3.1 Morphometric analyses
PC1 for linear body measurements described 86.4% of the variation and based on factor loadings, was used as a proxy for overall body size (Table S3). For head shape, PC1 explained 53% and was retained for further analyses. We found no significant correlation between pairwise differences in PC1 for linear body measurements and geographic distance between sites (Mantel r = 0.22; mantel simulated p-value > 0.05). Similarly, there was no relationship between geographic distance and relative leg length (Mantel r = 0.26; mantel simulated p-value > 0.05) or PC1 for head shape (Mantel r = 0.10; mantel simulated p-value > 0.05).
The RF model for body size explained 49% of the total variation. Significant environmental variables for this RF model comprised two measures of temperature variation (isothermality and minimum temperature of the coldest month), two measures of precipitation (precipitation seasonality and precipitation of the coldest quarter), elevation and latitude (Fig. S1). Site origin was also significant in the model, although it was the least important predictor. The results of the follow-up linear regression indicated that four predictors explained 50% of the variance in body size (R2 =0.50, F(7,143)=22.91, p<.001). These were: isothermality (β= 0.81, p<0.001), precipitation seasonality (β= -0.53, p<0.001), precipitation of the coldest quarter (β= -0.003, p<0.05), and site (β= -0.44, p<0.01). The RF model for relative leg length explained 44% of the variation with isothermality, precipitation seasonality, precipitation of the coldest quarter, precipitation of the warmest quarter, treecover and latitude as significant predictors. The RF model for head shape explained 59% of the variation and was explained by the following variables: isothermality, mean temperature of the driest quarter, minimum temperature of the coldest month, precipitation of the warmest quarter, wettest quarter, and coldest quarter, latitude and longitude (Fig. S1).
Projected variation in body size is greatest across the Cameroon highlands and forest-savanna ecotone as well as between the coast and interior of Gabon (Fig. 2a). Body size increases with distance from the equator with the exception of the southern coast of Gabon where body size decreased sharply towards the coast. Relative leg length also generally increased with distance from the equator, however it is relatively uniform in areas of continuous forest (Fig. 2b). This phenotypic variable also exhibits the greatest variation from the southern coast of Gabon moving inland. Head shape shares a similar pattern of variation across the forest-savanna ecotone of Cameroon as does body size but also exhibits a shift from wider to narrow heads moving away from the equator (Fig. 2c). In contrast to body size and relative leg length, there was little variation in head shape between the coast and interior of Gabon.