Most states in the U.S. are currently developing methods for assessing the integrity of aquatic habitats through the development of regional biocriteria. While multimetric indices have been used to show community composition, pollution tolerance, species diversity, and trophic structure with a combined index, the specific environmental factors that drive biological communities may be better explained through the use of multivariate statistical techniques. Macroinvertebrate and fish assemblages were sampled along with water quality, landuse and qualitative and quantitative habitat assessments from forty-nine sites throughout the Choctawhatchee-Pea, a southeastern U.S. watershed. Multivariate statistical analyses of habitat, water quality, and land-use data were used to determine the relationship between environmental variables and the dependent biological variables, macroinvertebrate and fish community structure. Sampling of biological and environmental data showed that there was a great deal of homogeneity within the watershed, which complicated the task of identifying environmental influences on biological assemblages. Macro-invertebrate and fish assemblages of the Choctawhatchee-Pea watershed were similar in their response to environmental conditions with water chemistry having the greatest relationship to macro-invertebrate and fish community structure followed by instream habitat and land use.
Land-use activities occurring within a watershed often severely alter stream morphology and water quality of individual catchment areas as well as complete river systems. These activities change the natural balance of flow, sediment movement, temperature, and other important variables, which ultimately impact the physical, chemical, and biological processes that occur within a stream system. Recently, studies have focused on quantifying biological response signatures for specific anthropogenic stressors occurring within entire watersheds and ecoregions (Cairns, 1992; Stewart et al., 2000; Carlisle et al., 2003). Unfortunately, the effects of nonpoint source pollution on aquatic biota can be difficult to quantify since most pollution occurs during heavy rain events (Petersen et al., 1987; Morris et al., 2003). Understanding the primary factors that have the most influence on differing communities of stream biota is often complex, requiring multivariate analysis of environmental variables along with species data (Dyer et al., 1998; Stewart et al., 1999).
As natural vegetation surrounding streams is converted to urban or agricultural uses, instream habitat and stream biota are negatively affected (Harding et al., 1998; Stewart et al., 2000). Streams located within predominately agricultural areas often contain excessive nutrients, lack diverse habitat complexity, and tend to be wide and shallow offering poor habitat for biota (Henley et al., 2000). Among the causes of habitat degradation, sedimentation and turbidity are identified as important contributors to declines in aquatic assemblages (Stewart and Swinford, 1995; Henley et al., 2000). Sources of stream sedimentation include agriculture, forestry, mining, road construction, and urban activities (Waters, 1995; Morris et al., 2003).
Since macro-invertebrate and fish assemblages are integrally linked to instream physical and chemical characteristics, they have frequently been used as indicators of water quality (Karr, 1981; Simon and Stewart, 1999). These assemblages respond to physical and chemical variables in specific geographical areas (Fausch et al., 1990; Richards et al., 1993). Estimation that fish and invertebrate communities are adversely affected by loss of habitat and other nonpoint sources might seem fairly straightforward, but closer examination of these organisms often reveals complex interconnections that make identification of specific environmental stressors difficult. Identifying factors that influence stream biota can be difficult due to the geophysical qualities of a particular watershed with varying land-use types (Allen et al., 1997; Richards et al., 1997). Quantifying nonpoint source pollution occurring within a watershed can also be problematic, and calculation of land use may be the best estimate of pollution potential (Schleiger, 2000). The objective of this study was to determine the response of macroinvertebrate and fish assemblages of a coastal plains watershed to their surrounding environmental factors, specifically, water chemistry, habitat, and land use.
The Choctawhatchee-Pea watershed is located in the Florida panhandle and southeastern Alabama and is part of the Eastern Gulf Coast Plain ecoregion. This watershed has a total drainage area of 13,855 km2 as it flows south into Florida and eventually empties into Choctawhatchee Bay. The coastal plains ecoregion is composed of mostly sand, clay, marl, and limestone substrates and is characterized by gentle rolling hills, sharp ridges, prairies, and broad floodplains. This study focused on the Alabama portion of the Choctawhatchee-Pea watershed that comprises 7,242 km of rivers and streams that flow through ten counties (Figure 1).
The watershed is divided into numerous land-use types with most of the region dominated by pinewood silviculture operations, livestock and row crop agriculture, and expanding urban development including a large proportion of unpaved roads. The conversion of mostly forested areas to agricultural land was an event completed in less than 200 years (Fajen and Layzer, 1993). Before agricultural practices, natural vegetation protected soil from intense erosion and problems of stream sedimentation and turbidity were probably insignificant (Fajen and Layzer, 1993). Deforestation of riparian areas has caused numerous changes in many stream habitats, including destabilization of stream banks and increased sedimentation from eroding fields and unpaved roads (Cook and O'Neil, 2000). This sedimentation has in many cases altered the geomorphology of streams decreasing stream depth and flow rates and increasing channel width and uniformity.
Sampling took place from early April to May in 2001 during normal (after the spring rains had abated) flow conditions and included 34 random and 15 targeted sites throughout the watershed. Random sites were selected for an accurate representation of the entire watershed, and specific sites were targeted for differing land-use practices and habitat quality in order to reduce uniformity among samples. For the watershed, U.S. Geological Survey 1:24,000 topographic maps were used to calculate watershed area and stream order (Strahler, 1957), and to establish all possible access points (roads and bridges). All access points were assigned a number, and sites were grouped depending on order. Since this study was designed for wadeable streams, all streams higher than fifth order were eliminated from the random selection. A proportion was then used to determine how many sites would be randomly selected from each order. This method of site selection allowed sampling effort to be concentrated on those streams that comprised the majority of the total river area without sacrificing randomness.
Physical habitat assessment
The intensive habitat sampling protocol of the U.S. Environmental Protection Agency's Environmental Monitoring and Assessment Program (EMAP) (Kaufmann et al., 1999) and Wisconsin Department of Natural Resources' Habitat Sampling Protocol (Wang et al., 1997a) were modified to assess habitat variables in a short time but with sufficient and accurate measurements. These qualitative and quantitative sampling methods focused on various spatial scales including catchment area, stream segment, and stream reach. Reach-scale assessment was completed for both habitat and biological sampling and consisted of sampling a segment 35 times the mean stream width, but no less than 150 m and no more than 500 m. Entire segment variables focused on counting large woody debris (LWD) and the number of riffle, run, and pool complexes for a particular sampling reach. Five individual transects spaced in proportion to reach length were used to measure the components of stream width, depth, substrate type, depth of fines, bank stability, canopy cover, fish cover, and percent riparian buffer.
Qualitative assessment of habitat quality for each site was measured using U.S. Environmental Protection Agency's Rapid Bioassessment Protocol (RBP) for low gradient streams (Barbour et al., 1999) and Ohio's Qualitative Habitat Evaluation Index (QHEI) (Rankin, 1989). Both of these multimetric habitat indices are based on qualitative evaluation of habitat integrity and focus on channel morphology and alteration, substrate type, riparian vegetation, and stream flow. Scores from RBP and QHEI are reflected by the presence or absence of physical characteristics that constitute quality habitat. Comparison of these habitat indices has shown that the QHEI and RBP are highly correlated (Rankin, 1995). The major difference between these two methods lies in their weighing of habitat factors in comparison to biological indices. Ohio's QHEI explains more of the variation in biological indices developed for fish communities than does U.S. Environmental Protection Agency's RBP.
Catchment area land use
Land use was calculated for the catchment area surrounding each sampling point. The U.S. Environmental Protection Agency's assessment system BASINS, integrated with Arcview® GIS was used to delineate areas of differing land-use types. This program was developed to assist in the examination of environmental information by providing a watershed and modeling framework. Land-use percentages were calculated for each catchment area and combined into categories of agriculture, forest, urban, and wetland. In addition to GIS analysis, land use occurring within 30 m of the stream edge was calculated by qualitative habitat assessment. This included a visual inspection of differing habitat types occurring from the stream's edge up to 30 m of riparian zone.
Water samples were collected at each site during biological sampling to provide baseline water chemistry data. Ambient water temperature (°C), dissolved oxygen (mg l− 1), percent saturation (% sat), specific conductance (μ S cm− 1), and pH (standard units) were measured in situ with a Hydrolab® multiprobe. Total alkalinity (mg l− 1 as CaCO3) and total hardness (mg l− 1 as CaCO3) were measured in the field using buret titrations. Water samples were collected and transported to the laboratory on ice for analysis of nutrients and suspended solids and analyzed within 24 hours (APHA, 1992). Known standards, duplicates, and spikes were used for quality assurance purposes.
Macroinvertebrate and fish collection
Benthic macro-invertebrate sampling followed standard U.S. Environmental Protection Agency's Rapid Bioassessment Protocols for genus-level identification (Barbour et al., 1999). Macroinvertebrate sampling consisted of a multihabitat composite from each site. A D-frame net with 600-μ mesh was jabbed among different available habitat types (e.g., woody debris, aquatic vegetation, vegetated banks, undercut banks, and substrates) in proportion to their occurrence within the reach. Fish collection was done using a Smith-Root® battery powered backpack electrofisher to sample all available microhabitats within the stream reach. All fish were counted and identified to species.
Several biological indices were used to measure the difference in macro-invertebrate and fish community structure among sites. The total number of individuals (N), the total number of different species (S), Margalef's index (d), Pielou's evenness index (J′), and Shannon-Wiener diversity index H′ (log10) were calculated for each site. Further analysis of fish data included determining the number and diversity of sensitive species such as darters (Etheostoma sp. and Percina sp.) and madtoms (Noturus sp.) present at each site (Mettee et al., 1996). Also, the number of tolerant and intolerant species and abundance was determined for each site. For macroinvertebrate data, the Ephemeroptera, Plecoptera, and Trichoptera (EPT) index (Plafkin et al., 1989) was calculated at each site. Spearman's correlations were used to correlate univariate environmental variables to these biological indices.
Due to the diverse units of measurement between physical and chemical data, it was necessary to transform data to eliminate skewness and approximate normality. Differing transformations were used for habitat, land use, and water chemistry data to achieve normality. Chemical and land-use data were normalized using common log (log10) transformations and habitat data were normalized using fourth root transformations.
Because community patterns respond to abiotic gradients, multi-dimensional scaling (MDS) was used to group samples with similarity and to graph them in two dimensions. Secondly, environmental variables were reduced to factors that explained the most variation within the data set using factor analysis (principal component analysis with varimax rotation). Further analysis focused on correlating these environmental factors against biological community structure. For these analyses, environmental data were separated into categories of instream habitat, catchment area land use, and water chemistry. The matching of abiotic variables to macroinvertebrate and fish community structure was done using PRIMER's BIO-ENV program (Clarke, 1993). This procedure takes into consideration that often more than one environmental variable explains the structure of biological communities, similar to multiple regression. The similarity and dissimilarity among sites according to biotic and abiotic data are matched by Spearman's correlations. Each environmental category, instream habitat, land use, and water chemistry, was analyzed separately using this procedure. This helped to illustrate which environmental category had the greatest influence on macroinvertebrate and fish communities. Principal component scores for each environmental factor were also analyzed using the BIO-ENV procedure.
Correlations of environmental variables to biological indices
Results of univariate correlations of environmental variables to macroinvertebrate indices found significant but low correlations. Total suspended solids had a negative relationship with the number of taxa (S) (r = −0.316, p = 0.029), species diversity (H′) (r = −0.381, p = 0.008), and EPT (r = −0.288, p = 0.047). Turbidity also had negative correlations to both macroinvertebrate H′ (r = −0.359, p = 0.012) and evenness (J′) (r = −0.345, p = 0.016). Mean stream width had the highest correlations with (S) (r = 0.480, p = 0.001), richness (d) (r = 0.475, p = 0.001), H′ (r = 0.416, p = 0.003), and EPT (r = 0.566, p ≤ 0.001). Bank full depths of streams were correlated to richness (d) (r = 0.468, p = 0.001) and H′ (r = 0.455, p = 0.001). For land-use data, percent wetland occurring within catchment area was positively correlated to (S) (r = 0.401, p = 0.005) and richness (d) (r = 0.445, p = 0.002). The percent forest occurring within 30 m of the stream edge was positively correlated to (S) (r = 0.297, p = 0.040) and H′ (r = 0.289, p = 0.046).
Results of univariate correlations of environmental variables to fish indices showed that total suspended solids was negatively correlated to richness (d) (r = −0.337, p = 0.019) and the number of intolerant fish (NIT) (r = −0.301, p = 0.038). Dissolved oxygen concentration was positively correlated to number of individuals (N) (r = 0.343, p = 0.017). Mean stream width had the highest correlations to the number of darter and madtom species (r = 0.551, p ≤ 0.001) and H′ (r = 0.435, p = 0.002). Percent canopy cover was positively correlated to evenness (r = 0.428, p = 0.002). For land-use data, percent urban occurring within 30 m of the stream edge was negatively correlated with the number of darter and madtom species (r = −.328, p = 0.023).
Results of correlation of habitat indices, RBP and QHEI, with biological indices showed that the RBP was positively correlated to the number of macroinvertebrate taxa (r = 0.530, p ≤ 0.001), species richness (r = 0.550, p ≤ 0.001), species diversity H′ (r = 0.521, p ≤ 0.001), and EPT (r = 0.526, p < 0.001). The RBP was also correlated to fish diversity (r = 0.387, p = 0.007) and the number of darter and madtom species (r = 0.504, p ≤ 0.001). The QHEI was positively correlated to the number of individual fish (r = 0.373, p = 0.009), species diversity H′ (r = 0.300, p = 0.039), darter and madtom species (r = 0.317, p = 0.028), and the number of intolerant species (r = 0.389, p = 0.006).
Environmental variables reduction
Spearman's correlations were used to eliminate variables that were highly correlated to one another. Variables were eliminated using professional judgment if they had correlations higher than (r = 0.80, p ≤ 0.001). From the nineteen water chemistry variables, four variables were highly correlated, considered as redundant variables and eliminated from the data set. These variables included air temperature, total dissolved solids, percent dissolved oxygen saturation, and hardness. No correlations were found within the habitat and land-use data that constituted elimination of variables. The mean, median, and standard deviation of all remaining environmental variables for the 49 sites included in principal component analysis (PCA) are reported (Table 1).
Principal component analysis
Environmental data were analyzed by factor analysis (principal components with varimax rotation), which identified environmental variables that explained the greatest proportion of data set variability. Principal component (axes) scores for each site were used in further analyses. From the 15 chemical variables measured, components 1, 2, and 3 explained cumulatively 53% of the total variance and were composed of conductivity, pH, alkalinity, and chloride for “ChemFactor1,” reactive phosphorus, “ChemFactor2,” and ammonia and turbidity for “ChemFactor3” (Table 2).
Principal components 1, 2, and 3, for the nine instream physical habitat variables, explained 71% of the cumulative variance and were comprised of mean width, bank full width, and catchment area for “DrainageFactor,” percent canopy cover and large woody debris for “CoverFactor,” and percent sand for “SubstrateFactor” (Table 3). A PCA of land-use data derived three principal components that explained 76% of the cumulative data set variability. Component 1 was composed of percent urban within 30 m of the stream as well as percent urban within the catchment area for “UrbanFactor1.” Component 2 was comprised of primarily percent agriculture occurring within 30 m of the stream for “AgFactor2,” and component 3 was comprised of mostly percent forest within the catchment area for “ForestFactor3” (Table 4).
Principal components 1 and 2 explained the majority of variation within the data set, and scores from these axes were graphed to show clustering of sites for each environmental category. Results from graphing principal components of chemical data showed little variation among sites based on chemical parameters alone, and only a few sites stood out as having extreme differences from other sites (Figure 2). Sites with positive scores for both axes one and two showed clustering based on primarily higher suspended solids and low conductivity. Negative scores represented clusters based primarily on sites with high conductivity and low suspended solids.
Principal component axis 1 and 2 for instream habitat data showed more variation within the data set than did chemical data. Four main habitat clusters were revealed within the data set (Figure 3). Group one consisted primarily of sites containing large catchment areas, low sediment loadings, and marl bottom streams. Group two was comprised of sites that were very similar in stream width, catchment area, and sediment loadings and included 77% of sites sampled. Group three was made up of sites that have little or no riparian vegetation, canopy cover, or large woody debris. Group four consisted of two sites that had the smallest catchments and highest sediment levels.
The PCA results for land-use data revealed three primary clusters of forest, agriculture, and urban (Figure 4). Group one consisted of primarily urban sites and made up only 6% of the total sites sampled. Group two consisted of forested sites and made up 79% of the total sites sampled. Group three consisted of agricultural sites and made up 15% of the total sites sampled.
Correlations of PCA factors to biological indices
Factors one, two, and three for habitat, chemistry, and land use were correlated to macro-invertebrate and fish diversity indices using Spearman's correlation to determine those factors that have the highest relationship to community structure (Table 5). The “DrainageFactor” for instream habitat had the highest correlation to both macro-invertebrate and fish diversity. This factor, composed of stream width and catchment area, was correlated to the number of EPT taxa (r = 0.588, p ≤ 0.001) as well as to the number of darter and madtom species (r = 0.527, p ≤ 0.001). This was the only habitat factor used in PCA that had any correlations to biological indices. For the correlation of biological indices to chemical factors, factor three, composed of primarily ammonia and turbidity loadings, was negatively correlated to macroinvertebrate diversity H′ (r = −0.358, p = 0.012). This was the only correlation of water chemistry factors to biological indices. Land use factor two, which consisted primarily of percent agriculture occurring within 30 m of a stream edge, was correlated to sensitive EPT taxa (r = 0.452, p = 0.006). This was the only land-use factor that was correlated to any of the biological indices.
Multidimensional scaling of biological data
Because the RBP and QHEI habitat indices showed the highest correlations with macroinvertebrate and fish diversity, scores from these habitat indices were used along with species data in Multi-Dimensional Scaling (MDS). Sites were assigned a code of one through four based on lowest to highest RBP and QHEI scores. Using Bray-Curtis similarity of square root transformed fish and macroinvertebrate abundance data, MDS was used to illustrate that sites with similar habitat scores contain similar species community composition. Results of MDS showed very little grouping of biological data according to RBP and QHEI scores.
Results of the bio-environmental procedure for macro-invertebrate assemblages showed that specific variables in each environmental category consisting of instream habitat, land use, and water chemistry were individually correlated to species assemblage data (Figure 5). Results of BIO-ENV for macroinvertebrate assemblages showed that a combined set of water chemistry variables including total phosphorus, dissolved oxygen, water temperature, and chloride had the highest correlation to species data (r = 0.321). A combined set of instream habitat variables including stream width, riparian vegetation, and catchment area were correlated to macroinvertebrate assemblages (r = 0.231). Finally, only one land use variable, percent urban occurring within 30 m of the stream, was correlated to macroinvertebrate assemblages (r = 0.225).
In addition to using individual environmental variables within the BIO-ENV procedure, combinations of variables were analyzed by correlating PCA factor scores to macro-invertebrate abundance data. Results showed that a combination of agriculture occurring within 30 m of a stream, total phosphorus concentrations, and catchment area were related to macro-invertebrate assemblages (r = 0.279). This was slightly higher than individual categories of instream habitat and land use but was lower than water chemistry.
Results of BIO-ENV for fish assemblages (Figure 6) showed that a combined set of water chemistry variables including dissolved oxygen, total nitrogen, total solids, and turbidity had the highest correlation to species data (r = 0.393). A combined set of instream habitat variables including percent riparian vegetation and percent canopy cover were correlated to fish assemblages (r = 0.313). Finally, only one land-use variable, agriculture occurring within 30 m of the stream edge, was correlated to fish assemblages (r = 0.202). Combinations of these categories were analyzed by correlating PCA factor scores to fish abundance data. Of these, results showed that a combination of the ammonia and turbidity factor with the percent urban factor had the highest relationship to fish assemblages (r = 0.324).
Results of univariate correlations of environmental variables to biological indices suggest that macroinvertebrate and fish communities of the Choctawhatchee-Pea watershed may be responding to increasing levels of suspended sediments. Correlations suggest that macroinvertebrate diversity and sensitive species or EPT taxa numbers have decreased in areas that have high concentrations of suspended sediments. These results are supported by several studies that have found decreased macro-invertebrate diversity to be related to increased levels of suspended sediments (Stewart and Swinford, 1995; Richards et al., 1997). Fish assemblages of the Choctawhatchee-Pea watershed also responded negatively to increased levels of suspended sediments and turbidity, particularly with a fall in the number of intolerant species. Studies have shown similar results from other regions indicating that sedimentation decreases available spawning habitat, increases egg and larvae mortality, and can decrease feeding success of species that rely on visual search strategies (Berkman and Rabeni, 1987; Henley et al., 2000).
Vannote et al. (1980) proposed the river continuum concept to explain the general increase in biological productivity at mid-order streams due to flow regime, nutrient concentrations, and habitat complexity. Our study found that as catchment area and mean stream width increased, the number and diversity of macroinvertebrate and fish assemblages also increased. Sites sampled within these larger catchment areas generally contained less sediment and were comprised of mostly marl and hardpan substrates. The geomorphology of these larger streams appeared to be more conducive to the survival of sensitive species. Because the Choctawhatchee-Pea watershed study consisted of sampling wadeable streams only, the largest streams sampled were mid-order and were consistent with the river continuum concept.
Land-use data had lower correlations to macro-invertebrate and fish indices than did catchment area and stream width, but had higher correlations than instream water chemistry. Percent forest occurring within 30 m of the stream and percent wetland were positively correlated to macroinvertebrate diversity H′ suggesting that these sites had higher diversity than non-forested sites. Fish assemblages of the watershed were more responsive to percent urban occurring within 30 m of the stream edge than any other land-use category. This was best illustrated by the lack of sensitive darter and madtom species as well as intolerant species at urban sites. The negative effect of urban development on fish diversity has been well documented (Schleiger, 2000; Warren et al., 2000), and results from this study suggest that urban areas within the watershed have a reduced number of sensitive species. Due to the low number of urban streams tested in this study, this conclusion requires further testing in this watershed.
The positive relationship among habitat indices, RBP and QHEI, to both macro-invertebrate and fish indices showed that higher habitat scores generally had higher macro-invertebrate and fish diversity. These results followed initial watershed expectations in that the biological potential was expected to be higher in areas where physical habitat was better preserved. Studies using these habitat indices have found similar results, which support that assessing physical habitat along with biological sampling is essential for establishing overall watershed health (Barbour et al., 1999; Rabeni, 2000).
Principal component analysis of environmental variables showed that water chemistry samples taken throughout the watershed were largely similar among sites. There were two general clusters for chemistry data, which were largely separated by the presence or absence of increased suspended solids. Only three sites fell out of the two main clusters, one site having high phosphate loading and the other two sites having low nitrate. The site with unusually high phosphorus concentrations was the only site sampled below the discharge of a municipal wastewater treatment plant.
The correlation of PCA factors to macroinvertebrate and fish indices revealed complex interconnections that were not represented through univariate correlations. Correlation of the ammonia and turbidity factor to macroinvertebrate diversity showed that as these concentrations increased, species diversity decreased. The negative affects of ammonia and sedimentation on these communities have been documented in other regions (Henley et al., 2000). Because ammonia often enters streams attached to sediment particles, these variables combined had a significant influence on macro-invertebrate diversity (Fajen and Layzer, 1993). Correlation of habitat factors to biological indices supported univaritate correlations in that stream width and catchment area had a relationship to macroinvertebrate and fish diversity.
Correlation of land-use factors to macro-invertebrate indices gave different results than expected for the watershed. The correlation of sensitive EPT taxa to agricultural land use suggests that these species may actually increase in spite of poor instream habitat quality as measured by quantitative variable and qualitative indices (RBP, QHEI). Barbour and Stribling (1994) found similar results wherein streams subjected to organic enrichment created an artificial elevation in the biological condition of macroinvertebrate assemblages.
The BIO-ENV procedure for macroinvertebrates showed that the major response of these communities within the watershed was to instream water chemistry, more specifically total phosphorus, dissolved oxygen, total nitrogen, and chloride. Because the watershed lies within an agriculturally dominated region, total phosphorus and total nitrogen were expected to have a significant influence on these communities. Richards et al. (1993) found similar results when studying the predominant environmental factors structuring macroinvertebrate communities within an agricultural catchment in Michigan. They found that nitrogen and phosphorus compounds had the strongest influences on these communities, which occurred through various indirect pathways.
Using PCA factors in the BIO-ENV procedure revealed combinations of instream habitat, land use, and water chemistry that were related to macro-invertebrate and fish communities. Factors related to macro-invertebrate community structure included the agriculture factor, drainage factor, and phosphorus factor. This provides additional evidence that agricultural impacts and nutrient levels may be important structural forces for these communities.
The BIO-ENV procedure for fish assemblages provided results similar to that of macro-invertebrate assemblages. Comparison of the two generally showed that environmental variables had higher correlations for fish than for macroinvertebrates. This procedure suggested that water chemistry had the greatest relationship to fish assemblages, specifically turbidity, dissolved oxygen, total nitrogen, and total solids. These results supported univariate correlations, which suggested that increased levels of turbidity had negative effects on the number of sensitive fish taxa. In contrast to a comparison between macroinvertebrate and fish as biomonitors of stream quality in agricultural areas, Berkman and Rabeni (1987) found the response of sedimentation was more directly related to macro-invertebrates than fish assemblages. They found that fish communities were more affected by aspects of feeding and reproduction. Results of the correlation procedures using the PCA factors in the BIO-ENV procedure found that the ammonia and turbidity factor and the urban factor were related to fish data. Again, these results emphasize that turbidity along with ammonia had a significant influence on fish community structure. The instream habitat variables, percent canopy cover and riparian vegetation, had a lower relationship with fish assemblages than did the PCA factors and chemical data, yet these variables were considered to influence fish community structure more than any other habitat variable. Wang et al. (1997b) also found that riparian buffers and instream canopy cover had the greatest influence on fish assemblages of Wisconsin streams.
Results from this study suggest that increased turbidity and suspended sediments had negative effects on macroinvertebrate and fish community structure. Increased levels of sediments were related to habitat alteration through urban development and agricultural practices. Future management and restoration efforts should focus on watershed-level activities that reduce the amount of riparian degradation and sediment contamination. Quantifying the amount of disturbance that has occurred to this watershed is difficult without historical data. This study has provided baseline environmental and biological data with particular attention given to quantitative measurements of instream habitat and surrounding land use. This data can be used along with continuous monitoring to predict ecological recovery of the watershed.
The substrate characteristics of streams found to be generally unimpaired are determined largely by local geology and may vary significantly throughout catchments. Because a moderate amount of instream sediment and bank erosion are natural processes of the coastal plain ecoregion (Waters, 1995), restoration and protective efforts in this watershed should focus on restoring substrate conditions to that of non-impaired streams that lie within predominately forested areas.
Macroinvertebrate and fish assemblages of the Choctawhatchee-Pea watershed were similar in their response to environmental conditions. Correlations of habitat indices to biological data illustrated that the quality of instream habitat had a significant influence on both macroinvertebrate and fish assemblages. Based on correlations of univariate environmental variables to biological indices, macroinvertebrate and fish assemblages were significantly related to chemical variables such as suspended solids, nutrients, and dissolved oxygen. Habitat measurements that had the highest correlations to macroinvertebrate and fish assemblages consisted of stream width, catchment area, bankfull depth, and riparian vegetation. Analysis of land-use data showed that agricultural practices and urbanization occurring within 30 m of the stream had higher correlations to macroinvertebrate and fish community structure than catchment area land use. The BIO-ENV procedure showed that water chemistry showed the greatest relationship to macroinvertebrate and fish community structure followed by instream habitat and land use.
Regional qualitative and quantitative habitat sampling protocols, along with GIS analysis and periodical chemical monitoring, have become effective tools for predicting changes in biological communities caused by anthropogenic influences (Stewart et al., 2000). This watershed-level study specifically targeted sites within a fairly small drainage system compared to an entire ecoregion. Sampling of biological and environmental data showed that there was a great deal of homogeneity within the Choctawhatchee-Pea watershed, which made identifying different environmental influences on biological assemblages difficult and probably contributed to the overall relatively low correlations that we found.
Financial contribution were provided by ALFA Insurance Corporation, The Choctawhatchee, Pea, and Yellow Rivers Watershed Management Authority, Alabama ALERT grant through Troy State University, and the Department of Biological and Environmental Sciences at Troy State University.