Morphometric analyses were used to characterise all taxa. High-resolution pictures of wings and surstyli were made using a Leica DFC320 video camera attached to a Leica MZ16 stereomicroscope. A video camera was connected to a PC computer in which the software to make pictures was installed. Landmarks (wings) and semi-landmarks (surstyli) were drawn on every picture using TpsDig 2.05 software (Rohlf, 2006). One-way analysis of variance (ANOVA) and Tukey’s post hoc test were used to test differences in wing centroid size between sexes and taxa. Wing centroid size is the square root of the sum of the squared distances between the centre of the wing and each landmark (Zelditch et al., 2004). Multivariate analysis of variance (MANOVA), canonical variate analysis (CVA) and discriminant function analysis (DA) were used to test differences in wing and surstylus shape between sexes and taxa. Statistical analyses were made using Statistica® for Windows (StatSoft, 2012). For males and females of morphotypes A and B analyses were carried out separately.
Variation of wing size and shape was studied in 249 specimens of all three taxa following Bookstein (1991). The right wing of each specimen was taken off by means of micro-scissors and then mounted in Hoyer’s medium (Anderson, 1954) on a microscopic slide. Wings are archived and labelled using unique codes shown in Table S1 (Supplementary information). Sixteen homologous landmarks were chosen at vein intersections and terminations throughout the wing; landmarks were selected in positions of the wing that could be easily recognisable at any time (Fig. 1). To reduce the statistical errors due to the low number of specimens (Arnqvist and Mårtensson, 1998) wings of Chrysotoxum orthostylum Vujić sp. nov., were digitized 10 times. Generalised least squares Procrustes superimposition (GLS) was used to minimise non-shape variations in location, scale and orientation of wings, and also to superimpose the wings in a common coordinate system (Rohlf and Slice, 1990; Zelditch et al., 2004). GLS, wing centroid size and partial-warp scores were computed using CoordGen7.14 and CVAgen7.14a, which are part of IMP package (Sheets, 2012). MorphoJ v2.0 was used to visualize the thin-plate spline deformation (Klingenberg, 2011).
For surstyli, outline shape was studied. Surstyli of 51 Chrysotoxum males were analyzed: 21 of morphotype A and 30 of morphotype B. The right surstylus was taken off using pins. Surstyli were mounted in Hoyer’s medium on a microscopic slide and immobilized with a cover slip. In the absence of clearly-identifiable, homologous, anatomical loci, 30 semi-landmarks were generated (Fig. 2). For each specimen semi-landmarks were drawn twice to reduce statistical error. To superimpose semi-landmarks, CoordGen7.14 and an integrated Semiland module were used following a distance-minimising protocol to minimize the shape differences due to the arbitrary nature of semi-landmark positions along the curve (Bookstein, 1997; Zelditch et al., 2004).