Australian Journal of Ecology (1999) 24, 495–502

What influences the structure of frog assemblages at forest

streams?
KIRSTEN M. PARRIS*† AND MICHAEL A. MC CARTHY*
Centre for Resource and Environmental Studies, The Australian National University, Canberra
ACT 0200, Australia

Abstract We investigated habitat variables influencing the species richness and composition of frog assemblages
at forest streams in southeast Queensland, Australia, to (i) determine whether these assemblage attributes can be
predicted from the nature of the local habitat, and (ii) assess the value of species richness as a measure of the
structure of frog assemblages. Stream size (as measured by the annual volume of precipitation in the upstream
catchment) was found to be a significant predictor of the species richness of an assemblage, while stream size,
understorey floristics in the riparian zone, the presence of palms and elevation were significantly correlated with
the composition of assemblages. No relationship was found between broad forest type and the species richness or
composition of frog assemblages. Species richness was not an informative description of the composition of frog
assemblages, as assemblages with identical species richness had a low average compositional similarity. Results of
this study indicate that frogs are responding to local, ground-level stream and vegetation variables, a finding of
both ecological interest and practical significance for the conservation management of forest-dwelling frogs.

Key words: amphibians, Australia, community ecology, conservation biology, forests, Mantel tests, multivariate
analysis, Queensland, species richness.

INTRODUCTION Zimmerman and Bierregaard (1986) found the

species-area relationship predicted by island bio-
A number of ecological theories have been developed geography theory to have little practical relevance to
to explain the species richness of plant and animal forest frog communities in the Brazilian Amazon.
assemblages. These theories propose that habitat, Rather than area of forest, the amount and distribu-
species interactions, disturbance, extinction, coloniz- tion of suitable breeding habitat was a critical factor
ation and evolution at local, regional and continental limiting the distribution of the majority of frog species
scales may all influence patterns of species richness in their study area (Zimmerman & Bierregaard 1986).
(Ricklefs & Schluter 1993). For example, island bio- This indicates that local habitat, in particular breed-
geography theory states that species richness increases ing habitat, is a primary determinant of the structure
with the area of an island and its proximity to the main- of frog assemblages. Furthermore, the species richness
land. The theory can be applied to both actual oceanic of an assemblage or community may not be a particu-
islands and metaphorical islands of habitat such as a larly informative measure of its structure. Assemblages
patch of forest surrounded by pasture. Species richness with identical species richness can have different com-
may increase with area because larger islands support ponent species and variable conservation significance
a larger range of habitats, and are therefore suitable for (Pearman 1997).
more species (Lack 1976), or because larger islands In this study, we investigated habitat variables influ-
suffer fewer extinctions and may be colonized more encing the species richness and composition of frog
readily than small islands, leading to a higher equilib- assemblages at forest streams in southeast Queensland,
rium species richness (MacArthur & Wilson 1967). Australia, in order to address the following questions:
Island biogeography theory also contends that islands (i) can the species richness of the frog assemblage at a
closer to the mainland are more likely to be colonized site be predicted from the nature of the local habitat;
than those further away, regardless of size (MacArthur (ii) what is the compositional similarity of frog assem-
& Wilson 1967). blages with identical species richness; (iii) can the com-
position of the frog assemblage at a site be predicted
from the nature of the local habitat; and (iv) what
habitat variables are influencing the species richness
*Present address: National Center for Ecological Analysis and
Synthesis, 735 State Street, Suite 300, Santa Barbara CA 93101-
and composition of frog assemblages? Answers to these
3351, USA. questions are not only of ecological interest but also of
†
Corresponding author: [email protected] practical significance for the conservation management
Accepted for publication December 1998. of forest-dwelling frogs.
496 K . M . PARRIS AND M. A. McCA RT HY

METHODS Field survey

From topographic maps, we selected 19 one-ha survey

Study area
sites in the Mapleton, Kenilworth and Jimna State
Forests. Each site was located on a stream, and incor-
The study area is in the Blackall and Conondale
porated a 100-m section of stream and the area 50 m
Ranges, southeast Queensland, Australia, approxi-
either side of it. We selected sites representative of a
mately 100 km northwest of Brisbane (Fig. 1). Thirty-
range of stream sizes (first order to fifth order at a scale
one native frog species are known from the area
of 1 : 50 000, using the method of Strahler 1964) and
(Czechura 1991), including the southern gastric brood-
broad forest types. In an effort to ensure spatial inde-
ing frog Rheobatrachus silus. Rheobatrachus silus has not
pendence, sites were located a minimum of 1 km apart
been seen in the wild since 1980, just seven years after
if they were in different catchments, and a minimum
its discovery (Ingram & McDonald 1993). The south-
of 4 km apart if in the same catchment. We surveyed
ern dayfrog Taudactylus diurnus, once a common
for frogs over two breeding seasons using nocturnal
species in the area, has not been encountered since
searching, automatic tape recording of advertisement
1979 (Czechura & Ingram 1990). Four other frog
calls, and pitfall traps with drift fences. Parris et al.
species found in the Blackall and Conondale Ranges,
(1999) provide details of these sampling techniques and
Litoria pearsoniana, Assa darlingtoni, Mixophyes fleayi
a comparison of their effectiveness. Each site was sur-
and M. iteratus, are of conservation concern (Ingram
veyed six times between January and March 1995 using
& McDonald 1993). The introduced cane toad, Bufo
all three sampling techniques concurrently. We also sur-
marinus, is also present in the study area. The majority
veyed the sites twice in January 1996, once with noc-
of frogs in the Blackall and Conondale Ranges are
turnal searching and once with automatic tape
spring or summer breeders, active in the warmer, wetter
recorders. This gave a total of eight nights of data from
months of the year between October and March.
each site.
Twenty-four of the 31 native species in the study area
We recorded 17 habitat variables describing the
are known to breed at streams, although only one quar-
structure and composition of the vegetation, stream
ter of these do so exclusively (Tyler 1992; Robinson
characteristics and evidence of disturbance at each site
1993; Cogger 1994; Barker et al. 1995). The other
(Table 1). We established a transect 20 m wide and
seven species breed only at dams, ponds, or temporary
100 m long, perpendicular to the stream at the mid-
pools that form after rain (Tyler 1992; Cogger 1994;
point of each survey site, to survey the vegetation 50 m
Barker et al. 1995).
either side of the stream. This transect was divided into
three sections of 33 m, with the middle section defined
as the riparian zone. We identified plants in the over-
storey, midstorey and understorey (ground layer) of
each section, to species level where possible, and

Table 1. The habitat variables recorded for each survey site

Habitat variable

Stream size (small or large)

Logarithm of volume of precipitation in the
catchment upstream from the site
Stream width (≤ 4 m or > 4 m)
Stream substrate (soil or rock)
Composition of the understorey vegetation
(riparian zone)
Presence of palms
Broad forest type (riparian zone)
Broad forest types within 50 m of the stream
Overstorey vegetation (riparian zone)
% Crown cover (riparian zone)
Elevation
Location (Blackall or Conondale Range)
Evidence of fire
Evidence of logging
Evidence of cattle grazing
Fig. 1. Location of the study area in southeast Queensland, Evidence of pigs Sus scrofa
Australia, with the survey sites in the Conondale Range (m) Presence of lantana Lantana camara
and the Blackall Range (1) marked; nearby towns (d).
 FROG A S S EM BLAG ES AT FOR ES T S T R EA M S 497

recorded the presence or absence of each species at presence or absence of frog species at each site to
each site. One of three broad forest types, dry sclero- calculate the dissimilarity measure.
phyll (open) forest, wet sclerophyll (tall open) forest or These dissimilarity values were used to construct
subtropical rainforest (closed forest), was assigned to a site by site dissimilarity matrix for the 19 survey
each section of the transect on the basis of the forest sites. In order to determine the usefulness of species
overstorey. We recorded stream width and stream sub- richness as a measure of the composition of frog assem-
strate where the vegetation transect crossed the stream, blages, we calculated the average compositional dis-
and all signs of disturbance in the whole 1-ha site. similarity of assemblages with identical species richness.
Stream size was divided into two categories: small We then clustered the sites on the dissimilarity of their
(first or second order stream) and large (third order frog assemblages using the UPGMA (unweighted pair-
stream or larger). As stream order varies with the scale group method using arithmetic averages) algorithm,
of a map and the method used to calculate it, the loga- with b 5 0 (Sneath & Sokal 1973). A similar procedure
rithm of the average annual volume of precipitation in was followed to construct a dissimilarity matrix for
the upstream catchment was included as another mea- each of six habitat variables: the four used in the
sure of stream size. We calculated this by multiplying Poisson regression analyses plus stream size (small
the catchment area upstream from the survey site, mea- or large) and composition of the riparian understorey
sured from topographic maps, by the average yearly vegetation. The Bray–Curtis measure was used to
precipitation in the catchment determined using BIO- calculate the dissimilarity between sites for all the habi-
CLIM, a bioclimatic prediction system (Busby 1991; tat variables except elevation and logarithm of the
Nix & Switzer 1991). We then took the logarithm of average annual volume of precipitation in the upstream
the precipitation volume (in GL), as the values ranged catchment. A simple difference measure was used
over several orders of magnitude. for these variables (e.g. elevation at site A – elevation
at site B).
We investigated correlations between the matrix
Data analysis representing the frog assemblages and matrices repre-
senting habitat variables, using the Mantel test of
Species richness of assemblages matrix correspondence (Mantel 1967). This method
has been used in a number of previous ecological stud-
After identifying intercorrelated habitat variables and
ies (Sokal et al. 1980; Douglas & Endler 1982;
excluding them from further analysis, we investigated
Burgman 1987) and was recommended by Diniz-Filho
relationships between the species richness of frog
and Bini (1996). The Mantel test calculates the cor-
assemblages at our survey sites and habitat variables
relation between two dissimilarity matrices. The stat-
using Poisson regression. Four variables appropriate
istical significance of this correlation is calculated using
for univariate analysis were included: logarithm of
a Monte Carlo method, where the rows and columns
upstream catchment volume, broad forest type,
of one matrix are permuted randomly while those of
presence of palms and elevation. Both forward and
the other are held constant (Cliff & Ord 1981). After
backward step-wise selection of variables were used for
each permutation, the correlation between the matri-
the regression analyses.
ces is recalculated, and the resulting values form an
empirical distribution that is used for the significance
Composition of assemblages test. We used one million Monte Carlo permutations
We calculated the dissimilarity between survey sites in to test the significance of correlations. A positive cor-
terms of the composition of their frog assemblages relation indicates that sites that are similar in terms of
using the Bray–Curtis measure of compositional dis- a given habitat variable have similar frog assemblages.
similarity (Bray & Curtis 1957). This has been shown
to be a robust measure of the ecological distance
between sites (Faith et al. 1987). It is calculated as RESULTS
follows:
b1c
D5 Correlations between habitat variables
2a 1 b 1 c
where a is the number of matches (shared presences, Within the 17 habitat variables measured at each
or 1,1), and b and c are the number of (1,0) and (0,1) survey site were groups of intercorrelated variables.
mismatches, respectively. The Bray–Curtis measure Broad forest type in the riparian zone was correlated
gives a dissimilarity value (D) between 0 and 1 for each with broad forest types within 50 m of the stream,
pair of sites, with sites supporting identical assemblages composition of the overstorey vegetation and crown
having a dissimilarity of 0, and sites with no shared cover in the riparian zone. Broad forest type was also
species having a dissimilarity of 1. We used data on the correlated with a number of disturbance variables:
498 K . M . PARRIS AND M. A. McCA RT HY

evidence of fire, logging and grazing, and presence of compositional dissimilarity of frog assemblages with
the weed Lantana camara. Drier forests tended to show identical species richness were all > 0.5, and were
more evidence of these disturbances while rainforest clustered closely to the average dissimilarity of all pairs
showed less. Location (Blackall or Conondale Range) of assemblages. Although we can predict the species
was correlated with elevation, composition of the under- richness of a frog assemblage from the size of a stream,
storey vegetation, evidence of pigs and upstream catch- species richness is not an informative measure of the
ment volume. Sites in the Conondale Range were at a composition of an assemblage (Fig. 3).
higher elevation, had larger upstream catchment vol-
umes and more feral pigs than those in the Blackall
Range. There were also differences in understorey veg- Composition of frog assemblages
etation between the two locations. Stream size was cor-
related with stream width and substrate. Larger streams Cluster analysis of the 19 survey sites based on the
were wider and tended to have a rocky substrate, as compositional dissimilarity of their frog assemblages
higher flows of water wash away fine sediments from the produced four groups (A–D), although the lower two
stream bed. (groups C and D) were not well defined (Fig. 4). No

Species richness of frog assemblages

Table 2. Frog species detected during the survey, and the
number of sites at which they were detected
We detected 14 native species of amphibians during the
survey (Table 2), with the species richness at a site vary- No. of
ing from zero to eight. Poisson regression analysis Species Common name sites
revealed a significant relationship (P < 0.01) between
frog species richness at a site and the logarithm of the Adelotus brevis Tusked frog 9
average annual volume of precipitation in the upstream Assa darlingtoni Hip-pocket frog 3
catchment (Fig. 2). This variable is a measure of the Limnodynastes peronii Striped marsh frog 4
Mixophyes fasciolatus Great barred frog 10
size of the stream at a survey site. Larger streams tended
Pseudophryne raveni Copper-backed brood frog 2
to support a greater number of frog species. We found Litoria chloris Red-eyed green tree frog 4
no relationship between species richness at a site and Litoria fallax Eastern dwarf tree frog 2
elevation, presence of palms or broad forest type. Litoria gracilenta Dainty green tree frog 1
Forward and backward step-wise selection of variables Litoria latopalmata Broad-palmed frog 2
resulted in the same final regression model. Litoria lesueuri Stony creek frog 5
We compared the average compositional dissimilarity Litoria pearsoniana Cascade tree frog 8
Litoria peronii Emerald-spotted tree frog 1
of frog assemblages with identical species richness to Litoria tyleri Laughing tree frog 3
the average dissimilarity of all pairs of assem- Litoria verreauxii Whistling tree frog 1
blages (n 5 171; Fig. 3). The values for the average
Total sites 5 19.

Fig. 2. Results of Poisson regression analysis of the rela- Fig. 3. A comparison between the average compositional
tionship between frog species richness at a survey site (l) and dissimilarity of frog assemblages with identical species rich-
the annual volume of precipitation in the catchment upstream ness (circles) and the average dissimilarity between all pairs
–
from the site (x): ln (l) 5 – 2.194 1 0.503 log(x), P < 0.01. of assemblages (D 5 0.75, n 5 171), represented by a line.
 FROG A S S EM BLAG ES AT FOR ES T S T R EA M S 499

both the size of the stream and the adjacent forest were
influencing the suitability of a site for different frog
species. Of all the habitat variables tested, we found
stream size (small or large) was most strongly correl-
ated with the composition of frog assemblages at the
survey sites. Sites at streams of similar size tended to
have similar frog assemblages. The logarithm of the
average annual volume of precipitation in a stream’s
catchment, an alternative measure of stream size, was
also significantly correlated with the composition of
frog assemblages. No relationship was found between
broad forest type and the composition of frog assem-
blages (Table 3), indicating that ground layer vegetation
is more important for frogs than the nature of the forest
overstorey.

Fig. 4. Cluster analysis of the 19 survey sites on the dis-

similarity of their frog assemblages, showing four groups DISCUSSION
A–D. Sites B1–B10 are in the Blackall Range and sites
C1–C9 are in the Conondale Range.
Species richness of frog assemblages
Table 3. Correlations between the composition of frog
assemblages and habitat variables, using the Mantel test of Upstream catchment volume, a significant explanatory
matrix correspondence variable for the species richness of frog assemblages at
forest streams (Fig. 2), is a measure of the size of the
Habitat variable r P stream at a survey site. It is correlated with stream
width, stream substrate, permanence of water and rates
Stream size (small or large) 0.408 0.002 of flow. Stream-breeding frogs require a stream that is
Log (upstream catchment volume) 0.247 0.009
large enough to hold water for a sufficient length of time
Elevation 0.166 0.036
Composition of understory vegetation 0.191 0.041 for their tadpoles to develop to metamorphosis. Despite
(riparian zone) being more likely to contain fish that prey on frog eggs
Presence of palms 0.146 0.046 and larvae (Petranka 1983; Kats et al. 1988), large
Broad forest type (riparian zone) 0.062 0.171 streams provide suitable breeding habitat for a greater
range of frog species than small streams because they
The probabilities are based on one million Monte Carlo contain water for longer. Many amphibian species that
permutations.
breed in permanent streams or ponds are excluded
from temporary waterbodies because their larval life
frogs were found at the three sites in group A. Eleven span is longer than the duration of water in these
species were found across the four sites in group B, with habitats (Kats et al. 1988). For example, Litoria pear-
Litoria fallax, L. gracilenta, L. peronii and L. verreauxii soniana and Limnodynastes peronii, two species recorded
restricted to sites in this group. Nine species were found only at large streams during this study, have larval life
across the nine sites in group C. This group contained spans of 2–2.5 months and up to 11 months, respect-
all four records of Litoria chloris, and nine of ten ively (McDonald & Davies 1990; Tyler 1992).
and five of eight records of Mixophyes fasciolatus and The smaller streams in the Blackall and Conondale
L. pearsoniana, respectively. We found five species Ranges are regularly dry, only flowing for short periods
across the three sites in group D, four of which after heavy rainfall. They provide suitable breeding
were present at sites in group C. The exception was habitat for species with short larval life spans such as
Litoria latopalmata, which we also found at one site in Pseudophryne raveni. This frog lays its eggs in burrows
group B. on land, where some larval development takes place
We found five of the six habitat variables to be sig- before rain washes the eggs into a stream or temporary
nificantly correlated with the composition of frog pond and the larvae are released (Martin 1967).
assemblages at forest streams during the breeding Tadpoles of the congeneric P. australis have been
season: stream size (small or large), the logarithm of observed to metamorphose after 28–40 days (Tyler
the average annual volume of precipitation in a stream’s 1992). We detected P. ravenii at two small streams with
catchment, species composition of the forest under- intermittent seasonal flows. The three sites where no
storey in the riparian zone, presence of palms, and frogs were found during our survey (group A in Fig. 4)
elevation (Table 3). Thus, habitat variables relating to were all at small streams that rarely contain water.
500 K . M . PARRIS AND M. A. McCA RT HY

A study of amphibian communities in the Upper- Composition of frog assemblages

Rhone floodplain in France found the species richness
of assemblages varied between types of breeding The composition of frog assemblages at forest streams
habitat (Morand & Joly 1995). Species richness was in southeast Queensland was significantly correlated
lowest at sites characterized by frequent flood events with stream size, understorey vegetation in the riparian
but short water duration, and permanent ponds con- zone and elevation (Table 3). Both measures of stream
taining predatory fish. The highest species richness was size (small or large, and annual upstream catchment
observed at temporary ponds protected from flooding volume) were significant. Sites at streams of similar size
and without fish (Morand & Joly 1995). Morand & Joly had similar frog assemblages. This corresponds with a
(1995) interpreted these patterns of species richness as study of amphibian communities at forest streams in
a function of disturbance by flooding and drying, with Borneo, which found stream width and stream gradient
sites experiencing intermediate levels of disturbance strongly influenced the compositional similarity of
having the highest species richness. However, flood assemblages on both a local and regional scale (Inger
events are likely to occur with similar frequency at small & Voris 1993). Stream width is correlated with stream
and large streams in our study area in southeast size, as large streams are wider than small streams. All
Queensland. Consequently, water duration may have the streams studied by Inger & Voris (1993) were
a greater influence on the richness of the frog assem- perennial, and their width varied from 3 to 25 m. In
blages they support. Drying of the smaller streams is contrast, only half the streams included in our
a seasonal, predictable disturbance rather than a study were perennial, and they varied from 1 to 42 m
stochastic one, so the lower frog species richness at wide.
smaller streams is likely to be due to the lack of suit- Species composition of the understorey vegetation in
able breeding habitat for many species rather than the riparian zone and the presence of palms were both
higher levels of random disturbance. significantly correlated with the composition of frog
The significant relationship between upstream catch- assemblages. Understorey vegetation in the riparian
ment volume and the species richness of frog assem- zone is important for frogs, providing them with mois-
blages at forest streams in our study area is an example ture, shelter and calling sites. For example, Litoria pear-
of a local habitat variable influencing local species rich- soniana calls from palms, sedges or ferns adjacent to
ness. This can be contrasted with studies demonstrating or within streams (McDonald & Davies 1990). Palms
a relationship between fish species richness and the total such as picabeen Archontophoenix cunninghamiana are
catchment area of rivers, a regional scale variable (e.g. a characteristic feature of subtropical rainforest (Adam
Hugueny & Paugy 1995; Oberdorff et al. 1997). This 1994), and their presence at a site indicates mesic con-
relationship has been explained by the species-area ditions in the understorey. The correlation between the
hypothesis of island biogeography, with each river composition of frog assemblages and the presence of
system viewed as an island of habitat for fish. Larger palms at a site was of a similar magnitude and statis-
rivers consequently support fish assemblages with tical significance to that between the frog assemblages
higher species richness (Oberdorff et al. 1997). and the species composition of the riparian understorey
We found species richness to be a poor description (Table 3). It is much easier to determine the presence
of the structure and composition of frog assemblages or absence of palms at a site than to identify all the
at forest streams. Assemblages with identical species species of plants in the riparian understorey. The pres-
richness had low average compositional similarity (Fig. ence or absence of palms may be a useful indicator of
3). The composition of assemblages may have more the character of the forest understorey.
biological significance than species richness, as the Broad forest type does not significantly influence the
latter measure can exclude important information on suitability of a stream for frog assemblages in our study
the differences between assemblages. For example, area. This is consistent with information that the
Pearman (1997) found patterns of amphibian majority of frog species in the Conondale and Blackall
species richness in forests in Ecuador poorly Ranges have been recorded from a range of forest types
represented the distribution of Eleutherodactylus (Czechura 1991). We recognized three forest types on
species, a group of frogs sensitive to forest disturbance. the basis of the forest overstorey: dry sclerophyll forest,
He concluded that species richness was not a good wet sclerophyll forest and subtropical rainforest.
indicator of the conservation value of assemblages However, considerable intergrading of forest types
(Pearman 1997). Studies that only investigate the occurs, and the composition of the overstorey is not a
species richness of assemblages, perhaps in an effort to good indication of the nature of the understorey. In
test ecological theories such as the species-area southeast Queensland, forests with a sclerophyll
hypothesis of island biogeography, may have little (Eucalyptus or Lophostemon) overstorey often have a
practical value for conservation biology (Zimmerman rainforest understorey and in some cases, a rainforest
& Bierregaard 1986). midstorey as well (Adam 1994).
 FROG A S S EM BLAG ES AT FOR ES T S T R EA M S 501

Elevation was found to be significantly correlated descriptions of the habitat of these frogs are expressed
with the composition of frog assemblages at a site in terms of broad forest type (Tyler 1992; Robinson
(Table 3). Our survey sites ranged in elevation from 100 1993). The assumption that diversity of forest types can
to 580 m, with those in the Conondale Range having act as a surrogate measure for faunal diversity (Austin
a higher average elevation than those in the Blackall & Meyers 1996; JANIS 1997), and that reservation of
Range (x– 5 475 m and 260 m, respectively). Elevation the range of broad forest types in a region or across
was therefore correlated with location. Three frog Australia will be sufficient to conserve the diversity of
species, Litoria tyleri, L. peronii and Pseudophryne forest-dwelling animals, is questionable. The agreed
raveni, were found only at sites in the Blackall Range, criteria for establishment of a representative reserve sys-
with L. tyleri and P. raveni at more than one site. Assa tem for Australian forests include use of information
darlingtoni, Litoria fallax, L. gracilenta and L. verreauxii on the distribution and genetic variation of individual
were found only in the Conondale Range, with Assa faunal species as well as the distribution of forest types
and L. fallax at more than one site (Table 2). The (JANIS 1997).
similarity of frog assemblages at similar elevations may Results of our study indicate that reservation of a
be a function of these interlocality differences rather range of stream sizes across the range of variation in
than an effect of elevation per se. Elevation has been understorey vegetation may be a better way to conserve
shown to have an effect on amphibian assemblages stream-breeding frogs in the forests of southeast
across broader geographical and elevational scales, such Queensland than reservation of the range of broad
as the whole of the American tropics (Duellman 1988). forest types. Data pertaining to stream size would be
Duellman’s (1988) study sites represented a 4000-m available in digital format and easy to incorporate into
range in elevation, compared to the 480 m range in our the GIS systems used by forest managers to assist with
study area. planning and reserve selection. However, mapping the
composition of the forest understorey would currently
require extensive field survey. This kind of information
Implications for management may be obtainable in the future with remote sensing
techniques such as airborne videography, which was
Conservation management of Australia’s frogs, as with recently used to assess and predict habitat complexity
other faunal groups, has largely focused on individual in eucalypt forests (Coops & Catling 1997).
species rather than on assemblages or communities.
Statistical models of habitat preferences have been
developed for some frog species in an effort to improve ACKNOWLEDGEMENTS
understanding of the environmental variables that influ-
ence their distribution (NSW National Parks & Wildlife We thank Peter West, Adrian Caneris, Ursula Grott,
Service 1994; Holloway 1997). While this approach is Denise Elias, Greg Czechura, Tony Norton, Scott
important, particularly for rare and endangered species, Osborn, Wyn Boon, Randall Donohue, Natasha West
it is not practical to manage every species individually. and Derek Cleland for their assistance in the field. Scott
Collating information from numerous single species Osborn assisted enormously with plant identifications.
models (e.g. 328 models for forest vertebrates in north- Andrew Claridge and David Lindenmayer provided
ern New South Wales, Australia; NSW National Parks valuable comments on the manuscript. This study was
& Wildlife Service 1994) into a manageable form could conducted under Q.D.P.I. permits number 788, 860,
be complicated, time-consuming and ultimately unsat- and 919, Q.N.P.W.S. permit number 2001, and
isfactory, because each species has an individual and Q.D.E.H. permit number HO/000139/95/SAA.
potentially conflicting set of habitat requirements.
Consequently, the aim of our study was to identify habi-
REFERENCES
tat variables influencing the structure of frog assem-
blages as a whole rather than the presence of their Adam P. (1994). Australian Rainforests. Oxford Biogeography
individual component species. Series no. 6. Oxford University Press, Oxford.
Species richness is not an informative measure of the Austin M. P. & Meyers J. A. (1996) Current approaches to mod-
structure of frog assemblages at forest streams in south- elling the environmental niche of eucalypts: implication for
east Queensland. Knowledge of the compositional management of forest biodiversity. For. Ecol. Manag. 85,
similarity of assemblages, and an understanding of the 95–106.
Barker J., Grigg G. C. & Tyler M. J. (1995). A Field Guide to
habitat variables that influence this, are likely to be
Australian Frogs. 2nd edn. Surrey Beatty and Sons, Chipping
more valuable for conservation management. Although Norton.
the composition of frog assemblages at forest streams Bray J. R. & Curtis J. T. (1957) An ordination of the upland forest
in southeast Queensland appears to be unrelated to the communities of southern Wisconsin. Ecol. Monog. 27,
character of the forest overstorey, many qualitative 325–49.
502 K . M . PARRIS AND M. A. McCA RT HY

Burgman M. A. (1987) An analysis of the distribution of plants MacArthur R. H. & Wilson E. O. (1967). The Theory of Island
on granite outcrops in southern Western Australia using Biogeography. Princeton University Press. Princeton.
Mantel tests. Vegetatio 71, 79–86. Mantel N. A. (1967) The detection of disease clustering and a
Busby J. R. (1991) BIOCLIM—A bioclimate analysis and pre- generalized regression approach. Cancer Res. 27, 209–20.
diction system. In: Nature Conservation: Cost Effective Martin A. A. (1967) Australian anuran life histories: some
Biological Surveys and Data Analysis (eds C. Margules & evolutionary and ecological aspects. In: Australian Inland
M. Austin), pp. 64–8. CSIRO, Canberra. Waters and Their Fauna: Eleven Studies (ed. A. H. Weatherley),
Cliff A. D. & Ord J. K. (1981). Spatial Processes: Models and pp. 175–91. Australian National University Press, Canberra.
Applications. Pion, London. McDonald K. R. & Davies M. (1990) Morphology and biology
Cogger H. G. (1994). Reptiles and Amphibians of Australia. 5th of the Australian tree frog Litoria pearsoniana (Copland)
edn. Reed, Sydney. (Anura: Hylidae). Trans. Roy. Soc. S. Aust. 114, 145–56.
Coops N. C. & Catling P. C. (1997) Utilising airborne multi- Morand A. & Joly P. (1995) Habitat variability and space util-
spectral videography to predict habitat complexity in euca- ization by the amphibian communities of the French upper-
lypt forests for wildlife management. Wildl. Res. 24, 691–703. Rhone floodplain. Hydrobiologia 301, 249–57.
Czechura G. V. (1991) The Blackall–Conondale Ranges: frogs, Nix H. A. & Switzer M. A. (1991). Rainforest Animals. Atlas of
reptiles and fauna conservation. In: The Rainforest Legacy, Vertebrates Endemic to Australia’s Wet Tropics. Kowari 1.
Vol. 2: Flora and Fauna of the Rainforests (eds G. Werren & Australian National Parks and Wildlife Service, Canberra.
P. Kershaw), pp. 311–24. Australian Government Printing NSW National Parks and Wildlife Service (1994) Fauna of north-
Service, Canberra. east NSW forests. North East Biodiversity Study Report no.
Czechura G. V. & Ingram G. J. (1990) Taudactylus diurnus and 3, New South Wales National Parks and Wildlife Service
the case of the disappearing amphibians. Mem. Queensl. Mus. (unpubl.).
29, 361–5. Oberdorff T., Hugueny B. & Guegan J. (1997) Is there an influ-
Diniz-Filho J. A. F. & Bini L. M. (1996) Assessing the relation- ence of historical events on contemporary fish species rich-
ship between multivariate community structure and environ- ness in rivers? Comparisons between Western Europe and
mental variables. Mar. Ecol. Prog. Ser. 143, 303–6. North America. J. Biogeog. 24, 461–7.
Douglas M. E. & Endler J. A. (1982) Quantitative matrix Parris K. M., Norton T. W. & Cunningham R. B. (1999) A
comparisons in ecological and evolutionary investigations. comparison of techniques for sampling amphibians in the
J. Theor. Biol. 99, 777–95. forests of southeast Queensland, Australia. Herpetologica 55,
Duellman W. E. (1988) Patterns of species diversity in anuran 271–83.
amphibians in the American tropics. Ann. Missouri Bot. Gard. Pearman P. B. (1997) Correlates of amphibian diversity in an
75, 79–104. altered landscape of Amazonian Ecuador. Cons. Biol. 11,
Faith D. P., Minchin P. R. & Belbin L. (1987) Compositional 1211–25.
dissimilarity as a robust measure of ecological distance. Petranka J. W. (1983) Fish predation: a factor affecting the spatial
Vegetatio 69, 57–68. distribution of a stream-breeding salamander. Copeia 1983,
Holloway S. (1997) Survey protocols for the stream-breeding 624–8.
frogs of Far East Gippsland: the application of habitat mod- Ricklefs R. E. & Schluter D. Eds. (1993). Species Diversity in
elling and an assessment of techniques. MSc thesis, The Ecological Communities. University of Chicago Press, Chicago.
University of Canberra, Canberra. Robinson M. (1993). A Field Guide to Frogs of Australia.
Hugueny B. & Paugy D. (1995) Unsaturated fish communities Australian Museum/Reed, Sydney.
in African rivers. Am. Nat. 146, 162–9. Sneath P. H. A. & Sokal R. R. (1973). Numerical Taxonomy.
Inger R. F. & Voris H. K. (1993) A comparison of amphibian Freeman, San Francisco.
communities through time and from place to place in Sokal R. R., Bird J. & Riska B. (1980) Geographic variation in
Bornean forests. J. Trop. Ecol. 9, 409–33. Pemphigus populicaulis (Insecta: aphididae) in eastern North
Ingram G. J. & McDonald K. R. (1993) An update on the decline America. Biol. J. Linn. Soc. 14, 163–200.
of Queensland’s amphibians. In: Herpetology in Australia: a Strahler A. N. (1964) Quantitative geomorphology of drainage
Diverse Discipline (eds D. Lunney & D. Ayres), pp. 297–303. basins and channel networks. In: Handbook of Applied
Royal Zoological Society of New South Wales, Mosman. Hydrology (ed. V. T. Chow), pp. 4–39. McGraw-Hill, New
JANIS (1997) Nationally agreed criteria for the establishment of a York.
comprehensive, adequate and representative reserve system for Tyler M. J. (1992). Encyclopedia of Australian Animals: Frogs.
forests in Australia. Commonwealth of Australia, Canberra. Angus and Robertson, Sydney.
Kats L. B., Petranka J. W. & Sih A. (1988) Antipredator defenses Zimmerman B. L. & Bierregaard R. O. (1986) Relevance of
and the persistence of amphibian larvae with fishes. Ecology the equilibrium theory of island biogeography and species-
69, 1865–70. area relations to conservation with a case from Amazonia.
Lack D. (1976). Island Birds. Blackwell Science, Oxford. J. Biogeog. 13, 133–43.