HABITAT SELECTION BY FEMALE RATTUS LUTREOLUS DRIVES ASYMMETRIC COMPETITION AND COEXISTENCE WITH PSEUDOMYS HIGGINSI

Transcription

1 HABITAT SELECTION BY FEMALE RATTUS LUTREOLUS DRIVES ASYMMETRIC COMPETITION AND COEXISTENCE WITH PSEUDOMYS HIGGINSI VAUGHAN MONAMY AND BARRY J. Fox School of Biological Science, University of New South Wales, Sydney, New South Wales, Australia, 2052 Two native Australian rodents. the velvet-furred swamp rat (Rattus lutreolus velutinus) and the long-tailed mouse (Pseudomys higginsi). are sympatric in wet sclerophyll forests on Mount Wellington. Tasmania. Australia. Their relatively large macrohabitat overlap implies that these species partition one or more resources to coexist. We used census data to examine the role of microhabitat structure in resource partitioning. At trap stations where only one species or the other was trapped. significant differences in microhabitat structure existed. R. lutreolus was trapped in areas of densest vegetation ::51 m in height. Male and female R. lutreolus exhibited macrohabitat separation during the non-breeding season (winter). At that time. adult and subadult female rats occupied areas of thickest ground cover; males occupied surrounding. more open areas of forest at lower densities. After the onset of breeding. male rats visited areas occupied by female rats. P. higginsi was never trapped in areas where most female R. [utreolus were captured but was most often trapped in the same macrohabitat as male R. lutreolus. After the onset of breeding. relative rate of capture of P. higginsi increased in these areas as male R. lutreolus shifted their trap-revealed ranges to incorporate areas where females were resident. We propose that the seasonal intersexual habitat separation exhibited by R. lutreolus, and particularly the habitat use by females. may be an essential component in structuring the rodent community in wet sclerophyll forests. Key words: Rattus [utreolus, Pseudomys higginsi, niche dimensions, macrohabitat, microhabitat, intersexual habitat selection Many reports of habitat selection by small mammals have revealed complex relationships between habitat heterogeneity and species coexistence, resource partitioning, and competition among sympatric populations. Consequently, empirical studies have emphasized the importance of floristic and structural variation within plant associations in determining animal distributions (Adler, 1985; Fox, 1982; Mazurkiewicz, 1994; M'Closkey and Fieldwick, 1975; M'Closkey and Lajoie, 1975; McMurry et al., 1996; Morris. 1984; Ostfeld et al., 1985; Seagle, 1985). Small mammals may perceive heterogeneous habitat. and its suite of resources, in two ways. Fine-grained species may use subsets within such habitat in direct pro- portion to availability of those subsets. Conversely, coarse-grained species may use heterogeneous habitat differentially (Mac Arthur and Levins, 1964; cf. Morris, 1987). This realization has led to careful examination of both macrohabitat and microhabitat components of complex vegetational mosaics when assessing habitat use by small mammals. However, in most studies of habitat selection by coexisting small-mammal species, the emphasis has been only on interspecific differences in habitat use. Differential intersexual habitat use has been often ignored, or regarded as having an insignificant influence on structuring of multispecies communities. In Tasmania, Australia, the velvet-furred Journal of Mammalogy, 80(1): ,

2 1999 MONAMY AND FOX-RODENT HABITAT SELECTION AND COEXISTENCE 233 rat, Rattus lutreolus velutinus (Rodentia: Muridae), and the long-tailed mouse, Pseudomys higginsi (Rodentia: Muridae), coexist in wet sclerophyll forests. Both species occupy heterogeneous forest macrohabitat differentially. Female R. lutreolus were trapped significantly more often than male rats in areas where understory was densest (Monamy, 1995a). Male R. lutreolus occupied macrohabitat peripheral to the areas inhabited by females until the onset of breeding, when they moved into the areas occupied by females. P. higginsi was never caught in areas where female R. lutreolus densities were highest but were trapped more often in areas where male R. lutreolus were captured (Monamy, 1995b). A role for asymmetric interspecific competition in the coexistence of these species was inferred by Monamy (1995b) and tested by Luo et al. (1998) using regression techniques. R. lutreolus had higher competition coefficients than P. higginsi in forested areas where ground cover was thickest (Luo et al., 1998). We quantified macrohabitat- and microhabitat-niche dimensions of R. lutreolus and P. higginsi in a wet sclerophyll forest and showed that coexistence between these species was best understood when seasonal intersexual differences in habitat use by R. lutreolus were considered. MATERIALS AND METHODS Species.-The velvet-furred swamp rat is a common mostly-nocturnal native murid of Tasmania. It occurs in a variety of vegetation types including sedgelands, coastal heathlands, dry sclerophyll forests, and wet sclerophyll forests. This species requires dense ground cover to 1 m in height (Monamy, 1995a; Norton, 1987), a resource requirement it shares with a mainland Australian subspecies, R. lutreolus lutreolus (Braithwaite and Gullan, 1978; Braithwaite and Lee, 1979; Fox, 1984; Haering and Fox, 1995). R. l. velutinus attains a mean adult body mass of g (Monamy, 1995c) and breeds from October through April (summer to autumn). More than one litter per season may be reared (Green, 1967; Taylor and Homer, 1973). The long-tailed mouse is a strictly nocturnal, endemic Tasmanian rodent that occupies temperate rainforest, wet and dry sclerophyll forests, and alpine boulder fields at low densities. Few studies detail geographic distribution (Rounsevell et ai., 1991) and habitat requirements (Monamy, 1995b). Green (1968) and Watts and Aslin (1981) observed that P. higginsi preferred areas of high rainfall (>2,500 mml year) where forest floors were well covered with rotting vegetation. Norton (1987) and Stoddart and Challis (1993) confirmed that it also occurs in drier parts of Tasmania, including east coast dry sclerophyll forests. Adults attain a mean body mass of 65 g (one-half that of adult R. lutreolus) during the summer breeding season (November-March). More than one litter of three young may be reared during this time. Study area and sampling procedures.-a 4- ha 10- by lo-trapping grid (20-m spacing) was surveyed in an area of heterogeneous forest near Shoobridge Bend in the Mount Wellington Flora and Fauna Reserve, near Hobart, Tasmania, Australia (42 56'S, 'E). The site for the grid was selected to incorporate four vegetation types apparent from visual inspection, and the structural and floristic heterogeneity of the grid was confirmed using classification and ordination techniques. The grid was located on the southeastern slopes of Mount Wellington (1,270 m) at m above mean sea level in an area burned by wildfire in Temperatures range between 4.4 C (mean annual minimum) and 11.0 C (mean annual maximum). Mean annual rainfall exceeds 1,200 mm, and snow may be recorded in any month. Trapping was conducted in 12 trapping sessions in winter (May-October 1989) and in the summer breeding season (November 1989-April 1990; 6 sessions/season, 4 weeks apart; 4 nights/ session). An aluminium live trap (33 by 10 by 10 cm-elliott Scientific Equipment, Upwey, Victoria, Australia) baited with a mixture of rolled oats and peanut butter was positioned within 2 m of each permanent trap point. Plastic bags were placed over traps as protection from rain and snow, and shredded paper was placed in all traps as insulation against cold. At first capture, all individuals were eartagged with a numbered fingerling fish tag (Size FF-Salt Lake Stamp Company, Salt Lake City, UT). At every capture, species, sex, reproduc-

3 234 JOURNAL OF MAMMALOGY Vol. 80, No.1 tive condition, body mass, and grid position of each individual were recorded. To provide a quantitative description of the vegetation, an intensive survey of structural and floristic variation over the trapping grid was conducted. We centered a 4- by 4-m quadrat on each of the 100 trap points. Quadrat size was based on a minimal-area per species relation curve (Goldsmith and Harrison, 1976). Presence or absence of vascular plant species rooted in or projecting over each quadrat was recorded at each of seven height strata; 0-20 cm, cm, cm, 1-2 m, 2-5 m, 5-10 m, and >10 m. Cover was estimated visually (Braun-Blanquet scale-mueller-dombois and Ellenberg, 1974). Cover afforded by decaying logs and boulders also was estimated in each quadrat. Two-way indicator species analysis (TWIN SPAN-Hill, 1979a) was used, combining floristic and structural attributes, to identify and classify four discrete macrohabitat groups. Ordination using detrended correspondence analysis (DECORANA-Hill 1979b) confirmed structural and floristic relatedness between quadrats within each macrohabitat group. Detailed descriptions of structure and floristics within these four macrohabitat groups were provided by Monamy (1995a). Briefly, habitat group 1 (HG1; nine trap points) was the wettest area of the grid where a tall canopy of Eucalyptus regnans and E. delegatensis shaded a dense fern understory (Blechnum wattsii, Dicksonia antarctica, and Polystichum proliferum). Habitat group 2 (HG2; 39 trap points) was floristically similar to HG 1 but structurally more open. Habitat group 3 (HG3; 28 trap points) was characterized by an overstory of E. delegatensis and presence of large dolerite boulders. Ground cover was dominated by the ferns Microsorum diversifolium and P. proliferum and two grasses, Holcus lanatus and Deyeuxia rodwayi. Habitat group 4 (HG4; 24 trap points) was the driest part of the grid with little or no fern or grass understory. Large numbers of rotting logs afforded some disjunct cover for small mammals, particularly where pockets of fern (P. proliferum) were growing. Data analyses.-presence or absence of R. lutreolus and P. higginsi was recorded at each trap station during each trapping session. If an individual entered the same trap on more than one occasion in a trapping session, it was recorded as one capture for that trap station. All capture locations of an individual within a trapping session were scored (once for each trap it entered). We pooled presence or absence data over the 6 winter trapping sessions and over the 6 breeding-season sessions to make seasonal comparisons of macrohabitat and microhabitat use. Relative rates of capture in each macrohabitat group (percentage of total captures) were examined to determine if they fell within 95% or 99% CI given the different number of trap stations in each habitat group (Rohlf and Sokal, 1981: ). This was a simple method of determining if each cohort or species was using any particular habitat group more or less than expected from the number of trap stations in each habitat group. An index of capture (I) was calculated as the percentage of captures in each habitat divided by the percentage of traps in that habitat. An index score of 1 represented a rate of capture equivalent to that expected given the number of trap stations in each macrohabitat group. A score < 1 represented fewer captures than expected; a score> 1 represented more captures than expected. Proportional use of the trapping grid by P. higginsi and R. lutreolus, with each macrohabitat group as a separate resource state, was used to determine the level of sympatry on the trapping grid. We calculated Pianka's (1974) symmetrical measure of niche overlap (Oij): Oij = I(Pik,pjk)/(Ipik 2 Ipjk 2 )05 where Pik and Pjk were the proportions of the two rodent species i and j captured in each habitat group k. Microhabitat-niche overlap was determined in the same way, using each trap station as a separate resource state. Preliminary analyses revealed marked seasonal differences in intersexual habitat overlap for R. lutreolus, and we treated data for males and females separately. No such differences were evident for P. higginsi, and data for both sexes were pooled. When applied to assess macrohabitat-niche dimensions, Pianka's symmetrical niche-overlap equation does not consider the relative size of each resource state (Le., dependent on the number of trap stations in each macrohabitat group for our study). To gain greater insights into the microhabitat relationships of R. lutreolus and P. higginsi, the following methods were used. First, a factor analysis (principal components analysis of the correlation matrix in its original form) was used to reduce the number of structural at-

4 1999 MONAMY AND FOX-RODENT HABITAT SELECTION AND COEXISTENCE 235 tributes measured at each trap station to three factors. Previous analyses using stepwise-multiple regression, with captures as the dependent variable, demonstrated that both rodent species respond more strongly to structural microhabitat attributes than to floristic variables (Monamy, 1995a, 1995b). Factor scores for each trap station where each species was captured were used to determine the relative microhabitat-niche positions of the two species. Factor scores for each trap station were not weighted by the number of captures of each species. To ensure independence of data, analyses were conducted only between groups of mutually-exclusive trap stations. Six groups were compared in winter and in the breeding season: trap stations where I, only P. higginsi; 2, only R. lutreolus; 3, both species; 4, neither species; 5, only male R. lutreolus; and 6, only female R. [utreolus, were caught. The Euclidean distance (Ed) between sample populations was calculated using: Ed = {(Fla - FIb) + (F2a - F2b) + (F3 a - F3 b ) los where F I, Fz and F3 were the first, second and third mean factor scores, respectively, for population samples a and b. We tested for differences in overall microhabitat-niche position between each sample population (Le., the Euclidean distance between sample populations) by using the nonparametric multivariate Mantel test (Luo and Fox, 1996a; Mantel, 1967). The Mantel test compared a square symmetrical similarity matrix (S) with a square symmetrical hypothesis matrix (H) to determine if pairs of elements within the two matrices were correlated significantly (Sokal and Rohlf, 1994). The test statistic, Z, is defined as: n Z = 2: 2: S;jH;j where n was the size of each matrix in the comparison; S;j and Hu were the elements in matrices S and H for each sample i and j respectively. The observed Z was compared against an altered null distribution of Z generated by randomly changing the order of columns and rows of one of the matrices. With the option of weighting the hypothesis matrix (Luo and Fox I 996a, I 996b), the Mantel test was particularly well-suited to studies such as ours where sample sizes were low or unequal. All significance tests were two-tailed. We quantified the microhabitat-niche position n of each species to determine relative levels of habitat specialization using a formula proposed by MacNally and Doolan (1986) and modified by Luo and Fox (1996b). Niche position (-y) was defined as the mean of displacements from the hypervolume origin of all positions at which a member of a sample population occurred (MacNally and Doolan, 1986). The hypervolume had reduced dimensionality so that structurally similar trap stations occur close together in the hypervolume. Niche position for sample k was calculated as: where w; was the amount of variance explained for factor i, and X;k was the mean of the ith factor score of sample k. By calculating niche position in this way, the relative contribution of each factor in explaining the total variance was taken into account. That measure represented a weighted Euclidean distance of the trivariate mean from the origin in factor space (Luo and Fox, 1996b). A high value for -y indicated specialized habitat use; a low value indicated more generalized use of habitat. RESULTS Habitat use and movements.-overall, capture success varied between species, seasons, sexes, and heterogeneous habitat groups. In winter, female R. lutreolus were trapped in all habitat groups, but those captures most often occurred in HGI (38% of captures) and HG2 (36%). Female R. lutreolus strongly preferred HG 1 and significantly avoided HG3 and HG4 (Table 1). Female rats dominated HGl, an area where no P. higginsi were ever trapped, and where only seven captures of one male rat (10%) were recorded prior to the onset of breeding. The remaining 90% of winter captures of male R. lutreolus were made in HG2 (56%) with significant preference and HG3 (34%). No male rats were ever trapped in HG4. In winter, P. higginsi showed significant preference for HG3 (58%) but significantly avoided HG2 (27%) and HG4 (15%). In the breeding season, preference for HG3 and HG4 declined to 44% and 11 %, respective-

5 236 JOURNAL OF MAMMALOGY Vol. 80, No.1 TABLE I.-Index of captures for P. higginsi and male and female R. lutreolus for each of four macrohabitat groups in winter 1989 and the summer breeding season. A score of 1.0 represents a rate of capture equivalent to that expected given the number of trap stations in each macrohabitat group; a score <1.0 represents fewer captures than expected; a score >1.0 represents more captures than expected. Significance based on percent captures lying outside confidence limits for percentages of trapsites available. Numbers in parentheses denote number of captures. Sample HGl population Season (9 traps) P. higginsi Winter 0.00** (0) Breeding 0.00** (0) R. lutreolus Winter 3.00** (52) Breeding 4.16** (71) Male Winter 1.10 (7) Male Breeding 4.39** (45) Female Winter 4.20** (45) Female Breeding 3.80** (26) * p < 0.05, ** P < ly, while a rise was recorded in HG2 (27% to 45%). Examination of where individual animals were most often trapped throughout the study showed that changes in relative rates of capture and preference for each macrohabitat group after onset of breeding could be attributed, in part, to movements between habitat groups of resident individuals. Samples were small, but in all cases where movement occurred, it was in the direction toward habitat groups with greater ground cover. In HG1, five female and only one male R. lutreolus were trapped repeatedly during winter (Fig. la). The male and three females were adults; two females were subadults. After the onset of breeding, one female rat left the trappable population, leaving four females that remained until the end of the study. None was trapped in any other habitat group. At the same time, 13 previously untagged male R. lutreolus moved onto the trapping grid and were most often caught in HGI (Fig. Ib). Three of four subadult male rats that were trapped in HG2 and HG3 in winter also moved to include HG 1 in their trap-revealed home range as they became sexually mature. Examination of data for individual P. higginsi revealed movements of similar magnitude, Macrohabitat group HG2 HG3 HG4 (39 traps) (28 traps) (24 traps) 0.68* (21) 2.08** (12) 0.63 (12) 1.16 (29) 1.56* (7) 0.45** (7) 1.12 (83) 0.68* (36) 0.42** (19) 1.05 (78) 0.68* (36) 0.11** (5) 1.44** (40) 1.21 (24) 0.00** (0) 1.00 (45) 0.75 (24) 0.00** (0) 0.93 (43) 0.36** (12) 0.67* (19) 1.11 (33) 0.56** (12) 0.28** (5) and in the same direction, as those seen for R. lutreolus (Fig. 1). Habitat group 3 was the area of the grid where most P. higginsi were trapped and for which they showed the greatest preference (Table 1). Eleven individuals (four females and seven males) were trapped in HG3 in winter. Niche overlap.-a relatively high macrohabitat-niche overlap (0.66 in winter; 0.72 in the breeding season) was found between the two species for the entire study confirming the sympatric relationship of these species in wet sclerophyll forest (Table 2). Comparison of overlap between each sex of R. lutreolus with P. higginsi in each season revealed differences. In winter, there was low macrohabitat-niche overlap between P. higginsi and female R. lutreolus (0.49). That appeared to be less than overlap between P. higginsi and male R. lutreo Ius (0.79), but we could not test that. During the breeding season, overlap between both sexes of R. lutreolus and P. higginsi was very similar (Table 2). This was attributable to the relative rate of capture of P. higginsi in HG2 increasing after the onset of breeding. At that time, niche overlap between male and female rats also increased from 0.74 to Microhabitat-niche overlap between P.

6 1999 MONAMY AND FOX-RODENT HABITAT SELECTION AND COEXISTENCE 237 'C Female R. /utreo/us 'C Female P. higginsi GI GI C. C. C. C (1) (1) III III ca 15 ca 15 iii 10 iii 10 ::J ::J 'C 'C.s;.s; (1) (1) 5 5 :0 :0.E.E 0 0 HG1 HG2 HG3 HG4 HG1 HG2 HG3 HG4 Macrohabitat group Macrohabitat group! 20 (b) Male R. /utreo/us (3) (1) 'C c. c ca 15 ca 15 III III (d) Male P. hlgglnsi iii 10 iii 10 (1) ::J ::J 'C 'C.s;.s; 5 5 :0 :0.E.E 0 0 HG1 HG2 HG3 HG4 HG1 HG2 HG3 HG4 Macrohabitat group Macrohabitat group FIG. I.-Changes in locations of individual a) female R. lutreolus, b) male R. lutreolus, c) female P. higginsi, and d) male P. higginsi in the trappable population after the onset of breeding. Arrows show the direction of this movement from the habitat group (HG) in which individuals over-wintered to a HG of denser ground cover in the breeding season. The number denotes the number of individuals that moved in this way. Other differences in abundance are explained by movement of previously unmarked individuals into, or the loss of individuals from, the trappable population. higginsi and R. lutreolus was low (Table 2). As with macro habitat comparisons, greatest overlap (0.37) occurred between P. higginsi and male R. lutreolus in winter. That overlap declined after the onset of breeding, reflecting a seasonal change in the location of most male R. lutreolus captures from trap stations in HG2 and HG3 to trap stations located in HG 1 and HG2 where relative rates of captures of female R. lutreolus were high. Intersexual microhabitat-niche overlap in R. lutreolus rose, in consequence of that shift, from 0.14 in winter to 0.83 in the breeding season. P. higginsi and female R. lutreolus showed almost no microhabitatniche overlap during winter (0.07) or the breeding season (0.16). Microhabitat-niche.-The factor analysis TABLE 2.-Habitat-niche overlaps for P. higginsi and male and female R. lutreolus based on Pianka's (1974) symmetrical overlap equation in macro- and microhabitats in Tasmania, Macrohabitat Microhabitat Season Overlap pairings overlap overlap Winter R. lutreolus-p. higginsi Male R. lutreolus-p. higginsi Female R. lutreolus-p. higginsi Male R. lutreolus-female R. lutreolus Breeding season R. lutreolus-p. higginsi Male R. lutreolus-p. higginsi Female R. lutreolus-p. higginsi Male R. lutreolus-female R. lutreolus

7 238 JOURNAL OF MAMMALOGY Vol. 80, No.1 TABLE 3.-Factor-score matrix/or trap-revealed structural microhabitat use by P. higginsi and R. lutreolus in Tasmania, Habitat variables Factor 1 Factor 2 Factor 3 Ground cover 0-20 cm 0.820' Ground cover cm 0.910' Ground cover cm 0.800' Canopy stratum > 10 m height ' Logs >50 cm diameter ' ' Rocks > 10 cm diameter ' Eigenvalues Cumulative variance (%) a Significant correlation coefficients; r = 0.244, d.f = 2, 44, P < used on a matrix of structural attributes and 100 trap points yielded three factors that explained 74% of the variance (Table 3). Factor 1 explained 38.1 % of the variance and correlated strongly with the three groundcover strata to 1 m in height. The second factor explained 18.7% of the variance and correlated with canopy stratum and presence of logs. The third factor explained 17.2% of the variance and correlated strongly and positively with presence of boulders but negatively and much less strongly with presence of logs. That relationship between logs and boulders resulted from the survey-quadrat size. Fallen logs were large (:::::2 m diameter) and occupied a considerable portion of any quadrat in which they occurred. It was not possible to have a high score for boulders, even when they occurred in such quadrats, if they were completely hidden by the logs. The Mantel test tested for significant differences in the microhabitat use between paired combinations of six sample populations (e.g., traps in which only female R. lutreolus were captured versus traps that only caught P. higginsi; Fig. 2) in winter and in the breeding season. Sample sizes for each comparison are given in Appendix I. Large separation occurred in canonical space between traps used exclusively by P. higginsi and those used exclusively by R. lutreolus. Largest separation occurred on factor 1 representing a gradient of increasingly-dense ground cover to 1 m in height. Negative values for that factor were found for P. higginsi in winter (Fig. 2a) and in the breeding season (Fig. 2b); positive values were recorded for male and female R. lutreolus in both seasons. Both species avoided trap stations where ground cover was most sparse. Differences in microhabitat structure (as revealed by the mean factor scores) were evident in winter between trap stations where only P. higginsi and only R. lutreo Ius were trapped (P < 0.01) and between stations where only P. higginsi and only female R. lutreolus were caught (P < 0.05; Fig. 2a). No significant differences occurred in comparisons between stations at which only male rats and only P. higginsi were trapped (indicating a similar use of microhabitat) or in comparisons between stations where only male and only female R. lutreo Ius were caught. During the breeding season, differences in position in canonical-space occurred between those traps recording only P. higginsi and all sample populations involving R. lutreolus (i.e., only R. lutreolus, P < 0.05; only female R. lutreolus, P < 0.05; and only male R. lutreolus, P < 0.05; Fig. 2b). Significant differences were apparent for all comparisons in winter and the breeding season between trap stations where animals of either species were caught and those trap stations where no individuals were caught. That indicated that areas of less suitable microhabitat do exist within a heterogeneous area of wet sclerophyll forest. Densities of both species were low in this study, how-

8 1999 MONAMY AND FOX-RODENT HABITAT SELECTION AND COEXISTENCE 239 (a) 1.0 U/ CI N..J >- "' os 0 u.. c os '" ~ I t (b) I 1.0 Neither species PH ~ 0.5 Neither N.3 spe~ies l) + b ~ 0.0 If g fi 8l 0.5 t=. Male RL Both :pecies RL<1.285-) Female RL(1.288 ) Dense vegetation to 1 m Factor 1 PH. Female RL (1.046') RL--'... (1.062') Male RL Both (1.122') species ~-~~-+-~-~~~ Dense vegetation to 1 m Factor 1 FIG. 2.-Bivariate means for six population samples in a) winter and b) the breeding season. Factor 1 describes a gradient based on an increasing availability of dense vegetation in the three ground strata to 1 m in height. Factor 2 describes the presence of decaying logs and an increasingly closed tree canopy. Greatest separation is seen with changes in ground-cover density (factor 1). Numbers in parentheses denote significant separations measured by Euclidean distances from PH to samples with positive values for factor 1 (* P < 0.05, ** P < 0.01). Samples for each comparison are given in Appendix I. PH = P. higginsi only; RLV = R. lutrealus only. ever, and it was not possible to state that such areas would not be used if densities increased. Analysis of microhabitat-niche position ("I) revealed that R. lutreolus displayed a greater degree of microhabitat specialization in both seasons than did P. higginsi (0.53 and 0.47, respectively, in winter). However, comparisons between P. higginsi and male and female R. lutreolus revealed a subtle seasonal relationship between the two species. In winter, the lowest "I-values were recorded by male R. lutreolus (0.26), being about one-half those of values for female rats (0.55) and for P. higginsi (0.47). In the breeding season, however, "I-values for male and female R. lutreolus combined (0.70) greatly_ exceeded those for P. higginsi (0.17), highlighting an increased degree of microhabitat selection. DISCUSSION Quantification of macrohabitat use by R. lutreolus and P. higginsi conformed to what is known about general habitat preferences of these species in wet sc1erophyll forests (Monamy, 1995a, 1995b). Throughout the study, R. lutreolus were trapped most often in areas of densest ground cover to 1 m in height. P. higginsi were significantly more likely to be caught in areas of boulder scree with a patchy understory. What was not known previously was that the sympatric relationship between these two murids appears to involve a subtle seasonal component that may be driven by female R. lutreolus. An overall macrohabitat-niche overlap (Pianka, 1974) of 0.66 in winter masked intersexual differences in overlap between male rats and P. higginsi (0.79) and. between female rats and P. higginsi (0.49). At that time, male and female R. lutreolus were using forest macrohabitat differentially, with females most often trapped in areas with the densest ground cover. Male R. lutreolus tended to be trapped in the macrohabitat groups where P. higginsi were most active (i.e., HG2 and HG3). Those areas were peripheral to the area occupied by the female rats (HGl). There is no empirical evidence to suggest that areas of macrohabitat such as HG 1, where capture rates for female R. lutreolus were consistently and significantly higher than expected (Table 1), afford superior nutrition, protection from predation, or other advantage. However, persistence of highest capture rates throughout the year suggests

9 240 JOURNAL OF MAMMALOGY Vol. 80, No.1 that these areas best fulfill resource requirements of female rats. The significant increase in the capture rate for male rats in HG 1 after the onset of breeding suggests that their greatest requirement at that time is access to females. This is consistent with predictions that habitat resources are less important than mate receptivity in determining male fitness (Ims, 1987, 1988; Trivers, 1972). Sexual differences in habitat use by rodents are reported infrequently, but competing theories exist to explain them. Bowers and Smith (1979) recorded that female deer mice, Peromyscus maniculatus, occupied more favorable microhabitats than males in xeric plant communities. They observed that females and males only were caught in close proximity to one another during periods of sexual activity. They concluded that sexual differences in habitat use, with females occupying areas of best resource, maximized reproductive effort and survival of offspring by reducing predation on nest sites and lowering resource overlap. That selective advantage implied that there was intersexual competition because males were displaced to less favorable areas. Morris (1984) also found sexual differences in habitat use by P. leucopus, but offered an alternative hypothesis based on resource requirements. Put simply, females must choose safe nest sites and must spend considerable time near those sites during rearing of young. Therefore, rather than reflecting an evolutionary strategy, differential habitat use reflected reproductive constraints that limited females to areas with suitable nesting sites. Ostfeld et al. (1985) reported female-biased sex ratios, increased survivorship, and higher rates of juvenile recruitment in California voles, Microtus californicus, occupying high-quality meadow habitat. They recorded a tendency for females to aggregate in the best areas and concluded that the manner in which females responded to spatial and temporal variation in resource quality reflected a reliance on resource acquisition to enhance reproductive success. It is interesting to speculate that, given that HG 1 best fulfills resource requirements of female R. lutreolus, then females that survive beyond one breeding season may remain in these areas. In our study, three adult females born before 1989 were trapped repeatedly in HG 1 along with two sub adult females. Subadults may have been the offspring of one or more of the adults, and an advantage may have been afforded to these young. It appears as if sub-adult male R. lutreo Ius are relegated to less-preferred habitat, which they occupy prior to the breeding season. At the same time, P. higginsi occupy open rocky areas at greater than expected rates. P. higginsi increase their traprevealed ranges after the onset of breeding to include areas of macrohabitat that had been occupied at greater than expected rates by male rats in winter (Table 1). Asymmetric interspecific competition between these species (Luo et al., 1998), determined by using standardized data in a rejuvenated regression technique (Fox and Luo, 1996), was greatest during the breeding season, with female rats most likely defending core areas of habitat and being visited by male rats. It may be that R. lutreolus and P. higginsi prefer similar areas of microhabitat, but P. higginsi is unable to compete with individuals of twice their body mass. P. higginsi may be able to coexist only with subadult male R. lutreolus during winter when interspecific differences in body mass are smallest. During the breeding season, they occupy areas vacated by male R. lutreolus (now adult) that are searching for mating opportunities. We acknowledge that our study was conducted with species that occur at low densities and consequently samples are small. We know of no small-mammal studies that have demonstrated that intersexual differences in habitat use by one species have had direct implications for a sympatric species. R. lutreolus is a relatively recent col-

10 1999 MONAMY AND FOX-RODENT HABITAT SELECTION AND COEXISTENCE 241 onizer of Australian forests «1 X 10 6 years ago). Pseudomys have been present for possibly five times as long (Watts and Aslin, 1981). There may have been independent niche evolution for R. lutreolus and P. higginsi in Tasmania's wet sc1erophyll forests, but it is more likely that the realized niche of P. higginsi is being constrained by the differential and specialized use of habitat by R. lutreolus. ACKNOWLEDGMENTS We thank J. Luo for expert advice on the analysis of niche data and for the use of his computer program for running the Mantel test. The M. A. Ingram trust provided generous financial assistance to V. Monamy, which enabled collection of field data while he was a student in the Department of Zoology, University of Tasmania, Hobart, Australia. Animals were trapped with the permission of the University of Tasmania Animal Experimentation Ethics Committee, the Tasmanian Department of Parks and Wildlife, and the Hobart City Council. M. H. Gott and J. E. Taylor provided constructive criticism that improved an early draft of this paper. Data analyses and manuscript preparation were supported by Australian Research Council grant A to B. J. Fox. LITERATURE CITED ADLER, G. H Habitat selection and species interactions: an experimental analysis with small mammal populations. Oikos, 45: BOWERS, M. A., AND H. D. SMITH Differential habitat utilization by sexes of the deermouse, Peromyscus maniculatus. Ecology, 60: BRAITHWAITE, R. W., AND P. K. GULLAN Habitat selection by small mammals in a Victorian heathland. Australian Journal of Ecology, 3: BRAITHWAITE, R. W., AND A. K. LEE The ecology of Rattus lutreolus. I. A Victorian heathland population. Australian Wildlife Research, 6: Fox, B. J Fire and mammalian secondary succession in an Australian coastal heath. Ecology, 62: Small scale patchiness and its influence on our perception of animal species' habitat requirements. Pp , in Survey methods for nature conservation, Volume 1 (K. Myers, C. Margules, and I. Musto, eds.). Commonwealth Scientific and Industrial Research Organisation, Canberra, Australia. Fox, B. J., AND J. Luo Estimating competition coefficients from census data: a re-examination of the regression technique. Oikos 77: GOLDSMITH, E B., AND C. M. HARRISON Description and analysis of vegetation. pp , in Methods in plant ecology (S. B. Chapman, ed.). Blackwell Scientific Publications, Oxford, United Kingdom. GREEN, R. H The murids and small dasyurids in Tasmania. Parts I and 2. Records of the Queen Victoria Museum, 28: The murids and small dasyurids in Tasmania. Parts 3 and 4. Records of the Queen Victoria Museum, 32:1-19. HAERING, R. AND B. J. Fox Habitat utilization patterns of sympatric populations of Pseudomys gracilicaudatus and Rattus lutreolus in coastal heathland: a multivariate analysis. Australian Journal of Ecology, 20: HILL, M. O. 1979a. TWINSPAN: a FORTRAN program for arranging multivariate data in an ordered two-way table by classification of the individuals and attributes. Ecology and Systematics, Cornell University, Ithaca, New York b. DECORANA: a FORTRAN program for detrended correspondence analysis and reciprocal averaging. Ecology and Systematics, Cornell University, Ithaca, New York. IMS, R. A Male spacing systems in microtine rodents. The American Naturalist, 130: Spatial clumping of sexually receptive females induces space sharing among male voles. Nature, 335: Luo, J., AND B. J. Fox Diet of the eastern chestnut mouse (Pseudomys gracilicaudatus): II. Seasonal and successional patterns. Wildlife Research, 21: I 996a. A review of the Mantel test in dietary studies: effect of sample size and inequality of sample sizes. Wildlife Research, 23: I 996b. Seasonal and successional dietary shifts of two sympatric rodents in coastal heathland: a possible mechanism for coexistence. Australian Journal of Ecology, 21: Luo, J., V. MONAMY, AND B. J. Fox Competition between two Australian rodent species: a regression analysis. Journal of Mammalogy, 79: MACARTHUR, R. M., AND R. LEVINS Competition, habitat selection, and character displacement in a patchy environment. Proceedings of the National Academy of Sciences, 51: MACNALLY, R. c., AND J. M. DOOLAN An empirical approach to guild structure: habitat relationships in nine species of eastern-australian cicadas. Oikos, 47: MANTEL, N. A The detection of disease clustering and a generalized regression approach. Cancer Research, 27: MAZURKIEWICZ, M Factors influencing the distribution of the bank vole in forest habitats. Acta Theriologica,39: M'CLOSKEY, R. T., AND B. FIELDWICK Ecological separation of sympatric rodents (Peromyscus and Microtus). Journal of Mammalogy, 56: M'CLOSKEY, R. T., AND D. T. LAJOIE Determi-

11 242 JOURNAL OF MAMMALOGY Vol. 80, No.1 nants of local distribution and abundance in white footed mice. Ecology, 56: McMURRY, S. T., R. L. LocHMILLER, J. E BOGGs, D. M. LESLIE, JR., AND D. M. ENGLE Demography and condition of populations of white-footed mice (Peromyscus leucopus) in late and early successional habitats. Journal of Mammalogy, 77: MONAMY, V a. Population dynamics of, and habitat use by, Australian native rodents in wet sclerophyll forest, Tasmania. I. Rattus lutreolus velutinus (Rodentia: Muridae). Wildlife Research, 22: b. Population dynamics of, and habitat use by, Australian native rodents in wet sclerophyll forest, Tasmania. II. Pseudomys higginsi (Rodentia: Muridae). Wildlife Research, 22: c. Ecophysiology of a wild-living population of the velvet-furred rat, Rattus lutreolus velutinus (Rodentia: Muridae), in Tasmania. Australian Journal of Zoology, 43: MORRIS, D. W Sexual differences in habitat use by small mammals: evolutionary strategy or reproductive constraint? Oecologia, 65: Ecological scale and habitat use. Ecology, 68: MUELLER-DoMBOIS, D., AND H. ELLENBERG Aims and methods in vegetative ecology. John Wiley & Sons, New York. NORTON, T. W The ecology of small mammals in north-eastern Tasmania. I. Rattus lutreolus velutinus. Australian Wildlife Research, 14: OSTFELD, R. S., W. Z. LIDICKER, AND E. J. HESKE The relationship between habitat heterogeneity, space use, and demography in a population of California voles. Oikos, 45: ApPENDIX I. PIANKA, E. R Niche overlap and diffuse competition. Proceedings of the National Academy of Sciences, 71: ROHLF, E J., AND R. R. SOKAL Statistical tables. Second ed. W. H. Freeman and Company, New York. ROUNSEVELL, D. E., R. J. TAYLOR, AND G. J. HOCKING Distribution records of native terrestrial mammals in Tasmania. Wildlife Research, 18: SEAGLE, S. W Patterns of small mammal microhabitat utilization in cedar glade and deciduous forest habitats. Journal of Mammalogy, 66: SOKAL, R. R., AND E J. ROHLF Biometry: the principles and practice of statistics in biological research. Third ed. W. H. Freeman and Company, New York. STODDART, D. M., AND G. CHALLIS Habitat use and body form of the long-tailed mouse (Pseudomys higginsi). Wildlife Research, 20: TAYLOR, J. M., AND B. E. HORNER Reproductive characteristics of wild native Australian Rattus (Rodentia: Muridae). Australian Journal of Zoology, 21: TruVERS, R. L Parental investment and sexual selection. pp , in Sexual selection and the descent of man (B. Campbell, ed.). Aldine, Chicago, Illinois. WATTS, C. H. S., AND H. J. ASLIN The rodents of Australia. Angus and Robertson, Sydney, Australia. Submitted 9 April Accepted 28 May Associate Editor was James C. Hallett. Overall differences in trap-revealed microhabitat use between P. higginsi and R. lutreolus in winter (June-October 1989) and the summer breeding season (November 1989-April 1990; 6 trapping sessions per season) in Tasmania, Euclidean distances between trivariate means for each population sample comparison in canonical space are given. Winter Breeding Comparisons between trap stations where the Mantel test Euclidean Mantel test Euclidean following sample populations were caught n (g)a Distance n (g)a Distance R. lutreolus only and P. higginsi only 30, ** , * Male R. lutreolus only and P. higginsi only 11, , * Female R. lutreolus only and P. higginsi only 12, * , * Male R. lutreolus only and Female R. lutreolus only 11, , R. lutreolus only and both species 30, , P. higginsi only and both species 14, * , R. lutreolus only and neither species 30, *** , *** Male R. lutreolus only and neither species 11, ** , *** Female R. lutreolus only and neither species 12, ** , *** P. higginsi only and neither species 14, * , Both species and neither species 24, *** , *** * p < 0.05, ** P < 0.01, *** P < " A weighting procedure, in which all 'I' values in the hypothesis matrix were divided by the size of the corresponding sample, applied due to inequalities of sample size (see Luo and Fox, 1994).