EXECUTIVE SUMMARY

In small lotic systems, lungless salamanders of the Family Plethodontidae can exist in surprisingly high numbers, serving the dual function of invertebrate predator and vertebrate prey.  Unlike vernal pool breeding amphibians, populations of most stream-dwelling salamanders also tend to be remarkably stable over time. In Pennsylvania, up to seven species, representing two subfamilies and a diverse range of life histories that include both aquatic and terrestrial egg-laying habits, can compose a stream bank assemblage.  Interspecific differences in tolerance to adverse water conditions among stream salamanders may also exist, regardless of life history, as demonstrated in vernal-pool breeding species.  In either situation, acidified or otherwise impaired stream conditions are likely to be important elements in shaping stream bank salamander assemblages either through direct toxicity or indirect ecological interactions among community members. Ecological indicators, if properly selected and evaluated, can provide a reliable expression of environmental stress or change that can help scientists, managers, and policy makers document trends, prioritize issues, and target restoration activities. In general, amphibians, are considered to be valuable response indicators. By virtue of their diverse and complex life histories, and abundant, stable, and geographically widespread populations, stream salamanders offer the opportunity of providing another biological tool to assess headwaters. Forested headwater streams comprise 60-75% of the total stream length and watershed area in the Mid-Atlantic states.  A variety of environmental stressors can impact these small watersheds.

The objectives of this research were to: 1) determine if stream Plethodontid assemblages respond to acidified or otherwise degraded streams and to 2) describe or characterize the effects of the response(s). The composition and abundance of free-ranging, naturally-occurring stream Plethodontid assemblages in 14 degraded and non-degraded watersheds in the Pennsylvania Central Appalachians were studied by intensive sampling. Ecologically dissimilar conditions or varying gradients and types of anthropogenic disturbances were represented by four stream condition categories associated with episodic acidification (n = 3), acid mine drainage (AMD; n = 3) , and moderate agricultural and rural development (n = 4). Minimally disturbed, relatively pristine, forested watersheds served as reference sites or Acontrols@ (n = 4). Streams were carefully chosen and assigned to a category based on past watershed and stream studies.

Streams were sampled with 4-8 transects (sections), approximately 100-200 m long. Upper, middle, and lower sections of each stream were surveyed when possible. Transects were sampled by thoroughly searching 5 rectangular plots measuring 4 m². Plots were flagged to include both terrestrial and aquatic stream bank habitat. Seeps were also sampled in 6 watersheds. Searches were timed to the nearest minute. Abundance, biomass, species composition, lifestage, and physical abnormalities served as response variables. Relationships between stream Plethodotid assemblage attributes and stream habitat variables (e.g., water quality, stream physical habitat, watershed land-cover) were examined by a variety of methods including canonical correspondence analysis (CCA). Sample metrics are presented as well.                   

To a lesser or greater extent, all reference, episodically acidified, and AMD watersheds were underlain by acidic sandstones and shales. Limestone bearing rock covered greater amounts of reference streams. In the valley, fragmented streams flowed primarily through freshwater and marine limestones. Stream chemistry largely reflected these geological settings. The four categories of stream condition represented a fairly wide water quality gradient, particularly with respect to pH and alkalinity. Forest cover ranged from 44 - 64% in fragmented watersheds; all others were almost entirely forested (84 - 99%).

A total of 340 plots representing 68 stream channel and bank, and seep transects were sampled during a 630 person-hrs effort in the spring and summer of 1997 - 1998. Plot sampling effort differed significantly across streams and stream condition categories (0 = 1.8 person hrs). Searches in acidified streams were on average shorter than in reference or fragmented streams; the latter were the longest.

A total of 4,065 salamanders were collected. Total abundance ranged from 20 - 700 individuals/stream and 0 - 117/plot. Approximately 85% of all salamanders were captured in non-acidified streams. Median plot densities were lowest in AMD-impacted and episodically acidified streams, followed by reference and fragmented streams (0.4, 0.5, 3, 4.7 individuals/m², respectively). Although sampled differently, salamander densities were generally higher in seep habitats than elsewhere in the same watershed, particularly in degraded watersheds. Variance component estimates by nested ANOVA indicated that the effect of stream condition was significant and accounted for about 58% of the variability of two models. Stream condition, individual streams, transects, and sampling plots represented the 4 sequentially, fully nested factors. Mean plot biomass for the 14 study streams ranged from 0.02 - 0.42 g/m² and were lowest in AMD-impacted and episodically acidified streams. In general, biomass in acidic streams was mostly attributable to a few large animals; greater numbers of smaller individuals characterized the biomass in non-acidic streams.

Seven species were captured during the plot-sampling study. The northern two-line (Eurycea b. bislineata) accounted for 52% of captured individuals. Species composition in the 14 study streams ranged from 1 - 7, but at the watershed scale, species composition was generally ineffective and provided poor resolution. The mountain dusky (Desmognathus ochrophaeus) was by far the most widespread and persistent species. The abundance of the northern dusky (D. f. duscus) and northern two-line was substantially reduced in acidified watersheds.  The northern spring (Gyrinophilus p. porphyriticus) occurred in comparable densities in reference and episodically acidified streams, but became scarce in AMD and fragmented streams. The northern red (Pseudotriton r. ruber) was primarily documented in fragmented streams. Other species were rarely found, regardless of stream condition. The absence of one of more species is not as striking as the subtle shift in their dominance or relative abundance. This latter trait will probably limit the usefulness of any efforts to detect stream plethodontid assemblage responses by examination of species composition alone.

Larval individuals comprised 74% of all specimens. Not all taxa were evenly represented by all lifestages; 91- 98% of northern two-line, northern red, and northern spring were larvae. Generally, an abundance of early lifestages was associated with reference and fragmented streams; juveniles and adults were the dominant groups in acidified streams. Seeps also emerged as potentially important refugia for early lifestages, especially in impaired streams.

Of the 4,065 specimens examined, 160 (3.9%) exhibited signs of limb loss, abnormality or other major injury. The northern spring (10.6%) and Appalachian seal (9.9%) were the species with the highest incidence. The incidence of injuries was independent of transect type and was also highly variable across watersheds, ranging from 2- 27%.

Ordination analyses were applied to transect-compiled data. This approach was thought to be a desirable scale for the investigation of relationships between assemblage attributes and stream habitat correlates. The species matrix was constructed from 11 lifestages representing 5 of the more widely distributed species. Twenty (20) variables reflecting stream water chemistry, climatic conditions, stream and bank morphology, and watershed land-cover were initially screened. Sampling effort was used as a covariable on all analyses. Although 65 transects were ordinated, incomplete data prevented active participation of 7. Data were analyzed by detrended correspondence analysis (DCA), canonical correspondence analysis (CCA), and detrended canonical correspondence analysis (DCCA) by the computer program CANOCO 4.

The subset chosen by the automatic forward selection algorithm consisted of 5 variables: watershed forest, pH, distance to source, canopy closure, and bank vegetation. The second subset, based exclusively on stream water and stream bank habitat variables consisted of 6 predictors: alkalinity, pH, canopy closure, cobble cover, stream temperature, and bank vegetation. The total variance in the species data, or inertia, for both subsets was 0.844. The covariable explained 9.7% of the total inertia (l₁ =  0.082).

In both subsets, eigenvalues for the first 4 ordination axes were relatively low, ranging from 0.202 - 0.016. Despite this, the resulting gradients were sufficiently long to span the range of scores (optima) of all 11 species-lifestages. Gradient length in all axes, a gauge of the unimodality of a species response along an ordination axis, ranged from 2.347 - 1.728 for axes 1 and 3, respectively. Correlation coefficients for the species-environment relation, were appreciably high for axis 1 of both subsets (0.868, 0.832), but not comparatively so for other axes (0.57 - 0.737). Evidently, the species-environment relationship is strongest for the first axis, and consequently, for variables expressed by this axis. Nonetheless, two-dimensional ordination plots for the subsets investigated depict, about 85 and 82% of the species-environment variance, and 34 and 32% of the species derived variance. More importantly, the plots reflect relationships that emerge in other comparisons and descriptions.

Several potentially useful metrics are presented and discussed. The metrics are based on counts or relative abundance of one of more of the 11 lifestages used in the ordination analyses. Where present, the scope for detection was generally narrow and impairment specific.

Stream Plethodontids do appear responsive to the gradients investigated. Data presented provides a starting point for the description of the effects of impairment on a potentially valuable bioindicator that is abundant, widespread, ecologically important to several trophic levels, and interfaces between aquatic and terrestrial components of riparian areas.