Grantee Research Project Results
Final Report: Land Use and Geomorphic Indicators of Biotic Integrity in Piedmont Streams
EPA Grant Number: R826597Title: Land Use and Geomorphic Indicators of Biotic Integrity in Piedmont Streams
Investigators: Leigh, D. S. , Kramer, E. A. , Freeman, Mary C. , Rosemond, A. D. , Freeman, Byron J. , Pringle, Catherine M.
Institution: University of Georgia
EPA Project Officer: Packard, Benjamin H
Project Period: January 1, 1999 through December 31, 2001
Project Amount: $780,834
RFA: Ecological Indicators (1998) RFA Text | Recipients Lists
Research Category: Ecological Indicators/Assessment/Restoration , Aquatic Ecosystems
Objective:
The main objective of this research project was to evaluate the predictive capabilities of spatially and temporally variable land cover and geomorphic elements with respect to the biotic integrity and habitat quality of stream ecosystems. The study area included the Etowah River Basin (ERB) north of Atlanta, Georgia. The southern part of the ERB includes some of the most rapidly urbanizing counties in the nation, whereas the nonurban northern part of the basin is more pristine and exemplary of the exceptional biotic diversity and endemism that is characteristic of the Southern Appalachian Highlands. The ERB exhibits a mosaic of land cover types and historical changes that have influenced water and sediment inputs (stressors) over a wide spectrum of spatial and temporal scales.
Given these various aspects of landscape change, our research project investigated the following two main questions:
· How well can stream biota and habitat descriptors be predicted from measures of geomorphic and land cover conditions?
· What are the most important spatial and temporal aspects of geomorphology and land cover with respect to predicting stream ecosystem condition?
To answer these questions, we split our research into two main phases of data collection. The first phase involved a comprehensive survey of land cover, geomorphic, water quality, habitat, and biological (fish and macroinvertebrates) conditions in 31 wadeable streams that encompass a wide spectrum of environmental settings. We developed a suite of bivariate and multiple linear regression models to predict biotic and habitat descriptors using basinwide, riparian, and reach scales of predictor variables using these data. The second phase was a detailed analysis of site-specific geomorphic and biotic conditions in extended reaches of streams to evaluate within site variability. This analysis was designed to help us understand some of the unexplained variance found in the first phase of correlation and regression models. In addition, it allowed us to highlight the influence of local-scale heterogeneity in channel morphology, particularly where it overrides basinwide scales of influence on stream ecosystems.
Summary/Accomplishments (Outputs/Outcomes):
Our research approach involved the use of descriptors of stream habitat and biota (fish and macroinvertebrates), and the prediction of the descriptors using ordination, correlation, and multiple regression techniques. Stream habitat descriptors included scores from the U.S. Environmental Protection Agency (EPA) visual assessment protocol (Barbour, et al., 1999) and the U.S. Department of Agriculture (USDA) stream visual assessment protocol (Bjorkland, et al., 2001). Macroinvertebrate descriptors included Ephemeroptera, Plecoptera, and Trichoptera (EPT) richness; the Ohio EPA's (1989) Invertebrate Community Index (ICI); and axis scores derived from nonmetric multidimensional scaling (NMDS) of macroinvertebrate assemblages. Fish descriptors included a provisional version of an index of biotic integrity for the Coosa River Basin (Coosa IBI), an endemic to cosmopolitan ratio (E:E+C) characterizing homogenization of the fish assemblages (Walters, 2002), and axis scores derived from NMDS of fish assemblages. We collected hundreds of geomorphic and land cover variables at a variety of different scales and they were evaluated to derive predictors of the descriptors.
Our ultimate goal was to identify key elements of the landscape that can be used to model and predict stream habitat and biological conditions. Toward this end, we identified a short list of key "indicator" variables (see Table 1) that should be useful in other basins of the Piedmont and possibly in other physiographic provinces. Success in achieving our goal of identifying key ecological indicators is earmarked by robust linear regression models that commonly explain 60-90 percent of the variation in descriptors of stream habitat and biotic assemblages. We believe that our indicator models can be tested, modified, and used in other parts of the Appalachian Piedmont province, thereby serving as a way to quantitatively assess stream conditions, while minimizing the investment of time and labor. We found that multivariate models are more effective than bivariate models in terms of predicting the quality of stream ecosystems, but bivariate models have great value for determining the influence of individual variables.
We developed several categories of multivariate models, including: (1) models using only land cover data; (2) models using only geomorphic variables, showing equal or better predictability than land cover models; (3) Geographic Information Science (GIS) models that require no field data using only land cover plus topographic map data, providing the lowest cost remote sensing options for stream assessments; (4) models requiring minimal field data associated with GIS data providing intermediate cost options, which we refer to as "management models;" and (5) models using a comprehensive set of variables, providing the highest cost option. Although we do not expect the constants in these models to precisely apply to other basins, we do expect the identified variables and their relative magnitudes, as expressed by the constants, to be broadly applicable to other parts of the Piedmont. This summary presents the GIS and management models (see Tables 2 and 3) as examples, and the acronyms for these models are explained in Table 1.
The results of this study contribute to a body of literature showing correlations between landscapes and the quality of aquatic ecosystems (Allen, et al., 1996; Roth, et al., 1996; Richards, et al., 1996; Wang, et al., 1997; Kennan and Ayres, 2002), and the Etowah data are some of the first to characterize these relationships in the Southern Piedmont. Finally, this study provides insightful data about processes and functional relationships between landscape and biota that strongly support the Process Domains Concept (Montgomery 1999), which is an alternative to the River Continuum Concept (Vannote, et al., 1980). The Process Domains Concept (PDC) states that "process domains are spatially identifiable areas characterized by distinct geomorphic processes," and ecological conditions and biotic assemblages respond to unique settings influenced by the geomorphic template.
Key Indicators and Their Roles as Stressors to Stream Biota
We presented a list of variables considered strongly indicative of stream habitat and biotic conditions (see Table 1), particularly when used in multivariate models. Although this list presents individual variables that worked well in our forward stepwise regression models, those variables represent broader landscape components that can be discussed in the context of physical and chemical stress related to biotic integrity. If a landscape component (land cover, geomorphic, or water quality) negatively impacts habitat or biota because of human influence, then we consider it a stressor. Landscape components identified in our study include basinwide geomorphology, channel geomorphology, sedimentology, current land cover, land cover change, and water quality variables. These components vary considerably with respect to their status as anthropogenic stressors to stream ecosystems, as discussed below.
Basinwide Geomorphology
Basinwide geomorphology is a static variable over timescales of seconds to millions of years, which cannot be significantly influenced by humans. Thus, it cannot be considered a stressor to biota, but instead, it is an inherent template of the landscape that influences biotic assemblages. Basinwide geomorphic variables typically resulted as secondary or minor predictors in our models, but particularly were useful for improving the predictive capabilities of GIS models that relied solely on remotely sensed and map data. We recognize the basinwide geomorphic variables as important elements of the bedrock and topographic template that ultimately influence channel form and sedimentology.
We identified topographic ruggedness, local relief, and length of the trunk stream to be key indicators. Ruggedness and local relief exert strong influences on the localized morphology of the stream reach and physical processes operating within it. Rugged, high relief terrain is most conducive to a high frequency of riffles and shoals that tend to favor both high levels of habitat quality and habitat heterogeneity. Biotic assemblages in such streams tend to have high species richness as well as imperiled fish and macroinvertebrates that are positive indicators of biotic integrity.
Our finding that the length of the trunk stream is a key variable indicates that the size of the stream is an important factor, even within the relatively small size range of wadeable streams examined in the ERB (10-130 km2). Thus, size matters to stream ecosystems, and it should be factored into assessments of biotic integrity. This is no surprise in view of the widely accepted River Continuum Concept (Vannote, et al., 1980), but it is a significant finding within the rather narrow size range examined in our study. However, in the ERB, the stream size generally was found to be a minor indicator, which fell behind better indicators (channel morphology, sedimentology, land cover) that have little or no relationship with stream size.
Channel Morphology
Channel morphology is an indicator category that may or may not have much of a relationship to land use, depending on which variable is considered. We identified stream slope, channel entrenchment, variation in thalweg depth, average water depth, and riffle frequency as key indicators within the channel morphology category (see Table 1).
Stream slope (EGL) most likely is the single most important channel morphology variable, because it establishes the template for velocity, stream power, and tractive forces that shape the channel morphology. In addition, slope is the key determinant of the particle size composition on the streambed. Unlike textbook relationships that show stream slope steadily decreasing with basin size (Knighton, 1998), we found poor correlations between basin area and slope (r2 = 0.15). This lack of a correlation most likely is because of the relatively narrow size range of the basins sampled and is an indication of the overriding importance of localized topographic variation. It is not likely that land use has had much influence on channel slopes, because many of our sites have their slopes controlled by bedrock or they are in alluvial settings, where no evidence for historical changes in slope can be observed. An exception is that large changes in slope can be associated with direct modification of channels (i.e., channelization and straightening), which can steepen channel slopes, and a few of our sample sites most likely had been channelized several decades ago. In general, we do not view slope as a human-induced stressor, but find it to be a very important element of the physical template that influences stream communities, and it should be considered in evaluations of biotic integrity. We found that habitat and biotic assemblage characteristics in high gradient streams typically were of better quality than in low gradient streams. Thus, low slope streams may never achieve the same levels of quality as high slope streams, and slope appears to limit the capacity for habitat quality and biotic integrity as they currently are measured. We suggest that an alternative approach to the assessment of biotic quality may be to determine the maximum achievable levels of biotic quality as dictated by slope conditions, and to rank biotic quality within the range of values dictated by slope.
Our results indicated that steeper slopes correlated with better habitat and biotic conditions, largely because of coarser particle sizes and more extensive riffles. We also found that surveyed reach slope (EGL) was far superior to map slope (MAPSLOPE). In fact, map slope should only be used when surveyed slope cannot be obtained. Slope is very highly correlated with particle size composition of the streambed (i.e., r2 = 0.79). Thus, both slope and particle size cannot be used in the same predictive model.
Another channel morphology variable that has little relation to land use is the proportion of riffle space in a reach (%RIFFLEzz5). Riffle space mostly was explained by a positive correlation with surveyed slope, such that slope explained 66 percent of the variation in the proportion of riffles. Thus, the percentage of riffles is another variable that is difficult to describe as a stressor. Higher proportions of riffle space typically were correlated with better habitat and biotic integrity.
Other important elements of channel morphology (entrenchment, variation in thalweg depth, average water depth) revealed statistically significant influences of land use. Our stepwise regression models of elements influencing these variables all involved a significant component of land cover or land cover change, except in the case of thalweg depth heterogeneity (twegSTERR), where the land cover linkage is indirectly associated with particle size (cenPHI). This suggests that human alteration of the landscape has some effect on these variables through processes of erosion and sedimentation within the channels. However, the statistical inference is not particularly strong, indicating that other factors relate to these elements of channel morphology.
Sedimentology
Sedimentology, the particle size composition of the channel bed, was found to be an extremely influential variable group with respect to habitat and biota. Larger particle sizes correlated with better habitat and biotic conditions. Channel bed sediment can be considered a stressor, partly because we found that land cover variables explained additional variance in particle size variables beyond the primary correlations with slope. There is a very strong correlation between slope and particle size, suggesting that most sedimentology is determined by nonanthropogenic controls, but we found that land cover variables added significantly to explaining additional variance in particle size, as well. We found that the phi transformation (phi = -log2 mm) was the most suitable expression of average particle size rather than raw data or log10 transformations. The key sedimentology parameters that we identified included point counts of average particles size of the centerline of the streambed in phi units (cenPHI), and percent fines (percent <2 mm particles by weight) in riffles sieved from grab samples (percent FINESRIFFSV). Both of these variables revealed statistically significant correlations to recent land cover after the slope explained most of the variance. This suggests that current land cover and land cover changes within the last decade have influenced the particle size composition of stream beds to some extent. The models indicate that decreases in forest cover or increases in urban land lead to an increase in fine sediment, most likely because of an increase in sediment yield from soil erosion. Increases in fine sediment negatively influence biota and their habitat because the fines tend to fill interparticle spaces (embeddedness) that would otherwise provide habitat for fish and macroinvertebrates. Also, we found that coarse sediment (gravel) is much more likely to result in well-developed riffle and pool morphology than in fine sandy sediment, thus facilitating a wider diversity of morphological units and habitat diversity within a stream reach.
Although fine particles tend to be detrimental to biota (Waters, 1997), we concluded that some streams are inherently prone to fine-textured beds because of the control that slope exerts on the ability to transport coarse sediment. Low-slope streams (0.001-0.003 m/m) most likely will never become gravel bed streams, because they will remain as low-energy channels incapable of transporting gravel. Thus, slope exerts inherent controls on sedimentology that, in turn, dictates limits to biotic integrity. Our multiple regression models indicated that urban land cover significantly correlated with smaller average particle sizes in streams. However, only a small part of variation in sedimentology can be viewed as stressor-related, because the dominant influence is stream slope. The Etowah River and Lake Allatoona serve as the local base levels for streams in the basin, and it is very unlikely that reach-scale slopes will change enough to greatly transform stream sedimentology.
In addition to particle size variables, we developed sediment mobility ratios (i.e., Vb/Vc0.5), which combined particle size composition with channel morphology to express the tendency for channel beds to be mobilized (disturbed) during floods. The mobility ratios are multimeteric variables that integrate flood discharge, slope, channel width, and particle size into one variable. They are excellent predictors of biotic integrity that allow assessment of the magnitude and frequency of channel bed disturbance (bed sediment transport) that alters habitats and negatively impacts stream fauna. A drawback of the sediment mobility variables is that they require intensive field survey and data collection.
Current and Past Land Cover
We found that current land cover was a very significant predictor of habitat and biota, and was somewhat more related to habitat and macroinvertebrates than to fish. In fact, we were able to predict 80 percent of the variation in the EPA's rapid habitat assessment score (EPAHAB) using only two variables: percent woodland cover and percent water impounded (artificially) in the riparian zone (WATRIP). The fact that habitat and biota are responsive to land cover is not surprising, given the findings of previous studies (Allen, et al., 1996; Roth, et al., 1996; Richards, et al., 1996; Wang, et al., 1997; Kennan and Ayres, 2002), but we found that the reliability of prediction can be improved by incorporating geomorphic variables to the predictive models. Also, we found impoundments (ponds and reservoirs) to be excellent indicators of habitat and biota. We view the impoundments as proxies for many sorts of stresses to aquatic ecosystems, because they represent signs of agricultural and urban development, they typically are associated with livestock within and close to the stream, and they directly affect water temperature, chemical conditions, and connectivity of stream systems. We found impoundments to be one of the easiest land cover indicators to measure, because water has a very distinctive signature on Landsat images and thus exhibits high levels of accuracy and reproducibility.
Current land cover clearly is a stressor, because it significantly is correlated to water quality, habitat, and geomorphic variables that negatively influence biota. We typically found that current land cover variables were better, or at least equal, predictors of habitat and biota compared to antecedent land cover of 1938, 1973, and 1987. Furthermore, our forward stepwise regression models typically selected current land cover variables preferentially over land use change variables. This indicates that while antecedent land use may be important with respect to biotic integrity of streams (Harding, et al., 1998) the present land cover is the most suitable indicator with respect to aquatic habitat and biota. We found it to be very difficult, if not impossible, to determine the exact influence of antecedent land cover on biota, because present land cover patterns typically overlaid and paralleled past land use patterns with respect to their status as stressors. For example, gently sloping land that was in cropland in 1938 may have been reforested or converted to pasture by the 1960s, but in the latter part of the 1980s, the same gently sloping land was the preferred site for commercial and residential urban development. Another problem that we encountered with past land cover, particularly with the 1938 data, was that it could not be normalized for parametric statistical analysis.
Land cover variables that we found to be most indicative of habitat and biota for inclusion on our short list (see Table 1) were the 1998 woodland cover (98WOODS), 1997 low-density urban cover (97LDU), 1997 low-density urban cover along the stream network's riparian zone (LDURIP), open water along the stream network's riparian zone (WATRIP), 1997 forest cover along the stream network's riparian zone (FRIP), and mixed forest in the riparian zone along a 1-km reach (MF1KM). This list indicates that adequate assessment of the land cover as a stressor can be made by simply measuring forest cover, low-density urban cover, and impoundments (ponds) along the stream network. Without human influences, the ERB would be almost 100 percent forest; the loss of forest cover is an appropriate indicator of human impact in the basin. Low-density urban cover is a fitting indicator for the Etowah sites because the suburban sprawl north of Atlanta currently is the most widespread type of land cover conversion in the basin.
Land Cover Change
We found that the 1987-1997 land cover change variables were the easiest to work with, and generally were the most predictive of all the land cover change variables ('38-'97, '73-'97, '87-'97). This partly is because the 1987 and 1997 land cover variables were the most compatible in terms of the collection and classification methods used. Like the current land cover, land cover change variables clearly are stressors that influence changes in habitat and biotic conditions. The fact that many of our models selected 1987-1997 land cover change variables illustrates the effectiveness of those data as indicators, and again illustrates that much of the stress to aquatic habitat can be attributed to recent changes in land cover conditions. The only land cover change variable that we selected for the short list (see Table 1) is the 1987-1997 change in low density urban land cover (8797LDU), which is the variable reflecting the most widespread type of land cover change occurring in the ERB. The fact that we find this variable to be highly correlated to habitat and biota indicates that recent land cover change is an important stressor to aquatic ecosystems, and that degradation of stream conditions cannot be entirely attributed to past impacts from intensive agriculture that occurred circa 1880-1930.
Water Quality
Although the primary goal of our research project was to develop landscape indicators of biotic integrity, we measured water quality conditions to try to identify some of the linkages between land cover, water quality, and stream ecosystems. Land cover conditions correlated with water quality, which validates land cover as a stressor to biota, but the correlations were not particularly strong. This suggests that other factors influenced by land cover (e.g., stream sedimentology) also are important stressors to biota, and that water chemistry is not the only direct stressor.
We identified three water quality parameters on our short list (see Table 1) that relatively are easy to collect and contribute significantly to prediction of habitat and biota, including: dissolved oxygen (DO), specific conductivity (SC), and turbidity (NTU6GEOMEAN). All of these can be measured with field-based sensors that commonly are used in rapid assessments of water quality. In addition, soluble reactive phosphorus (SRP) and ammonium (NH4) were significant stressors to biota. These chemistry variables (SRP and NH4) are more expensive and labor intensive to collect than DO, NTU, and SC, but they are not superior indicators of stream biotic integrity.
Spatial Scales of Indicators
Our approach to identifying ecological indicators involved several different scales of analysis, including: (1) basinwide land cover and geomorphology; (2) whole-stream network riparian land cover within 100 m buffers; (3) 1 km segments of riparian land cover within 100 m buffers; (4) reach scales at 20-times stream width; and (5) extended reaches at about 100-times stream width. Our results indicate that it is desirable to have data at all of these scales to maximize predictive capabilities, and that strong models can be developed using basinwide and stream network riparian land cover combined with reach-scale geomorphic variables. Even our models with only land cover variables (see Table 3) indicate that the best predictions are made by combining both basinwide and riparian measures of land cover.
We found that the basinwide and whole-stream network riparian scales of analysis equally were favorable. In fact, our "short list" (see Table 1) of key indicator variables includes all three land-cover scales. However, we generally consider the 1 km riparian scale to be the weakest of the three land-cover scales. The whole-stream network scale of riparian land cover was a particularly good method to measure artificial impoundments in the stream network (WATRIP), and we strongly advocate using ponds along the stream network as an indicator of human impacts.
Geomorphically, both the basinwide and reach scales were both found to be important. The reach scale is generally more predictive of ecosystem quality, but it had dependencies on the basinwide scale with respect to lithology and topography. In the absence of field data, we strongly advocate the use of basinwide geomorphic variables from maps and digital elevation models (DEMs) to improve predictions associated with land cover data. In general, the geomorphic variables (at any scale) allow inherent elements of the landscape to be factored into quality assessments, and reveal the importance of lithology, topography, channel form, and sedimentology in structuring the template for stream ecosystems. In fact, models that we developed revealed that prediction of biotic assemblages solely from geomorphic variables is essentially as good as predictions developed from combined data sets of land cover, water quality, and geomorphology.
Our analysis of the extended-reach scale (100-times channel width) solidified the fundamental importance of channel morphology and sedimentology as inherent, and often nonanthropogenic, elements of the stream that exert tremendous influence on the capacity for biotic integrity. Within the extended reaches, the overarching land cover and basinwide geomorphic conditions were constant. However, we observed pronounced shifts in the quality of both fish and macroinvertebrates in response to very localized changes in habitat and geomorphic conditions. This observation, and previous stated observations related to the geomorphic template, led us to conclude that the Process Domains Concept (Montgomery, 1999) is an excellent conceptual framework for understanding landscape dependencies of aquatic ecosystems. Thus, our extended-reach analysis underscores the importance of incorporating geomorphic variables into assessments of ecosystem quality.
The Process Domains Concept (PDC)
The main tenet of the Process Domains Concept (Montgomery, 1999) is that spatial variability in geomorphic processes governs stream habitat and disturbance regimes that influence ecosystem structure and dynamics. Montgomery (1999) proposed the PDC as an alternative to the River Continuum Concept (Vannote, et al., 1980) and patch dynamics models (Pringle, et al., 1988), because neither of those models explicitly addresses the spatial structure of geomorphic controls on physical stream attributes. Process domains are predictable areas of the landscape, within which distinct geomorphic processes operate and thereby impart spatial variability to lotic communities at landscape scales. Montgomery (1999) supported the PDC with published studies of riparian plant, macroinvertebrate, and fish communities, but noted that few data existed to directly test the model. The PDC has received little attention from stream ecologists, and to our knowledge had not been objectively evaluated with stream community data until our study in the ERB. We believe that the Etowah data strongly supports the PDC, because much of the local variation we observed in geomorphic conditions strongly is related to biotic assemblages, yet is not related to basin size.
In conclusion, it is appropriate to recall our two general research questions concerning: (1) how well stream habitat and biota can be predicted from geomorphic and land cover conditions; and (2) the most important spatial and temporal aspects with respect to the predictions of stream habitats and biota. Our results clearly indicate that strong predictions (r2 of 50-70 percent) of stream ecosystems can be made with multivariate models of either land cover or geomorphic variables, but that the best models (r2 of 70-90 percent) involve a combination of both land cover and geomorphology to capture local "process domains" that structure stream ecosystems. Spatially, we found that the best predictions typically involve a combination of multiple scales of data collection (stream reach, drainage network, watershed), and if possible, other researchers and managers should try to collect data at multiple scales to produce the best predictive models. Temporally, we found that the most recent land cover data was indicative of stream ecosystems and that modern land use appears to be functionally related to ecosystem quality. Recent land cover change (last 10-15 years), particularly the recent conversion of forest and agricultural land to urban land, also was a strong indicator of stream quality. In contrast, we found the older land cover data sets (1938, 1973) to be much more difficult to work with, because of problems with methods of data collection and statistical normality of the data. There is no doubt that antecedent land use influences present stream conditions, but the functional relationships and strategies that account for the relative importance of antecedent land use remains elusive. However, from a predictive standpoint, we found that antecedent land cover data were not needed to produce robust models of stream ecosystems. In fact, modern land cover was often the best predictor of stream conditions.
One may question the regional applicability of our models, given that they were derived for a single river basin, but we specifically chose the ERB because it contained a wide range of land cover characteristics and a wide range of topographic variation. Thus, we believe that a relatively reasonable level of regional applicability should exist. Tests of our models in other regions will be the only way to validate their general applicability.
| Variables | Definition | Transform | Descriptors Predicted and Comments |
| Geomorphology | |||
| RUGGTR | ruggedness of the trunk stream basin (drainage density x relief of the trunk stream) | log(10) | Fish. GIS-based measure of relief. Basins with steep terrain have steep, coarse-textured channel beds and tend to have more forest cover. |
| LOCREL | local relief (m) | log(10) | Fish and macroinvertebrates. GIS-based measure of relief adjacent to stream reaches. High local relief indicates steep stream reaches and more forested riparian zones. |
| LENTR | length of the trunk stream (km) | log(10) | Fish and macroinvertebrates. GIS-based measure of stream size. Better predictor than basin area, a commonly used scale variable. |
| BKF/UQ2* | entrenchment ratio channel-full flood to urbanized 2-year RI flood(cms/cms) | log(10) | Fish. Requires topographic surveys of stream cross-sections, regional flood frequency data, and modeling software. Deeply entrenched streams experience high tractive force during large floods (another predictor of fishes), and most have coarse-textured beds. |
| EGL* | surveyed stream slope or Energy Grade Line measured from riffle top to riffle top (m/m) | log(10) | Fish and macroinvertebrates. Determines key elements of stream habitat and stream power. Requires topographic survey. Important driver of bed texture, riffle habitat, tractive force and other variables that predict stream assemblages and habitat quality and distinguish reaches. |
| MAPSLOPE | map slope (m/m) | log(10) | Fish and macroinvertebrates. GIS-based surrogate for slope measured it in the field. |
| cenPHI | average particle size of the channel centerline (average phi size of 34 observations along the centerline of the stream) | none | Fish and habitat. Easily collected bed texture data that is a strong predictor of stream assemblages. Particle size is a key element of bed mobility and depth heterogeneity. |
| %FINESRIFFSV | Percent fines in riffles (percent sieved by weight averaged from grab samples from three riffles) | Asin-Sqrt | Macroinvertebrates and habitat. Easily collected, quantitative measure of riffle embeddedness and one of the best overall predictors of macroinvertebrates. |
| Vb/Vc0.5* | velocity-based bed sediment mobility index (ratio of the average bed velocity during a 0.5-year flood divided by the actual velocity needed to move average particle size) | none | Fish and habitat. Requires topographic survey and modeling software. Mobility is highly dependent on texture and slope. Low slope, sand textured streams have high mobility and are frequently disturbed by small floods. |
| twegSERR* | standard error of thalweg elevations (standard deviation of surveyed elevations) | log(10) | Fish and habitat. Measure of depth heterogeneity that requires survey of bed elevation. |
| DMIDavg | average depth of the centerline of the stream (mean of 34 centerline observations) | log(10) | Macroinvertebrates. Simple measure of depth from centerline survey. Highest in sites with repeating riffle-pool morphology and lowest in streams dominated by shallow run habitat. |
| %RIFFLEzz5 | Percent of riffle space in the channel (percent of channel bed area) | Asin-Sqrt | Fish. Can be easily measured in centerline survey. Many sensitive, imperiled, and endemic fishes are riffle habitat specialists. |
Table 1. Continued.
| Variables | Definition | Transform | Descriptors Predicted and Comments |
| Land Cover | |||
| 98WOODS | 1998 supervised woodland cover (percent of basin area) | Asin-Sqrt | Fish and habitat. Easily measured from Landsat TM images available online. Indicator of total catchment disturbance. |
| 98WOOD10 | 1998 unsupervised woodland cover (percent of basin area) | Asin-Sqrt | Fish and habitat. Easily measured from Landsat TM images available online. Indicator of total catchment disturbance. |
| 97LDU | 1997 low-density urban cover (percent of basin area) | Asin-Sqrt | Macroinvertebrates. Low-density urban cover is the primary cover in suburban areas. |
| LDURIP | 1997 low-density urban cover in the riparian zone of the 1:100,000 stream network (percent of 100 m buffer) | Asin-Sqrt | Macroinvertebrates and fish. Indicator of development intensity in riparian zones. |
| WATRIP | 1997 open water cover in the riparian zone of the stream network (percent of 100 m buffer) | Asin-Sqrt | Fish, macroinvertebrates, and habitat. Indicator of catchment disturbance and stream network fragmentation resulting from artificial ponds and lakes. Open water easily is classified, so estimates have high accuracy. |
| FRIP | 1997 forest cover in the riparian zone of the 1:100,000 stream network (percent of 100 m buffer) | Asin-Sqrt | Habitat and water quality. Effective indicator of riparian condition. |
| MF1KM | 1997 mixed forest in the riparian zone extending 1 km upstream of the reach (percent of 100 m buffer) | Asin-Sqrt Fish. | Effective indicator of local riparian condition. |
| 8797LDU | basin land cover converted to low density urban from 1987-1997 (percent of basin area) | none | Fish and bed texture. Effective indicator of the pace of development and disturbance in urbanizing areas. |
| Water Quality | |||
| NTU6GEOM | baseflow turbidity (geometric mean of ntu) | log10 | Fish. Easily collected field data and effective indicator of ongoing, chronic disturbance. |
| SC | specific conductivity (mean) | none | Macroinvertebrates. Easily collected field data and good indicator of chronic, nonpoint source pollution from urbanized areas. |
| DO | dissolved oxygen (mean) | none | Fish. Related to steep, turbulent streams and riparian condition. |
| SRP* | soluble reactive phosphorus (mean) | none | Macroinvertebrates. Requires water sample collection and laboratory analysis. Indicator of eutrophication and nonpoint source pollution. |
| NH4* | ammonium (mean) | log10 | Fish, macroinvertebrates, and habitat. Requires water sample collection and laboratory analysis. Indicator of eutrophication and nonpoint source pollution and the best overall water quality variable for predicting across categories of descriptor variables. |
| Note: * Indicates variables that require extensive field and/or laboratory work. | |||
Table 2. Multiple Linear Regression Models Using Only GIS-Based Variables of Land Cover and Basin Morphometry. Acronyms are explained in Table 1. These models do not require field data.
| Descriptor | Parameters | Estimates | p value | Cumulative R2 | F ratio |
| ICI9 | Intercept | 100.139 | <0.0001 | - | 23.23 |
| TLDURIP | -116.839 | <0.0001 | 0.57 | ||
| TWATRIP | -68.482 | 0.0584 | 0.62 | ||
| EPT | Intercept | 31.921 | 0.0007 | - | 11.91 |
| T97U | -24.925 | 0.0003 | 0.42 | ||
| TLENTR | 12.495 | 0.005 | 0.50 | ||
| TMAPSLOPE | 8.038 | 0.0533 | 0.57 | ||
| invert Axis1* | Intercept | -4.111 | <0.0001 | - | 37.58 |
| T97LDU | 5.045 | <0.0001 | 0.68 | ||
| TMAPSLOPE | -1.399 | 0.0006 | 0.72 | ||
| TLENTR | -1.223 | 0.0021 | 0.81 | ||
| invert Axis2* | Intercept | -3.113 | 0.0013 | - | 19.76 |
| TMAPSLOPE | -1.398 | 0.0001 | 0.45 | ||
| TMFRIP | -1.380 | 0.0051 | 0.59 | ||
| Coosa IBI | Intercept | -32.167 | 0.0040 | - | 20.45 |
| T98WOOD10 | 44.774 | <0.0001 | 0.48 | ||
| TLENTR | 16.285 | 0.0013 | 0.60 | ||
| TMF1KM | 12.204 | 0.0081 | 0.69 | ||
| E/E+C | Intercept | -0.102 | 0.9629 | - | 18.24 |
| T98WOODS | 4.678 | 0.0176 | 0.42 | ||
| TWATRIP | -10.360 | 0.0008 | 0.62 | ||
| 8797LDU | -0.095 | 0.0480 | 0.67 | ||
| E/E+C | Intercept | -0.907 | 0.6352 | - | 22.20 |
| T98WOOD10 | 5.271 | 0.0024 | 0.41 | ||
| TWATRIP | -9.457 | 0.0011 | 0.59 | ||
| 8797LDU | -0.125 | 0.0025 | 0.71 | ||
| Fish Axis2 | Intercept | -4.877 | <0.0001 | - | 19.91 |
| T98WOODS | 3.361 | <0.0001 | 0.49 | ||
| TMF1KM | 1.048 | 0.0037 | 0.59 | ||
| TLENTR | 1.009 | 0.0064 | 0.69 | ||
| Fish Axis2 | Intercept | -4.930 | <0.0001 | - | 18.54 |
| T98WOOD10 | 3.366 | 0.0001 | 0.48 | ||
| TMF1KM | 1.124 | 0.0023 | 0.60 | ||
| TLENTR | 0.891 | 0.0181 | 0.67 | ||
| Fish Axis3 | Intercept | -2.435 | 0.0040 | - | 23.06 |
| TMAPSLOPE | -0.646 | 0.0699 | 0.49 | ||
| 8797U | 0.062 | 0.0001 | 0.63 | ||
| DA | 0.006 | 0.0075 | 0.72 | ||
| SVAP^2 | Intercept | -5.135 | 0.8346 | - | 29.36 |
| TFRIP | 72.809 | 0.0009 | 0.57 | ||
| TWATRIP | -137.939 | 0.0058 | 0.68 | ||
| EPAHAB | Intercept | -2.443 | 0.4800 | - | 56.07 |
| T98WOODS | 18.932 | <0.0001 | 0.54 | ||
| TWATRIP | -31.860 | <0.0001 | 0.80 | ||
| EPAHAB | Intercept | 0.914 | 0.8455 | - | 30.40 |
| T98WOOD10 | 15.214 | 0.0006 | 0.40 | ||
| TWATRIP | -33.723 | <0.0001 | 0.68 | ||
| Note: - * High positive values indicate low quality, which is the inverse of other descriptors. |
|||||
| - "T" at the beginning of the acronym indicates that the variable is numerically transformed. | |||||
Table 3. Multiple Linear Regression Models Based on GIS and Easily Collected
Field Data, Considered as "Management Models." These models were created
with and without water quality variables in the data used in the forward stepwise
selection process to illustrate the additional prediction offered by water quality
variables.
| Descriptor | Parameters | Estimates | p value | Cumulative R2 | F ratio |
| ICI9 | Intercept | 52.588 | 0.0155 | - | 30.36 |
| TLDURIP | -91.714 | <0.0001 | 0.57 | ||
| T%FINESRIFFSV | -22.660 | 0.0035 | 0.72 | ||
| TDMIDavg | 29.396 | 0.0218 | 0.77 | ||
| EPT | Intercept | -2.098 | 0.8262 | - | 12.87 |
| T97LDU | -30.541 | 0.0003 | 0.40 | ||
| TDMIDavg | 14.159 | 0.0160 | 0.52 | ||
| TLENTR | 7.146 | 0.0509 | 0.59 | ||
| EPT | Intercept | -13.673 | 0.1161 | - | 23.63 |
| SC* | -0.118 | <0.0001 | 0.49 | ||
| TLOCREL | 16.853 | 0.0003 | 0.67 | ||
| TLENTR | 6.285 | 0.0343 | 0.72 | ||
| invert Axis1* | Intercept | -0.106 | 0.9125 | - | 37.92 |
| T97LDU | 5.494 | <0.0001 | 0.68 | ||
| TDMIDavg | -1.327 | 0.0269 | 0.73 | ||
| invert Axis2* | Intercept | -3.286 | <0.0001 | - | 49.08 |
| T%FINESRIFFSV | 1.426 | <0.0001 | 0.66 | ||
| TMAPSLOPE | -0.947 | 0.0006 | 0.78 | ||
| Coosa IBI | Intercept | -45.876 | 0.0015 | - | 14.75 |
| T98WOODS | 36.059 | 0.0022 | 0.37 | ||
| TLENTR | 16.271 | 0.0030 | 0.54 | ||
| TDMIDavg | 19.855 | 0.0216 | 0.62 | ||
| CoosaIBI | Intercept | 6.457 | 0.6215 | - | 19.79 |
| TNTU6GEOM* | -23.870 | 0.0001 | 0.44 | ||
| TDMIDavg | 23.697 | 0.0021 | 0.59 | ||
| TLENTR | 13.525 | 0.0066 | 0.69 | ||
| E/E+C | Intercept | -3.707 | 0.0494 | - | 26.73 |
| CenPHI | -0.389 | 0.0002 | 0.56 | ||
| TLOCREL | 3.285 | 0.0019 | 0.69 | ||
| TWATRIP | -6.524 | 0.0205 | 0.75 | ||
| E/E+C | Intercept | 1.319 | 0.3073 | - | 59.40 |
| TNTU6GEOM* | -4.345 | <0.0001 | 0.67 | ||
| CenPHI | -0.308 | <0.0001 | 0.82 | ||
| TDMIDavg | 2.349 | 0.0041 | 0.87 | ||
| Fish Axis2 | Intercept | -4.419 | <0.0001 | - | 37.97 |
| T98WOODS | 1.852 | 0.0089 | 0.49 | ||
| cenPHI | -0.231 | <0.0001 | 0.59 | ||
| TLENTR | 1.608 | <0.0001 | 0.81 | ||
| Fish Axis2 | Intercept | -10.080 | <0.0001 | - | 23.26 |
| T98WOODS | 2.730 | 0.0009 | 0.49 | ||
| DO* | 0.730 | 0.0008 | 0.63 | ||
| TLENTR | 0.936 | 0.0072 | 0.72 | ||
| Fish Axis3 | Intercept | -3.835 | <0.0001 | - | 17.52 |
| TMAPSLOPE | -0.879 | 0.0195 | 0.49 | ||
| TLDURIP | 2.520 | 0.0031 | 0.60 | ||
| TLENTR | 0.747 | 0.0400 | 0.66 | ||
| SVAP^2 | Intercept | -55.093 | 0.0006 | - | 35.88 |
| TFRIP | 85.748 | <0.0001 | 0.57 | ||
| cenPHI | -4.660 | 0.0007 | 0.72 | ||
| EPAHAB | Intercept | 7.298 | 0.0341 | - | 74.16 |
| T%FINESRIFFSV | -5.637 | <0.0001 | 0.64 | ||
| TWATRIP | -25.955 | <0.0001 | 0.82 | ||
| T98WOODS | 11.633 | 0.0002 | 0.89 | ||
| Note: - * High positive values indicate low quality, which is the inverse of other descriptors. |
|||||
| - "T" at the beginning of the acronym indicates that the variable is numerically transformed. | |||||
References:
Allan JD, Erickson DL, Fay J. The influence of catchment land use on stream integrity across multiple spatial scales. Freshwater Biology 1997;37(1):149-161.
Barbour MT, Gerritsen J, Snyder BD, Stribling JB. Rapid bioassessment protocols for use in wadeable streams and rivers: periphyton, benthic macroinvertebrates, and fish. Second Edition. U.S. Environmental Protection Agency, Office of Water, Washington, DC, 1999.
Bjorkland R, Pringle CM, Newton B. A stream visual assessment protocol (SVAP) for riparian landowners. Environmental Monitoring and Assessment 2001;68(2):99-125.
Harding JS, Benfield EF, Bolstad PV, Helfman GS, Jones EBD. Stream biodiversity: the ghost of land use past. Proceedings of the National Academy of Sciences of the United States of America 1998;95(25):14843-14847.
Kennen JG, Ayers MA. Relation of environmental characterstics to the composition of aquatic assemblages along a gradient of urban land use in New Jersey, 1996-98. Presented at the U.S. Geological Survey Water Resources Investigations, Denver, CO, 2002, 77 pp.
Knighton D. Fluvial forms and processes: a new perspective. London: Edward Arnold, 1998, 383 pp.
Montgomery DR. Process domains and the river continuum. Journal of the American Water Resources Association 1999;35(2):397-410.
Ohio Environmental Protection Agency (Ohio EPA). Biological criteria for the protection of aquatic life. Vol I-III, Surface Water Section, Division of Water Quality Monitoring and Assessment, Ohio Environmental Protection Agency, Columbus, Ohio, 1989.
Pringle CM, Naiman RJ, Bretschko G, Karr JR, Oswood MW, Webster JR, Welcomme RL, Winterbourn MJ. Patch dymanics in lotic systems-the stream as a mosaic. Journal of the North American Benthological Society 1988;7:503-524.
Richards C, Johnson LB, Host GE. Landscape-scale influences on stream habitats and biota. Canadian Journal of Fisheries and Aquatic Sciences 1996;53(Suppl 1):295-311.
Roth NE, Allan JD, Erickson DL. Landscape influences on stream biotic integrity assessed at multiple spatial scales. Landscape Ecology 1996;11(3):141-156.
Vannote RL, Minshall GW, Cummins KW, Sedell JR, Cushing CE. River continuum concept. Canadian Journal of Fisheries and Aquatic Sciences 1980;37(1):130-137.
Walters DM. Influence of geomorphology and urban land cover on fish assemblages in the Etowah River system, Georgia. Ph.D. Dissertation. The University of Georgia, Athens, GA, 2002, 221 pp.
Waters TF. Sediment in streams: sources, biological effects, and control. Presented at the American Fisheries Society, Bethesda, MD, 1997.
Wang LZ, Lyons J, Kanehl P, Gatti R. Influences of watershed land use on habitat quality and biotic integrity in Wisconsin streams. Fisheries 1997;22(6):6-12.
Waters TF. Sediment in streams: sources, biological effects, and control. American Fisheries Society, Bethesda, MD, 1997.
Journal Articles on this Report : 3 Displayed | Download in RIS Format
| Other project views: | All 55 publications | 4 publications in selected types | All 4 journal articles |
|---|
| Type | Citation | ||
|---|---|---|---|
|
|
Roy AH, Rosemond AD, Leigh DS, Paul MJ, Wallace JB. Habitat-specific responses of stream insects to land cover disturbance: biological consequences and monitoring implications. Journal of the North American Benthological Society 2003;22(2):292-307. |
R826597 (2001) R826597 (Final) |
|
|
|
Walters DM, Leigh DS, Bearden AB. Urbanization, sedimentation, and the homogenization of fish assemblages in the Etowah River Basin, USA. Hydrobiologia 2003;494(1-3):5-10. |
R826597 (Final) |
Exit Exit |
|
|
Walters DM, Leigh DS, Freeman MC, Freeman BJ, Pringle CM. Geomorphology and fish assemblages in a Piedmont river basin, U.S.A. Freshwater Biology 2003;48(11):1950-1970. |
R826597 (2001) R826597 (Final) |
Exit |
Supplemental Keywords:
ecology, geography, fluvial geomorphology, watersheds, ecological effects, ecosystem, indicators, aquatic, habitat, Landsat, remote sensing, southeast, EPA Region 4, Appalachian., RFA, Scientific Discipline, Water, Geographic Area, Ecosystem Protection/Environmental Exposure & Risk, Water & Watershed, Ecology, Ecosystem/Assessment/Indicators, Ecosystem Protection, exploratory research environmental biology, Chemical Mixtures - Environmental Exposure & Risk, Environmental Chemistry, Ecological Effects - Environmental Exposure & Risk, Monitoring/Modeling, Ecological Effects - Human Health, Ecological Risk Assessment, Biology, Geology, Watersheds, Ecological Indicators, EPA Region, risk assessment, aquatic biota , landscape indicator, biodiversity, land use effects, stream ecosystems, ecosystem indicators, geomorphic indicators, soil, aquatic ecosystems, water quality, GIS, land useProgress and Final Reports:
Original AbstractThe perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.