Preliminary results of the shape analysis of pine processionary moth scales in Turkey

Larvae of the two pine processionary moth (PPM) sister species, Thaumetopoea wilkinsoni and T. pityocampa, causes severe allergies in mammals and are dangerous defoliators of Mediterranean pines. While T. pityocampa occurs mainly in Europe, T. wilkinsoni occurs in Turkey and the Middle East. Recent studies showed that ranges of the two species are in contact in Turkey. Female moths of the two species cover their eggs with scales on their abdomens. These scales are believed to protect eggs from parasitism. Although T. wilkinsoni and T. pityocampa exhibit highly similar morphologies, few studies have suggested that scale morphology could be used for species identification. However, our field observations in Turkey do not confirm this suggestion. Nevertheless, statistical approaches have never been used on this issue. In this study, we aimed at solving this contradiction by digitizing PPM scale shape photographs taken standardly from 60 individuals collected from different locations in Turkey and Cyprus and by subjecting them to geometric morphometry approaches. We conducted shape analysis by using TPS software and Elliptic Fourier Outline Analysis which is a sensitive method for morphometric computations. Results did not reveal any correlation between scale morphology and species taxonomy. This result could be interpreted as an insufficiency of scale morphology being able to be used as a taxonomic character to separate T. wilkinsoni and T. pityocampa in Turkey. However, it should be kept in mind that these two species have a contact zone in Turkey, which might cause hybridization. This phenomenon could completely change morphological patterns. In order to reach more accurate results, this study should be repeated with samples from the entire ranges of the two species.

PPM is a univoltine organism. Its gregarious larvae feed on the pine host during the winter, descend to the soil in spring for pupation, and adults fly in late summer or early autumn depending mostly to latitude and altitude; and females lay their eggs on pine needles immediately after mating (Avcı and Oğurlu, 2002;Avcı, 2000;Démolin, 1969;Halperin, 1990). Female moth covers its oviposited eggs with the scales on its abdomen. These scales probably protect the eggs from parasitoids and desiccation (Avcı, 2000;Floater, 1998, Mirchev et al., 2004, Özkazanç, 2002, Schmidt, 1990. T. wilkinsoni and T. pityocampa exhibit highly similar morphologies, but few studies have suggested that scale morphology could be used for species identification (Doğanlar et al., 2005;Tsnakov et al., 1991). Although our field observations do not confirm this suggestion (İpekdal, 2012) statistical approaches have never been used so far to solve this ambiguity.
In this study we aimed to quantify shapes of the scales collected from different localities in Turkey and Cyprus. This study is the first attempt to investigate scale morphology statistically and to reveal its suitability as a taxonomic character. Our results did not reveal any correlation between scale morphology and species taxonomy.

Material and methods
We collected 60 PPM egg batches from 13 localities in Turkey and one in Cyprus. According to İpekdal (2012) 45 of them were identified as T. wilkinsoni, 12 as hybrids, and 3 as T. pityocampa. We picked off 20 scales per egg batch and fixed them on microscope glasses by using Entellan (Merck). We took photographs of the fixed scale preparations under a Leica DC300-MZ75 digital dissection microscope at Hacettepe Uni. Dept. of Biology and Ecological Sciences Research Lab. by using a magnification of 0.32 and scaling of 1mm. Then the two-dimensional outline of the scale was digitized. The starting point of the outline was defined at the peak point of the scale (Figure 1).
For each scale, 64 points at equally spaced intervals along the outline were sampled using TpsDig2 (Rohlf, 2010). This set of x, y-coordinates was then analysed using an Elliptical Fourier Analysis (EFA) using the software EFAwin (Ferson et al., 1985). This method is based on separate Fourier decompositions of the incremental changes along x and y as a function of the cumulative length along the outline (Kuhl and Giardina, 1982). The outline is approximated by a sum of trigonometric functions of decreasing wavelength: the harmonics. Any harmonic corresponds to four coefficients: A n and B n for x, and C n and D n for y, defining an ellipse in the xy-plane.
The first ellipse corresponds to the best-fitting ellipse to the outline. Its major axis was taken as new x-axis to adjust the orientation of the outline (Rohlf, 1990). Its area was used to standardize the Fourier coefficients (FCs) for size differences in order to eliminate isometric size effects and to concentrate on shape information only. Since the coefficients A 1 , B 1 and C 1 correspond to residuals after standardization (Crampton, 1995;Renaud et al., 1996) they were not included in the subsequent statistical analysis. The coefficient D 1 still retains information about the elongation of the outline (Michaux et al., 2007). It was thus included in the statistical analyses.
In order to evaluate the adequate threshold harmonic for the analysis of PPM scale, the shape of one UM1 was measured ten times, providing an estimate of the measurement error for each harmonic (Figure 2). Measurement error was estimated as the percentage of error for the amplitude (square root of the sum of the squared FCs). In our case, measurement error was low (<5%) until the fifth harmonics and abruptly increased thereafter ( Figure  2). The content of information of each harmonic provided information about the amount of shape information provided by that harmonic (Crampton, 1995). The amplitude of each harmonic wass cumulated over the total range of harmonics, and the information brought by each harmonic wass then estimated as the percentage of this sum represented by its harmonic. Each of the five first harmonics increased significantly the amount of shape information up to 97.07% of the total information ( Figure 2). The plateau reached afterwards showed that the subsequent harmonics brought almost no further relevant shape information. Hence, considering the set of the first five harmonics appeared as a good compromise between measurement error, information content, and the number of variables to be considered. A data set of 17 variables (20 FCs minus A 1 , B 1 and C 1 ) was thus retained for subsequent analyses.
Statistical analyses: Principal component analysis (PCA) was performed on the set of 17 FCs in order to test differences between the two species T. wilkinsoni and T. pityocampa. All statistical analyses were performed by using SYSTAT v.11.

Results and discussion
The first axes of principal component (PC1), based on 17 variables came from elliptic fourier outline analysis of scale measurements, accounted for 94.6% of the total morphometric variation and seemed to express the overall shape of the scale with the populations of T. wilkinsoni and T. pityocampa and their hybrids.
Our results did not reveal any significant correlation between scale morphology and taxonomy. However, we found a slight differentiation between T. wilkinsoni and T. pityocampa on the first axis of PCA ( Figure 3); yet it is not enough to seperate the two PPM species and their hybrids. Thus we can conclude that scale morphology is an insufficient character to be used for seperating T. wilkinsoni and T. pityocampa in Turkey. However, it should be kept in mind that these two species have a contact zone in Turkey. Hybridization could completely change morphological patterns. Furthermore, our sampling size could cause biased results. Therefore, in oreder to reach more accurate results, this study should be repeated with a larger sampling size from the entire ranges of the two species.