Ecological Applications, 8(4), 1998, pp.

1118–1132
q 1998 by the Ecological Society of America

STREAM AMPHIBIANS AS INDICATORS OF ECOSYSTEM STRESS:

A CASE STUDY FROM CALIFORNIA’S REDWOODS
HARTWELL H. WELSH, JR. AND LISA M. OLLIVIER
USDA Forest Service, Pacific Southwest Research Station, Redwood Sciences Laboratory, 1700 Bayview Drive,
Arcata, California 95521 USA

Abstract. Road construction of the Redwood National Park highway bypass resulted
in a large accidental infusion of fine sediments into pristine streams in Prairie Creek State
Park, California, during an October 1989 storm event. This incident provided a natural
experiment where we could measure, compare, and evaluate native stream amphibian den-
sities as indicators of stream ecosystem stress. We employed a habitat-based, stratified
sampling design to assess the impacts of these sediments on the densities of aquatic am-
phibians in five impacted streams by comparing them with densities in five adjacent, un-
impacted (control) streams. Three species were sampled in numbers sufficient to be infor-
mative: tailed frogs (Ascaphus truei, larvae), Pacific giant salamanders (Dicamptodon te-
nebrosus, paedomorphs and larvae), and southern torrent salamanders ( Rhyacotriton var-
iegatus, adults and larvae). Densities of amphibians were significantly lower in the streams
impacted by sediment. While sediment effects were species specific, reflecting differential
use of stream microhabitats, the shared vulnerability of these species to infusions of fine
sediments is probably the result of their common reliance on interstitial spaces in the
streambed matrix for critical life requisites, such as cover and foraging. Many stream-
dwelling amphibians are highly philopatric and long-lived, and they exist in relatively
stable populations. These attributes make them more tractable and reliable indicators of
potential biotic diversity in stream ecosystems than anadromous fish or macroinvertebrates,
and their relative abundance can be a useful indicator of stream condition.
Key words: Ascaphus truei; bioindicators; California; Dicamptodon tenebrosus; ecosystem stress;
redwood ecosystem; Rhyacotriton variegatus; sedimentation; stream amphibians.

INTRODUCTION Odum 1985, Steedman and Regier 1987). Among these

The condition of the physical habitat is critically symptoms of ecosystem dysfunction are: (1) alteration
important in stream (lotic) ecosystems and can change in biotic community structure to favor smaller life
more easily and quickly than in most other ecosystems forms; (2) reduced species diversity, (3) increased dom-
(Power et al. 1988). Sedimentation of aquatic ecosys- inance by ‘‘r’’ selected species, (4) increased domi-
tems is a common outcome of many land management nance by exotic species, (5) shortened food-chain
activities, including timber harvesting, road building, length, (6) increased disease prevalence, and (7) re-
mining, and grazing (Meehan 1991, Reid 1993, Waters duced population stability (Rapport 1992). While
1995). Consequently, stress due to increased sedimen- stressed ecosystems do not always manifest all of the
tation is one of the most common causes of ecological above symptoms, in the majority of cases, most do
dysfunction in lotic ecosystems (Waters 1995). The appear (Rapport et al. 1985). The major challenge in
negative impacts of sediments on stream-dwelling or- ecosystem diagnosis is to identify early warning signs
ganisms, including fishes, stream and benthic inver- of incipient pathology (Rapport 1992, Rapport and Re-
tebrates, and periphyton, are well documented (New- gier 1995). Odum (1992) noted that ‘‘the first signs of
combe and MacDonald 1991, Meehan 1991, Waters environmental stress usually occur at the population
1995). However, few studies have examined the direct level, affecting especially sensitive species’’ (see also
effects of sediments on stream-dwelling amphibians Rapport and Regier 1995). Such sensitive species are
(see Hall et al. 1978, Hawkins et al. 1983, Bury and obvious candidates for indicator species. The use of
Corn 1988, Corn and Bury 1989). indicator species is fraught with pitfalls and must be
In the developing lexicon of ecosystem ‘‘health’’ based on precise definitions and procedures to be ef-
(see Suter 1993 for a critique of the health analogy fective and credible (Landres et al. 1988). However,
applied to ecosystems), there is consensus that ‘‘un- the approach of finding and monitoring early indicators
healthy’’ or stressed ecosystems manifest common of ecosystem stress has the advantage of shortening the
symptoms of degradation (Godron and Forman 1983, relatively slow response time of the whole ecosystem
to stress by shifting attention to the much quicker re-
Manuscript received 21 February 1997; revised 14 Feb- sponse time of sensitive species (Rapport 1992). Such
ruary 1998; accepted 6 March 1998. indicators would ideally have the combined attributes
1118
November 1998 AMPHIBIANS AS BIOINDICATORS 1119

of being holistic, early warning, and diagnostic (Rap- planet, and even the relatively dynamic lotic environ-
port 1992). Furthermore, these indicators need to be ment (Power et al. 1988) within late seral redwood
abundant and tractable elements of the system whose forest is comparatively stable. Contrasting the potential
natural perturbations can be distinguished from states life-spans of the native amphibians relative to that of
indicative of ecosystem dysfunction. the trees that define this ecosystem, it is certainly a
Amphibians are thought to be sensitive to pertur- highly stable environment from the perspective of the
bations in both terrestrial and aquatic environments be- amphibians. We believe that it is reasonable to assume
cause of their dual life histories, highly specialized that in such a stable system, natural population per-
physiological adaptations, and specific microhabitat re- turbations within the amphibian assemblage would be
quirements (Bury 1988, Vitt et al. 1990, Wake 1990, minimized, and marked changes in their numbers over
Olson 1992, Blaustein 1994, Blaustein et al. 1994 a, a short period of time could confidently be considered
Stebbins and Cohen 1995). During their aquatic stages, an indication of ecosystem dysfunction. Even with
many stream-dwelling amphibian larvae are highly metamorphosis and the consequent movement of in-
specialized in their uses of lotic microhabitats for both dividuals from aquatic to terrestrial environments, pop-
foraging and cover. Such specialized adaptations can ulations of long-lived species with multiyear larval pe-
render them susceptible to even minor environmental riods would remain relatively stable. Any pulses of
changes that alter their ability to seek cover from pred- newly hatched larvae entering the system could easily
ators and to forage for phytoplankton, zooplankton, be accounted for in analysis by removing the first year
insects, and other invertebrates. In lotic habitats these class if that were appropriate given the question being
specializations are shared with early life stages of both addressed. While we can offer no direct evidence from
anadromous and freshwater fishes, as well as many the Pacific Northwest in support of our assumption of
stream invertebrates. Amphibians are relatively long- stable amphibian populations in stable environments,
lived compared with invertebrates and fishes (e.g., there are relevant data from forested ecosystems of the
Moyle 1976, Groot and Margolis 1991). Daugherty and eastern United States. Hairston (1987) indicated that
Sheldon (1982a) reported a tailed frog with a known
stream salamander populations from the Appalachian
age of 14 yr, and Hairston (1987) reported longevity
Mountains (Desmognathus spp.) have remained stable
records for six families of salamanders that ranged from
for up to seven years (length of time studied). He also
10 to 55 yr. Amphibians are also highly philopatric
reported stable populations in pond and terrestrial en-
compared to most fishes (see Daugherty and Sheldon
vironments (see also Hairston and Wiley 1993), and
1982b, Welsh and Lind 1992), can occur in relatively
concluded that salamander populations are apparently
stable numbers (Hairston 1987), and are readily sam-
minimally affected by stochastic events, unless these
pled. Thus, we believe they are potentially more trac-
events are destructive of the habitat (Hairston 1987).
table and reliable environmental indicators than these
A combination of natural and anthropogenic events
other taxa. Few studies have been designed specifically
during the fall of 1989 created a natural experiment,
to examine the responses of amphibians to environ-
mental perturbations in aquatic ecosystems (but see which afforded us an opportunity to test the response
Moyle 1973, Hall et al. 1978, Hawkins et al. 1983, of amphibians to ecosystem stress in streams of an old-
Hayes and Jennings 1986, Corn and Bury 1989, Welsh growth redwood ecosystem. The Redwood National
1990, Blaustein et al. 1994b). In this paper we report Park bypass project was a large highway construction
the results of a study of amphibian population re- project adjacent to the eastern border of Prairie Creek
sponses to alterations of the physical habitat in streams Redwoods State Park, Humboldt County, California.
due to abnormal infusions of fine sediments and eval- This area received .12.7 cm of precipitation during a
uate the use of amphibians as indicators of stream eco- major storm 20–23 October 1989, which resulted in
system dysfunction. large infusions of sediments from the ongoing road
The primary challenge with indicator species, or any construction into seven stream channels in the Prairie
study where causal arguments are being made about Creek drainage. The fine sediment layer deposited on
shifts in presence or abundance, lies in separating any affected streambeds measured 0.3–5.0 cm in depth
natural fluctuations in numbers from those attributable (Anonymous 1991).
to anthropogenic environmental stresses (Pechmann et Here we provide an analysis of the effects of this
al. 1991, Blaustein 1994, Blaustein et al. 1994b, Pech- combination of shallow mass wasting and surficial ero-
mann and Wilbur 1994). The coast redwood (Sequoia sion (hereafter the erosion event) on densities of the
sempervirens) ecosystem (Zinke 1977) where our study three most abundant native, stream-dwelling amphib-
was conducted is self-perpetuating and in a late-seral ians in five of these streams. Our approach was to ex-
or old-growth stage (i.e., in a steady state; Bormann amine and compare these densities with those of the
and Likens 1979; see also Franklin and Hemstrom same species in five unimpacted (control) streams in
1981, Veirs 1982). Based on the resistance–resilience the same basin. We also examined fine-scale micro-
model of ecosystem stability (Waide 1995), the coastal habitat relationships within the unimpacted streams to
redwood ecosystem is among the most stable on the help interpret any differences in amphibian numbers
1120 HARTWELL H. WELSH, JR. AND LISA M. OLLIVIER Ecological Applications
Vol. 8, No. 4

FIG. 1. Locations of impacted (*) and un-

impacted streams in Prairie Creek State Red-
woods and Redwood National Park, Humboldt
County, California. All drainages were sampled
for amphibians from June through August 1990.
(Modified from Welsh et al. 1997.)

that might be revealed between the impacted and un- morphic forms of this salamander are strictly aquatic
impacted sets of streams. and general accounts of their habitat describe them as
bottom dwellers in mountain streams, lakes, and ponds
Site and species accounts (Nussbaum et al. 1983, Leonard et al. 1993) where they
For our study of the impacts of the erosion event on are often found under cobble-size substrates (Parker
the amphibian community we selected the five of seven 1991, Welsh 1993). This salamander can be extremely
streams affected by the event that drained westward abundant in small streams of the Pacific Northwest,
into Prairie Creek (Anonymous 1991). Our five control accounting for as much as 99% of the predator biomass
streams were selected from those unimpacted streams in such systems (Murphy and Hall 1981, Hawkins et
in the same drainage, with a similar westward aspect, al. 1983). Larvae of this species typically require two
that were interspersed among the impacted streams complete summers of growth before metamorphosis oc-
(Fig. 1). The two sets of streams (five unimpacted and curs (Leonard et al. 1993).
five impacted) were of similar size and orientation, and Tailed frog.—Welsh (1993) summarized the niche
vegetative cover. Of the total set of 10 streams, nine for the larval tailed frog as ‘‘. . . clear, cool, fast-flow-
were located within Prairie Creek Redwoods State Park ing streams in coniferous forests of the Pacific North-
and one control stream (Little Lost Man Creek) was west.’’ Conditions within streams with larvae ‘‘. . . con-
located in the same drainage basin in adjacent Redwood sisted of fast current over coarse gravel, pebble, cobble,
National Park (Fig. 1). or boulder substrates, with little fine sediment’’ (Welsh
Three species of amphibians were sufficiently abun- 1993). These conditions included intermediate to high
dant in these streams to enable our study. water velocity and cold water temperatures (Welsh
Pacific giant salamander.—The larval and paedo- 1990, 1993; see also H. H. Welsh and A. J. Lind,
November 1998 AMPHIBIANS AS BIOINDICATORS 1121

unpublished manuscript). The strong association with Comparisons of physical habitat in unimpacted and
fast-flowing, cold water habitats probably reflects the impacted streams
evolutionary history of this frog (sensu Holt 1987).
We lumped similar mesohabitat types into five com-
Tailed frogs are unique among temperate anurans in
posite categories (after Hawkins et al. 1993), in order
being specifically adapted to these unusual and extreme
to increase sample sizes and simplify analyses: (1) all
conditions (cf. deVlaming and Bury 1970, Gradwell
pools, including main channel, backwater, and second-
1971, Claussen 1973, Brown 1975). Larvae from low-
ary channel pools; (2) glides and runs; (3) riffles; (4)
land populations of the tailed frog typically require 1–
step runs; and (5) step pools. These five categories are
2 yr before metamorphosis occurs (Leonard et al.
hereafter referred to as the primary mesohabitat types
1993).
(Appendix).
Southern torrent salamander.—General descriptions
In order to insure that any differences in amphibian
of the habitat of this small, secretive salamander in-
densities detected between the unimpacted and im-
dicate that it occurs in and along small streams, spring
pacted streams could not be attributed to differences
heads, and seepages (Anderson 1968, Nussbaum and in stream reach type (alluvial, braided, or confined) or
Tait 1977, Nussbaum et al. 1983, Good and Wake 1992, differences in the composition of primary mesohabitat
Welsh 1993, Welsh and Lind 1996). Larval individuals types, we tested for differences in these parameters
can be found in the loose substrates of small stream- between the two stream sets. We performed unpaired
beds. Adults are both stream and streamside dwellers, Student’s t tests (Zar 1995) of the mean proportions of
occurring where water flows through a matrix of un- stream length by reach type and primary mesohabitat
sorted rock substrates (J. Baucom, personal commu- type for each set of streams. The significance level (a)
nication). Typical habitats include the splash zones of was set at 0.05 with a Bonferroni adjustment (Stevens
rocky tumbling brooks and waterfalls. Adults often oc- 1986) applied for multiple tests (a for mesohabitat type
cur side-by-side with larvae within coarse substrates tests 5 0.01; a for reach type tests 5 0.017).
in streams (Welsh and Lind 1992, 1996). The southern To evaluate sediment loads in each stream we sam-
torrent salamander has a four and one-half to five year pled the pool mesohabitats where fine sediments (,2
larval period (Leonard et al. 1993). mm) tend to collect (Lisle and Hilton 1992). Fine sed-
iment depths were measured at three locations in each
METHODS pool bowl (the upstream end, the middle, and at the
From June to August 1990, we sampled five impacted downstream end) (Appendix), with the three measure-
ments averaged for analysis. We also visually estimated
(subjected to a mass sediment infusion) and five un-
the percentage of embedded coarse substrate at the pool
impacted streams. Our study design assumed that am-
tail (Appendix). The two pool sediment variables were
phibian community composition and densities in the
employed to evaluate differences in fine sediments be-
unimpacted streams resembled the composition and
tween the two sets of streams but were not used in the
densities present in the impacted streams had the ero-
analyses of amphibian densities. Unpaired Student’s t
sion event not occurred. The similarities and proximity
tests were used to test differences in the mean sediment
of these 10 streams, the stability of the coast redwood
depth and the mean percentage of pool tail substrate
ecosystem, and the lack of any documented historical
embedded for each set of streams (a 5 0.05).
perturbations that impacted any of these streams prior
to the highway construction project, all support this Amphibian sampling
assumption. We alternated sampling between impacted
and unimpacted streams to ameliorate the effects of Stream habitats for amphibian sampling were se-
any recruitment of newly hatched larval amphibians on lected using a random systematic design based on
the density estimates. In addition, we tested the sup- stream length and ratios of primary mesohabitat types
position that the two stream sets were geomorphically along each stream (Welsh et al. 1997). Working from
similar (see Methods: Comparisons of physical habi- west to east (upstream) and beginning at Highway 101
tat). (Fig. 1), we sampled the first unit of every mesohabitat
type encountered, then a randomly selected unit of each
Habitat typing of streams type between the second and the sixth, then every fifth
unit of each type thereafter. This provided a propor-
Our sampling design was stratified by mesohabitat tional sampling effort of each mesohabitat type relative
type (e.g., pool, run, riffle, and other types; Welsh et to its availability in each stream.
al. 1997). Prior to sampling for amphibians, each Within each selected stream mesohabitat unit, we
stream was mapped from Highway 101 east to its head- systematically placed one or more amphibian sampling
waters (Fig. 1). The mapping included the subdivision units (cross stream belt transects) based on habitat
and classification of streams at the level of geomor- length, placing one belt transect for every 10 m of
phological reach type (braided, alluvial, or confined) habitat (Fig. 2). Belt transects (hereafter belts) were
and stream mesohabitat composition (Appendix). 0.6 m wide and extended from bank to bank so that
1122 HARTWELL H. WELSH, JR. AND LISA M. OLLIVIER Ecological Applications
Vol. 8, No. 4

combined for analysis. Only six adult tailed frogs were

captured. Because of this small sample and their pri-
marily terrestrial habitat associations, they were omit-
ted from the analyses. Four adult torrent salamanders
were found, and because they occur in the same aquatic
microhabitats as the larvae, the two life stages were
combined for analyses. Histograms of snout–vent
length indicated that our sampling occurred after the
recruitment of Pacific giant and southern torrent sala-
mander larvae, and before the recruitment of tailed frog
larvae to our stream set.

Biotic and abiotic measurements associated with

amphibian sampling
In order to characterize fine-scale or microhabitat
attributes associated with amphibian captures, we es-
timated or measured 28 microhabitat parameters as-
sociated with the individual belt samples (Fig. 2) (Ap-
pendix: microhabitat attributes).

Statistical analyses
We used Statistical Analysis System (SAS version
6.12; SAS Institute 1997) to conduct all data analyses.
In contrast to the stricter a 5 0.05 used in testing for
differences in geomorphology and pool fine sediment
levels among the sets of streams, we set a 5 0.10 for
our analysis of variance (ANOVA), analysis of co-
variance (ANCOVA), and correlation analysis. This
FIG. 2. Schematic representation of random-systematic moderate a provides a criterion more appropriate for
belt placement within selected mesohabitats (see Methods: the detection of ecological trends and it increases sta-
Amphibian sampling). Modified from Welsh et al. (1997). tistical power (Toft and Shea 1983, Toft 1991) (see
Schrader-Frechette and McCoy [1993] for a thorough
sampling unit length varied with stream width. The justification and evaluation of this methodological ap-
length of each mesohabitat unit was divided by the total proach in ecology). Dependent variables were natural
number of belts desired (approximately one every 10 m) log-transformed, and some independent variables were
to determine exact spacing, and a random distance be- arcsine-transformed, to meet the assumptions of nor-
tween 0 and 10 m was used to determine placement of mality and homogeneity of variance.
the first belt (Welsh et al. 1997). Each belt was then Analysis of variance.—We used partial hierarchical
thoroughly searched for amphibians. The area was first ANOVA to test for differences in densities of each
scanned for visible animals and then all cover objects amphibian species (the dependent variables) between
were removed working from bank to bank and upstream impacted and unimpacted streams. Within each impact
until the entire area was searched. Animals were cap- category (impacted and unimpacted) there are five
tured using a metal mesh net, identified, sexed (if pos- streams, and within those streams five mesohabitat
sible), measured (snout–vent and total length), and re- types are possible. This method permits us to partition
leased after sampling was completed. Cover objects the total variability into three components while ad-
were returned to their original positions. We are con- justing for unequal sample sizes within the different
fident that our searches captured all amphibians present levels. The unit of analysis was the mesohabitat unit
in the open, and probably most of those under the first (i.e., mesohabitat types within streams within impacts).
layer of large substrate (.16 mm diameter). The effect for impact was calculated using streams
Three species were detected and sampled in numbers within impact as the mean square error (MSE). The
sufficient for statistical analyses: larval and paedo- effects for mesohabitat type and impact by mesohabitat
morphic (animals with larval morphology and sexual type interaction were calculated using mesohabitat type
maturity) Pacific giant salamanders (Dicamptodon te- within stream within impact as the MSE. The mean
nebrosus), larval tailed frogs (Ascaphus truei), and lar- squares were calculated using Type I sums of squares
val and adult southern torrent salamanders (Rhyaco- (SS) as all mesohabitats were sampled in proportion to
triton variegatus). We did not differentiate larval and their occurrence in the population (Milliken and John-
paedomorphic Pacific giant salamanders, and they were son 1984). This was not the case with the overall model
November 1998 AMPHIBIANS AS BIOINDICATORS 1123

F; thus it was not used to determine model significance. a nonsignificant F for the covariate, failed to explain
The following null hypotheses were tested: additional effects beyond those detected in the ANO-
H01: There are no significant differences between im- VA.
pact and no impact for any of the three species; Correlation analysis.—We performed correlation
H02: There are no significant differences among me- analyses of 28 microhabitat attributes measured or es-
sohabitat types for any of the three species; timated within each belt sample (Appendix). We re-
H03: There is no interaction between mesohabitat stricted this analysis to those data from belts in the
type and impact for any of the three species. control streams with captures of each of the three spe-
When ANOVA provided evidence of differences cies in order to address the question ‘‘what measured
among mesohabitats, we used Tukey’s studentized mul- or estimated microhabitat variables best characterized
tiple range test to compare the means. the fine-scale ecological relationships of the resident
Analysis of covariance.—In order to more closely amphibians under pristine stream conditions?’’
examine the effects of specific fine sediment parameters
(Appendix: fine aquatic substrates) on individual spe- RESULTS
cies we employed ANCOVA. We used this method to Comparisons of physical attributes between impacted
look for evidence of other possible effects of the ero- and unimpacted streams
sion event that were not measured during our sampling
(e.g., chronic suspended sediment load, bed instability) We surveyed and habitat typed 3.6 km of impacted
that may be indirectly related to sediment transport. streams and 3.2 km of unimpacted streams (Fig. 1).
This allowed us to adjust the ANOVA models by each Comparisons of mean proportions of stream length by
fine sediment variable measured (Appendix). The mod- reach type and primary mesohabitat type indicated that
el structure for the ANCOVA is the same as that of the there were no significant differences between the im-
ANOVA described (partial hierarchical) and employed pacted and unimpacted sets of streams (Table 1). As-
Type I SS. The following null hypotheses were tested: suming that the relative amount of available habitat is
H01: There are no significant differences between im- a reasonable indicator of the number of organisms that
pact and no impact for any of the three species, when may be supported there (Southwood 1977, 1988), we
densities are adjusted by each of the fine sediment co- consider that this lack of difference in geomorpholog-
variates; ical composition supported our assumption that the am-
H02: There are no significant differences among me- phibian assemblages in the two sets of streams probably
sohabitat types for any of the three species, when den- would have had similar species composition and den-
sities are adjusted by each of the fine sediment covari- sities had the erosion event not occurred.
ates; Mean fine sediment (,2.0 mm) depths in the im-
H03: There are no significant differences for the pacted pools ranged from 0.1 to 25.0 cm compared with
interaction of mesohabitat type and impact for any of 0.0–4.0 cm in the unimpacted pools (Fig. 3). Percent-
the three species, when densities are adjusted by each age embeddedness of pool tails ranged from 10 to 100%
of the fine sediment covariates. in the impacted streams and from 0 to 85% in the un-
The five fine sediment parameters consisted of two impacted streams. Tests between the two sets of streams
visual estimates of substrate composition, one estimate for both the mean sediment depth in pool bowls, and
of substrate condition, and two measures of fine sed- the percentage of substrate embeddedness at the pool
iment derived from grab samples collected immediately tails, showed significantly greater amounts of sediment
adjacent and upstream of the belts (Appendix: fine in the impacted streams (Fig. 3). This clearly demon-
aquatic substrates). In order to simplify the ANCOVA strates an impact effect of the 1989 erosion event still
by eliminating redundancy among closely related vari- remained when we sampled in 1990.
ables, we used correlation analysis to select one vari-
Comparisons of amphibian densities between
able from those pairs that described a similar parameter
impacted and unimpacted streams
(percentage fines and silt volume, r 5 0.466, P ,
0.0001; percentage sand and sand volume, r 5 0.303, We sampled a total of 267 belts in 179 mesohabitat
P , 0.0001). From each of these pairs we chose the units, with 93 habitat units (137 belts) in the impacted
variable with the highest correlation with our depen- streams and 86 habitat units (130 belts) in the unim-
dent variables (percentage fines), or if the significant pacted streams. We captured a total of 540 amphibians;
r values were equivocal relative to the dependent vari- larval and paedomorphic individuals of the Pacific gi-
ables, we chose the measured variable (sand volume). ant salamander were the most common (n 5 296), fol-
Significant covariates were determined using Type lowed by larval tailed frogs (n 5 205), and larval and
III sums of squares. Only those ANCOVA results with adult southern torrent salamanders (n 5 39).
a reduced error variance for our tests were meaningful. Analysis of amphibian densities.—Densities of the
Consequently, only those models with a decreased three species varied by mesohabitat type and impact
overall MSE were evaluated further. ANCOVA models (Fig. 4). The Pacific giant and southern torrent sala-
that failed to reduce the MSE over the ANOVA or had manders showed significant differences for impact
1124 HARTWELL H. WELSH, JR. AND LISA M. OLLIVIER Ecological Applications
Vol. 8, No. 4

TABLE 1. Comparison of reach types and mesohabitat composition for 10 streams sampled for aquatic amphibians in Prairie
Creek Redwoods State Park and Redwood National Park, Humboldt County, California, 1990.

Reach types Mesohabitat types

Stream Alluvial Braided Confined All pools Glide/run Riffle Step run Step pool
Unimpacted streams
Corkscrew 0 0 100.0 5.3 2.0 41.5 9.7 41.5
Good 0 67.7 32.3 4.4 5.6 40.6 46.2 4.2
Little Lost Man 67.7 0 32.3 7.6 1.4 3.4 37.0 51.1
S. fork Big Tree 0 0 100.0 10.8 6.2 14.9 0.0 68.1
Sweet 0 3.2 96.8 3.0 0.7 36.6 39.1 21.0
x̄ 13.5 14.2 72.3 6.2 3.2 27.4 26.4 37.2
(14.0) (13.0) (16.0) (1.4) (1.1) (7.7) (9.1) (11.0)
Impacted streams
Big Tree 6.0 27.7 66.3 20.6 3.9 19.3 11.4 45.6
Boyes 84.9 15.1 0 19.9 4.8 19.6 30.2 18.0
Brown 0 0 100.0 26.2 8.5 20.7 22.0 24.3
N. fork Big Tree 0 20.2 79.8 3.7 0.0 12.0 4.2 80.2
Ten Tapo 81.9 18.1 0 18.0 0.0 14.7 12.2 56.4
x̄ 34.6 16.2 49.2 17.7 3.4 17.3 16.0 44.9
(20.0) (4.6) (21.0) (3.7) (1.6) (1.7) (4.6) (11.0)
t 20.87 20.14 0.87 22.88 20.12 1.29 1.03 20.48
P† 0.41 0.89 0.41 0.02 0.91 0.27 0.33 0.64
Note: Percentage of stream length by reach and mesohabitat type, mean, and standard error (in parentheses) are reported.
Comparisons of impacted and unimpacted streams were made using Student’s t.†
† Significant t probability values were interpreted using a Bonferroni adjustment (Stevens 1986).

(sedimentation); in all cases the densities in unim-

pacted streams were greater (Table 2a). The tailed frog
and southern torrent salamander showed significant dif-
ferences among mesohabitat types (Table 2a). There
was also a significant interaction between impact and
mesohabitat type for the tailed frog (Table 2a).
In the impacted streams, there were no significant
differences among mesohabitat types for the Pacific
giant and southern torrent salamanders. However, tailed
frog density was significantly greater in riffles com-
pared to pools (Table 2b). In the unimpacted streams
tailed frog larvae showed strong habitat specialization
and were significantly more abundant in both riffles
and step runs compared with other mesohabitat types
(Table 2b). The torrent salamander also occurred more
often in riffle than pool habitat in the unimpacted
streams (Table 2b), although there were no captures in
pools, glides, or runs (Fig. 4). There were no differ-
ences in mesohabitat type for the Pacific giant sala-
mander in the unimpacted streams (Fig. 4).
Effects of fine sediment attributes.—The Pacific giant
salamander and tailed frog yielded significant covariate
models (Table 3), indicating additional variation was
explained beyond the ANOVA. For these same depen-
dent variables, percentage embedded caused the great-
est reduction in variability (Table 3). In the models that
were adjusted for percentage embedded, there were no
significant differences detected with respect to impact
(the first hypothesis test) for the Pacific giant sala-
FIG. 3. Comparisons of sediment depths (a) and pool tail mander, or the tailed frog, indicating that once the data
embeddedness estimates (b) in impacted and unimpacted
streams. Boxes indicate means (from three measures of sed-
were adjusted for this covariate no further differences
iment depth [cm] along the central axis at the top, middle, could be explained (Table 3a, b).
and bottom of each pool bowl), 6 1 SE . With respect to percentage embedded, the Pacific
November 1998 AMPHIBIANS AS BIOINDICATORS 1125

4). Nine attributes were correlated with Pacific giant

salamander density, two attributes were correlated with
tailed frog density, and five attributes were correlated
with southern torrent salamander density (Table 4). The
two salamander species responded differently to flow
rates within belts. The Pacific giant salamander den-
sities were lower in areas of high flow, while southern
torrent salamander densities increased with flow rate
(Table 4). Pacific giant salamander density increased
in belts with larger amounts of woody debris cover,
while southern torrent salamander density declined in
association with both wood cover and substrates (Table
4).
DISCUSSION
Our study indicated that the stream amphibian com-
munity was negatively impacted by the erosion event
caused by the bypass construction and the October
1989 storm (Table 3, Fig. 4). Our analysis indicated
that this response differed considerably by species (Ta-
ble 3). For example, the ANCOVA model for the Pa-
cific giant salamander suggests that it is less sensitive
than the other species to fine sediments (Table 3), but
it was negatively associated with sand (Table 4). Our
ANCOVA results for the tailed frog (Table 3) suggested
that the impact of the erosion event acted at the level
of microhabitat within streams and consisted primarily
of fine particles restricting access to the streambed ma-
trix (i.e., percentage embedded) (cf. Lisle 1989, Lisle
and Lewis 1992). However, the significant results for
both mesohabitat type and the interaction effects (Table
3) indicated that additional factors may be affecting
the tailed frog. For the Pacific giant salamander and
the tailed frog, we found significant positive associa-
tions with relatively coarse substrates (e.g., cobble; Ta-
ble 4), where matrix interstices can be reduced or elim-
inated by fine sediments (i.e., percentage embedded).
FIG. 4. Densities of three species of amphibians are
shown with respect to impact and mesohabitat type. Bars Pacific giant salamander
represent means (and one standard error) for the stream sets The Pacific giant salamander was the least habitat
(five streams in each). Numbers over bars are belts sampled.
specific, showing no clear association with any partic-
ular stream mesohabitat type (Fig. 4, Table 3). As a
giant salamander showed no differences for mesoha- habitat generalist, this species is most likely affected
bitat type or the interaction (Table 3a). In the model by sedimentation across all stream mesohabitat types,
adjusted for percentage embedded, the tailed frog but probably more so in pools where fine sediment
showed significant results in the tests for mesohabitat accumulation is greatest (Lisle and Hilton 1992).
type and its interaction with impact, indicating that Analysis of substrate associations indicated that
additional sediment effects were influencing the system higher relative amounts of gravel and cobble were the
beyond those explained by the ANOVA (Table 2) and best predictors of Pacific giant salamander abundance
the adjustment for percentage embedded (Table 3b). (Table 4; H. H. Welsh and A. J. Lind, unpublished
The Pacific giant salamander had one additional sig- manuscript). This outcome underscores the relative im-
nificant covariate, percentage fines. As with percentage portance of coarse, rocky substrates, which have a high
embedded above, no significant differences were found relative amount of interstitial space (see also Welsh
in the tests (Table 3a). There were no other significant 1993). Parker (1991) experimentally demonstrated the
covariate models for any of the three species (Table 3). importance of cobble-size substrates as cover for larval
Correlation analyses of microhabitat attributes.—Of Pacific giant salamanders in pool habitats in a stream
the 28 microhabitat parameters we examined, 14 were similar to ours in northwestern California. Concomi-
significantly correlated with amphibian density (Table tantly, we found fewer salamanders in areas with great-
1126 HARTWELL H. WELSH, JR. AND LISA M. OLLIVIER Ecological Applications
Vol. 8, No. 4

TABLE 2. (a) Partial hierarchical analysis of variance (ANOVA) of three amphibian species by impact (presence or absence
of fine sediment infusion), stream number, and mesohabitat type, and (b) Tukey pairwise comparisons of mesohabitat types.

a) ANOVA results
Factor df MSE F P Result
Dependent: Pacific giant salamander
Overall model 46, 132 0.2799 0.96 0.5588
Tests
Impact 1, 8 1.5010 3.95 0.0820 U . I†
Mesohabitat type 4, 29 0.2932 1.58 0.2050 NS
Impact 3 Mesohabitat type 4, 29 0.3058 1.65 0.1881 NS

Dependent: Tailed frog

Overall model 46, 132 0.1803 2.72 0.0001
Tests
Impact 1, 8 0.9252 2.06 0.1888 NS
Mesohabitat type 4, 29 2.2925 11.38 0.0001
Impact 3 Mesohabitat type 4, 29 0.7507 3.73 0.0145
Dependent: Southern torrent salamander
Overall model 46, 132 0.0568 2.78 0.0001
Tests
Impact 1, 8 0.7982 4.93 0.0572 U . I†
Mesohabitat type 4, 29 0.3144 2.67 0.0519
Impact 3 Mesohabitat type 4, 29 0.1258 1.07 0.3896 NS

b) Tukey pairwise comparison results‡

Pacific giant salamander Comparison: Mesohabitat type (with respect to Impact)
Impacted streams Glide/run Step Pool Pool Step Run Riffle
Unimpacted streams Step Pool Riffle Glide/Run Pool Step Run
Tailed frog Comparison: Impact 3 Mesohabitat type
Impacted streams Pool Glide/Run Step Run Step Pool Riffle
Unimpacted streams Pool Glide/Run Step Pool Riffle Step Run
Southern torrent salamander Comparison: Mesohabitat type (with respect to Impact)
Impacted streams Pool Step Run Glide/Run Step Pool Riffle
Unimpacted streams Pool Glide/Run Step Run Step Pool Riffle

† U 5 unimpacted streams, I 5 impacted streams.

‡ Amphibian mean density increases from left to right; lines indicate nonrejecting subsets.

er volumes of sand (Table 4), a condition that limits ciation with coarse substrates (cobble) (Table 4; see
available interstitial spaces (see also Hall et al. 1978, also Nussbaum et al. 1983, Welsh 1993; H. H. Welsh
Murphy and Hall 1981, Murphy et al. 1981, Hawkins and A. J. Lind, unpublished manuscript). Coarse sub-
et al. 1983, Corn and Bury 1989). However, none of strates provide the interstitial space important for cover
the fine sediment variables alone could explain the sig- from both predation and high winter stream flows (e.g.,
nificant differences we saw in giant salamander abun- Metter 1963, 1968), as well as providing abundant sur-
dances with respect to impact (Table 2, Table 3). Be- face area for diatom production, an important food
cause giant salamanders use more available stream me- source. Fast-water habitats are less prone to trapping
sohabitat types (Fig. 4), it is possible they are better sediment due to the higher, more uniform velocity of
able to compensate for habitat loss resulting from sed- water (Lisle and Hilton 1992). However, results for the
imentation (Table 2). Such adjustments might involve tailed frog showed a significant interaction between
changing habitat use patterns or even modifying pre- sediment impact and mesohabitat type (Table 3). This
ferred sites by excavating sediments as has been seen indicated that tailed frog larvae were adversely im-
with an ambystomatid salamander (e.g., Jennings pacted even in those high velocity habitats that are
1996), but these hypotheses are currently untested. likely to have lower sediment loads (Fig. 4). This result
suggests that something other than sediment filling the
Tailed frog larvae interstices was affecting tailed frog abundances in im-
Tailed frog larvae were the most specific in habitat pacted streams. Sediment may impact critical food re-
use, showing a strong association with step runs and sources, both in adjacent lower gradient areas and in
riffles vs. step pools and all other stream mesohabitat those mesohabitats occupied by tailed frog larvae.
types (Fig. 4). They also demonstrated a strong asso- When we examined data from across all streams, we
November 1998 AMPHIBIANS AS BIOINDICATORS 1127

TABLE 3. Partial hierarchical analysis of covariance of three species by impact (presence or absence of sediment), stream
number, and mesohabitat type. The covariates were sediment variables taken in association with animal sampling.

Factor df MSE F P
a) Dependent: Pacific giant salamander
i) Overall model 47, 130 0.268 1.11 0.3127
Covariate: Percentage embedded 1, 8 1.812 6.75 0.0105†
Tests
Impact 1, 8 0.071 0.19 0.6718
Mesohabitat type 4, 29 0.346 1.84 0.1484
Impact 3 Mesohabitat type 4, 29 0.313 1.66 0.1860
ii) Overall model 47, 131 0.269 1.11 0.3199
Covariate: Percentage fines 1, 8 1.711 6.36 0.0129†
Tests
Impact 1, 8 0.474 1.27 0.2928
Mesohabitat type 4, 29 0.311 1.61 0.1984
Impact 3 Mesohabitat type 4, 29 0.353 1.83 0.1510
iii) Overall model 47, 130 0.280 0.96 0.5512
Covariate: Sand volume 1, 8 0.248 0.89 0.3476
b) Dependent: Tailed frog
i) Overall model 47, 130 0.166 3.17 0.0001
Covariate: Percentage embedded 1, 8 2.255 13.61 0.0003†
Tests
Impact 1, 8 0.519 1.34 0.2808
Mesohabitat type 4, 29 0.933 3.82 0.0129‡
Impact 3 Mesohabitat type 4, 29 0.720 2.95 0.0367‡
ii) Overall model 47, 131 0.178 2.75 0.0001
Covariate: Percentage fines 1, 8 0.470 2.64 0.1065
iii) Overall model 47, 130 0.179 2.72 0.0001
Covariate: Sand volume 1, 8 0.435 2.42 0.1219
c) Dependent: Southern torrent salamander
i) Overall model 47, 130 0.058 2.69 0.0001
Covariate: Percentage embedded 1, 8 0.024 0.41 0.5212
ii) Overall model 47, 131 0.057 2.71 0.0001
Covariate: Percentage fines 1, 8 0.009 0.16 0.6881
iii) Overall model 47, 130 0.057 2.82 0.0001
Covariate: Sand volume 1, 8 0.008 0.14 0.7119
Note: Test results are not reported for those models lacking a significant covariate.
† Covariate models with a significant model F using Type III SS and reduction in the MSE in the overall model over that
of the ANOVA.
‡ Hypothesis tests with a significant effect detected using Type I SS after the covariate has been incorporated into the
model.

found highly significant negative correlations between other two species examined. Southern torrent salaman-
percentage of nonfilamentous algae and the three fine ders were absent from pools, and glides and runs. They
sediment variables used in our ANCOVA (percentage occurred predominately in riffles, step runs, and step
embedded, r 5 20.572, P 5 0.0001; percentage fines, pools (Fig. 4). Thus, all of the mesohabitat types where
r 5 20.476, P 5 0.0001; sand volume, r 5 2.393, P they did occur were comprised primarily of moving
5 0.0001). Welsh (1993) reported that the amount of and mixing waters. Even in these mesohabitats, south-
nonfilamentous algae (diatoms or periphyton) was a ern torrent salamanders were found in higher abun-
significant predictor of the presence and abundance of dance in the thalweg (main flow) and appeared to avoid
tailed frog larvae. Diatoms are the primary food for mesohabitats composed primarily of margin (Table 4).
larval tailed frogs (Metter 1964, Nussbaum et al. 1983), This meso- and microhabitat specificity may be related
so it follows that they would occur in greater abundance to physiological constraints resulting from their spe-
where periphyton is plentiful and avoid areas where it cialized, reduced gill-arch system that restricts them to
is sparse or absent. Even a thin layer of fine sediment habitats that are characterized by cold, highly oxygen-
can block sufficient light and inhibit the growth of algae ated water (i.e., mountain brooks, Valentine and Dennis
(Newcombe and MacDonald 1991). During high flows 1964). The specific meso- and microhabitat associa-
greater amounts of sediment might scour algae off tions of the southern torrent salamander could reflect
streambed substrates and thereby reduce periphyton a response to lower sediment loads in these habitats,
biomass (Alabaster and Lloyd 1982). but the lack of an interaction (Table 2) suggests that
Southern torrent salamander this habitat specificity is an ecological or evolutionary
The southern torrent salamander demonstrated in- adaptation (Holt 1987) rather than a temporary re-
termediate mesohabitat specificity compared with the sponse to adverse conditions. This species also ap-
1128 HARTWELL H. WELSH, JR. AND LISA M. OLLIVIER Ecological Applications
Vol. 8, No. 4

TABLE 4. Significant results of Pearson product-moment the possibility that the lack of an interaction effect may
correlations are reported for 14 microhabitat variables (Ap-
pendix). Correlations were performed using stream belts have resulted from the low number of belts with cap-
with captures in unimpacted streams. tures (10%) or high variability, which may have par-
tially compromised our ability to detect differences.
Pacific Southern In summary, our study indicated that sediment de-
giant torrent posits from the October 1989 storm event had a neg-
sala- sala-
Variable mander† Tailed frog‡ mander§ ative effect on amphibian populations, with a pro-
Aquatic conditions
nounced effect on two out of three species examined.
Water temperature 20.335 ··· ··· Furthermore, our ANCOVA results add new insight
Proportion margin ··· ··· 20.448 into the explanation for reduced abundances of tailed
Flow thalweg 20.282 ··· 0.311 frog larvae based on sedimentation of interstices of-
Cover types fered by Corn and Bury (1989). It appears that tailed
Woody debris cover 0.318 ··· 20.421 frog larval abundances were reduced by some factor
Riparian vegetation 0.197 ··· ··· other than the direct impact of embeddedness, possibly
Large rock cover 20.383 ··· ···
Without cover ··· ··· 0.502 as a result of the inhibition of periphyton growth, the
scouring of that growth from streambed substrates, or
Coarse aquatic substrates
Cobble 0.245 0.298 ···
both. Our results also documented differential use of
Large rock substrates 20.437 ··· ··· stream mesohabitats by two of these species, and dem-
Fine gravel volume 0.234 ··· ··· onstrate how fine sediments can differentially affect
Woody debris sub- 0.273 ··· 20.460 stream amphibians in accordance with their particular
strates
meso- and microhabitat associations.
Fine aquatic substrates
Embedded ··· 20.461 ··· Amphibians as bioindicators
Fines ··· ··· 20.654
Sand volume 20.272 ··· ··· Results of our analyses are consistent with other
† Correlations with salamander density are based on 78 studies that examined the habitat associations of these
belts with salamander captures; correlations . 0.188 are sig- species at finer spatial scales and in ecosystems other
nificant at P 5 0.10. than the redwoods (Murphy et al. 1981, Hawkins et al.
‡ Correlations with tadpole density using 49 stream belts 1983, Corn and Bury 1989, Bury et al. 1991, Parker
in the four primary mesohabitat types that had tailed frog
captures (step runs, step pools, runs/glides, riffles); correla- 1991, Welsh 1993; H. H. Welsh and A. J. Lind, un-
tions . 0.238 are significant at P 5 0.10. published manuscript). Bury and Corn (1988) dis-
§ Correlations with salamander density using 19 stream cussed the potential negative impacts of erosion events
belts in the three primary mesohabitat types that had southern
torrent salamander captures (step runs, step pools, and riffles); on stream amphibians of the Pacific Northwest. Such
correlations . 0.389 are significant at P 5 0.10. impacts have been documented for other stream sys-
tems in connection with timber harvesting activities
and associated road building (Burns 1972, Beschta
peared to use areas lacking large cover objects (Table 1978, Rice et al. 1979, Reid and Dunne 1984, Cham-
4). We suspect that their avoidance of wood cover and berlin et al. 1991, Furniss et al. 1991). Corn and Bury
substrates could be a means to elude predatory Pacific (1989) documented differences in amphibian species
giant salamanders, which were often found associated richness and in the density and biomass of southern
with this cover type (Table 4). Stebbins (1953) and torrent salamanders, tailed frog larvae, and Pacific gi-
Nussbaum (1969) also speculated that Pacific giant sal- ant salamanders in logged vs. unlogged streams in
amander presence may restrict southern torrent sala- southern Oregon. They attributed these declines to loss
mander distribution. of critical microhabitat due to infusions of fine sedi-
The lack of a significant covariate model for the ments. Populations of stream amphibians can be par-
southern torrent salamander indicated that no further ticularly sensitive to increased siltation because they
effects were detected over what was indicated by the frequent interstitial spaces among the loose, coarse sub-
ANOVA. However, the correlation analysis for this sal- strates that comprise the matrix of most natural stream-
amander showed a strong negative relationship with beds of the Pacific Northwest (Bury and Corn 1988,
percentage fines (Table 4). Previous research also con- Corn and Bury 1989). Sedimentation fills these spaces,
cluded that torrent salamanders are sensitive to fine reducing available cover and foraging area and, un-
sediments in, and substrate embeddedness of, the doubtedly, has similar impacts on other substrate-
streambed matrix (Welsh 1993, Welsh and Lind 1996). dwelling biota (cf. Lisle 1989, Lisle and Lewis 1992;
Nonetheless, the southern torrent salamander may be see also Waters 1995).
able to compensate to some degree for the negative As to the question of their applicability as bioindi-
effects of sedimentation by favoring shallow stream cators of environmental stress, we conclude that mea-
microhabitats with steady flow where they occur in suring and monitoring stream amphibian densities can
close association with cobble substrates (Welsh 1993, provide a highly suitable and extremely sensitive ba-
Welsh and Lind 1996). However, we cannot discount rometer of ecological stress resulting from fine sedi-
November 1998 AMPHIBIANS AS BIOINDICATORS 1129

ment inputs, arguably one of the most pervasive stres- Blaustein, A. R., P. D. Hoffman, D. G. Hokit, J. M. Kie-
sors of lotic systems worldwide (Waters 1995). Other secker, S. C. Walls, and J. B. Hayes. 1994 b. UV repair
and resistance to solar UV-B in amphibian eggs: a link
studies have indicated that the tailed frog and torrent to population declines? Proceedings of the National
salamander also show a marked sensitivity to another Academy of Sciences (USA) 91:1791–1795.
stressor in lotic systems, increased water temperature Blaustein, A. R., D. B. Wake, and W. P. Sousa. 1994 a.
(Brattstrom 1963, deVlaming and Bury 1970, Claussen Amphibian declines: judging stability, persistence, and
susceptibility of populations to local and global extinc-
1973, Welsh 1990, Welsh and Lind 1996). We believe
tions. Conservation Biology 8:60 –71.
that stream amphibians demonstrate strong potential as Bormann, F. H., and G. E. Likens. 1979. Pattern and pro-
‘‘sensitive species’’ (cf. Odum 1992), whose numbers cess in a forested ecosystem. Springer-Verlag, New
can change relatively quickly in response to a range of York, New York, USA.
environmental perturbations. Furthermore, use of Brattstrom, B. H. 1963. A preliminary review of the ther-
mal requirements of amphibians. Ecology 44:238–255.
streambed interstices by amphibians is a characteristic Brown, H. A. 1975. Temperature and development of the
shared with early life stages of both resident and anad- tailed frog, Ascaphus truei. Comparative Biochemistry
romous fishes, as well as many stream invertebrates. and Physiology 50:397– 405.
These other taxa, however, are either short-lived, ex- Burns, J. W. 1972. Some effects of logging and associated
plosive breeders, or subject to seasonal movements, all road construction on northern California streams. Trans-
actions of the American Fisheries Society 101:1–17.
of which can complicate their use as bioindicators. Bury, R. B. 1988. Habitat relationships and ecological
Many species of stream-dwelling amphibians are high- importance of amphibians and reptiles. Pages 61–76 in
ly philopatric, long-lived, and occur in relatively stable K. J. Raedeke, editor. Streamside management: riparian
populations in undisturbed ecosystems. These attri- wildlife and forestry interactions. College of Forest Re-
butes can make their relative numbers a useful and sources, University of Washington, Seattle, Washington,
USA.
reliable indicator of environmental perturbations, both Bury, R. B., and P. S. Corn. 1988. Responses of aquatic
from known causes (Corn and Bury 1989, Blaustein et and streamside amphibians to timber harvest: a review.
al. 1994b) and also possibly from causes that have yet Pages 165–181 in K. J. Raedeke, editor. Streamside man-
to be identified (e.g., Corn and Fogleman 1984, Wey- agement: riparian wildlife and forestry interactions. Col-
goldt 1989, Drost and Fellers 1996, Laurance 1996, lege of Forest Resources, University of Washington, Se-
attle, Washington, USA.
Laurance et al. 1996, Pounds et al. 1997, Woolbright Bury, R. B., P. S. Corn, K. B. Aubry, F. F. Gilbert, and L.
1997, Lips 1998). L. C. Jones. 1991. Aquatic amphibian communities in
ACKNOWLEDGMENTS Oregon and Washington. Pages 353–362 in L. F. Rug-
giero, K. B. Aubry, A. B. Carey, and M. H. Huff, tech-
We thank D. Waters and B. Twedt for collecting the data, nical coordinators. Wildlife and vegetation of unmana-
and D. Waters and D. Hankin for assistance with the sampling ged Douglas-fir forests. USDA Forest Service General
design. A. Lind helped with the analysis and she, R. Wilson, Technical Report PNW-285.
D. Reese, B. Bingham, J. Waters, and especially B. Harvey Chamberlin, T. W., R. D. Harr, and F. H. Everest. 1991.
made helpful comments on earlier drafts. We also thank two Timber harvesting, silviculture, and watershed process-
anonymous reviewers for their many useful comments on the es. Pages 181–205 in W. R. Meehan, editor. Influences
manuscript. We thank J. Baldwin for statistical guidance, and of forest and rangeland management on salmonid fishes
K. Shimizu for help with Figure 1. We are grateful to the and their habitats. American Fisheries Society Special
staffs of Prairie Creek State Park and Redwood National Park, Publication 19, Bethesda, Maryland, USA.
and especially Valerie Gizinski, for help and encouragement. Claussen, D. L. 1973. The thermal relations of the tailed
Funding was provided by the California Department of Trans-
frog, Ascaphus truei, and the Pacific treefrog, Hyla re-
portation under contract number 01C757; we thank Mark
gilla. Comparative Biochemistry and Physiology 44:
Moore of this agency for his assistance.
137–171.
LITERATURE CITED Corn, P. S., and R. B. Bury. 1989. Logging in Western
Alabaster, J. S., and R. Lloyd. 1982. Finely divided solids. Oregon: responses of headwater habitats and stream am-
Pages 1–20 in J. S. Alabaster and R. Lloyd, editors. phibians. Forest Ecology and Management 29:39 –57.
Water quality criteria for freshwater fish. Second edition. Corn, P. S., and J. C. Fogleman. 1984. Extinction of mon-
Butterworth, London, UK. tane populations of the northern leopard frog ( Rana pi-
Anderson, J. D. 1968. Rhyacotriton and R. olympicus. Cat- piens ) in Colorado. Journal of Herpetology 18:147–152.
alogue of American Amphibians and Reptiles 68.1–68.2, Daugherty, C. H., and A. L. Sheldon. 1982 a. Age-deter-
Society for the Study of Amphibians and Reptiles, Hays, mination, growth, and life history of a Montana popu-
Kansas, USA. lation of the tailed frog, Ascaphus truei. Herpetologica
Anonymous. 1991. Monitoring the impacts and persis- 38:461– 468.
tence of fine sediment in the Prairie Creek watershed Daugherty, C. H., and A. L. Sheldon. 1982 b. Age-specific
1989 –1990. U.S. Department of the Interior National movement patterns of the tailed frog, Ascaphus truei.
Park Service, Redwood National Park Report, Arcata, Herpetologica 38:468– 474.
California, USA. deVlaming, V. L., and R. B. Bury. 1970. Thermal selection
Beschta, R. L. 1978. Long-term patterns of sediment pro- in tadpoles of the tailed frog, Ascaphus truei. Journal of
duction following road construction and logging in the Herpetology 4(3– 4):179 –189.
Oregon Coast Range. Water Resources Research 14: Drost, C. A., and G. M. Fellers. 1996. Collapse of a re-
1011–1016. gional frog fauna in the Yosemite area of the California
Blaustein, A. R. 1994. Chicken little or Nero’s fiddle? A Sierra Nevada, USA. Conservation Biology 10:414 –
perspective on declining amphibian populations. Her- 425.
petologica 50(1):85–97. Franklin, J. F., and M. A. Hemstrom. 1981. Aspects of
1130 HARTWELL H. WELSH, JR. AND LISA M. OLLIVIER Ecological Applications
Vol. 8, No. 4

succession in the coniferous forests of the Pacific North- Lips, K. R. 1998. Decline of a tropical montane amphibian
west. Pages 222–229 in D. C. West, H. H. Shugart, and fauna. Conservation Biology 12:106–117.
D. B. Botkin, editors. Forest succession. Springer-Ver- Lisle, T. 1989. Sediment transport and resulting deposi-
lag, New York, New York, USA. tion in spawning gravels, north coastal California. Water
Furniss, M. J., T. D. Roelofs, and C. E. Yee. 1991. Road Resources Research 25(6):1303–1319.
construction and maintenance. Pages 297–323 in W. R. Lisle, T., and S. Hilton. 1992. The volume of fine sediment
Meehan, editor. 1991. Influences of forest and rangeland in pools: an index of sediment supply in gravel-bed
management on salmonid fishes and their habitats. streams. Water Resources Bulletin 28:371–383.
American Fisheries Society Special Publication 19, Be- Lisle, T., and J. Lewis. 1992. Effects of sediment transport
thesda, Maryland, USA. on survival of salmonid embryos in a natural stream: a
Godron, M., and R. T. T. Forman. 1983. Landscape mod- simulation approach. Canadian Journal of Fisheries and
ification and changing ecological characteristics. Pages Aquatic Sciences 49:2337–2344.
12–28 in H. A. Mooney and M. Godron, editors. Dis- Meehan, W. R., editor. 1991. Influences of forest and
turbance and ecosystems components of response. rangeland management on salmonid fishes and their hab-
Springer-Verlag, New York, New York, USA. itats. American Fisheries Society Special Publication 19,
Good, D. A., and D. B. Wake. 1992. Geographic variation Bethesda, Maryland, USA.
and speciation in the torrent salamanders of the genus Metter, D. E. 1963. Stomach contents of Idaho larval Di-
Rhyacotriton (Caudata: Rhyacotritonidae). University of camptodon. Copeia 1962:435– 436.
California Publications in Zoology 126:1–91. . 1964. A morphological and ecological compari-
Gradwell, N. 1971. Ascaphus tadpole: experiments on the son of two populations of the tailed frog Ascaphus truei
suction and gill irrigation mechanisms. Canadian Journal Stejneger. Copeia 1964:181–195.
of Zoology 49:307–332. . 1968. The influence of floods on population struc-
Groot, C., and L. Margolis, editors. 1991. Pacific salmon ture of Ascaphus truei Stejneger. Journal of Herpetology
life histories. University of British Columbia Press, Van- 1:105–106.
couver, British Columbia, Canada. Milliken, G. A., and D. E. Johnson. 1984. Analysis of
Hairston, N. G. 1987. Community ecology and salamander messy data. Vol 1: designing experiments. Van Nostrand
guilds. Cambridge University Press, Cambridge, UK. Reinhold Company, New York, New York, USA.
Hairston, N. G., and R. H. Wiley. 1993. No decline in Moyle, P. B. 1973. Effects of introduced bullfrogs, Rana
salamander populations: a twenty year study in the catesbeiana, on the native frogs of the San Joaquin Val-
southern Appalachians. Brimleyana 18:59 –64. ley, California. Copeia 1973:18–22.
Hall, J. D., M. L. Murphy, and R. S. Aho. 1978. An im- Moyle, P. B. 1976. Inland fishes of California. University
proved design for assessing impacts of watershed prac- of California Press, Berkeley, California, USA.
tices on small streams. Internationale Vereingung fur Murphy, M. L., and J. D. Hall. 1981. Varied effects of
theoretishe und angewandte Limnolgie Verhanlungen clear-cut logging on predators and their habitat in small
20:1359 –1365. streams of the Cascade Mountains, Oregon. Canadian
Hawkins, C. P., J. L. Kershner, P. A. Bisson, M. D. Bryant, Journal of Fisheries and Aquatic Sciences 38:137–145.
L. M. Decker, S. V. Gregory, D. A. McCullough, C. K. Murphy, M. L., C. P. Hawkins, and N. H. Anderson. 1981.
Overton, G. R. Reeves, R. J. Steedman, and M. K. Effects of canopy modification and accumulated sedi-
Young. 1993. A hierarchical approach to classifying
ment on stream communities. Transactions of the Amer-
stream habitat features. Fisheries 18:3–12.
ican Fisheries Society 110(4):469 – 478.
Hawkins, C. P., M. L. Murphy, N. H. Anderson, and M.
Newcombe, C. P., and D. D. MacDonald. 1991. Effects
A. Wilzbach. 1983. Density of fish and salamanders in
of suspended sediments on aquatic ecosystems. North
relation to riparian canopy and physical habitat in
American Journal of Fisheries Management 11(1):72–
streams of the northwestern United States. Canadian
Journal of Fisheries and Aquatic Sciences 40:1173– 82.
1185. Nussbaum, R. A. 1969. The nest site of the Olympic sal-
Hayes, M. P., and M. R. Jennings. 1986. Decline of ranid amander, Rhyacotriton olympicus. Herpetologica 25:
frogs in western North America: are bullfrogs ( Rana ca- 277–278.
tesbeiana ) responsible? Journal of Herpetology 20:490 – Nussbaum, R. A., and C. A. Tait. 1977. Aspects of the
509. life history and ecology of the torrent salamander, Rhy-
Holt, R. D. 1987. Population dynamics and evolutionary acotriton olympicus (Gaige). American Midland Natu-
processes: the manifold roles of habitat selection. Evo- ralist 98:176–199.
lution and Ecology 1987:331–347. Nussbaum, R. A., E. D. Brodie, and R. M. Storm. 1983.
Jennings, M. R. 1996. Ambystoma californiense (Califor- Amphibians and reptiles of the Northwest. University of
nia tiger salamander). Burrowing ability. Herpetological Idaho Press, Moscow, Idaho, USA.
Review 27(4):194. Odum, E. P. 1985. Trends expressed in stressed ecosys-
Landres, P. B., J. Verner, and J. W. Thomas. 1988. Eco- tems. Bioscience 35:419 – 422.
logical uses of vertebrate indicator species: a critique. . 1992. Great ideas in ecology for the 1990’s. Bio-
Conservation Biology 2:316–328. science 42:542–545.
Laurance, W. F. 1996. Catastrophic declines of Australian Olson, D. H. 1992. Ecological susceptibility of amphib-
rainforest frogs: is unusual weather responsible? Bio- ians to population declines. Pages 55–62 in R. R. Harris
logical Conservation 77:203–212. and D. C. Erman, technical coordinators. Symposium on
Laurance, W. F., K. R. McDonald, and R. Speare. 1996. biodiversity of northwestern California. Report 29.
Epidemic disease and the catastrophic decline of Aus- Wildland Resources Center, University of California,
tralian rainforest frogs. Conservation Biology 10:406– Berkeley, California, USA.
413. Parker, M. S. 1991. Relationship between cover avail-
Leonard, W. P., H. A. Brown, L. L. C. Jones, K. R. Mc- ability and larval Pacific giant salamander density. Jour-
Allister, and R. M. Storm. 1993. Amphibians of Wash- nal of Herpetology 25:355–357.
ington and Oregon. Seattle Audubon Society, Seattle, Pechmann, J. H. K., D. E. Scott, R. D. Semlitsch, J. P.
Washington, USA. Caldwell, L. J. Vitt, and J. W. Gibbons. 1991. Declining
November 1998 AMPHIBIANS AS BIOINDICATORS 1131

amphibian populations: the problem of separating human and indexes. Environmental Toxicology and Chemistry
impacts from natural fluctuations. Science 253:892–895. 12:1533–1539.
Pechmann, J. H. K., and H. M. Wilbur. 1994. Putting de- Toft, C. A. 1991. Reply to Seaman and Jaeger: an appeal
clining amphibian populations in perspective: natural to common sense. In Points of view: a controversy in
fluctuations and human impacts. Herpetologica 50:65– statistical ecology. Herpetologica 46:357–361.
84. Toft, C. A., and P. J. Shea. 1983. Detecting community-
Platts, W. S., W. F. Megahan, and G. W. Minshall. 1983. wide patterns: estimating power strengthens statistical
Methods for evaluating stream, riparian, and biotic con- inference. American Naturalist 122:618–625.
ditions. U.S. Forest Service General Technical Report Valentine, B. D., and D. M. Dennis. 1964. A comparison
INT-138. of the gill-arch system, and fins of three genera of larval
Pounds, J. A., M. P. L. Fogden, J. M. Savage, and G. C. salamanders, Rhyacotriton, Gyrinophilus, and Ambys-
Gorman. 1997. Tests of null models for amphibian de- toma. Copeia 1964:196–201.
clines on a tropical mountain. Conservation Biology 11: Veirs, S. D., Jr. 1982. Coast redwood forest: stand dy-
1307–1322. namics, successional status, and the role of fire. Pages
Power, M. E., R. J. Stout, C. E. Cushing, P. P. Harper, F. 119 –141 in J. E. Means, editor. Forest succession and
R. Hauser, W. J. Matthews, P. B. Moyle, B. Statzner, and stand development research in the northwest. Forest Re-
I. R. Wais De Badgen. 1988. Biotic and abiotic com- search Laboratory, Oregon State University, Corvallis,
munities. Journal of the North American Benthological Oregon, USA.
Society 7:1–25. Vitt, L. J., J. P. Caldwell, H. M. Wilbur, and D. C. Smith.
Rapport, D. J. 1992. Evaluating ecosystem health. Journal 1990. Amphibians as harbingers of decay. Bioscience
of Aquatic Ecosystem Health 1:15–24. 40:418.
Waide, J. B. 1995. Ecosystem stability: revision of the
Rapport, D. J., and H. A. Regier. 1995. Disturbance and
resistance–resilience model. Pages 372–396 in B. C. Pat-
stress effects on ecological systems. Pages 397– 414 in
ten and S. E. Jorgensen, editors. Complex ecology, the
B. C. Patten and S. E. Jorgensen, editors. Complex ecol-
part–whole relation in ecosystems. Prentice Hall, En-
ogy, the part–whole relation in ecosystems. Prentice-
glewood Cliffs, New Jersey, USA.
Hall, Englewood Cliffs, New Jersey, USA. Wake, D. B. 1990. Declining amphibian populations. Sci-
Rapport, D. J., H. A. Regier, and T. C. Hutchson. 1985. ence 253:860.
Ecosystem behavior under stress. American Naturalist Waters, T. F. 1995. Sediment in streams: sources, biolog-
125:1248–1255. ical effects and control. American Fisheries Society
Reid, L. M. 1993. Research and cumulative watershed Monograph 7. Bethesda, Maryland, USA.
effects. U.S. Forest Service General Technical Report Welsh, H. H., Jr. 1990. Relictual amphibians and old-
PSW-141. growth forests. Conservation Biology 4:309 –319.
Reid, L. M., and T. Dunne. 1984. Sediment production . 1993. A hierarchical analysis of the niche rela-
from forest road surfaces. Water Resources Research 20: tionships of four amphibians from forested habitats of
1753–1761. Northwestern California. Dissertation. University of
Rice, R. M., F. B. Tilley, and P. A. Datzman. 1979. A California, Berkeley, California, USA.
watershed’s response to logging and roads: south fork Welsh, H. H., Jr., and A. J. Lind. 1992. Population ecology
of Caspar Stream, California, 1967–1976. U.S. Forest of two relictual salamanders from the Klamath Moun-
Service Research Paper PSW-146. tains of Northwestern California. Pages 419 – 437 in D.
SAS Institute. 1997. SAS user’s guide. SAS Institute, R. McCullough and R. H. Barrett, editors. Wildlife 2001:
Cary, North Carolina, USA. populations. Elsevier Applied Science, London, UK.
Schrader-Frechette, K. S., and E. C. McCoy. 1993. Meth- Welsh, H. H., Jr., and A. J. Lind. 1996. Habitat correlates
od in ecology: strategies for conservation. Cambridge of the southern torrent salamander ( Rhyacotriton var-
University Press, Cambridge, UK. iegatus ) (Caudata: Rhyacotritonidae), in northwestern
Southwood, T. R. E. 1977. Habitat, the templet for eco- California. Journal of Herpetology 30:385–398.
logical strategies? Journal of Animal Ecology 46:337– Welsh, H. H., Jr., L. M. Ollivier, and D. R. Hankin. 1997.
365. A habitat-based design for sampling and monitoring
. 1988. Tactics, strategies, and templets. Oikos 52: stream amphibians with an illustration from Redwood
3–18. National Park. Northwestern Naturalist 78:1–16.
Stebbins, R. C. 1953. Southern occurrence of the Olympic Weygoldt, P. 1989. Changes in the composition of moun-
salamander, Rhyacotriton olympicus. Herpetologica 11: tain stream frog communities in the Atlantic Mountains
of Brazil: frogs as indicators of environmental deterio-
238–239.
ration? Studies on Neotropical Fauna and Environment
Stebbins, R. C., and N. W. Cohen. 1995. A natural history 243:249 –255.
of amphibians. Princeton University Press, Princeton, Woolbright, L. L. 1997. Local extinctions of anuran am-
New Jersey, USA. phibians in the Luquillo Experimental Forest of north-
Steedman, R. J., and H. A. Regier. 1987. Ecosystem sci- eastern Puerto Rico. Journal of Herpetology 31:572–
ence for the Great Lakes: perspectives on degradative 576.
and rehabilitative transformations. Canadian Journal of Zar, J. H. 1995. Biostatistical analysis. Fourth edition.
Fisheries and Aquatic Sciences 44(Suppl 2):95–103. Prentice-Hall, Englewood Cliffs, New Jersey, USA.
Stevens, J. 1986. Applied multivariate statistics for the Zinke, P. J. 1977. The redwood forest and associated north
social sciences. Lawrence Erlbaum Associates, Hills- coast forests. Pages 679 –697 in M. G. Barbour and J.
dale, New Jersey, USA. Majors, editors. Terrestrial vegetation of California.
Suter, G. W. 1993. A critique of ecosystem health concepts John Wiley and Sons, New York, New York, USA.
1132 HARTWELL H. WELSH, JR. AND LISA M. OLLIVIER Ecological Applications
Vol. 8, No. 4

APPENDIX
Definitions of primary mesohabitat types, pool sediment measures, and microhabitat attributes measured or estimated in
association with belt samples.

Term Definition
a) Mesohabitat attributes
i) Primary mesohabitat types†
All pools Reaches with water depths from shallow to deep with evidence of scour. Cause of scour
may be an obstruction, blockage, merging of flows, or constriction. This type includes
main channel, lateral, backwater, and secondary channel pools. Flow velocities range
from very low to swift. Substrate size is highly variable.
Run/glide Wide shallow reaches flowing smoothly, with little surface agitation and no major flow
obstructions. Velocities are low to moderate. These often appear as flood riffles. Typical
substrates are gravel, cobble, and boulders.
Riffle Shallow to moderately deep, swift, turbulent water. Amount of exposed substrate will
vary. Substrates are usually cobble or boulder dominated.
Step run A sequence of runs separated by short riffle steps. Substrates are usually cobble and
boulder dominated.
Step pools A sequence of pools separated by short riffle steps. Substrates are usually cobble and
boulder dominated.
ii) Pool sediment measures
Pool tail embedded Visual estimate (percentage) of vertical surfaces of large substrates buried in fines and/or
sand in pool tail.
Pool bowl sediment Depth of sediment to the nearest tenth of a centimeter is taken at three points along the
depth midline of the pool bowl. These measures are then averaged.
b) Microhabitat attributes Measures and estimates of microhabitat attributes taken in association with amphibian
sampling.
i) Aquatic conditions
Proportion margin‡ Visual estimate (percentage) of channel composed of margin flow (percentage).
Proportion intermediate Visual estimate (percentage) of channel composed of intermediate flow.
Proportion thalweg Visual estimate (percentage) of channel flow composed of thalweg flow.
Flow margin Flow rate in channel margin measured with a flowmeter in centimeters per second.
Flow intermediate Flow rate in intermediate channel flow measured with a flowmeter in centimeters per
second.
Flow thalweg Flow rate in channel thalweg measured with a flowmeter in centimeters per second.
Canopy open‡ Measured by densiometer at center of the belt (percentage).
Water temperature Measured by thermometer (8C).
Density of other Density (captures per square meter) of the two other species of amphibians present in the
amphibians§ belt.
ii) Cover estimates Visual estimate of instream cover (percentage) in a series of categories.
Undercut banks‡ Overhang of stream banks, within 30 cm of water surface.
Woody debris‡ Woody debris of any size, including leaf litter overhanging water surface or underwater.
Riparian vegetation‡ Vegetation growing on the banks or in the stream. Must overhang within 30 cm of the
water surface.
Large rock‡ Comprised of boulders and bedrock ledges. Only those portions that provide an overhang
capable of hiding an amphibian are counted in this estimate.
Without cover‡ Portion of the belt lacking any of the above cover types.
iii) Coarse aquatic Visual estimate of belt surface area comprised of coarse substrates (percentage) in the
substrates\ following categories.
Gravel 2.0–32.0 mm in diameter
Pebble 32.0–64.0 mm in diameter
Cobble 64.0–256.0 mm in diameter
Large rock .256.0 mm in diameter and bedrock
Woody debris‡ Woody debris of any size and leaf litter. Must be in or surrounded by water.
Fine gravel volume Proportion of mass of sediment sample taken at each belt (2.0–16.0 mm diameter).
Coarse gravel volume Proportion of mass of sediment sample taken at each belt (16.0–32.0 mm diameter).
iv) Fine aquatic substrates\
Embedded Visual estimate (percentage) of vertical surfaces of large substrates buried in fines and/or
sand in the belt.
Fines‡ Visual estimate (percentage) of belt surface area comprised of substrates ,0.06 mm
diameter.
Sand‡ Visual estimate (percentage) of belt surface area comprised of substrates 0.06–2.0 mm
diameter.
Silt volume‡ Proportion of mass of sediment sample taken at each belt (samples are dried before sifting
and weighing; ,0.063 mm diam).
Sand volume‡ Proportion of mass of sediment sample taken at each belt (0.063–2.0 mm diameter).
Nonfilamentous algae Visual estimate (percentage) of belt substrates covered by nonfilamentous algae growth.
† Modified from Hawkins et al. (1993).
‡ Variable is transformed using arcsine to meet assumptions of normality.
§ Variable is transformed using natural log to meet assumptions of normality.
\ Particle size based on Platts et al. 1983.