The association between the age, sex and reproductive subclasses of Rattus fuscipes, Rattus tunneyi, and Melomys burtoni and “wallum” vegetation was studied in two areas of Fraser island. Box traps and wire cages were employed in trapping program that totalled 7920 trapnights, over twenty-six months, for the southern study area (AREA1) and 5310 trapnights, over eighteen months, for the northern study area (AREA2). In AREA 1 a fire razed much of the local vegetation twelve months after trapping commenced. Consequently, the study of AREA 1 was extended for twelve months to encompass the immediate post-fire period. Field trips to both areas were partitioned into “Breeding” ad “Non-breeding” subsets on the basis of the number of newly independent subadults captured on each trip. Floristic and structural vegetation quadrats were taken at all trap sites (180) in AREA 2 and at 135 sites before the fire, and 64 sites after the fire in AREA 1.
A combined classification and ordination analysis independently patterned the trapping data and the floristic and structural vegetation data. The degree of overlap between the respective patterns was assessed by canonical correlation analysis. These methods have long been used by plant ecologists; they were developed specifically to handle large multivariate data sets. Furthermore, they are free from the restrictive theoretical assumptions that underlie popular statistical methods.
Floristically, the vegetation in both AREA 1 and AREA 2 aligns along a soil moisture gradient, although this interaction is modified by localised disparities in fire regime. When the vegetation is burnt, taxa relying on vegetative regrowth dominate the immediate post-fire flora, effectively circumventing the initial stages of Clementsian succession.
The vegetation in AREA 1 is structurally more complex than that in AREA 2. This enhanced structural pattern arises from a greater floristic diversity and differences in vegetation age caused by recent fire mosaics. In AREA 2, differences in vegetation age are not as obvious because the vegetation has been too extensively burnt in the past to leave a substantial portion of it intact for comparison. Although the vegetation understorey was emphasised in the structural analysis the tree layer consistently influenced the structural pattern in both areas.
Site preference for R. fuscipes in AREA 1 and R. fuscipes, R. tunneyi, and M. burtoni in AREA 2 are highly structured between age, sex, and reproductive subclasses. Social interactions appear to direct these site preferences. Subadult males are particularly alienated, especially from adult males. Although the distribution of all three species is relatively stable through time, the internal distribution of subclasses within each population fluctuates markedly. As such, the distribution of species subclasses in any trapping period is a poor predictor of the distribution of that subclass at other times.
The fire in AREA 1 did not kill many small mammals directly. Rather, the animals were affected indirectly through the razing of the vegetation understorey. In particular, the destruction of the swamp vegetation led to the local extinction of M. burtoni in AREA 1. In contrast, the fire apparently created more favourable conditions for R. fuscipes.
The relationship between small mammal distribution and vegetation pattern in AREA 1 and AREA 2 varies between Breeding and Non-breeding periods. Both R. fuscipes and R. tunneyi have their highest correlations with vegetation pattern, particularly floristic pattern, during the Breeding periods. M. burtoni, however, displays its strongest correlations in the Non-breeding periods. The vegetation attributes most highly correlated with R. fuscipes are similar in both AREA 1 and AREA 2. In contrast, sites frequented by M. burtoni differ markedly between AREA 1 and AREA 2. Unlike the previous two species, R. tunneyi is most abundant in sites that have been extensively burnt in the past.
After the fire in AREA 1, adult female R. fuscipes supplanted the males in showing the highest correlations with vegetation pattern. At this time when other R. fuscipes subclasses were responding to floristic patterns, the adult females were displaying a high correlation with vegetation structure. The strength of the relationship between R. fuscipes and the vegetation declined in the second year after the fire.
Three important implications concerning the design of small mammal “habitat” studies arise from this study. First, the vegetation preferences of age, sex, and reproductive subclasses need not coincide. Hence, the analysis of trapping data at the species level can confound these different associations and lead to vague and confusing results. Second, the relationship between small mammal subclasses is, at least in some areas, dynamic between breeding and non-breeding seasons. Hence, the pooling of data on an annual basis, or other independent chronological criteria, as is common practice, can also reduce the effectiveness of an analysis. Third, this study has shown that pattern analyses, in comparison with conventional statistical methods, can be very successful in identifying vegetation and trapping patterns that are not immediately obvious, either from the raw data, or from inspection in the field. Furthermore, they are powerful in identifying vegetation attributes that have a predictive role in explaining small mammal distribution.