1997), and the rare British butterfly tral Sweden, east of Arvika (Varmland¨ province) in the Plebejus argus where bottlenecked populations did not summers of 1997 and 1998. This population was one have elevated levels of FA (Brookes et al. 1997). of the larger and consisted of about 200 adults. Total In this paper we evaluate asymmetry in wings sample size consisted of 25 males and 21 females. and eyespots of three Coenonympha species. Poulsen The small heath (Coenonympha pamphilus)isone (1996) reported that asymmetry in wing length in of the most abundant butterfly species in Swedish decreasing Danish butterfly populations tended to be grasslands. Two of the present authors (JJW and SN) higher than in stable populations. She included among sampled it in the national city-park of Stockholm in others pairs of Coenonympha species and reported ele- the summer of 1997. There are many extensively man- vated levels of FA in decreasing populations. Here we aged grasslands (predominantly mown once a year) in examine whether asymmetry differs between abundant this park. It has a large population, of several thou- and decreasing Swedish Coenonympha species, taking sand, possibly more than ten thousand individuals, of into account the effects of measurement error, trait size Coenonympha pamphilus. Total sample size consisted and type of asymmetry. We end with a discussion of the of 27 males and 24 females. usefulness of analysing FA for conservation purposes The pearly heath (Coenonympha arcania) is still in butterflies. rather abundant in Sweden. PTR sampled it near Lovsta¨ (Uppland province near Uppsala) in the summer of 1998. Population size is probably a few hundred indi- Methods viduals. Total sample size consisted of 26 males and 11 females. Species Traits and measurements Three species were analysed. The scarce heath (Coenonympha hero) is placed on the red list of A total of seven fore and hind wing traits were measured Swedish butterflies as vulnerable (Ehnstrom¨ et al. 1993). (Figure 1). These were: area of fore and hind wing, Figure 1. Schematic representation of the three species analysed and measurements performed. Besides measurements shown, wing area for both fore and hind wing were measured as well. Fluctuating asymmetry in Coenonympha butterflies 255 length of fore and hind wing, width of fore and hind the consequence of a slight bias in the system. Change wing, and total area of the eyespot on the ventral side in the order left and right wings were measured did not of the fore wing. The fringe of hairs at the edge of the influence the asymmetry, nor did the position of the wings was excluded from all measurements since it was wings in the image, or whether wings were measured worn off in some individuals. Several width and length on the dorsal or ventral side. When wings were dam- measurements were tried, and the following were cho- aged the damage was either corrected manually, e.g. a sen that gave the smallest error. The length of the fore small part missing at the edge of the wing was drawn wing was measured from the wing base to where vein on the screen with the mouse when the original contour M1 (nomenclature, see Chinery 1978) reached the edge of the wing was clear, or in most cases the wing was of the wing. The length of the hind wing was measured discarded for that measurement. from the angle at the edge just above the base of the wing to where vein Cu1a reached the edge. The width Statistical analysis was measured from the edge below the vein 1A to vein R3+4 in the fore wing, and from vein 2A to M1 in the For a better overview and to reduce the number of hind wing. The eyespot measured was the only eyespot statistical tests, we used principal component analy- that was present in all individuals in all three species. sis (PCA) to transform wing size measurements into Measurements were made with an image analyser. two traits. PCA combines, based on correlation coef- An image analyser consists of a video camera coupled ficients, traits into new traits (principal components) to a computer, and allows for very precise measure- so that the first principal component (PC1) describes ments of butterfly wings (Windig 1991). The image most of the variation originally separately present in analyser used here consisted of a JVC TK 5066E the old traits. Each following principal component Camera and a Leica image analysis system. Measure- describes the largest part of variation that is not corre- ments were performed semi-automatically with the lated to previous principal components. PCA was run help of a program written by JJW in the Qwin image on a data set of the three species and two sexes com- analysis software (produced by Leica). Wings were bined. Wing areas were log-transformed before PCA, detached from the body before measurement, by care- to obtain normally distributed variables. PCA resulted fully cutting their connection to the body at the root in the first two components describing over 95% of of the wings. The program automatically thresholded a the variation. If for an individual a measurement was wing from the background for the area measurements missing for one trait only, a dummy measurement was (= number of pixels converted to mm2). Start and end- filled in for this measurement with the help of mul- points for width and length were indicated manually on tiple regression, before the PCA was performed. If the monitor. Both were more than 300 pixels in length more than one measurement was missing, individuals so that pixellation error (warned for in Palmer 1994) were discarded. Results were hardly influenced by the was not large. Area of eyespots were measured by first inclusion of dummy measurements. Only the statistical thresholding black (white) colour from the rest of the power would have been lower, and not enough females image, and then manually indicating which black and of C.