Concern has been acute over world-wide declines in amphibians (Blaustein and Wake, 1990; Wake, 1991). These declines include reductions in abundance and extinction of populations, increased isolation of extant populations (e.g., Bradford et al., 1993), and restrictions in geographic ranges of many species. There is particular concern over amphibian declines in areas that have been relatively undisturbed by human activities, such as high elevation locations in the west (Blaustein and Wake, 1990) and national parks (Bradford, 1989; Bradford et al., 1993).

Of the many possible causes for amphibian declines (e.g., habitat destruction, pollution, global climate change, increased UV radiation), the impact of exotic species on native amphibians has received considerable attention (Blaustein and Wake, 1990; Wake, 1991). Bullfrogs (Rana catesbeiana) have been implicated in declines of native ranids in the west (Moyle, 1973). Hayes and Jennings (1986) suggested that predation by introduced fish was a likely cause of decline in western ranids, but they concluded that evidence was inadequate to distinguish between fish, bullfrogs, and other possible causes of decline. Bradford (1989) found that Rana muscosa did not co-occur with introduced trout in Sequoia and Kings Canyon National Parks, CA. Bradford et al. (1993) suggested that continued declines in R. muscosa in fishless waters may be a consequence of isolation of extant populations due to the presence of fish in connecting waters.

Salamanders are important aquatic predators in many fishless high mountain lakes in the western US (e.g., Dodson 1970, 1974; Dodson and Dodson, 1971; Sprules, 1972; Giguere, 1979; Taylor, 1983). Several studies have reported reduced abundance or absence of ambystomid larvae from habitats with fish. Sprules (1974a) found 16 times more paedogenic Ambystoma gracile in fishless ponds than in ponds with fish. He also reported observing paedogenic Ambystoma tigrinum only in small fishless ponds in the Rockies in central Colorado. Similarly, Taylor (1983) found much greater abundances of A. gracile in fishless lakes than in lakes with fish in the Oregon Cascades. In the east, Semlitsch (1988) found that Ambystoma talpoideum in the upper coastal plain and sandhill regions of South Carolina occurred primarily in ponds that lacked fish. He demonstrated that this was likely a consequence of fish predation on eggs. In Kentucky streams, Petranka (1983) reported that eggs and larvae of Ambystoma barbouri were largely confined to sections of streams that lacked fish. He showed that this disjunct distribution of salamanders and fish was likely a result of fish predation on larvae.

Fish have been shown to inhibit growth (Semlitsch, 1987; Figiel and Semlitsch, 1990) and reduce survival of larval ambystomids (Semlitsch, 1987; Sih et al., 1988) in artificial ponds. In the presence of fish, larvae may exhibit reduced activity and remain confined to refuges for longer periods of time than when fish are absent (Taylor, 1983; Stangel and Semlitsch, 1987; Semlitsch, 1987; Sih et al., 1988, Figiel and Semlitsch, 1990). However, factors other than the presence of fish especially physical conditions such as temperature and lake drying (Snyder, 1956; Anderson, 1967, 1968; Sprules, 1 974a,b; Sexton & Bizer, 1978; Figiel & Semlitsch, 1990), availability of food (Anderson, 1967; Dodson & Dodson, 1971), and winter kill from freezing in shallow lakes can influence salamander life histories, distribution, and abundance.

The amphibian section of this report is divided into three parts. In the first part, we report the influence of trout and other biological and physical variables on the abundance and distribution of long-toed salamander (Ambystoma macrodactylum) larvae in high mountain lakes. Nearly all of the lakes included in this study are eastslope lakes. Although A. macrodactylum is widespread throughout the northwest (Ferguson, 1961; Nussbaum et al., 1983; Leonard et al., 1993), demographic data on high elevation populations is sparse and, although distributions of A. macrodactylum and introduced trout overlap geographically, effects of trout on this salamander are unknown. In the second part, we examine patterns of distribution and abundance of fish and salamanders (A. macrodactylum and A. gracile) in westslope lakes. To date, only a limited number of westslope lakes have been sampled for salamanders. Consequently, we consider the observed patterns only as suggestive, but they serve as a basis for hypotheses that will direct field work on the westslope in 1994 and 1995. In the last section, we report observations on anuran taxa in the Complex.

METHODS

We sampled 13 lakes (six fishless lakes and seven lakes with fish) from 1990- 1992. Twelve of the lakes were east of the Cascade crest (within the Stehekin River drainage). Eastslope fishless lakes and lakes with fish were geographically interspersed. Fishless lakes ranged in elevation from 1504-1873 m and lakes with fish ranged from 1629-1988 m. Ambystoma macrodactylum was the only salamander found in eastslope lakes. Pyramid Lake, a lower elevation lake (802 m), is the only westslope lake we sampled with allopatric A. macrodactylum.

Some of the study lakes that are presently fishless lakes have a history of fish stocking. MR 11, an eastslope lake, was fishless in 1990 and probably for several previous years. It was stocked with a low density of fingerling rainbow trout in late fall of 1990. The fish had reached a total length of 250-350 mm by 1992. In our analysis, MR 11 was treated as a fishless lake in 1990 and a fish lake in 1991 and 1992. Two lakes that are presently fishless had been stocked with fish at least once in the past. Pyramid Lake was last stocked in 1968 and MR13-1 in 1984 (Jarvis, 1987).

Each lake was sampled one-three times in summer and early fall, depending on the length of the ice-free period. Larval densities were estimated by snorkeling the lake perimeter during daylight. Initially we attempted to estimate larval salamander densities with funnel traps (Carpenter, 1953), searches from the shoreline, and snorkeling. Snorkeling proved to be the most effective technique for detecting larvae. We captured very few larvae with funnel traps. Although we were able to detect larvae with searches from shoreline, when snorkeling we could more effectively search a wider area of the lake perimeter, more easily locate larvae hidden in bottom substrates, and see larvae more clearly underwater.

During sampling the snorkeler carefully searched through talus, woody debris, fine organic material, and emergent vegetation to determine number and location of larvae. The area from the lake edge to 2-3 m offshore (up to a depth of about 1 m) was searched. Behavior, particularly whether salamanders were concealed in the bottom substrates or moving freely in the open, was noted. Quantitative data on actual number of larvae hidden or in the open was not recorded. Lake area and lineal meters of lake perimeter were estimated with a digitizer from 7 1/2 min USGS topographic maps. Relative density was expressed as number of larvae observed/100 m of shoreline.

Normally the entire lake perimeter was snorkeled. If a lake was large or the water very cold it was sometimes not possible to snorkel the entire shoreline. For partial surveys the number of meters of shoreline sampled was estimated by multiplying the total length of the lake perimeter by the estimated proportion of shoreline sampled. When partial surveys were conducted at least 50% of the lake perimeter was snorkeled. Sampling was usually conducted by a single snorkeler, although in very large lakes it was sometimes necessary to employ two snorkelers.

Crustacean zooplankton, an important food resource for larval salamanders (Anderson, 1968; Dodson and Dodson, 1971), were sampled with a #25 (64 micron mesh) net with collecting basket. Counts from three vertical tows taken over the deepest point in the lake were averaged to provide an estimate of density on each sample date. Presence of fish was initially determined from park records (Jarvis, 1987) and verified by angling, gill-netting, snorkeling, and visual observation from shore and from an inflatable boat. Water temperature was measured at the lake surface and at 1 m intervals to the lake bottom with an Omega 871 thermo-couple. Lake elevation was determined from 7 1/2 min USGS topographic maps. Maximum depth was determined with a hand-held sonar gun.

The Mann-Whitney U test (Snedecor and Cochran, 1989) was used to determine if fishless lakes differed statistically (=0.05) from lakes with fish in average summer densities of salamander larvae, average summer densities of crustacean zooplankton, average maximum water temperature 1 m below the lake surface, and lake depth, surface area, and elevation. When a lake was sampled for two or more years, densities of larvae and zooplankton and maximum temperature were averaged over all years.

To investigate relationships between larval salamander density and other variables, log transformed larval density was regressed on each variable separately for fishless lakes (N=6) and for lakes with fish (N=7). Initially both a linear and quadratic term were used in each regression. If the regression coefficient of the quadratic term was not significantly different from zero (=0.05), the quadratic term was omitted and the linear term was tested for significance (=0.05).

RESULTS

Larval salamander densities in lakes with fish were significantly lower than in fishless lakes (Table 1; Mann-Whitney U test, P=0.0059). When larval densities in eastslope fishless lakes (Pyramid omitted) were compared with lakes with fish, differences were still significant (Table 1; P=0.01 04). Salamander larvae were present in several lakes with fish and were moderately abundant in MR 11 in 1991 and 1992. However, we never observed larval densities in lakes with fish that approached the highest densities (>70/100 m) found in fishless lakes. Fishless lakes did not differ significantly from lakes with fish in zooplankton density, water temperature, or lake depth, surface area and elevation, either when Pyramid Lake was included in the analysis or when it was omitted (Table 2).

Table 1. Average and range interval (in parenthesis) of larval densities of A. macrodactylum, years each lake was sampled and total number of snorkel surveys in each year (in parenthesis) in North Cascades National Park Service Complex, WA ,USA.

a: One adult found in fish stomach
b: Five eggs identified
c: Two larvae observed while sampling benthos

Table 2. Average and 95% confidence interval (in parenthesis) of physical and biological variables in fish and fishless lakes in North Cascades National Park Service Complex, WA, USA. P-values are for comparisons of fish and fishless lakes using the Mann-Whitney U test (a=0.05). P-values for comparisons among all lakes and for eastslope lakes (Pyramid Lake omitted) are given.

Larval densities were highly variable among fishless lakes and ranged from near 170/100 m in Pyramid to 2.2/100 m in MR11(1990) and Waddell. However, in fishless lakes and lakes with fish, regression analysis revealed no statistically significant relationship between larval density and zooplankton density (fishless lakes, P=0.35; fish lakes, P=0.53), water temperature (fishless lakes, P=0.12; fish lakes, P>0.99), elevation (fishless lakes, P=0.50; fish lakes, P=0.660), surface area (fishless lakes, P=0.07; fish lakes, P=0.50), or depth (fishless lakes, P=0.30; fish lakes, P=0.66). However, when Pyramid Lake was omitted (eastslope lakes only), there was a significant negative relationship between maximum lake depth and larval density (r²=0.82; N=5; P=0.03) in fishless lakes. On the eastslope, the highest densities of salamander larvae occurred in smaller, shallower lakes (MR 2, MR 3, MR13-1; maximum depth 3m and surface area <=0.3 ha). Pyramid, a westslope lake with the highest density of larvae observed in this study, was small (0.3 ha) but relatively deep (8.8 m).

In fishless lakes larvae were observed moving freely in the open across the substratum, swimming in the water column, and concealed on the lake bottom in woody debris, conifer needles, organic detritus, cobble, and talus boulders. In contrast, most larvae found in lakes with fish were well hidden in bottom materials, although a few were observed in open areas.

DISCUSSION

Trout are not indigenous to many high elevation lakes of the west. Bahls (1992) reported that nearly 60% of all high mountain lakes and 95% of larger, deeper lakes in the western US support fish. About 95% of lakes now supporting fish were estimated to have been originally fishless.

Taylor (1983) considered A. gracile to be the top carnivore in many fishless Cascade mountain lakes. In North Cascades National Park Service Complex, A. macrodactylum is a major vertebrate predator, particularly in eastslope lakes. The species is widespread in the Pacific Northwest and occurs from sea level to high elevations (Ferguson, 1961; Nussbaum et al., 1983; Leonard et al., 1993).

Salamander distribution and abundance in high mountain lakes is determined by a complex of biological and physical factors. Population variation and periodic population extinction in amphibians may be related to natural fluctuations in local environments (Corn and Fogelman, 1984; Pechmann et al., 1991; Wissinger and Whiteman, 1992), often making it difficult to distinguish anthropogenic from natural causes of local declines (Pechmann et al., 1991; Wake, 1991). In North Cascades National Park Service Complex, we found significantly higher abundances of A. macrodactylum larvae in fishless lakes than in lakes with fish. Furthermore, we found no significant differences between fishless lakes and lakes with fish in zooplankton density, temperature, and lake elevation, surface area, and depth. Our results are consistent with several other studies that have shown that fish can reduce survival and abundance of larval salamanders (Sprules, 1974a; Taylor, 1983; Petranka, 1983; Semlitsch, 1987, 1988; Sih et al., 1988, 1992).

Variability in larval abundance among fishless lakes in our study indicates that factors other than introduced trout may also affect larval abundance. In particular, on the eastslope, A. macrodactylum tended to reach greatest abundance in smaller, shallower lakes. Some of these lakes, although permanent, may be too shallow to support trout and, thus, they may provide productive refuges for A. macrodactylum. Furthermore, we have observed A. macrodactylum larvae in shallow pools in lake inlet and outlet streams that may not be accessible to trout, and in small temporary ponds and seeps near lakes with fish. Apparently A. macrodactylum is able to persist in the same geographic area as introduced trout by utilizing habitats that may be relatively unfavorable or inaccessible to fish.

Ambystoma macrodactylum also occur in larger, deeper lakes (Kezer and Farner, 1955; Anderson, 1967). Of all lakes sampled in the Complex through 1992, the highest densities of A. macrodactylum larvae occurred in Pyramid Lake, a relatively small (0.3 ha), deep (8.8 m), low elevation (802 m) lake. However, larval densities were relatively low in two larger, deeper lakes, Waddell (surface area = 4.1 ha, maximum depth = 11.9 m) and MR11 (surface area 1.3 ha, maximum depth 8.8 in). Many deeper, permanent lakes can support fish and in these lakes A. macrodactylum larvae may be vulnerable to fish predation.

In both field and laboratory studies, fish have been shown to alter salamander behavior (Stangel and Semlitsch, 1987; Semlitsch, 1987; Figiel and Semlitsch, 1990). Ambystoma gracile appear to be more reclusive in lakes with fish, staying hidden at least during daylight hours and possibly foraging for food mainly at night (Efford and Mathias, 1969; Sprules, 1974a; Taylor, 1983). This is generally consistent with our observations of A. macrodactylum behavior in lakes with fish, although larvae in fishless lakes were often hidden in talus, woody debris, and line organic detritus accumulated on the lake bottom. Avoidance behavior by larvae in fishless lakes may have been related, in part, to intraspecific predation because larger A. macrodactylum larvae may consume smaller larvae (Leonard et al., 1993; personal observation).

Sih et al. (1988) discuss the importance of the dynamics of refuge use in determining prey survival. They emphasize that, even though few prey may be observed outside of refuges, prey populations may be rapidly reduced by predators if emergence rates of prey from refuge are high and the prey are quickly killed after emerging. Moreover, shifts in behavior and habitat use to avoid predation, while increasing probability of larval survival (Figiel and Semlitsch, 1990), may reduce food consumption and growth by decreasing foraging efficiency (Semlitsch, 1987; Figiel and Semlitsch, 1990). Eventually larvae may need to leave refuge to forage and so become vulnerable to predation. Thus refuge use within a lake may not necessarily ensure long-term survival.

Little is also known of the ability of A. macrodactylum to recolonize high mountain lakes where they have been eliminated. Howard and Wallace (1981) suggested that dispersal may be overland as well as along waterways. Salamanders are abundant in two lakes (Pyramid, MR 13-1) that had been stocked with fish in the past, but are presently fishless. We do not know if stocked trout fry survived long enough in either lake to have had an effect on larval abundance. If the life expectancy of adult A. macrodactylum is long, as documented for other ambystomatids (Bowler, 1977), the adult population may be buffered against short-term catastrophic losses of larvae. Many years may be required for adult populations to be locally extirpated. Furthermore, the ability of A. macrodactylum to utilize habitats near trout lakes that may be unsuitable for fish may favor recolonization and salamander population recovery if fish were removed from the lake.

In 1991 and 1992, larval salamander densities in MR 11 were relatively high compared to other lakes with fish. The reasons for these comparatively high densities are unknown. The lake apparently has had a rather sporadic history of fish stocking. Trout fry were stocked at low density in fall 1990. The fish were relatively small during summer 1991, but had grown considerably (250-350 mm total length) by summer 1992, probably from feeding on abundant amphipods, and were certainly large enough to consume salamanders. Fish density in 1992 is not known (we were unable to estimate density by mark-recapture due to difficulty in capturing fish by angling or gill-netting). Perhaps, by 1992, fish density had declined so much that there was little impact on salamander larvae or, possibly, more adequate refugia were present in MR 11 than in other lakes with fish.

Diving techniques have been used previously to census ambystomids in lakes (Neish, 1971; Taylor, 1983). Of the census techniques we tried (including funnel traps and searches from shoreline), snorkeling was the most effective in detecting larvae. In both fish and fishless lakes, by careful searching we were routinely able to detect larvae hidden in leaf litter and other organic debris and in talus along the shoreline. However, larvae hidden deeply in heavy cover such as dense submerged log jams are very difficult to detect. It is possible that larvae may be more deeply hidden in lakes with fish and thus, with snorkeling and other techniques that rely on visual detection, larval densities may be underestimated more in lakes with fish than in fishless lakes. In lakes with reproducing trout we found no significant differences in larval densities between surveys conducted during the day and at night (unpublished data).

Failure to detect any larvae in lakes with fish by snorkeling may not necessarily imply that salamanders were extirpated from these lakes. In some lakes where no larvae were observed, even after several years of sampling, there was occasionally evidence of salamander use (an egg mass was observed, an adult was found in a fish stomach; Table 1). However, it is unlikely that failure to observe larvae due behavioral differences was sufficient to account for the orders-of-magnitude differences in densities between lakes with fish and fishless lakes with the highest larval densities.

METHODS

Salamanders in westslope lakes were sampled following the procedure described in the previous section. In lakes with A. gracile, both embryo masses and larvae were counted while snorkeling. No attempt was made to distinguish mature from immature forms of neotenic A. gracile. Both mature and immature forms are grouped as larvae in this report. In LS-1 in 1993, larvae were counted during the day by snorkeling and at night from the shoreline using a flashlight. In a few lakes, species identifications were made, but larval density was not assessed quantitatively.

Some westslope lakes (LS-1, LS-2, Upper Panther, Lower Panther) were sampled annually from 1990 through 1993 (Table 3). Quantitative sampling was conducted in Pyramid Lake in 1990,1991, and 1993, but not in 1992. All fish were removed from Upper Panther, Lower Panther, and LS-1 in late summer in 1990. Lower Panther and LS-1 were restocked with fry (Table 4). Upper Panther was not restocked with fish.

Table 3. Physical characteristics of westslope lakes and number of times each lake was sampled annually, North Cascades National Park Service Complex, WA, U.S.A.

Table 4. Density estimates (see Fish section) and average total length (TL) of trout in four westslope lakes, North Cascades National Park Service Complex, WA, U.S.A.

a: estimate using fish removal technique (Ricker, 1975)
b: stocking density of fry
c: estimate using mark-recapture, fish > 170 mm
*: three fish observed during September snorkel for salamanders

Distinguishing species of small larval salamanders in the field was difficult. When possible, larvae were taken from the field and reared to metamorphosis under controlled conditions to distinguish A. macrodactylum from A. gracile.

Larval size and characteristics of embryos were also used to distinguish species. Ambystoma macrodactylum are not known to be paedomorphic and have been reported to metamorphose when their snout-vent length (SVL) reaches 30 - 46 mm (Nussbaum et al., 1983). Ambystoma gracile are paedomorphic at high elevations (Snyder, 1956; Sprules, 1974a; Leonard et al., 1993). Neotenes have been reported to reach sexual maturity between 50 and 60 mm SVL (Efford and Mathias, 1969; Neish, 1971; Sprules, 1974a). We commonly found A. gracile larvae 50-80mm SVL. Thus, the presence of A. gracile was indicated by occurrence of large larvae ( 50 mm SVL) having the external morphological characteristics of this species (Nussbaum et al., 1983).

Embryos of A. gracile have a thick outer jelly coat and are often deposited in grapefruit-sized clusters. A symbiotic algae frequently imparts a green hue to the clusters (Nussbaum et al., 1983). Ambystoma gracile embryo masses are often attached to woody debris or vegetation and are highly visible when snorkeling. Ambystoma macrodactylum eggs are laid singly or in smaller, loose clusters (Stebbins, 1985). The embryos are attached to various substrates and do not have a green hue (Slater, 1936; Nussbaum et al., 1983; Leonard et al. 1993). We had difficulty detecting A. macrodactylum embryos.

RESULTS

While A. macrodactylum occurred in both east- and westslope lakes, A. gracile and Taricha granulosa were found only in westslope lakes (Table 5). Taricha granulosa larvae were found in 1991 in Pyramid Lake (802 in). These distributions are consistent with other reports of distributions of these species (Dunn, 1944; Nussbaum et al., 1983; Stebbins, 1985; Leonard et al., 1993).

Table 5. Occurrence of Ambystoma and Taricha in westslope lakes, North Cascades National Park Service Complex, WA, U.S.A.

S=indicates snorkel survey; V=indicates visual observation from shore

a: very low density (0.79 individuals/100 m shoreline)
b: last stocked in 1988, fish removed in fall 1990

Ambystoma gracile was found more commonly in westslope lakes with fish than A. macrodactylum. Three of the five westslope lakes with A. macrodactylum were fishless (Table 5). Ambystoma macrodactylum was present in very low densities in LS2 (Table 6), a lake with high fish density (Table 4). All six lakes with A. gracile support trout or have a recent history of stocking.

In westslope lakes sampled between 1990 and 1993, A. gracile and A. macrodactylum co-occurred only in LS-1 (Table 5). In LS-1, the two species appear to be spatially segregated. Ambystoma gracile larvae were found only in the main body of the lake. A. macrodactylum larvae also occurred in the lake (confirmed by rearing larvae), but their relative densities were unknown since it is difficult to distinguish small A. gracile and A. macrodactylum larvae while snorkeling. However, densities of A. macrodactylum larvae in the main body of the lake may be low. In 1993, a few small larvae, which could be A. macrodactylum, were observed during daylight snorkeling, while during the 1993 night survey we observed only large larvae which were likely A. gracile. However, in 1993 in LS-1, A. macrodactylum larvae were abundant (180 larvae/100 m) in shallow pools of slow-moving water in the lake outlet. No A. gracile or fish were observed in these pools during the day or at night.

Ambystoma gracile may be able to maintain higher larval densities in lakes with fish than A. macrodactylum. Larval A. macrodactylum densities in eastslope lakes with fish were very low, averaging between 0.0 and 0.4 larvae/100 m shoreline with the exception of MR11, where the average larval density was 8.4 larvae/i 00 m shoreline (see subsection I of Amphibian section). The density of larval A. macrodactylum could not be determined for LS-1 but, in LS-2, a westslope lake with high fish density, larval density was low (Table 6). In contrast, in fishless lakes, A. macrodactylum larvae can reach high densities. Pyramid Lake, a westslope fishless lake, had the highest densities of A. macrodactylum larvae of any surveyed lake (average of 113.6 larvae/100 m shoreline). On the eastslope, average larval densities in fishless lakes ranged from 2.2 to 73 larvae/100 m shoreline (see subsection I of Amphibian section).

Quantitative samples of larval A. gracile density are available for only three lakes, all of which supported fish (Table 6). Ambystoma gracile density in Upper Panther in 1990, prior to fish removal, was 23 larvae/100 m shoreline. In Lower Panther, densities averaged nearly 18 larvae/100 m shoreline over the four years in which the lake was studied. In LS-1, a night survey revealed nearly 21 larvae/100 m shoreline. These densities are much higher than the highest densities of A. macrodactylum larvae in lakes with fish. Moreover, numerous embryo masses (average 38.6 embryos/mass, range 17-56 embryos/mass, N=38) were observed in Lower Panther and in Upper Panther prior to fish removal, suggesting substantial reproductive potential.

Table 6. Annual average densities of salamander life stages determined by snorkel surveys in five westslope lakes, North Cascades National Park Service Complex, WA, U.S.A.

Although A. gracile may be able to maintain higher larval densities in the presence of fish than A. macrodactylum. A. gracile larvae may still be affected by trout. In Upper Panther, the density of embryo masses and larvae increased considerably following removal of fish in 1990 (Table 6). In 1993, we observed nearly twice as many embryo masses and over five times as many larvae in Upper Panther than in 1990 and many larvae were in the open during daylight. Although we do not know what proportion of the increase in observed larvae from 1990 to 1993 resulted from actual density increases rather than behavior changes, the increase in density of embryo masses suggests an increase in number of reproductive individuals. In LS-2, which has a relatively high density of reproducing trout, no A. gracile larvae or egg masses were observed in four years of sampling even though this species is present in LS-1, which is about 100 m away. The outlet streams of the two lakes discharge into a common tributary.

There was considerable interannual variation in densities of embryo masses and larvae in Lower Panther (Table 6). This variation may be related to changes in fish size and density of the cohort of trout stocked in fall 1990. In 1990, prior to fish removal, large fish were present at low density in Lower Panther (Table 3). Larval A. gracile density also was relatively low. Following removal of large trout and stocking of fry in fall 1990, larval density appeared to increase in 1991. Larval density declined from 1991 to 1992. This decline may be related to increased size of fish (Table 3) and, thus, intensity of fish predation. By 1992 fish from the 1990 cohort had reached a relatively large size (263 mm) and fish density was high (320 fish/ha). Larger fish may more effectively prey on a greater size range of larvae than smaller fish. Larval A. gracile density increased from 1992 to 1993. In 1993 we observed only three fish in Lower Panther. Density was undoubtedly much lower in 1993 than in 1992, due both to removal of nearly half the estimated number of fish by the mark-recapture procedure in September 1992 and to natural and angling mortality. In LS-1, a similar relationship between changes in larval density and fish density and size was not evident.

DISCUSSION

The number of westslope lakes that were sampled for salamanders was limited. Thus, we emphasize that conclusions concerning interactions of trout and salamanders are tentative and are intended to serve as the basis for generating hypotheses that will direct more intensive and detailed research on interactions of fish and salamanders in westslope lakes in 1994 and 1995.

Density and behavior of larval A. gracile may be affected by fish. Other studies have reported reduced densities of A. gracile in the presence of fish (Sprules, 1974a; Taylor, 1983). However, A. gracile may be less vulnerable to fish predation than A. macrodactylum. In westslope lakes, A. gracile were found more frequently and reached higher densities in lakes with fish than A. macrodactylum. We know of no studies that have addressed the relative vulnerability of these two species to fish predation. Differences in vulnerability to fish predation of A. gracile and A. macrodactylum may relate to differences in patterns of refuge use, body size, and defense mechanisms.

Sih (1987) categorized anti-predator mechanisms as responses to avoid encounters with predators and responses to ensure survival after encounters. Refuge use is a response to avoid encounters with predators. Sih et al. (1988) argue that patterns of refuge use and survival of prey can be understood mechanistically in terms of attack and capture rates by predators, rates of emergence of prey from refuge, and rates of re-entry of prey into refuge (a function of attack and capture rates). For example, increased short-term survival could be a consequence of reduced emergence rates (i.e., more time spent in refuge) and higher re-entry rates (i.e., less time spent outside refuge) and result in reduction in proportion of prey observed outside the refuge. However, emergence and re-entry rates may be influenced not only by predation risk but also by feeding and reproductive requirements. If food resources are scarce inside the refuge, larvae may eventually have to leave the refuge to feed or to deposit eggs and so become susceptible to predation (Sih et al., 1988).

Ambystoma gracile and A. macrodactylum may differ in dynamics of refuge use. Perhaps A. gracile are more reclusive in the presence of fish than A. macrodactylum. Patterns of refuge use by A. gracile may differ temporally. Taylor (1983) showed that neotenic A. gracile were more active at night than during the day in lakes with fish. Both Efford and Mathias (1969) and Sprules (1974a) also suggest that A. gracile tend to be more nocturnal in the presence of fish. In NOCA lakes with fish, larval A. gracile were found almost exclusively in refuges (e.g., woody debris, talus, and undercut banks) during daylight. However, in lakes where fish density was very low (e.g., Nert Lake) or absent (e.g., Upper Panther in 1993), many larvae were observed in the open during daylight. The night survey at LS-1 suggests that A. gracile tend to be more active at night (or at least more visible to the human observer) in lakes with fish. The night survey estimate of A. gracile density was more than ten times greater than the estimate derived from daylight snorkeling. In contrast, in MR-i 1, an eastslope lake with fish, a daylight snorkeling survey revealed over 2 1/2 times more larval A. macrodactylum than a night survey, perhaps indicating that diel patterns of refuge use are not as prevalent in A. macrodactylum as in A. gracile. Increased nocturnal activity may reduce attack and capture rates, allowing emergence rates to increase and re entry rates to decrease (sensu Sih et al., 1988), thus increasing time spent foraging for food.

Vulnerability to predation may be influenced by larval body size. Semlitsch (1987) found that bluegill (Lepomis macrochirus) predation was directed at the smallest individuals in an experimental salamander population. He also suggested that predation by a "gape-limited" predator was probably rare after larvae attained a large body size. Larval A. gracile, particularly paedomorphs, may reach larger sizes than A. macrodactylum larvae (Efford and Mathias, 1969; Neish, 1971; Stebbins, 1985) and so may be less susceptible to fish predation (i.e., attack, capture, and re-entry rates may be reduced and emergence rates inc eased). However, predation on larvae may also depend upon fish size. Efford and Mathias (1969) suggested that rainbow trout (Oncorhynchus mykiss) reached about 250 mm fork length before they were able to effectively prey upon neotenic A. gracile in Marion Lake, British Columbia. Thus, there may be a positive relationship between maximum size of larvae consumed by fish and fish body size. Populations of fish with larger individuals may be able to prey effectively on a greater size range of larvae. Consequently, attack and capture rates may vary with body size of both fish and larvae.

Mechanisms that help prey survive encounters with predators can be considered anti-predator responses (Sih et al., 1987). Development of noxious skin secretions has been postulated as an anti-predator mechanism (Voris and Bacon, 1966; Brodie and Gibson, 1969; Heyer et al., 1975; Formanowicz and Brodie, 1982; Sih, 1987). Brodie and Gibson (1969) suggested that A. gracile may have the most highly developed noxious secretions among Ambystomids and that the glands responsible for the secretions are present in larvae. Leonard et al. (1993) suggest that both the larvae and adults of A. gracile are toxic and distasteful and this may allow A. gracile to persist in lakes with fish. Noxious secretions as a defense against fish predation have not been reported for A. macrodactylum. As with increased body size, noxious secretions may reduce attack, capture, and re-entry rates, and increase emergence rates.

Ambystoma gracile and A. macrodactylum may have different strategies for persisting in the same geographic area as introduced trout. Ambystoma macrodactylum may be highly vulnerable to fish predation in deeper, permanent lakes with fish. However, it appears to be able to utilize habitats that are either unsuitable or inaccessible to fish and can reach high densities in some of these habitats (see subsection I of Amphibian section). Ambystoma macrodactylum are known to inhabit temporary ponds as well as deeper, permanent lakes (Kezer and Farner, 1955). We have observed larval A. macrodactylum in permanent ponds and lakes, in temporary ponds, in pools and shallow areas in inlet and outlet streams, and in small seeps adjacent to lakes with fish. The use of temporary ponds by amphibians has been suggested as a response to avoid predation (Heyer et al., 1975; Heyer, 1976).

Ambystoma gracile, on the other hand, apparently requires deeper, permanent lakes to complete its long larval period (Sprules, 1974a). These are the same kinds of lakes that typically support trout. Ambystoma gracile is probably able to persist in the same lakes with fish through behavioral and morphological adaptations (e.g., refuge use, large body size, noxious skin secretions) that enable it to reduce impacts of predation.

A. macrodactylum and A. gracile rarely appeared to co-occur in westslope lakes. The caveat is that the number of lakes sampled was relatively small and, additionally, it is difficult to distinguish smaller larvae of the two species. LS-1 is the only lake in which both species were definitively identified. Both A. gracile and A. macrodactylum occurred in the main body of the lake but, in 1993, A. macrodactylum larvae were found primarily in small, shallow pools in the outlet.

Apparent lack of co-occurrence in most westslope lakes and habitat partitioning in LS-1 may result from predation and/or competition. Cortwright and Nelson (1990) found significant predation on Ambystoma jeffersonium larvae by Ambystoma opacum larvae, but no significant competition between the two Ambystomids. They suggest that salamander predation largely determines the composition of amphibian communities.

Competition between fish and larval salamanders and between species of salamanders has been postulated as a mechanism affecting co-occurrence. Morin (1981) suggested that larger larvae in an experimental salamander community competed with Lepomis macrochirus for available food resources. Efford and Tsumura (1973) found an overlap in diets for O. mykiss and A. gracile in Marion Lake, British Columbia. They suggest that the species may compete for food. Anderson and Graham (1967) postulated that competition may occur between two larval Ambystomids for food and habitat resources. Taylor (1984) suggested competition for food resources between larval A. gracile and T. granulosa in lakes in the Oregon Cascades.

Distribution and abundance of A. gracile and A. macrodactylum in westslope lakes is undoubtedly influenced by abiotic factors that may include elevation (Snyder, 1956; Stebbins, 1985; Leonard et al., 1993), lake area and depth (Kezer and Farner, 1955; Sprules, 1974a,b), lake temperature (Snyder, 1956; Anderson, 1968), and conditions in the terrestrial habitat (Sprules, 1974a,b). We did not attempt analysis of the influence of abiotic factors since the number of lakes with quantitative information on larval densities was small.

Five anuran taxa were found in NOCA (Table 7). The Western or Boreal Toad, Bufo boreas, was present in or near five lakes (Battalion, Lower Thornton, PM 5-3, Trapper, Willow). The lakes varied in size (<2.5 ha. to 59.0 ha.), depth (<3.7m to 49.0m), and elevation (870m to 1629m). Adults were typically seen in talus slopes and near ponds associated with the larger lakes. Tadpoles were observed at Trapper Lake. Adult Pacific Treefrogs, Hyla regilla, were found in a variety of habitats near three low elevation (807m to 958m) lakes (Pyramid, Willow, Ridley). Tadpoles were captured from a pond above Pyramid Lake. The Spotted Frog, Rana pretiosa, was identified at one low elevation (Coon, 662m) and three high elevation sites (McAlester Creek, 1674m; McAlester Lake, 1679m; McAlester Pass, 1846m) east of the hydrologic crest. At two locations (Coon and McAlester Lakes) adults and tadpoles were found primarily in wet-meadow habitat through which the outlet stream flowed, and around the lake perimeter. They were also found in a ephemeral stream channels associated with a meadow spring seep (McAlester Pass) and a marshy area created by the outflow from a nearby lake (East Fork of McAlester Creek from Dagger Lake). Tadpoles of the Tailed Frog, Ascaphus truei, were found at two stream sites; one within the Complex (East Fork of Pyramid Creek), and another one mile from the Complex boundary (Bridge Creek). A. truei adults and tadpoles were also observed near or in the outlet streams of Nert and Waddell Lakes.

Table 7. Anuran taxa found in NOCA Lakes (A=adult, L=larva, T=tadpole)