1Department of Biology, Center for Biodiversity Studies, Western Kentucky University, Bowling Green, KY 42101
²Department of Agriculture, Western Kentucky University, Bowling Green, KY 42101

     Abstract. Buck Creek, an Order III tributary to the Trammel Fork of Drake's Creek, receives organic-waste irrigation water runoff along part of its watershed. Fish community samples were made in the fall and spring from 1992 to 2000, above and below the treated pasture area, using portable electro-fishing methods. A total of 17,734 fish representing 29 species and two hybrids were taken in the study. Except for two drought season collections at the upstream station (I), species occurrences were similar at both stations in both seasons, averaging 64% overlap. Significantly more fish were taken at the downstream Station II (68.3% vs 31.7%) and in the Fall (63.6% vs 36.4%). Small, short-lived fishes dominated the assemblages by station and season. The largescale stoneroller minnow, Campostoma oligolepis, was the only herbivore taken and represented 38.7% of the total; the bluntnose minnow, Pimephales notatus, striped shiner, Luxilus chrysocephalus, and creek chub, Semotilus atromaculatus, were major omnivores making up 8.9%, 5.7%, and 5.1%, respectively, of the total. The northern studfish, Fundulus catenatus, (13.2%) and orangethroat darter, Etheostoma spectabile, (9.6%) were the major insectivores, while the green sunfish, Lepomis cyanellus, (4.2%) was the major top carnivore. Seasonally, significantly greater numbers of herbivores occurred during the Fall (P = 0.031) while carnivores were more abundant in the Spring (P = 0.023). Insectivores were most abundant at Station I (P = 0.006) while omnivores (P = 0.001) and carnivores (P = 0.002) occurred in greater numbers at Station II. Principal component analyses indicated a clear site effect on patterns of species composition; PC I (15.68% of the total variation) was driven by variation in the rosefin shiner, Lythrurus ardens, orangefin darter, Etheostoma bellum, and bluntnose minnow, all more common in samples from the upstream site. PC II (12.87%) showed some separation of fall and spring samples; greenside darter, Etheostoma blennioides, rainbow darter, Etheostoma caeruleum, and splendid darter, Etheostoma barrenense, tended to be more abundant, and the largescale stoneroller less abundant, in spring samples. A reduced PCA excluding rarely-occurring taxa produced similar clusters, and patterns of loadings on the first two corresponding pairs of axes were highly correlated (p < 0.00002). However, the observed site effect did not appear to result from anthropogenic input; samples from two unimpacted sites in Mays Branch clustered with their spatial counterparts in Buck Creek in all analyses. As a result, community differences in Buck Creek were likely the result of habitat differences between the two sites rather than anthropogenic land-use practices.

     Much has been written about stream fish assemblages since 1980 including works questioning contemporary ecological theory (Fisher 1983, Minshall et al. 1983, Schoener 1987) and the application of fresh analytical techniques (Felley and Hill 1983; Matthews 1985, 1990; Ross et al. 1987; Bianchi and Hoisaeter 1992; Hansen and Ramm 1994; Taylor et al. 1996; among others). Accompanying this renewed discussion has been considerable debate and divergent views regarding such topics as persistence and stability of stream fish communities (Hansen and Ramm 1994) and physical vs biological processes in regulating stream-fish community structure (Schlosser 1987, Freeman et al. 1988). From these studies and discussions has emerged a set of considerations to be used in conducting stream fish surveys; namely, simultaneous treatment of spatial and temporal features (Meador and Matthews 1992), long-term baseline data set acquisition (Ross et al. 1987, Taylor et al. 1996), identification of the influence of upstream-downstream environmental variation in lotic systems (Schlosser 1990, Capone and Kushlan 1991), and the recovery of fish populations and communities following natural and cultural disturbances (Detenbeck et al. 1992).

     In 1992, an opportunity was realized to apply some of the above directives to an anthropogenic land-use practice on the fish community in a small stream in southern Kentucky. A Best Management Practice swine rearing production facility was established in south central Kentucky that has adopted the practice of irrigating concentrated animal waste effluents onto pasture lands at rates of 52 to 95 kg N/ha, 12 to 15 kg P/ha, and 91 to 95 kg K/ha per application, two times in the spring and two times in the summer. The wastes are taken from the second of a two cell, in-series lagoon system and applied to the pasture land. Buck Creek, an Order III tributary to the Trammel Fork of Drake's Creek, lies adjacent to the treated land and receives pasture land runoff via two effluent ditches.

     The objective of this study was to compare the fish community in an unaffected reach of the stream, above the land-use plan, with that in a potentially affected stream reach, immediately below the most downstream land-use effluent over an eight-year sampling regime. The spatial separation between the two potential enrichment sites on the stream (1 km) was insufficient to consider the stream collecting sites to be representative of upstream versus downstream locales. On the basis of a different stream bed physiography at each locale, sampling the stream fishes above and below the inflows essentially compared the fish assemblages in an upper riffle/bedrock chute habitat with those in a lower riffle/pool habitat. Additional inferences were drawn regarding the stability of sampled communities in a small stream over an extended time frame. Interpretations were proposed to explain any differences that might occur within and between the fish communities.

STUDY AREA

     Buck Creek is a small, Order III, perennial flow stream receiving spring inflows and groundwater runoff in Allen County, Kentucky. It is approximately 7.25 km long and empties into the Trammel Fork of Drake's Creek in south-central Kentucky (Figure 1). Buck Creek is canopied by riparian forest of varying width and drains land used mostly for cattle pasture.

     Sampling stations were located in relation to groundwater inflows from an intensive swine rearing facility which utilizes animal waste, pasture fertilization-irrigation practices. Two sloping ravines from the treated pastures surrounding the rearing facility form intermittent effluents into Buck Creek less than 1.5 kilometer from the point of waste application. The most upstream station (Station I) was just above a gully-tributary draining one side of the watershed while the most downstream station (Station II) was just below a similar gully draining the other side of the property. The average stream gradient along the reach of both sampling stations was 4.5m/km.

     Station I was a shallow water station (avg 0.1m), 35m long containing an upstream riffle and two, bedrock pool/chute areas separated by a small cataract with visible water movement. This station had a karstic, bedrock substrate with little gravel and represented an average 2,347 ft² (218.1 m²) surface area. Station II was 87m long and consisted of two riffles and one large and one small pool representing an average 6,393 ft² (594.2 m²) surface area. During the study period, Station II changed in size and configuration in response to several high-water episodes and the construction of a concrete, low-water bridge 100 m upstream from the head of the station. It's average depth was 0.3m with some bedrock but mostly gravel substrate. Station II was 1.04km downstream from Station I.

     A second stream, Mays Branch, an Order III tributary of the Sulphur Fork of the Middle Fork of Drake's Creek, was sampled as a reference stream beginning in June 1995. Mays Branch was 10.5km west of Buck Creek and was not influenced by swine rearing practices. Two stations were

sampled on Mays Branch in order to compare station and stream effects on the fish communities in an affected and unaffected stream. Station I on Mays Branch was a shallow water station, 42m long, with mostly bedrock substrate. Mean depth was 0.1m and an average 2,314 ft² (215.1m²) surface area were sampled. Station II was a larger station, 70m long, consisting of one long pool with entering and exiting riffle. Mean depth was 0.3m, the substrate was mixed bedrock and large rocks and gravel, and represented an average of 4,811 ft² (447.1 m²) surface area. Station II was 0.75km downstream from Station I.

     Fish were collected from the Buck Creek stations in May-June and September-October from 1992 to 2000. Collections were made with electro-fishing gear consisting of portable electrodes powered by 120v AC current supplied by an industrial grade portable generator. Block nets were placed above and below the sampling stations and electrical current applied until fish were no longer observed. Fish were hand-picked from the substrate to maximize sampling effectiveness. Mean sampling time at each station was approximately 60 minutes. Fish were fixed in 10% formaldehyde upon capture, sorted, identified, measured, and weighed in the laboratory. All fish have been curated and reposited in the Biodiversity Center fish collection at Western Kentucky University.

Community Parameters

     Various stream fish community metrics including the Shannon-Wiener diversity index (Brower and Zar 1984), index of biotic integrity (Karr 1981, Kentucky Division of Water 1993), biomass in pounds per acre, number of fish collected per sample, and species richness were determined for each collection.

Univariate Analyses

     The Paired t-test was used to compare species richness, fish density, biomass, species diversity and Index of Biotic Integrity , for station and season effects. Because there were only single replications from each season/station, combined data were pooled across seasons in testing for station effects and vice-versa. This is a conservative approach as it increases the variance within groups. Significance was assessed using Bonferroni adjusted levels which were calculated separately for station and season tests. P_critc values were determined on the basis of .05/5 = .01.

Multivariate Analyses

     All multivariate analyses were conducted using SYSTAT 9.0 (SPSS, Inc. 1999). Variation in species composition among samples from Buck Creek was summarized through principal components analysis (PCA) of the correlation matrix derived from species abundance data. For all analyses, the Fall 1999 Station I sample was excluded, as only 1 fish was captured in that sample). Use of the correlation rather than variance-covariance (VCV) matrix was deemed preferable given the two order of magnitude difference in abundance among the various taxa. In fact, while preliminary PCAs based on the variance-covariance matrix explained a higher proportion of the total variation in the data set (as expected), lower axes were all dominated by among-sample variation in the most abundant species and provided less evidence of among-sample structure.

     The influence of rarely-occurring taxa on the relationships among samples was evaluated by comparing the PCA utilizing all Buck Creek taxa to one based on a reduced number of species. Here, species occurring in fewer than 40% of the samples were excluded; thirteen taxa were eliminated using this criterion: Hypentelium nigricans, Catostomus commersoni, Cyprinella spiloptera, Notemigonus crysoleucas, Phoxinus erythrogaster X Semotilus atromaculatus hybrid, Ameiurus nebulosus, Noturus elegans, Chologaster agassizi, Ambloplites rupestris, Lepomis megalotis, Pomoxis annularis, Lepomis cyanellus X L. macrochirus hybrid, Etheostoma squamiceps (Table 1). The correspondence in the ordinations resulting from the two analyses was qualitatively determined by comparing patterns of loadings on the first two principal axes. In addition, for each analogous pair of axes we computed the Pearson correlation coefficient of the loadings on those axes and evaluated their significance using Bonferroni-adjusted criteria.

     To separate the possible impact of anthropogenic input into Buck Creek Station II from effects due to the more downstream position of that station, species abundance data from Mays Branch samples were projected onto the first two principal components derived from the Buck Creek complete data. Three taxa occurring only in Mays Branch were excluded. This approach allowed us to interpret the similarity of upstream and downstream stations in the two streams along the axes of variation important in Buck Creek.

     A total of 17,734 fish representing 29 species and 2 hybrids was collected from Buck Creek during the study (Table 1). One cyprinid cross, probably Semotilus X Phoxinus (Burr and Warren 1986) represented by six individuals and a single centrarchid cross, probably Lepomis cyanellus X L. macrochirus, were taken at Station II (Table 1). Species occurrences were similar between stations and seasons during the study ranging from 5% to 93% overlap and averaging 64% (Table 2). Likewise, the numerical make-up of the assemblages by station and season consisted mostly of small, short-lived, minnow/darter type species, 87% and 94% in the Spring and Fall, respectively, and 93% and 91% at Station I and II, respectively (Table 1).

Station Effects

     Twenty-one species were taken at Station I during the study while 28 species including both hybrids were taken at Station II (Table 1). With the exception of the northern hogsucker, spotfin shiner (Cyprinella spiloptera), and longear sunfish, which favored the deeper water of Station II, all other species not occurring at Station I were rare species represented by only 1 or 2 individuals at Station II (Table 1). Species richness was greater at the downstream station (17.6, SE 0.394) than at the upstream station (12.1, SE 1.02) (P = 0.001) (Table 2). Likewise, Shannon-Weiner species diversity was greater at Station II (2.764, SE 0.108) than at Station I (2.347, SE 0.209) (P = 0.018) (Table 2).

     Although a greater total number of fish was collected at the downstream site than at the upstream station, when adjustments were made for the surface area difference between the two stations, slightly more fish occurred at the upstream station per unit area, 1.5 fish/m² (SE 0.202) vs.1.2 (SE 0.118) fish/m² (P = 0.215). Average biomass estimates at the two stations were similar for the entire study, 9.8 kg/ha (SE 1.13) for Station I and 10.3 kg/ha (SE 0.70) for Station II (P = 0.63). The Index of Biotic Integrity was not significantly different between the two stations, 43.9, SE 7.846 at Station I and 46.1, SE 0.98 at Station II (P = 0.249).

Season Effects

     The total number of species collected in the study was similar in the Fall and Spring collections, 25 in the Spring and 28 in the Fall (Table 1). Average species richness per collection was similar in the Fall (14.2 spp, SE 1.234) and Spring (15.4 spp., SE 0.612), (P = 0.357). Greater seasonal variation in species richness was observed at Station I, 10.9, SE 1.83 in the Fall and 13.4, SE 0.532 in the Spring, although they were not significantly different (P = 0.200). Little variation in species richness was observed between seasons at Station II (17.25, SE 0.56 in Spring and 17.9, SE 0.56 in Fall) (Table 2).


     The total number of fish taken in Fall collections was significantly greater than in the Spring, 11,284 vs. 6,450, respectively (P = 006) (Table 1). When adjusted for area of stream sampled, however, only slightly more fish per square meter were taken in the Fall than Spring, 1.54 m² (SE 0.16) vs 1.14 m² (0.165), respectively (P = 0.188). Biomass estimates were also similar for the two season’s collections, 10.0 kg/ha (SE 0.684) in the Fall and 9.34 (SE 1.134) kg/ha in the Spring (P=0.585). Seasonal biomass estimates within the two stations were trended during the study, although not significantly so; Station I estimates were greater in the Spring than those from Station II, (10.95 kg/ha (SE 1.16) and 8.99 kg/ha (0.622), respectively, P = 0.087), while in the Fall estimates were greater at Station II than Station I (11.43 kg/ha (1.096) and 8.72 kg/ha (SE 1.867), respectively, P = 0.098) (Table 2).

     Shannon-Weiner species diversity was greater in the Spring than the Fall, 2.96 (SE 0.073) and 2.20 (SE 0.184), respectively (P = 0.002). Diversity indices were greater in the Spring than Fall at Station I, although not significantly so, 2.76 (SE 0.102) vs. 1.98 (SE 1.047), respectively, (P = 0.112), and also greater in the Spring than Fall at Station II, 3.16 (SE 0.088) vs. 2.41 (SE 0.114), respectively, (P = 0.001 ) (Table 2).

The Index of Biotic Integrity was similar for both seasons, 44.4 (SE 1.87) in the Fall and 45.8 (SE 0.892) in the Spring (P = 0.59).

Feeding Guilds

     Herbivore and insectivore species were the dominant feeding types in Buck Creek (Table 3). Herbivores represented 38.7% of the total number of specimens collected followed by insectivores 34.2%, omnivores 22.6%, and top carnivores 4.5% (Table 3). The largescale stoneroller, Campostoma oligolepis, was the only true herbivore (Table 1). Omnivores included seven species and one hybrid but were mostly represented by the bluntnose minnow, Pimephales notatus, striped shiner, Luxilus crysocephalus, creek chub, Semotilus atromaculatus, and southern redbelly dace, Phoxinus erythrogaster (Table 1). Sixteen species were insectivores and included predominantly shiner and darter species, namely the rosefin shiner, Lythrurus ardens, orangethroat darter, Etheostoma spectabile, and rainbow darter, Etheostoma caeruleum, and most notably the northern studfish, Fundulus catenatus. Six species were top carnivores or piscivores and included mostly the green sunfish, Lepomis cyanellus.

     Feeding preferences of fish in Buck Creek varied between stations and from season to season. Omnivores (0 25.4%, SE 2.77; P = 0.001) and top carnivores (0 5.0%, SE 1.418; P = 0.002) were significantly more common at Station II than Station I (0 16.9%, SE 2.77; and 0 3.3%, SE 1.135, respectively) while insectivores were most characteristic of Station I (0 49.3%; SE 5.46) than Station II (27.1%; SE 2.85) (P = 0.006). Herbivores were more common in the Fall (0 44.2%; SE 5.74; P = 0.031) than Spring (0 29.2%, SE 2.907) while carnivores were more common in the Spring (0 6.0%, SE 1.663; P = 0.023) than Fall (0 3.6%, SE 0.829). Although not significantly supported in all cases, numerical trends suggested a strong herbivore presence in the Fall at Station II, insectivores in the Spring at Station I, and carnivores in the Spring at Station II (Table 3). An obvious Cladophora and filamentous algal mat was present on the substrate at the downstream station during the Fall collections that was not obvious in the spring. The biomass of the respective feeding groups followed basically the same trends as fish numbers.

PCA Results

     Principal components analysis of fifteen samples from two stations along Buck Creek suggested the existence of station and season effects on patterns of variation in species composition among sites. The first three principal components accounted for 37.77 % of the total variation in the data set; 12 of 31 latent roots contained significant information (Reyment 1991). The first principal component (PC I - 15.68 % of the total variation) separated Station I and Station II samples (Fig. 2). Variation in abundance of Lythrurus ardens, Etheostoma bellum, Pimephales notatus, Catostomus commersoni, and Ambloplites rupestris loaded heavily on PC I (Table 4); these species were each more common at Station II (Table 1). Station I samples were also more variable than Station II samples, evidenced by the larger envelope encasing downstream sample scores (Figure 2). PC II (12.87 %) provided some separation of fall and spring samples; Campostoma oligolepis, E. caeruleum, and E. barrenense loaded heavily on this axis; C. oligolepis was more common in fall samples, while the other taxa were generally more abundant in spring samples (Table 1).

     Patterns of variation in incidentally-occurring taxa did not strongly affect the relationships among samples. Reduced PCA excluding thirteen taxa generated ordinations qualitatively similar to those based on all 31 taxa. In addition, patterns of loadings on the first two corresponding pairs of principal components were each highly correlated (P < 0.00002).

Station I and Station II samples from Mays Branch clustered in principal component space with their spatial counterparts in Buck Creek (Fig. 2). As in Buck Creek, Station II samples from Mays Branch showed greater temporal variability than did Station I samples.

Species Richness
     
     Species richness occurrences from Buck Creek in this study were consistent with those of other small streams in the southeastern U.S. (Ross et al. 1987, Freeman et al. 1988, Matthews 1990). In addition, the degree of temporal variation/overlap in species presence observed here agreed well with long-term sampling data from three Arkansas streams (Marsh-Matthews et al. unpubl.). In that study, individual samples contained 50-60% of the total number of species encountered, and approximately 70% of species observed in randomly-selected pairs of samples were common to both (E. Marsh-Matthews pers. comm.).

     Similar to the observations of Schlosser (1987), and Taylor et al. (1996), the variation in abundance and distribution patterns of the assemblages observed in this study appeared to be primarily due to environmental variables such as habitat heterogeneity and the life histories of the fishes involved, rather than anthropogenic influences. The greater species richness and diversity indices at the Buck Creek downstream station was a reflection of greater availability of fish habitats in the pool setting. Schlosser (1987, 1990) and Capone and Kushlan (1991) reported that stream reaches with poorly developed pool areas have shallow depths, low habitat availability, low species richness, and low numbers of fish. Diehl (1992) likewise suggested structurally complex habitats (pools) to hold more diverse communities than simple ones.

     The onset of a prolonged, two-year drought in 1999-2000 saw the number of fish collected at the upstream riffle station decrease dramatically during the last four collections of the study. Concurrent with the reduction in fish number at the riffle station was a proportional increase in the number of fish at the pool station. Meador and Matthews (1992) described similar occurrences in the temporal consistency of fishes in an intermittent stream in Texas. The proximity of the fish-rich pool area to the drought induced depauperate riffle area amplified the role of the pool as a local species refugia and voided the conclusion that this reach of Buck Creek is a harsh system for its resident fish species. Meador and Matthews (1992) reported a similar conclusion for the fish assemblage in an intermittent Texas stream.

     Insectivorous fish species were most abundant in Buck Creek during the Spring while herbivores were the dominant feeding category in the Fall. The increase in insectivores was greatest at the upstream riffle site and appeared to be a concurrent effect of the increased numbers of shiners, bluegill, and darter species that migrated to the upstream reaches in the spring as cited earlier. The herbivore increase in the Fall was predominantly a result of the increase in food availability for that feeding category at the pool station. Based upon PCA analysis of macroinvertebrate communities at the two fish collecting sites, VanStone and Jack (1997) reported taxa richness and total numbers of individuals to be higher in the downstream pool site. Although speculative, they suggested the greater macroinvertebrate communities in the downstream pool to result from an increased periphyton/filamentous algae growth as an effect of nutrient enrichment. The pattern of more herbivorous species at that site likewise suggests a feeding shift toward the algal forms present in the extensive Cladophora mats present (J. Jack pers. comm.). The Mays Branch samples showed a fish community pattern by station similar to that at the Buck Creek stations, however, suggesting at best a minor groundwater runoff influence at the downstream Buck Creek station.

Potential Influence of Groundwater Runoff

     The downstream Buck Creek site is distinct in its fish community structure from the upstream site, i.e., greater species richness, higher species diversity, and greater herbivore/omnivore representation, especially in the Fall. Nevertheless, it is difficult to attribute this pattern to agricultural runoff generated just upstream of that site, for several reasons. First, water quality parameters potentially indicative of agricultural impact (nitrates, nitrites, phosphates, etc.) did not show a significant correlation with patterns of species abundance between sites during the study (pers. obs.). Even if these nutrients were exerting a significant impact on the biological community, their uptake by primary producers would be sufficiently rapid such that detecting their signal through chemical analysis of water samples would be unlikely (W. Matthews pers. comm.). Second, Reash and Berra (1987) reported the rock bass and fantail darter to be missing from polluted waters and darter species as a group to be especially sensitive to anthropogenic disturbances. In Buck Creek, representatives of these species or species groups were all present in the pool station in the Fall, following the influence of any nutrient enrichment on summer algal growth or water quality alteration. However, with the exception of the orangefin darter, which was most abundant in the Fall at the pool site, each of these species did occur in the Fall in percentages of only 29% to 39% of their numbers at the riffle station at the same time. Whether these reductions in occurrence at the downstream station reflected an anthropogenic influence from the surrounding land use or were merely the spurious result of seasonal distribution patterns of the species could not be confirmed. And lastly, the overlap of test stream and reference stream collection sites as shown by the PCA suggests a strong station effect operating on both streams rather than a selective anthropogenic effect at the downstream Buck Creek site.

     In summary, data collected from Buck Creek suggest that it supports a fish fauna typical of small, upstream Order III streams in the south-east United States, that stream site differences explain most long-term variations in stream presence, that individual fish species biology is responsible for most seasonal community differences, and that anthropogenic influences were of minimal effect on the fish assemblages during the duration of the study.

The U. S. Environmental Protection Agency 319 is gratefully acknowledged for providing funds which made this study possible. The Kentucky Division of Conservation and Division of Water provided technical assistance at the initiation of the project which is very much appreciated. The Kentucky swine producer was a willing and supportive cooperator throughout the project in providing financial support and logistical assistance in making the field collections possible. The Department of Agriculture at Western Kentucky University provided willing and interested students who assisted in making all field collections. Dr. John Andersland is acknowledged for his kind and expert assistance in making data figures.

Bianchi, G., and T. Hoisaeter. 1992. Relative merits of using numbers and biomass in fish community studies. Mar. Ecol. Prog. Series 85:25-32.

Brower, J. E., and J. H. Zar. 1984. Field and Laboratory Methods for General Ecology. W. C. Brown Publishers, Dubuque, IA. 2nd Ed. 226 pp.

Burr, B. M., and M. L. Warren, Jr. 1986. A distributional atlas of Kentucky fishes. Kentucky Nature Preserves Commission, Scientific and Technical Series Number 4, Frankfort, KY. 398 pp.

Capone, T. A., and J. A. Kushlan. 1991. Fish community structure in dry season stream pools. Ecology 72:983-992.

Detenbeck, N. E., P. W. DeVore, G. J. Niemi, and A. Lima. 1992. Recovery of temperate-stream fish communities from disturbance: A review of case studies and synthesis of theory. Environ. Mgmt. 16:33-53.

Diehl, S. 1992. Fish predation and benthic community structure: the role of omnivory and habitat complexity. Ecology 73:1646-1661.

Felley, J. D., and L. G. Hill. 1983. Multivariate assessment of environmental preferences of cyprinid fishes of the Illinois River, Oklahoma. Am. Midl. Nat. 109:209-221.

Fisher, S. G. 1983. Succession in streams. pp 71-127. In J. R. Barnes and G. W. Minshall (eds.). Stream Ecology: Application and Testing of General Ecological Theory. Plenum Press, NY.

Freeman, M. C., M. K. Crawford, J. C. Barrett, D. E. Facey, M. G. Flood, J. Hill, D. J. Stouder, and G. D. Grossman. 1988. Fish assemblage stability in a southern Appalachian stream. Can. J. Fish. Aquat. Sci. 45:1949-1958.

Hansen, M. J., and C. W. Ramm. 1994. Persistence and stability of fish community structure in a southwest New York stream. Am. Midl. Nat. 132:52-67.

Karr, J. R. 1981. Assessment of biotic integrity using fish communities. Fisheries 6:21-26.

Kentucky Division of Water. 1993. Methods for Assessing Biological Integrity of Surface Waters. KY Division of Water, Water Quality Branch, Frankfort, KY. 139 pp.

Matthews, W. J. 1985. Distribution of midwestern fishes on multivariate environmental gradients, with emphasis on Notropis lutrensis. Am. Midl. Nat. 113:225-237.

Matthews, W. J. 1990. Spatial and temporal variation in fishes of riffle habitats: A comparison of analytical approaches for the Roanoke River. Am. Midl. Nat. 124:31-45.

Meador, M. R., and W. J. Matthews. 1992. Spatial and temporal patterns in fish assemblage structure in an intermittent Texas stream. Am. Midl. Nat. 127:106-114.

Minshall, G. W., R. C. Petersen, K. W. Cummins, T. L. Bott, J. R. Sedell, C. E. Cushing, and R. L. Vannote. 1983. Interbiome comparison of stream ecosystem dynamics. Ecol. Monogr. 53:1-25.

Reash, R. J., and T. M. Berra. 1987. Comparison of fish communities in a clean-water stream and an adjacent polluted stream. Amer. Midl. Nat. 118:301-322.

Reyment, R. A. 1991. Multidimensional Paleobiology. Pergamon Press, NY.

Ross, S. T., J. A. Baker, and K. F. Clark. 1987. Microhabitat partitioning of southeastern stream fishes. pp 42-51. In W. J. Matthews and D. C. Heins (eds.). Community and Evolutionary Ecology of North American Stream Fishes. University of Oklahoma Press, Norman, OK.

Schlosser, I. J. 1987. A conceptual framework for fish communities in small warmwater streams. pp 17-24. In W. J. Matthews and D. C. Heins (eds.). Community and Evolutionary Ecology of North American Stream Fishes. University of Oklahoma Press, Norman, OK.

Schlosser, I. J. 1990. Environmental variation, life history attributes, and community structure in stream fishes: Implications for environmental management and assessment. Environ. Mgmt. 15:621-628.

Schoener, T. W. 1987. Axes of controversy in community. pp 8-16. In: W. J. Matthews and D. C. Heins (eds). Community and Evolutionary Ecology of North American Stream Fishes. University of Oklahoma Press, Norman, OK.

SPSS, Inc. 1999. SYSTAT 9.0. SPSS, Inc. Chicago.

Taylor, C. M., M. R. Winston, and W. J. Matthews. 1996. Temporal variation in tributary and mainstem fish assemblages in a Great Plains stream system. Copeia 1996:280-289.

VanStone, J. and J. Jack. 1997. Using macroinvertebrate indicators to monitor the performance of agricultural best-management practices. pp. 153-161. In: A. F. Scott, S. W. Hamilton, E. W. Chester, and D. S. White (eds.). Proceedings of the Seventh Symposium on the Natural History of Lower Tennessee and Cumberland River Valleys. The Center for Field Biology, Austin Peay State University, Clarksville, TN.