Abstract

We censused and measured armadillo burrows in ten 10 m x 40 m plots in each of four habitat types at a study site in northern Florida and one in the Atlantic coastal rainforest of Brazil. The nine-banded armadillo (Dasypus novemcinctus) was the only species of armadillo found in Florida, but several additional species were present in Brazil. Burrows were more numerous but smaller in Brazil than in the U. S., probably due to the inclusion of burrows dug by the smaller congener D. septemcinctus. In Brazil, burrows were larger and more numerous in swamp and forest habitats than in grassland or disturbed areas, suggesting that D. novemcinctus is found primarily in forests and swamps while D. septemcinctus is located in the other areas. This was supported by data from sightings of live animals. In Florida, burrows were more numerous in hardwood hammocks than in wetlands, fields or upland pine areas, but burrow dimensions did not vary across habitat types. In Florida, armadillos were seen more frequently than expected in hammocks and wetlands and less frequently than expected in fields and upland pine areas. There were also age (juvenile versus adult), sex, and yearly differences in habitat use in Florida. Biomass, abundance, and species diversity of terrestrial invertebrates did not vary significantly between habitat types in Florida, suggesting that habitat associations of armadillos were not influenced by prey availability.
 

Key words

Dasypus, armadillos, Brazil, United States, burrows, habitat usage.
 

Landscapes are mosaics of different habitat types (Forman and Godron 1986, Pickett and Cadenasso 1995) and animals are expected to vary in their use of these habitats. For example, more individuals should be found in areas with abundant resources than in areas where resources are scarce (Sutherland 1996). However, determining habitat usage can be difficult for many species of mammals. For example, the nine-banded armadillo (Dasypus novemcinctus) is found throughout the southeastern United States (Humphrey 1974, Wetzel 1985, Taulman and Robbins 1996), but there are few quantitative data on which habitats these animals occupy within this broad geographic range. This is probably due to the fact that D. novemcinctus is relatively nocturnal and asocial (Galbreath 1982, McBee and Baker 1982), so obtaining large numbers of observations on individuals in different habitats is extremely time consuming. However, D. novemcinctus digs burrows (Newman 1913, Kalmbach 1943, Taber 1945, Clark 1951, Talmage and Buchanan 1954, Zimmerman 1990). As has been done with numerous species of burrowing animals, (e.g., King 1955, Schaller 1983, Burke 1989, Breininger et al. 1994, Moller et al. 1997), censusing burrows can provide data on population density and, if censused across different habitat types, habitat use. Caution is required though, as sole reliance on burrow counts can provide misleading data about population density (Hoogland 1981, Burke 1989, Vanhorne et al. 1997). In addition, animals may construct refuges in one area but travel to other areas to feed, so that presence of burrows may not accurately reflect the extent of habitat utilization. Consequently, observations of live animals are needed to confirm patterns of habitat usage and population size derived from burrow counts.

Burrow dimensions can be diagnostic of particular species of armadillos (Carter and Encarnaçao 1983). In the United States, where D. novemcinctus is the only armadillo present, differences in burrow dimensions would presumably reflect age differences, because juveniles might construct smaller burrows than adults due to their smaller body size. Thus, differences in burrow dimensions between habitat types could provide evidence about habitat usage by different age groups of armadillos. Along the same lines, knowing the dimensions of D. novemcinctus burrows in one habitat might allow identification of D. novemcinctus burrows in other habitats where other burrowing animals (such as other species of armadillos) may occur (Guyer and Hermann 1997).

In this study, we censused and measured armadillo burrows in each of four different habitat types in two different locales: one in northern Florida and one in the Atlantic coastal rainforest of Brazil. We used these data to compare burrow densities and dimensions between locales and between habitats within locales. In addition, because habitat types were similar in the two locales, we were able to compare burrow characteristics between matching habitat types.These data provided some insight into patterns of habitat usage by nine-banded armadillos.We used data from sightings of live animals to confirm these patterns and to examine age, sex, and yearly differences in the use of habitats. Finally, we measured resource availability (i.e., terrestrial invertebrates) in the four habitat types in Florida to examine the relationship between habitat usage and resource distribution.
 

Materials and Methods

Study sites: Data were collected during the summers (June-August) of 1992-1995 and 1997 at the Tall Timbers Research Station located just north of Tallahassee, Florida, and during the austral summer (January-June) of 1996 at the Poço das Antas Federal Biological Reserve, located approximately 100 km north of Rio de Janiero in Brazil (see Loughry and McDonough 1997, 1998a). Poço das Antas is 5200 ha in size and consists primarily of steep hills separated by narrow valleys.  There are four main habitat types (Camargo 1996, Dietz et al. 1997): (1) mata, or Atlantic coastal rainforest; (2) grassland, consisting primarily of Imperata brasiliensis and/or Melinus minultiflora; (3) disturbed woodland ("corridors" in Dietz et al. 1997), which contains some grass but is primarily dominated by the tree Gouchimatia polimorpha; and (4) swamp, which may have standing water during wet portions of the year, but not year round. The first three habitats are found on the slopes of hills while swamps comprise most of the valleys (Dietz et al. 1997).

The nine-banded armadillo is the most abundant species of armadillo at Poço das Antas, but its smaller congener, the seven-banded armadillo (D. septemcinctus) is also present, as is the southern naked-tailed armadillo (Cabassous unicinctus; Loughry and McDonough 1997). The yellow armadillo (Euphractus sexcinctus) used to occur but may now be locally extinct (Loughry and McDonough 1997). All of these species dig burrows and the possibility exists of misclassifying burrows constructed by the similarly-sized D. novemcinctus and C. unicinctus. However, this may not be a large problem because C. unicinctus was very rare at Poço das Antas (2 individuals sighted over the course of the study, Loughry and McDonough 1997) and, at least sometimes, appears to bury itself in the ground in a new location each night (Carter and Encarnaçao 1983, Eisenberg 1989, Emmons 1990, Redford and Eisenberg 1992), thus not producing a detectable burrow.

D. novemcinctus  is the only species of armadillo found at Tall Timbers Research Station (Humphrey 1974, Stevenson and Crawford 1974, Wetzel 1982, 1985, Taulman and Robbins 1996). Tall Timbers is 1600 ha in size and is situated along the north shore of Lake Iamonia. Brennan et al. (1998) identified three main habitat types at Tall Timbers: (1) hammocks, consisting primarily of hardwoods such as oak (Quercus spp.), American beech (Fagus grandifolia), and southern magnolia (Magnolia grandiflora), (2) fields, which were plowed annually and planted with corn (Zea mays), clover (Trifolium spp.), and browntop millet (Brachiaria ramosa), and (3) upland pine areas, consisting of loblolly (Pinus taeda), shortleaf (Pinus echinata) and some longleaf (Pinus palustris) pines. In addition, we included (4) wetlands, defined as areas within 25 m of permanent water and thus having the potential to be flooded at certain times during the year. These four habitat types were roughly similar to the habitat types found in Brazil, e.g., hammocks and rainforest were both characterized by a closed canopy of hardwoods with little understory, fields and grasslands were open areas with thick ground cover, upland pine and disturbed areas were open woodlands with considerable ground cover in the form of grasses and shrubs, and swamps and wetlands were both areas that could potentially be under water during certain times of the year.

Burrow characteristics: All data on burrows were collected during February 1996 at Poço das Antas and late June to early July, 1997 at Tall Timbers. At each locale we set up ten 10 m x 40 m plots in each of the four habitat types. The location of each plot was determined using a random numbers table to select x and y coordinates from a topographic map of each study site. Each plot was set up along polar coordinates; on hillsides, the long axis of the plot was oriented up the hill slope, in swamp/wetland plots, the long axis of the plot ran parallel to the edge of the water. In each plot, we counted all of the potential armadillo burrows we could find that looked large enough to contain at least a juvenile armadillo. To distinguish burrows from depressions created by armadillos while foraging (cf. Zimmerman 1990), burrows were defined as having a minimum depth of 25 cm. Other species of burrowing animals occur at Poço das Antas (e.g., paca, Agouti paca, and red-rumped agouti, Dasyprocta agouti) and it is possible that burrows dug by these animals could have been mistakenly classified as armadillo burrows. We feel this was not a large problem because we specifically avoided censusing areas where these species were known to dig burrows (e.g., in the sides of banks, Emmons 1990).

In addition to counting the number of burrows per plot, we obtained data on the following characteristics of individual burrows: (1) height was the maximum vertical size (in cm) of the burrow opening; (2) width was the maximum horizontal size of the burrow opening; (3) length was the mininum length of the burrow as determined by the insertion of a flexible rod as far into the burrow as possible; (4) burrows were characterized as active (= 1) or inactive (= 2) on the basis of a visual assessment of the recency of burrow occupation, e.g., a burrow that was filled with leaves or had spider webs across the entrance was classified as inactive, while one with fresh diggings was classified as active; and (5) the visibility of the burrow was ranked on a 3 point scale with 1 = a burrow that was clearly visible and conspicuous from a distance of several meters, 2 = a partially visible burrow, and 3 = a burrow that was hidden by foliage, debris, etc. We also noted the type of soil found in each plot but, because almost all plots had the same soil classification, these data are not presented here.

Population differences in burrow characteristics (pooled across habitat types) were analyzed with t-tests. In addition, we used t-tests to compare burrows between matching habitat types (e.g., hammock versus rainforest, grassland versus field, etc). Variation in burrow characteristics between habitat types within each locale was analyzed with ANOVA.

Habitat use: Burrow densities in different habitat types provided some indication of habitat usage by armadillos. We sought to confirm this using data from sightings of live animals (cf. Loughry and McDonough 1998a). At both sites we censused the study areas during both daylight and evening hours (250 days for a total of 2967 person-hours of field time in the U.S., in Brazil, 108 days and 958 person-hours). A daily census typically lasted 4-6 h and was conducted by walking or driving along trails or roads on each property. Spotlights and miner's lamps were used to observe animals after dark. The total linear distance censused was approximately 25 km at both sites. Cenuses at both sites passed through all habitat types. A rotating schedule of observations was used, such that a portion of the entire sampling area was censused each night, followed by another portion the next night, and so on until the entire area had been sampled. This schedule was then repeated for the duration of the field season.

We recorded the habitat in which every animal observed during these censuses was located. In many cases, we subsequently caught and marked these animals (for details, see McDonough and Loughry 1997a, Loughry and McDonough 1998a, b) so we were able to examine age (juvenile, i.e., young of the year, versus adult) and sex differences in habitat usage. We were able to assign individuals that were observed but not caught to age categories on the basis of body size and, if the genitalia were visible, we could classify the individual by sex as well. In some cases, we were unable to obtain any information on the identity of an animal. These individuals were included in the total count of animals observed in each habitat type, but were excluded from analyses of age or sex differences. Animals that were caught were marked with reflective tape for long-range identification. To avoid pseudoreplication, resightings of these animals on the same day were ignored. Although we tried to avoid sampling the same areas repeatedly on the same night, we can not be sure that all of our sightings of unmarked animals represented different individuals, so our estimates of habitat usage may be slightly inflated. In Brazil, we also noted the species of each animal sighted. However, adult D. septemcinctus are about the same size as juvenile D. novemcinctus (Wetzel 1985), so visual identification from a distance is difficult. In this study, we only included animals that we could positively identify.

We did not sample all four habitats at either site with equal frequency. To analyze habitat usage, we first determined the percentage of each habitat type found along our census routes. This was done by recording at 20 m intervals the type of habitat(s) found along each side of the road or path being sampled. These percentages allowed us to calculate the expected number of armadillos in each habitat type (total number of individuals observed across all habitats x proportion of entire study area comprised of that particular habitat), assuming the animals were distributed randomly. Comparisons of the overall distribution of armadillos observed with these expected values were analyzed with contingency tables, while comparisons between observed and expected frequencies for a particular habitat were analyzed by calculating Z scores (Bhattacharyya and Johnson 1977). In Florida, where we had a much larger sample to work with (see below), we further analyzed distributions of animals for each age/sex group separately and for each year of the study.

There is a potential problem with the above analyses. If armadillos were more detectable in some habitats than others, our results could reflect differences in detectability rather than differences in habitat usage. We consider this unlikely for the following reasons. First, while visibility does vary between habitats (Loughry and McDonough 1998a), armadillos are quite noisy as they forage, making them conspicous even when they are not initially visible. In many cases, we first detected armadillos aurally and followed them until they came into view. Thus, we are reasonably confident that, in all habitats, we were able to detect all armadillos within 25-30 m of the road or trail we were surveying. Second, if detectability was the primary determinant of our results, one would expect the most observations of armadillos in the habitats where they should be most detectable (e.g., open fields). In fact, this was not the case (see below), suggesting that our results can not be due solely to differences in detectability.

Prey availability: Nine-banded armadillos feed primarily on invertebrates they find while digging in the soil (Kalmbach 1943, Clark 1951, Fitch et al. 1952, Breece and Dusi 1985, Redford 1985, Sikes et al. 1990, White 1992, Lippert 1994). We sought to determine if habitat usage was related to prey availability by sampling terrestrial invertebrates in each of the plots censused for burrows. In each plot, a jar filled with 70% alcohol was placed in the soil and a 12.4 cm diameter funnel was placed on the top of the jar so that the top of the funnel was level with the ground. These pitfall traps were left in the plot for 48 h and then removed. Trap contents were identified to family for insects and order for non-insects. The total number (i.e., abundance), biomass (in g dry weight), and diversity of invertebrates found in each sample were recorded and compared between habitat types with ANOVA. Species diversity was calculated using the Shannon index (Smith 1992). Pitfall traps were only set out at Tall Timbers, so no data from Brazil were included in these analyses.
 

Results

Burrow characteristics: Armadillo burrows were more numerous, less visible, less active, and had smaller openings in Brazil than in the U.S. when data from all habitat types were combined (t-tests, all p < 0.005, Table 1). The difference in burrow density between Brazil and the U.S. was due to the larger number of inactive burrows per plot in Brazil (mean ± SD = 3.33 ± 3.25 versus 1.12 ± 1.22, t = 4.01, p < 0.0001, df =78). The proportion of active burrows (number of active burrows per plot divided by the total number of burrows in each plot) was significantly higher at Tall Timbers than at Poço das Antas (0.38 ± 0.35 versus 0.21 ± 0.28, t = 2.19, p = 0.03, df = 62; note that plots containing no burrows were excluded from this analysis).

Comparisons between comparable habitats showed that burrow densities were higher in Brazil in each case except for swamps versus wetlands (t-tests, all p < 0.05; for the swamp-wetland comparison, p = 0.78; Table 1). However, burrow dimensions (height, width and length) did not vary between forests and hammocks or between swamps and wetlands (t-tests, all p > 0.14; Table 1). Burrows appeared less active in swamps in Brazil than in wetlands in the U.S. (t = 3.23, p = 0.002), but not in any other comparison (all p > 0.06; Table 1). Burrows in grassland and disturbed areas in Brazil had smaller openings (height and width) than those in fields and upland pine areas in the U.S. (t-tests, both p < 0.04), but did not differ in any other measure (all p > 0.13; Table 1).

Burrow density varied significantly between habitats in Brazil (ANOVA, F = 3.64, p = 0.02, df = 3, 36; Table 1). Pairwise post-hoc Scheffé tests showed that burrows were more numerous in forest than in grassland. Burrow width (F = 4.69, p = 0.004, df = 3, 159), height (F = 9.47, p < 0.001) and visibility (F = 4.03, p = 0.009) also varied between habitats. Post-hoc Scheffé tests of these analyses showed that burrow entrances were larger and more visible in forests and swamps than in grasslands or disturbed areas.

Burrow densities also varied between habitats at Tall Timbers (ANOVA, F = 4.29, p = 0.01, df = 3, 36; Table 1). The only significant pair-wise comparison with post-hoc Scheffé tests showed that burrows were more numerous in wetlands than in fields. There were no significant differences in burrow characteristics between habitats (ANOVAs, all p > 0.22, df = 3, 76) with the exception of visibility (F = 3.27, p = 0.026). However, this comparison yielded no significant post-hoc pair-wise comparisons with Scheffé tests.

Habitat use: Nine-banded armadillos were not randomly distributed across habitat types in Brazil (= 26.7, p < 0.0001, df = 3, Table 2). Instead, they were found more frequently than expected in disturbed woodland and less frequently than expected in grassland (Table 2; distributions did not differ from expected in swamp and forested habitats). There were few confirmed observations of D. septemcinctus (n = 3), however all sightings were either in grassland or disturbed habitats.
 

Habitat use was also non-random in the U.S. (= 458.1, p < 0.0001, df = 3), with armadillos found more often than expected in hammocks and wetlands and less often than expected in upland pine and fields (Z scores, all p < 0.01, Table 3). These results were derived from data pooled across all individuals and all years of the study. However, there were significant differences in habitat usage between adults and juveniles (all years pooled, = 21.53, p < 0.0001, df = 3) and between years of the study (all ages pooled, = 27.06, p = 0.008, df = 12). Further examination of these effects showed that habitat usage by adults did not vary between years (= 14.16, p = 0.29, df = 12), but did for juveniles (= 50.15, p < 0.0001, df = 12). Consequently, we compared adult and juvenile habitat usage separately.

Adults were not randomly distributed across habitat types when data from all years were pooled (= 345.3, p < 0.0001, df = 3) or when each year was examined separately (Table 3). Instead, adults were more commonly observed in hammocks and wetlands and less frequently in upland pine (Z scores for data pooled across all years, all p < 0.0001, Table 3). There was no significant difference between the observed and expected frequencies of adults in fields (p = 0.395, Table 3).

Habitat usage by juveniles varied yearly, so we analyzed the distribution of juveniles across habitats separately for each year (except for 1994 when too few juveniles were observed to permit analysis). Juveniles were non-randomly distributed across habitats in 1992 (= 21.3, p < 0.0001, df = 3), 1993 (= 77.8, p < 0.0001), and 1995 (= 55.9, p < 0.0001), but not in 1997 (= 0.47, p = 0.93; Table 3). In years with significant non-random distribution, juveniles, like adults, were found more frequently than expected in hammocks (Z scores, p < 0.0001 in each year) and swamps (ns in 1992, p < 0.04 in 1993 and 1995) and less frequently than expected in fields (ns in 1992, p < 0.05 in 1993 and 1995) and upland pine areas (p < 0.003 each year, Table 3).