3.1.1.2. Wing venation pattern analysis Particular discrepancies exist in the measurement and analysis of wing venation, whose pattern variation is of specific interest in insect taxonomy in general. Due to their key role and the current interest, they are discussed here in greater depth. At present, three different main approaches are in use: classical wing morphometry as defined by Ruttner (1988), the DAWINO (Discriminant Analysis With Numerical Output) method (www.beedol.cz), and geometric wing shape analysis (Bookstein, 1991). [h=1]3.1.1.2.1. Classical wing morphometry[/h]Classical wing morphometry captures variation in wing shape by calculating 11 angles between 18 junctions in the wing venation (Fig. 1) which constitute a subset of a suite of 17 angles first introduced into bee morphometry by DuPraw (1965). The DAWINO method consists of the full set of DuPrawâ€™s angles, supplemented by 7 linear measurements, 5 indices, and one area (Table 3). All these angles and other parameters are considered as measurement characters in further analysis, where they can be combined with measurements of body characters. In the past decade, these somehow idiosyncratic morphometric methods for the bee-wing were increasingly replaced in a number of studies by "geometric morphometry", based on the theory of shape, which is explained here in more detail. Fig. 1. Wing angles in classical wing morphometry (Ruttner, 1988). [h=1]3.1.1.2.2. Geometric wing shape morphometry[/h]The geometric morphometric method is based on the coordinates of landmarks located at vein intersections of the wing (Bookstein, 1991; Smith et al., 1997). The fundamental advances of geometric morphometrics over traditional approaches include (i) the way the amount of difference between shapes can be measured (using Procrustes distance), (ii) the elucidation of the properties of the multidimensional shape space defined by this distance coefficient, (iii) the development of specialized statistical methods for the study of shape, and (iv) the development of new techniques for the graphical representations of the results (Bookstein, 1991; Rohlf, 2000; Mendes et. al., 2007). In the measuring process, landmark coordinates are superimposed using a Generalized least Squares (GLS) Procrustes Superimposition (Rohlf and Slice, 1990): Specimens are centered, normalized to unit centered-size (Bookstein, 1991) and interactively rotated to minimize the sum-of squared distances between each location and its sample mean. Shape differences are then analysed by multivariate analysis of variance (MANOVA), Canonical Variate Analyses (CVA) and Multiple Discriminate Analyses, and shape patterns along the canonical axes are estimated by multivariate regression (Monteiro, 1999). For this kind of analysis, the software packages Morpheus (Slice, 2002), NTSYS (Rohlf, 1990), MORPHOJ (Klingenberg, 2011) and DrawWing (Tofilslki 2004) are commonly used. The new methods of automated measures and geometric morphometry have been used to distinguish Africanized honey bees from African and European subspecies, and to characterize the evolutionary lineages of A. mellifera (Francoy et al., 2006, 2008; Baylac et al., 2008; Tofilski, 2008; Miguel et al., 2010; Kandemir et al., 2011). This method has also been used to analyse differences between three honey bee subspecies in Poland: A. m. mellifera, A. m. carnica, and A. m. caucasica (Tofilski, 2004, 2008). The high rates of correct classification obtained with the geometric method indicate that forewings carry sufficient information to distinguish between different groups of bees. Discrimination results obtained with this method proved superior to classical wing angles, although the degree of improvement was moderate (Tofilski, 2008). [h=1]3.1.1.2.3. Classical wing angles, geometric wing analysis, and full body character suites[/h]A main difficulty associated with this recent diversification of wing character suites is lack of downward compatibility. The morphometric definition of the currently recognized Apis mellifera subspecies has been based on traditional classical morphometry; consequently, studies based on a geometric character set cannot utilize reference data generated with traditional wing angle measurements accumulated in previous work by various authors, including the reference subspecies descriptions as given in Ruttnerâ€™s (1988) monograph. Fortunately, however, there is a high degree of consensus concerning the marking points between the different kinds of wing shape analysis. Fig. 2 shows the 20 landmarks predominantly used in geometric morphometry. Tofilski (2004, 2008) omitted landmark 15, while Kandemir et al. (2011), following Zelditch et al. (2004), moved point 15 from the apex of the radial cell to the junction of Rs5 and the costa, and located one additional point at the end of the vannal fold. Classical morphometry and the DAWINO method use the same landmarks, omitting point 15. However, the methods disagree in the sequence of numbering these points, which is of no major concern. To overcome the present sets of incompatible data and to avoid further parallel development of incompatible data sets in honey bee morphometry, our suggestion for a standardization of wing measurements is to store all future data as point coordinates (instead of the format of derived characters such as angles) to facilitate data exchange between different studies and research teams. We suggest using the point format exemplified in the description of Apiclass (http://apiclass.mnhn.fr), shown in Fig. 2. From these coordinates, used in a majority of geometric studies, all 30 DAWINO characters can be calculated, which include the Ruttner (1988) wing angles as a subset. Storing the point coordinates instead of calculated characters will also keep all options open for any future progress in analysis techniques. Unfortunately, however, the coordinates cannot be recreated from classical wing angles, but first attempts have been made to re-measure reference samples with the geometric method (Kandemir et al., 2011) to link geometric morphometry to subspecies characterizations obtained by the classical method. As a further issue the question arises whether geometric morphometry should replace "classical" morphometry for good, meaning that the accurate, powerful and labour-effective shape analysis based on wing geometry alone should replace the full set of classical characters, including all traditional body characters. Phylogenetically, the wing venation is more informative compared to the more environmentâ€“sensitive character categories of size, colour or pilosity (Diniz Filho et al., 2000), and thus represents a character set somewhat comparable to molecular characters. A high degree of consistency between wing morphometry and molecular information has been demonstrated by Miguel et al. (2010). Therefore, wing geometry is particularly suitable to track phylogenetic relationships between subspecies, where the full "classical" character set can be misleading. However, aiming at an inventory of honey bee variation as a numerical account of ecotype morphology, it appears indispensable to maintain classical morphology with a broad character set to represent the actual features of subspecies or ecotypes, apart from and in addition to the question of their phylogeny. However, geometric wing venation morphometry might replace the classical wing angles even within the classical morphometry set, but no attempt has been made so far to combine these methods. Fig. 2. Location of the nineteen landmarks on the fore-wing of honeybee workers considered in the geometric morphometric analysis (CC = cubital cell) http://apiclass.mnhn.fr; Miguel et al., 2010). http://www.coloss.org/beebook/I/subspecies-ecotypes/3/1/1/2