3. Results
3.1 Sample site characterisation
Linear permutation modelling of environmental data showed that elevation, isothermality, annual precipitation, and percent tree cover were significantly lower in G. m. morsitans than in G. m. centralis sampling sites (P < 0.0005 ). Annual temperature was observed to be higher in G. m. morsitans than inG. m. centralis range (P < 0.0005 ). Land surface temperature was higher in four of the five sampling sites of G. m. morsitans than in those for G. m. centralis (P < 0.0005 ). The LZP sample site for G. m. morsitans was observed to have LST 4 °C lower than all other sampling sites. Elevation and annual precipitation were observed to be the environmental variables contributing most of the variation for Principal Component (PC) 1, whereas annual precipitation, vegetation continuous field (percent tree cover) and elevation contributed the most for PC2 (Fig 3). Principal Component 1 accounted for 91.62% of the variation between sites.
Fig 3. Principal components analysis of sample sites based on environmental variables. (A) Scree plot showing that most of the variance in the data set could be explained by the first two principal components PC1 and PC2. (B) Score plot indicating that PC1 and PC2 accounted for 91.62 and 8.31% of the variation among sites, respectively. (C) Vector loading plot showing that elevation and annual precipitation were the variables that contributed the highest variance to PC1. (D) Vector loading plot showing that annual precipitation, vegetation continuous field and elevation contributed the highest variance to PC2. Abbreviations: Elev, Elevation; Bio1, Annual Temperature; Bio3, Isothermality; LST, Land Surface Temperature, VCF, Vegetation Continuous Fields indicating percent tree cover.
3.2 Wing Geometric Morphometrics
Significant wing CS differences were observed between male and femaleG. morsitans flies, the two subspecies G. m. centralis andG. m. morsitans , and between sample locations within the two subspecies ranges (Table 1). Male flies were observed to have an absolute size nine percent smaller than females and G. m. morsitans was two percent smaller than G. m. centralis . Within the G. m. centralis range, flies from KNP1 and KNP 2 were observed to be three percent smaller than those from the KSP site (P < 0.0075 and 0.013 respectively). In the G. m. morsitans range, flies from the LZP site were observed to be five percent larger than flies from all other sites (P< 0.0005 ).
Linear permutation regression showed that elevation, annual temperature, annual precipitation, and land surface temperature have a significant effect on G. morsitans wing CS (Table 2). The coefficients of the regression model indicated that land surface temperature had the largest per-unit effect on CS whose net effect was a reduction in fly size.
The covariation of wing shape on CS (allometry) was significantly different (P < 0.0005 ). Allometry was observed to account for an estimated 12% of shape variation in G. morsitans . Size-adjusted wing shape was found to be significantly different between male and female G. morsitans flies and among the two subspecies G. m. morsitans and G. m. centralis(Table 3). Thin-plate spline deformation grids and vector wireframe graphs showed that the observed difference in wing shape between the sexes was mainly due to the relative displacement of landmarks seven, ten, and 12 in the distal-anterior region of the wing blade, with reference to the mean shape of G. morsitans (Fig 4). In male flies, these landmarks exhibited an anterior displacement resulting in the costa margin of the wing blade having a more pronounced curvature (Fig 4A and 4B). In females, landmarks seven, ten, and 12 showed a posterior relative displacement resulting in the apparent flattening of the costa margin (Fig 4C and 4D). Further, the relative displacement of the medial landmarks one, three, and four resulted in the enlargement of the remigium region of the wing blade in males and a reduction in females. Thus, the female wing blade was narrower than in males anterior-posteriorly.
Figure 4. Thin-plate spline (TPS) deformation grid and wireframe graphs of landmark displacement of male and female flies relative to the mean shape of G. morsitans. (A) Male TPS deformation grid. (B) Male wireframe graph showing landmark vector displacement. (C) Female TPS deformation grid. (D) Female wireframe graph showing landmark vector displacement. To clarify shape changes figures were magnified 20 times.
Observed shape differences between the two subspecies were due to the relative displacement of landmarks three, four, and seven, with reference to the mean wing shape of G. morsitans (Fig 5). InG. m. centralis (Fig 5A and 5B), landmarks three and four are displaced posteriorly towards the posterior margin of the wing blade, while landmark seven showed proximal displacement along the costal margin towards the pleuron. In G. m. morsitans , the relative displacements of these landmarks were in the opposite direction (Fig 5C and 5D). Thus, the wing blade of G. m. centralis exhibited a posterior broadening of the remigium region such that the hatchet cell appeared larger than that of G. m. morsitans .
Figure 5. Thin-plate spline (TPS) deformation grid and vector plots of landmark displacement of G. m. centralis and G. m. morsitans relative to mean landmark configuration of G. morsitans. (A) G. m. centralis TPS deformation grid. (B)G. m. centralis wireframe graph showing landmark vector displacement. (C) G. m. morsitansTPS deformation grid. (D) G. m. morsitans wireframe graph showing landmark vector displacement. To clarify shape changes figures were magnified 20 times.
The size-adjusted wing shape was significantly different between sampling locations in both G. m. morsitans and G. m. centralis ranges (P < 0.0005). For both subspecies, the mean wing shape among flies from all sampling locations was significantly different (P < 0.0005 ). Thin-plate spline deformation grids and wireframe graphs with vectors showing the relative displacement of corresponding landmarks relative to other landmarks are shown in Fig 6 (G. m. centralis ) and 7 (G. m. morsitans ). For G. m. centralis , Fig 6A and 6B show that flies from the KNP1 site exhibited landmark displacements that resulted in the anterior-posterior narrowing of the wing blade relative to the mean shape of the subspecies. Flies from all other G. m. centralissites exhibit landmark displacements that result in a relative anterior-posterior broadening of the wing blade. Flies from KSP (Fig 6E and 6F) further showed landmark displacements that showed proximal-distal shortening of the wing blade relative to the mean shape of G. m. centralis . In the G. m. morsitans range, flies exhibited landmark displacements that resulted in the anterior-posterior narrowing (LZP, Figs 7E, and 7F), broadening (SLP, Figs 7G and 7H), and proximal-distal elongation (VNP, Figs 7I and 7J) of the remigium region of the wing blade relative to the mean shape of G. m. morsitans . Flies from CMR and LVA showed little or no displacement of landmarks that led to the remigium region shape change. For these flies, the major shape change observed was the enlargement of the basal stalk of the wing blade relative to that of the mean shape of G. m. morsitans (Figs 7B and 7D). This wing shape change was also observed in flies from the SLP site Figs 7G and 7H.
Fig 6 . Thin-plate spline (TPS) deformation grid and vector plots of landmark displacement of G. m. centralis from different locations relative to mean landmark configuration of G. m. centralis. (A) KNP1 TPS deformation grid. (B) KNP1 wireframe graph showing landmark vector displacement. (C) KNP2 TPS deformation grid. (D) KNP2 wireframe graph showing landmark vector displacement. (E) KSP TPS deformation grid. (F) KSP wireframe graph showing landmark vector displacement. (G) SNP TPS deformation grid. (H) SNP wireframe graph showing landmark vector displacement. To clarify shape changes figures were magnified 20 times.
Fig 7. Thin-plate spline (TPS) deformation grid and vector plots of landmark displacement of G. m. morsitans from different locations relative to mean landmark configuration of G. m. morsitans. (A) CMR TPS deformation grid. (B) CMR wireframe graph showing landmark vector displacement. (C) LVA TPS deformation grid. (D) LVA wireframe graph showing landmark vector displacement. (E) LZP TPS deformation grid. (F) LZP wireframe graph showing landmark vector displacement. (G) SLP TPS deformation grid. (H) SLP wireframe graph showing landmark vector displacement. (I) VNP TPS deformation grid. (J) VNP wireframe graph showing landmark vector displacement. To clarify shape changes figures were magnified 20 times.
Size-adjusted wing shape of G. morsitans was observed to be significantly associated with elevation, annual temperature, isothermality, annual precipitation, land surface temperature, and percent tree cover (Table 4). Collectively, these variables accounted for 10.7% of the observed variation in wing shape at species level. Land surface temperature, annual precipitation, and isothermality contributed the most to this environmental variation (Table 4).
Neighbour-joining trees derived from the analysis of Procrustes distances produced one cluster for G. m. centralis representing flies captured from the two KNP locations (Fig 7). Flies caught from the KSP and SNP sites were distant from this cluster and each other. The longest branch length in the G. m. centralis tree was that of flies from the KSP site. This indicated that these flies had undergone the largest amount of wing shape change relative to the mean shape ofG. m. centralis . For the G. m. morsitans tree, two clusters were observed with flies collected from the VNP site being distant from both clusters (Fig 8). Based on branch length, the largest amount of change was observed in G. m. morsitans flies captured from the LZP and CMR sites.
Fig 8. The phenetic trees of G. m. centralis and G. m. morsitans derived from Procrustes distances. Tsetse from KSP and SNP in the G. m. centralis showed the most divergence. In theG. m. morsitans range, two clusters were observed with flies from VNP showing the highest divergence.