Variability of Muskrat Skulls: Measurement Error, Environmental Modification and Size Allometry

Abstract

The biometrical consequences of measurement error, environmental effects caused by habitat quality, and effects of allometay transformation were studied in 22 morphometric traits of the skull of the muskrat, Ondatra zibethicus. Analyses were conducted at a regional level using four Finnish muskrat populations living 85–750 km apart, and at a local level using five subpopulations living 2–9 km apart in one Finnish lake. Only spring samples of animals born in the previous summer were used (n = 508), but sexes were pooled because they were not found to differ in variability. Reduced variability was found in traits connected with feeding. In larger variables the coefficient of variation (CV) was moderately negatively correlated with the mean, but in the smallest measurements the variability increased sharply. The dependence of the CV on the mean of the trait, thus, is hyperbolic. Corrected variabilities for each trait were calculated as the residuals from this function. Measurement error was estimated by repeated measurings of the same skulls, using two methods differing in acauracy. The proportion of measurement error in the CV varied from 3 to 23%, if measuring accuracy was 0.1 mm, and from 13 to 45%, if accuracy was only 0.5 mm (mean of the variables ranged from 4 to 63 mm). The proportion of measurement error in the CV increased with a decreasing mean value of the trait, but this cannot alone explain the great variability in the small traits. Environmental effects on variability were studied by grouping animals on the basis of their condition, classified by a growth index that measured body-weight gain per unit of molar wear. Improvement in animal condition decreased general morphological variability. We concluded that habitat quality, especially poor diet, was the main stress factor increasing variation, and specimens with the poorest condition were particularly exposed to this effect. In order to examine and compare allometric relationships on both a local and regional scale, each variable was divided by condylobasal length of the skull (CL; calculated by subtraction in log scale). If the corrected variability of a trait was lower or equal to that of CL, the transformation often induced a negative correlation with general size. In comparisons between local subpopulations, the effect of this bias was not serious: the allometric transformation reduced size differences that depend on habitat quality. If the correlation structure of the data was more heterogeneous, as in regional data, the transformation produced partly confusing results in univariaate and principal component analyses, but discriminant function analysis was not seriously affected.