DISCUSSION

The abundance of larval A. macrodactylum in NOCA lakes was related to both lake productivity, as indicated by TKN, and to presence of trout. According to McQueen et al. (1986), the potential productivity at all trophic levels in freshwater lakes is set by nutrient supply. High-elevation NOCA lakes are oligotrophic, or low productivity (Lomnicky 1996). Nevertheless, larval density in fishless lakes was positively related to the concentration of TKN. In turn, cell density of phytoplankton, which are fed upon by herbivorous zooplankters, was positively correlated with TKN in NOCA lakes (G. Larson unpublished). Total Kjeldahl-N was also positively correlated with total phosphorus concentration, another indicator of lake trophic state (Wetzel 1983).

Total Kjeldahl-N is a measure of ammonia plus all organically derived nitrogen (Lambou et al. 1983). Organic nitrogen is not readily utilized by algae and bacteria; thus TKN does not play an active role in the energetics of lakes (Goldman and Home 1983). However, TKN, when correlated with phosphorus concentration has been identified as a good predictor of lake productivity, as measured by chlorophyll density (Lambou et al. 1983) or by total plankton biomass (Paloheimo and Fulthorpe 1987). TKN was also positively correlated with water temperature. Higher water temperature may favor larval salamander growth and survival (Snyder 1956; Anderson 1968).

One link between TKN and larval salamander density appears to be through the pelagic food web. The density of crustacean zooplankton, an important food resource for larval A. macrodactylum, increased with increased TKN. Moreover, at high TKN, herbivorous cladocerans, which were identified in more larval stomachs than any other zooplankter, composed a greater proportion of crustacean zooplankton than did copepods. Other researchers also have demonstrated an association between the composition of crustacean zooplankton and nitrogen concentration in high-elevation lakes (Byron et al. 1984; Stoddard 1987; Liss et al. 1995).

Ambystomatid larvae utilize both crustacean zooplankton (Licht 1975; Brophy 1980; Branch and Altig 1981; Freda 1983; Taylor et al. 1988) and benthic macroinvertebrates (Henderson 1973; Licht 1975; Brophy 1980) as food resources. In particular, ambystomatids Dodson and Dodson 1971) and both zooplankton and benthic macroinvertebrates in later stages of development (Anderson 1968; Dodson 1970; Dodson and Dodson 1971; Freda 1983; Taylor et al. 1988; McWilliams and Bachmann 1989).

Although benthic macroinvertebrates, especially chironomid larvae, were also a large component of larval salamander diets in NOCA lakes, no significant relationship between TKN and either total benthic macroinvertebrate density or chironomid density was detected in NOCA lakes.

The effects of trout on larval salamander density was related to TKN concentration and the reproductive status of trout populations. Larval salamander densities were low (2.76 larvae/100 m) in lakes with TKN < 0.045 mg/L regardless of whether trout were present or absent. At TKN < 0.045, no statistically significant differences in mean larval densities were detected between fishless lakes and lakes with fish. In contrast, all fishless NOCA lakes with TKN 0.045 mg/L had significantly higher measured mean larval densities than did lakes with reproducing trout. Trout usually reach high densities ad trout populations have diverse age and size structures in NOCA lakes where fish reproduction occurs (Liss et al. 1995; Gresswell et al. 1997). These results suggest that fish predation is responsible for reducing larval A. macrodactylum abundance in lakes with reproducing trout. This conclusion is supported by laboratory and field studies that have demonstrated that fish can reduce the abundance of ambystomatid salamanders or eliminate them from aquatic systems when fish invaded bodies of water where salamanders were present (Sprules 1974a; Thompson et al. 1980; Petranka 1983; Taylor 1984; Semlitsch 1987, 1988; Sih et al. 1988, 1992; Dobler 1994).

Blaustein et al. (1994b) suggested that the fungus Saprolegnia ferax caused declines of Rana cascadae in the Oregon Cascades ad that S. ferax could be spread by introduced fish. Although we observed no A. macrodactylum eggs or larvae that appeared to be infected with S. ferax, the fungus cannot be eliminated as a possible factor contributing to reductions in larval salamanders in NOCA lakes.

Although all 10 fishless lakes with TKN 0.045 mg/L had significantly higher measured mean larval densities than did lakes with reproducing trout, only 4 of these 10 fishless lakes had significantly higher larval densities than lakes with non-reproducing trout. Lakes with non-reproducing trout were periodically stocked with low densities of fry (Liss et al. 1995). Thus, fish densities in lakes with non-reproducing trout are likely lower, ad fish population age and size structure less complex, than in lakes with reproducing trout. Moreover, significant differences in mean larval densities between fishless lakes and lakes with non-reproducing trout were detected only in lakes with high TKN concentrations (TKN 0.095 mg/L), where the highest mean larval densities in fishless lakes were predicted. Comparison of larval densities between fishless lakes and lakes with non-reproducing trout was hampered by small sample size (n =7) of lakes with non-reproducing trout. Further research on the effects of non-reproducing trout on larval salamanders in high-elevation lakes is needed.

Metapopulation processes may be important in regional persistence of amphibian populations (Sjogren 1991; Bradford et al. 1993). Metapopulations are spatially structured systems of local populations connected by dispersal (Haski and Gilpin 1991). Gill (1978) suggested that red-spotted newt (Notophthalmus viridescens) populations in the northeastern U.S. resemble a core-satellite metapopulation. In metapopulations with core-satellite structures, there is considerable variation in population abundance among local populations (Harrison 1991; 1994). Core populations are large populations that occupy high quality habitat and have relatively low probabilities of extinction. Satellite populations are smaller populations that are more susceptible to extinction than are core populations (Harrison 1991; 1994). Core populations can provide stable sources of dispersing individuals that recolonize satellite habitats where local extinction has occurred (Sjogren 1991; Harrison 1994). Although very little is known about metapopulation processes of high-elevation A. macrodactylum, it is possible that the relatively large local populations of A. macrodactylum that occur toward the upper end of the TKN gradient in NOCA lakes may function as core populations that provide stable sources of dispersing individuals capable of recolonizing habitats where smaller populations have gone extinct (Sjogren 1991; Harrison 1994). Therefore, we speculate that introduced trout could indirectly influence regional population distribution of A. macrodactylum by inhibiting recolonization of habitats where extinction has occurred either through reduction or elimination of critical core populations or by impeding dispersal between habitats (Sjogren 1991; Bradford et al. 1993).

The proportion of larvae hidden in benthic substrates tended to increase with both larval total length and fish presence, although the increases were not statistically significant. In both fishless lakes and lakes with fish, a greater proportion of larger salamander larvae were hidden than were smaller larvae. Anderson (1967) reported that metamorphosing larval A. macrodactylum became more secretive and congregated in nearshore areas of mountain lakes in California. The tendency for a greater proportion of larger larvae to be hidden may be related to approaching metamorphosis as amphibians may be particularly susceptible to predation at this critical stage of development (Wassersug and Sperry 1977; Arnold and Wassersug 1978).

Although not a statistically significant trend, it appeared that a greater proportion of larvae were hidden in lakes with fish than in fishless lakes. Ambystoma gracile appear to be more reclusive in lakes with fish, staying hidden during daylight hours and possibly restricting foraging to night (Efford and Mathias 1969; Sprules 1974a; Taylor 1983). Shifts in behavior and habitat use to avoid intra- and interspecific 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 (Sih et al. 1988). Thus, refuge use within a lake may not necessarily ensure long-term survival.

Our results imply that assessment of fish impacts on amphibians requires an understanding of natural abiotic ad biotic factors and processes influencing amphibian distribution and abundance (Pechmann et al. 1991; Blaustein et al. 1994b). In NOCA, detection of significant differences in larval salamander densities between fishless lakes and lakes with fish was related to TKN concentration which apparently influenced larval food resource availability and, thus, larval density. Significant differences were detectable only in lakes with high TKN where predicted larval densities were high. Furthermore, detection of differences in larval densities between fishless lakes and lakes with trout was related to the reproductive status of trout populations, which likely was indicative of trout density and age and size structures of trout populations.

ACKNOWLEDGEMENTS

We are grateful to the former and present members of our scientific advisory panel: S. Loeb, S. Dodson, R. Hughes, W. Neill, W. J. O'Brien, J. Petranka, W. Platts, ad H. B. Shaffer. J. Beatty and two reviewers provided helpful comments that improved the manuscript. This research would not have been possible without the co-operation and logistical support provided by the personnel of North Cascades National Park Service Complex. This work was funded by the National Park Service and the USGS-Forest and Rangeland Ecosystem Science Center.