Results - Cleveland State University



Biological Integrity in Urban Streams: Toward Resolving Multiple Dimensions of Urbanization

B. Michael Walton*a,

Mark Sallingb,

James Wylesb,

Julie Wolina,

aDepartment of Biological, Geological, and Environmental Sciences,

Cleveland State University, Cleveland OH 44115;

bNortheast Ohio Data and Information Center, Levin College of Urban Affairs, Cleveland State University, Cleveland, OH 44115;

November 19, 2004

*Corresponding author

Running head: Biological Integrity in Urban Streams

Abstract

Most studies of urban streams have relied on single variables to characterize the degree of urbanization, which may not reflect interactions among features of urban landscapes adequately. We report on an approach to the characterization of urbanization effects on streams that used principal components analysis and multiple regression to explore the combined, interactive effects of land use/land cover, human population demography, and stream habitat quality on an index of biological integrity (IBI) of fish communities. Applied to a substantially urbanized region in northeast Ohio, USA, the analysis demonstrated the interactive nature of urbanization effects. Urban land use and stream habitat quality were significant predictors of IBI, but were no better than and, in some cases, poorer predictors than other gradients and interactions among gradients. High integrity sites were characterized by low forest cover and high grassland cover at sub-catchment scale, but high forest cover within a 500 m radius local zone of the sample point, conditions often found in protected parklands in the region. The analysis also indicated that variability in stream habitat quality was unrelated to landscape or demographic features, a result we attribute to the interaction between the geological and urbanization histories of the region.

Keywords: biological integrity, fish, land use, urban streams, water quality

1. Introduction

Urbanization poses vexing challenges to the ecological sustainability and restoration of stream ecosystems. Stream habitat and biota in urban settings are often profoundly degraded in comparison to natural or less-impacted rural conditions, e.g., (Klein, 1979; Steedman, 1988; Schuler, 1994; May, et al., 1997; Boward, et al., 1999; Morley and Karr, 2002; Morse, et al., 2003; Miltner, et al., 2004), even at modest amounts urban development (Weaver and Garman, 1994; Booth and Jackson, 1997). Given these impacts and the accelerating pace of urbanization (Cohen, 2003), there has been great interest in describing quantitative relationships among the intensity of urbanization, constraints on stream recovery, and potential thresholds of degradation imposed by urban development, e.g, reviews by Paul and Meyer (2001) and Allan (2004).

However, considerable variability surrounds these general biological integrity-urbanization relationships, so that equivalent levels of urbanization can be associated with a wide range of biological indicator scores (Wang, et al., 1997; Klauda, et al., 1998). For example, streams sites in urban regions of Ohio can vary by more than four-fold in biological integrity within the same level of upstream urban land use (Yoder, et al., 2000; Miltner, et al., 2004). Further, slopes and thresholds of urbanization effects differ among urban regions (Yoder, et al., 1999, 2000; Coles, et al., 2004).

Some of this variability is surely attributable to identifiable, allied stressors affecting stream sites, such as point-source pollutants and combined or sanitary sewer outfalls that may exert effects in addition to the generalized impacts of urbanization (Yoder, et al., 2000; Miltner, et al., 2004). However, significant fractions may also be associated with complex interactions among features of urbanized landscapes and the effects of urbanization within regionally-specific contexts of geology, climate, or history of development and anthropogenic disturbance (Allan, 2004). Such complexity is unlikely to be captured by any single measure of urbanization (Booth and Jackson, 1997; Yoder, et al., 2000; Allan, 2004; Coles, et al., 2004).

Hence, a method for quantifying urban effects on streams that integrates multiple dimensions of urbanization and interactions among these factors is desirable. In this report, we describe such an approach and illustrate its application with an analysis of small stream sites in a highly urbanized region of northeastern Ohio, USA. The region has figured prominently in past analyses using single indicators of urbanization, e.g., Yoder et al. (2000), thereby facilitating comparisons with those earlier studies. The analysis integrates the influences of three sets of variables characterizing the urban environment: major land use / land cover features; human population and housing density; and stream habitat quality. We compared the relative impacts of these variables on measures of biological integrity based on fish communities to address the following questions: (1) Are multivariate descriptors of urbanization better predictors of biological integrity than a single variable measure of urban effects, e.g., % urban land, population or housing density?; (2) Does a multivariate approach provide useful, additional insight into effects of urbanization not revealed by single measures of urbanization, e.g., spatial interactions among variables that mitigate or exacerbate the general effects of urbanization, or spatial interactions that represent regionally-specific patterns of urban development?; (3) Do the impacts of urbanization and the interactions among landscape gradients differ with spatial scale?; and (4) Are stream habitat or landscape-level variables better predictors of biological integrity?

To accomplish this, we used principal components analysis to generate statistically-independent combinations of the original land cover/population demographic variables. Patterns of interaction among the original variables were interpreted from correlations and the magnitude and direction of factor loadings on the principal components. Multiple linear regression was used to test the value of the multivariate components for predicting biological integrity. The question of scale was addressed by comparing the relative influences of stream habitat and landscape-level effects on biological integrity, and by comparing analyses that aggregated the predictor variables at different scales, i.e., the catchment of the biological sampling point vs. 500 m radius “local zone” surrounding the biological sampling point.

2. Study Area

The analysis described herein focuses on small stream catchments, for the most part 20-52 km2 drainage area, tributary to the Cuyahoga River in the Cleveland-Akron metropolitan area, as well as a few small streams tributary to Lake Erie within the same geographic area. The Cuyahoga River consists of 1963 km of stream miles draining 2100 km2 in northeastern Ohio, USA. The catchment lies within the Erie/Ontario Lake Plain Ecoregion and parts of three different physiographic provinces: the Allegheny Plateau, till plains, and lake plains.

The Cuyahoga River watershed is one of the most densely populated, urbanized, and industrialized regions of Ohio. The basin accounts for 2% of total state land, but 17% of Ohio’s population, approximately 1.9 million people (921 people km-2) (Rybka et al., 2001). While the upper reaches are dominated by agriculture, the middle and lower reaches are heavily influenced by urbanization. Two of Ohio’s major cities, Akron and Cleveland, occur within the mid- and lower segments, respectively. The river is characterized by an unusual, U-shaped morphology, formed during the last glacial recession through the merging of several formerly separate drainages. Hence, the Cuyahoga flows southward from its upper reaches then changes direction near Akron, Ohio and flows northward to its terminus at Lake Erie in Cleveland, Ohio (Fig. 1). Because of this shape, eastward urban and suburban expansion is encroaching upon headwater regions that are currently dominated by forest and agriculture.

The region has a long history of urbanization, accelerated by completion of the Ohio and Erie Canal in 1825 (Cockrell, 1992). Subsequently, the Cleveland-Akron corridor became one of the major commercial-industrial centers of North America through the first half of the 20th century, with petroleum, steel, rubber, and manufacturing among the major industries (Cockrell, 1992; Rybka et al., 2001). The Cuyahoga Valley National Park, established in 1974, protects 134 km2 of the basin between Akron and Cleveland. However, even this parkland has a history of substantial disturbance, including agricultural, commercial, and industrial uses, as well as many contemporary impacts within and along its boundaries (Cockrell, 1992). Additional environmental challenges arise from economic and infrastructural decline and population loss from city-centers and older, inner-ring suburbs, and out-migration to the urban fringe (Bier, 1993, 2001).

3. Methods

3.1 Biological Integrity and Stream Habitat Data

Biological integrity and stream habitat data were extracted from a statewide database maintained by the Ohio Environmental Protection Agency (OEPA). These data have served as the basis for previously published analyses regarding the Cuyahoga and other Ohio watersheds (Yoder, et al., 1999, 2000; Miltner, et al., 2004) and are central component of the state-wide program for water quality assessment a. Biological data consisted of multimetric indices of biological integrity (IBI) and well-being (modified index of well-being, MIWB) based upon fish communities. The IBI is an aggregate index based upon 12 sub-metrics characterizing the taxonomic composition, trophic structure, abundance, and condition of the fish community (Karr, 1981; Karr, et al., 1986), as modified for Ohio streams and rivers (Ohio EPA, 1987b, 1989b; Yoder and Rankin, 1995). The MIWB is an index that incorporates measures of abundance, biomass, and diversity (Gammon, 1976; Gammon, 1980; Hughes and Gammon, 1987), as modified to increase sensitivity to conditions particular to Ohio’s streams and rivers (Ohio EPA, 1987b, 1989b; Yoder and Rankin, 1995). Stream habitat data were also extracted from the statewide database as the Qualitative Habitat Evaluation Index (QHEI). The QHEI is a qualitative assessment of major features of stream habitats presumed to influence the potential for healthy fish communities (Rankin, 1989, 1995).

3.2 Land Use and Census Data

Land use data were extracted from a statewide land cover inventory of Ohio produced by the Ohio Department of Natural Resources, based upon Landsat Thematic Mapper Data collected in September and October 1994 (Schaal and Motsch, 1997), the approximate mid-point of the time frame for the biological data used here. The data were classified into seven land cover categories of urban, agriculture/open urban areas, shrub/scrub, wooded, open water, non-forested wetlands and barren. Population density and housing density data were obtained from the U.S. Census Bureau, 2000 Census of Population and Housing, Summary File 1, 2001. Age of housing data were extracted from the U.S. Census Bureau, 2000 Census of Population and Housing, Summary File 3, 2002.

3.3 Delineation of sample point catchments

The basic spatial unit of analysis in this study was the catchment area for each of the biological sampling points. The sample-point catchment area was defined using digital elevation models (DEM), a vector hydrography database, and sample point locations. The DEM used here was extracted from the 1:24,000-scale seamless US Geological Service (USGS) National Elevation Dataset (NED) which was developed by merging USGS’s highest resolution, best quality elevation data available (NED is accessible on-line at gisdata.ned/default.asp).

To improve accuracy of stream and catchment delineation, we used a vector hydrography database, the Valley Stream Segment (VST Rivers) file, to adjust the DEM. The VST file is based upon a 1:100,000 base map from the National Hydrography Dataset (NHD) from the USGS and the U.S. Environmental Protection Agency (USEPA). The two main sources for information for this dataset are USGS digital line graphs and the USEPA Reach File Version 3 (). To a improve catchment delineation, raster cells were adjusted in elevation at or near the VST vector layer streams, thereby improving stream channel results. The adjusted raster elevation values were then used to create a new vector-based stream network which included only the streams recognized by the original hydrography stream layer but were precisely located in reference to the DEM and the slope, flow direction, and related information necessary to delineate catchment areas up stream and up slope from the sample points.

GIS surface analysis tools were used to create catchment area polygons. Figure 1 illustrates the points, VST Rivers hydrography layer, the revised DEM-based hydrography network, and the catchment polygons in a portion of the study area. In addition to the sample point catchments, we also delineated 500 m radius local zone sub-polygons for each sample point. These were defined by inscribing a 500 m radius around the sample point within the boundaries of the original catchment area (Fig. 1).

Population and housing unit counts are available at the census block level. Because catchments and local zones split census blocks and block groups, these census data were estimated by areal interpolation, specifically area apportionment. This method apportions the data based on the relative areas of the block or block group that are contained in each part split by catchments or zones. Geographic (polygon) boundary files in computerized GIS database structure for census blocks and block groups are available from the Census Bureau's Topologically Integrated Geographically Encoded Referenced (TIGER) database.

3.4 Statistical Procedures

The biological, land use, population demographic and stream habitat variables used in this study are listed in Table 1. Biological data (IBI and MIWB) served as dependent variables in these analyses. Also, IBI was decomposed into its 12 sub-metrics, and these also were analyzed as dependent variables. QHEI served as both a dependent and independent variable. QHEI was the dependent variable in regressions evaluating land use and demographic gradients as predictors of stream habitat quality, whereas QHEI was entered as an independent variable in regressions seeking to predict IBI, IBI sub-metrics, or MIWB.

All variables were evaluated for conformance to normality and transformed, if necessary, using appropriate transformations. Indices, counts, and density measures were transformed according to log10X or log10(X+1), whereas proportion variables were arcsin square-root transformed. Pearson-product moment correlations and linear regression analyses were used to explore urbanization-stream quality relationships urbanization, as measured using % urban land use and population or housing density.

To explore the contributions of landscape gradients in addition to urban land use and to assess the potential for interactions among landscape and demographic gradients, we employed principal components and multiple linear regression analyses. Principal components analyses reduced the dimensionality among the independent variables and produced gradients, i.e., principal components, which were independent and appropriate for multiple regression analysis. Principal components were obtained for the land use/land cover and population demography dataset. Distance of the sample point from stream terminus and area of the sample-point catchments were also included within these data since these two spatial variables were correlated with several of the landscape and demographic features. Only components with eigenvalues > 1 were retained for subsequent analyses. Principal components were interpreted based upon magnitude of factor loadings and inspection of bivariate plots of components against the original variables.

The predictive value of principal components for biological integrity or habitat quality was assessed using stepwise multiple regression (forward and backward selection procedures, P=0.05 for entry or removal from the model). R2-change was used to assess and rank the proportional contribution of each significant predictor to overall variance explained by regression models. Julian date of the biological sample was entered into these analyses to adjust for temporal changes in biological integrity. All statistical procedures were conducted using SPSS for Windows, Version 11.

4. Results

4.1 Single Indices of Urbanization

The index of biological integrity (IBI) was significantly correlated with variables that have been commonly used as measures of urbanization, e.g., % urban land use and population density, although the magnitudes of these correlations were generally low, and several other land use variables were more strongly correlated with IBI than urban land use, e.g., grassland cover (Table 2). Percent urban land use and population density were also found to be significant predictors, in combination with QHEI, of IBI by multiple linear regression analysis, although only population density retained statistical significance as a predictor of IBI in regressions that included two spatial covariates of IBI, distance from terminus and catchment area (Table 3).

4.2 Multivariate Indices of Urbanization: Sub-catchment Scale Analyses

Principal components based upon variables aggregated at the sample point sub-catchment scale yielded four components with eigenvalues > 1, which accounted for 67.5% of total variation in the data set. Factor loadings for the components are shown in Table 4, where principal components are labeled SP1, SPC2, etc., to indicate that they are based upon sub-catchment scale data. SPC1 was most strongly influenced by catchment area with lesser loadings associated with housing density, % barren, and % open water cover. SPC2 increased with % wetland and shrub/scrubland cover. SPC3 described a contrast between % forest and % grass/park/field cover (Fig. 2). SPC4 was strongly, positively correlated with urban land cover.

No component was correlated significantly with QHEI (Table 4), whereas two components, SPC1 and SPC3, were correlated significantly with IBI. Three principal components (1, 3, and 4) were significant predictors IBI according to stepwise, multiple linear regression (Table 5). IBI increased with SPC1, but declined with SPC3 and SPC4. Multiple regression also indicated that IBI increased significantly with QHEI and julian day (Table 5). The modified index of well-being (MIWB) was also significantly related to principal components and IBI. MIWB also increased with SPC1, as well as SPC2 (Table 5). As in the case for IBI, MIWB increased with QHEI and julian day.

The multiple regression relating principal components to IBI had a substantially higher coefficient of determination (R2) than any of the regressions relating principal components to submetrics of the IBI (Table 5). Nevertheless, each of the principal components was a significant predictor of at least two of the sub-metrics of the IBI. SPC3 showed the highest number of significant coefficients (it was a predictor of 6 of 12 sub-metrics, range of P-values = 0.017 - 1, accounting for 80% of the total variance within the dataset. Factor loadings for this analysis are shown in Table 6, where principal components are labeled “LPC” to indicate that 500 m local zone data were included in their calculation.

LPC1 was strongly related to population and housing density at both 500 m local zone and the overall sample sub-catchment scales. LPC1 also increased with urban land use at both scales. LPC2 was most strongly related to basin area at the sub-catchment scale. LPC3 was strongly related to wetland cover at both the sub-catchment and 500 m local scale. LPC4 was principally a descriptor of shrub/scrub cover at the two scales. LPC5 decreased strongly with increasing forest cover in the 500 m zone, but increased significantly with increasing grass/park/field cover and urban land cover in the 500 m local zone. LPC6 was essentially a descriptor of barren lands at both scales. LPC7 was strongly related to open water cover at the 500 m scale, and to a lesser extent to water at the sub-catchment scale. LPC8 described a contrast between grass/park/field cover and forest cover at the sub-catchment scale.

As in the analysis conducted at the sub-catchment scale exclusively, none of these components were correlated with QHEI (Table 6). Six of the eight components (LPC1, LPC2, LPC3, LPC4, LPC5, and LPC8) were correlated significantly with IBI (Table 6), but only five of these (LPC4 excluded) was predictive of IBI based upon stepwise multiple regression analysis (Table 7). LPC2 and LPC3 were predictive of MIWB (Table 7). IBI and MIWB also increased with QHEI and julian day.

Once again, the multiple regression predicting IBI had a substantially higher coefficient of determinantion (R2 = .44) than regressions for each of the IBI sub-metrics (range of R2 = .09 - .35), although each of the sub-metrics showed a significant regression with at least one of the principal components. In general, sub-metrics of the IBI responded to the same gradients as did overall IBI. However, LPC4, which did not emerge as a significant predictor of IBI, was associated 6 of 12 submetrics. LPC4 was negatively associated with measures of species richness and community composition and trophic structure, but positively associated with percent of individuals with deformities, eroded fins, lesions and tumors (Table 7). LPC7, which also showed no predictive value for IBI overall, was positively associated with the proportion of insectivores in samples.

5. Discussion

5.1 Are multivariate measures more informative than univariate measures of urbanization?

Several authors have indicated that single measures of urbanization are unlikely to be sufficient for assessing the ecological health of urban streams (Booth and Jackson, 1997; Karr and Chu, 1999; Yoder, et al., 2000; Morley and Karr, 2002), although no previous study has conducted a comparison of the relative power of bivariate and multivariate approaches for the same dataset. Previous analyses for northeast Ohio based upon a subset of the data used here found significant, negative relationships between urban land use and biological integrity based upon fish and invertebrate community quality and that a threshold for significant degradation of fish communities, as measured by IBI, occurred at 8% urban land use within the Cuyahoga River basin (Yoder, et al., 1999, 2000). Similarly, we found that IBI declined with % urban land use or population density (Tables 2 and 3, Fig. 3). Further, sites for which the multivariate combination of land use, population, housing, and stream habitat data predicted IBIs greater than 41 all had observed IBIs exceeding the minimum value required for attainment of Ohio EPA warm water habitat (WWH) use criterion (Fig. 4). The average % urban land use for this group of sites, 6.5% ± 2.3% (N= 7) was essentially indistinguishable from the 8% threshold value based on % urban land use alone. The majority of sites with predicted IBIs lower than 41 failed to meet WWH use attainment and showed poor IBIs overall. No sites with predicted IBIs below 26 achieved WWH status and the average % urban land use for this group was 24.6% ± 4.4% (N = 50). In an analysis of streams in the Columbus, Ohio area, Miltner et al. (2004) report a similar upper threshold of % urban land use (27.1%), above which stream sites failed to achieve WWH status.

Hence, the multivariate approach used here identified management and assessment thresholds largely equivalent to previous analyses based on bivariate approaches. However, the multivariate approach revealed interactions among landscape and demographic variables that could not be assessed with a single measure of urbanization. In particular, the importance of urban land use recedes in multivariate analyses, where other gradients and interactions among gradients emerge as more important predictors. At the sub-catchment scale, the fourth principal component (SPC4) most strongly represented urban land use. Although this component emerged as a significant predictor of IBI (Table 5), SPC4 accounted for only 1.8% of variance in IBI. In comparison, SPC3, which described a counter-gradient of forest versus grassland cover, accounted for 9.1% of variance in IBI.

When 500 m local zone land use/land cover was entered into the analysis, urban land use receded even farther into the background. Although urban land use at sub-catchment or 500 m scale loaded significantly on several principal components, loadings were relatively low (Table 6). In the local zone analysis, urban land use made its strongest contribution to LPC1, which was even more strongly related to population and housing density (Table 6). However, this component had predicted IBI only weakly, accounting for only about 3.3% of variance in biological integrity. Components 1, 2, 3, 5, and 8 describing other aspects of land use/land cover, including percent forest, wetland, and grasslands in the sub-catchment and within the 500 m local zone, all accounted for more variance in IBI (5.6-6.5%, accounting for 24% variance in total).

What accounts for the reduction in the influence of urban land use within these analyses? Certainly one important factor is that much of the area is either heavily urbanized or at least suburbanized to some degree, so that the effects of urbanization are pervasive but the gradient of urbanization is relatively small. Mean urban land use among the sites used in this study is relatively high (16%), even though many of the sites are outside the urban core or are found within parklands or forested ravines. This level of urbanization has been associated with substantial, and perhaps irreversible, biological degradation (Steedman 1988, Booth and Jackson 1997, Yoder et al.1999).

It is also likely that these analyses reflect the long history of anthropogenic disturbance within the region. Stream biota can reflect the historical legacy of past stressors and land uses long after those factors have changed (Harding, et al., 1998). Northeast Ohio has been a center for commerce and industry since early in the 19th century, when development of the region was accelerated substantially with the establishment of the Ohio and Erie Canal (Cockrell, 1992). Indeed, fish in many of the region’s streams had shown evidence of substantial decline for decades prior to the timeframe of the current study (Trautman, 1981).

5.2 Are there interactive and scale effects among land use/demographic gradients?

Our findings reinforce the notion that the mix and spatial juxtaposition of land uses within an urbanized basin are important determinants of biological integrity of streams (Wang, et al., 2003). For example, the principal components that emerged as most important in explaining variability in IBI in both the sub-catchment (SPC3, 9.1 % of total variance in IBI) and the local zone analyses (LPC8, 6.5 % of total variance in IBI) were components describing a spatial counter-gradient in forested and open, grassland cover (Fig. 2).

Further, the nature of land use effects changed profoundly with spatial scale and proximity to the biological sampling point. In particular, the polarity of forest cover effects on biological integrity changed between sub-catchment and local zone scales. Whereas high forest cover within the sub-catchment overall was associated with low IBI value, high forest cover within the local zone was associated with high IBI (Fig. 3, Table 7). Since forest cover in the local zone may represent riparian vegetation in large part, the positive effect on biological integrity at this scale is consistent with general findings that riparian vegetation can buffer upland effects (Steedman, 1988; Horner, et al., 1997; May, et al., 1997). On the other hand, the negative impact of forest cover at the sub-catchment scale seems counterintuitive at first glance.

However, these findings are interpretable in light of current and historical patterns of land use in northeast Ohio. Many sites characterized as having high forest cover at the sub-catchment scale are also associated with high population and housing density, as well as relatively high urban land use (Fig. 3). This combination of factors characterizes older, inner-ring suburbs in the region. In these neighborhoods, the canopies of large street trees overhang houses, streets and other impervious surfaces, and wooded parks are interspersed within densely populated residential areas. These suburbs have a long history of urban impact on local streams. By 1900, wealthy industrialists and merchants were leaving an increasingly industrialized city-center of Cleveland to establish new suburban neighborhoods just beyond the city limits (Cigliano, 1991). Out-migration from the city center and from older inner-ring suburbs has continued and, in fact, has accelerated in recent decades (Bier 1993, 2001). Within these older urban/suburban areas, population loss and economic decline are associated with ageing and inadequate waste water management infrastructure (Bier, 2001).

In addition, the counter-gradient of forest vs. grasslands at the sub-catchment scale, in combination with the positive effect of local zone forest cover on IBI, defines a landscape signature indicative of high biological integrity for the region. The sites with highest biological integrity in our dataset were those characterized by high open grassland cover and low forest cover at the sub-catchment scale, but high forest cover in the local zone (Fig. 3). This nexus of land cover categories is most often found in areas beyond the urban core where forest cover is associated with riparian strips adjacent to open parkland and/or agricultural fields. Within cities and suburbs, similar landscapes are found in protected and managed areas, including an extensive network of regional parks and the Cuyahoga Valley National Park.

5.3 Are stream habitat or landscape variables better predictors of biological integrity?

Our measure of stream habitat quality in these analyses, the Qualitative Habitat Evaluation Index (QHEI), was designed and calibrated as a measure of the potential for stream habitat to support healthy, native fish communities (Rankin, 1989). Hence, this variable was expected to covary significantly with IBI and MIWB. Indeed, QHEI was a significant predictor of IBI, MIWB, and a majority of the IBI sub-metrics. In this regard, the current findings are congruent with previous studies demonstrating that fish community health is associated with habitat quality (Schlosser, 1982; Roth, et al., 1996).

However, QHEI was a poorer predictor of IBI than were landscape variables overall. For the sub-catchment level analysis, QHEI accounted for 4.8 % of total variation in IBI, whereas the principal components describing landscape and demographic features combined to account for 19.5% of variance in IBI. In the analysis including variables describing the 500 m radius local zone, the landscape components combined to explain 27.3% of the variance in IBI, in comparison to 2.6% attributable to QHEI alone. Moreover, several landscape/demographic components explained more variability singly than did QHEI. For example, LPC 8 alone explained 3-fold more variance in IBI (8.6%) than did QHEI. Roth et al. (1996) also reported than habitat quality was no better as a predictor of fish community health than features of land cover. Overall, these finding emphasize the combined importance of both stream channel and conditions within the uplands as determinants of biotic quality of streams (Booth and Jackson, 1997).

We also found that QHEI was unrelated to any of the land cover or demographic variables, either alone or in combination as principal components, or when the land cover or demographic variables were aggregated at sub-catchment or local-zone scales. Given the links between landscape features and stream morphology, hydrology, and stream habitat quality that have been documented in a variety of studies (Richards and Host, 1994; Roth, et al., 1996; MacRae and DeAndrea, 1999), this finding is noteworthy, but it is not unique to the current analysis. Wang et al. (1997, 2003) reported little or no correlation between habitat quality variables and effective impervious surface cover among urban streams in Wisconsin. How then is it possible for stream habitat quality to vary independently of land use/land cover and demography, while biological integrity covaries significantly with both stream habitat quality and landscape level variables?

We suggest that the resolution of this apparent paradox lies in the interaction of the geological and urbanization histories of the region. In many cases, streams in northeast Ohio lie within ravines, often quite deeply incised, that were formed by the retreat of the last glaciation (White and Totten, 1982). During early settlement of the region, the deeply incised terrain made transportation and communication difficult, isolated settlements, and the steep, unstable hillsides were largely unavailable for building, cultivation or pasture land (Cockrell, 1992). Many of these ravines formed the template for city and suburban parklands, including the Cuyahoga Valley National Park. Hence, these areas preserved natural features precisely because they were not useful for other purposes. Thus, the ravines, and associated parks, have provided some degree protection from the worst effects of urbanization on the physical features of streams.

Nevertheless, biological degradation may proceed inexorably through a variety of urbanization effects that degrade biota but have lesser impacts on stream habitat (Allan, 2004), including stressors that short-circuit the riparian zone, e.g., sewer outfalls, thermal heat island effects, and atmospheric deposition. Further, stream biodiversity can reflect the impacts of devastating pulse events that may not necessarily have discernable long term effects on physical habitats. One local example is a large fire in a scrap tire yard in 1981 that released tens of thousands of liters of petroleum derivatives into the headwaters of a small stream that was otherwise largely contained within the Cuyahoga National Park (Cockrell, 1992). In these geological and historical contexts, contemporary assessments of stream habitat for this region may provide only limited guidance regarding the potential for streams to support healthy biological communities.

6. Conclusions

While our findings are consistent with previous studies indicating that urban land use has a negative association with biological integrity of streams (Klein, 1979; Steedman, 1988; Roth, et al., 1996; Dreher, 1997; May, et al., 1997; Boward, et al., 1999; Yoder, et al., 2000; Morse, et al., 2003; Roy, et al., 2003; Miltner, et al., 2004), this analysis also demonstrates that spatial interactions with other aspects of the urban landscape are important determinants of variability in stream biota. In fact, our results suggest that in regions with long histories of urban development such as northeast Ohio, other axes of landscape variability may emerge as even stronger predictors of variability in biological quality among stream sites. Further, multiple landscape features may have interactive effects on biological integrity which may vary both in magnitude and direction with spatial scale, e.g., forest cover in the current case.

Our analysis also emphasizes that the influence of urbanization on streams is shaped by regional geological and historical contexts. Within the Cuyahoga River basin, unstable ravines of glacial origin have impeded agricultural and urban development in some stream reaches, thereby preserving natural features of riparian zones and stream habitats, but not necessarily biological integrity. Rather, in northeastern Ohio, land use signatures indicative of parklands are better predictors of biological integrity of fish communities than measures of stream habitat quality.

Acknowledgements

We a thank Stuart Schwartz, Director of the Center for Environmental Sciences, Technology, and Policy (CESTP) at Cleveland State University, and his staff for their administrative and data management assistance on this project. We thank Elizabeth Whippo-Cline for her assistance with early stages of the project and Lester Stumpe of the Northeast Ohio Regional Sewer District for his advice and support. The project has also benefited from the work of the following graduate student assistants: Shawn Bleiler, Sonya Steckler and Cari-Ann Hickerson. This project was financed through a grant from the Ohio Environmental Protection Agency and the United States Environmental Protection Agency, under the provisions of Section 319(h) of the U.S. Clean Water Act, and through the U.S.E.P.A. National Risk Management Laboratory.

References

Allan, J.D., 2004. Landscapes and riverscapes: the influence of land use on stream ecosystems. Annu. Rev. Ecol. Evol. Syst. 35, 257-284.

Bier, T., 1993. Cuyahoga County Outmigration. Housing Policy Research Program, Levin College of Urban Affairs, Cleveland State University.

Bier, T., 2001. Moving Up, Filtering Down: Metropolitan Housing Dynamics and Public Policy. The Brookings Institution, Center on Urban and Metropolitan Policy.

Booth, D.B., Jackson, C.R., 1997. Urbanization of aquatic systems: degradation thresholds, stormwater detection, and the limits of mitigation. J. Amer. Wat. Resour. Assoc. 33, 1077-1090.

Boward, D., Kayzak, P., Stranko, S., Hurd, M., Prochaska, T., 1999. From the Mountains to the Sea: the State of Maryland's Freshwater Streams. Maryland Department of Natural Resources, Monitoring and Non-tidal Assessment Division. EPA 903-R-99-023. Annapolis, Maryland.

Cigliano, J., 1991. Showplace of America: Cleveland's Euclid Avenue, 1850-1910. The Kent State University Press, Kent, Ohio.

Cockrell, R., 1992. A Green Shrouded Miracle: The Administrative History of Cuyahoga Valley National Recreation Area, Ohio. .

Cohen, J.C., 2003. Human population: the next half century. 303, 1172-1175.

Coles, J.F., Cuffney, T.F., McMahon, G., Beaulieu, K.M., 2004. The effects of urbanization on the biological, physical, and chemical characterisitcs of coastal New England streams. U.S. Geological Survey Professional Paper 1695.

Dreher, D.W., 1997. Watershed urbanization impact on stream quality indicators in northeastern Illinois, in: Murray, D., Kirschner, R. (Eds.), Assessing the Cumulative Impacts of Watershed Development on Aquatic Ecosystems and Water Quality. Northeastern Illinois Planning Commission, Chicago, Illinois, pp. 129-135.

Gammon, J.R., 1976. The Fish Populations of the Middle 340 km of the Wabash River, Tech. Report No. 86. Purdue University, Water Resources Research Center. West Lafayette., IN.

Gammon, J.R., 1980. The use of community parameters derived from electrofishing catches of river fish as indicators of environmental quality. Seminar on Water Quality Management Tradeoffs. U.S. Environmental Protection Agency. Washington, D.C.

Harding, J.S., Benfield, E.F., Bolstad, P.V., Helfman, G.S., Jones, E.B.D., III, 1998. Stream biodiversity: the ghost of land use past. Proc. Natl. Acad. Sci. U.S. Am. 95, 14843-14847.

Horner, R.R., Booth, D.B., Azous, A., May, C.W., 1997. Watershed determinants of ecosystem functioning, in: Roesner, C. (Eds.), Effects of Watershed Development and Management on Aquatic Ecosystems. American Society of Civil Engineers, New York, New York, pp. 251-274.

Hughes, R.M., Gammon, J.R., 1987. Longitudinal changes in fish assemblages and water quality in the Willamette River, Oregon. Trans. Am. Fish. Soc. 116, 196-209.

Karr, J.R., 1981. Assessment of biotic integrity using fish communities. Fisheries. 6, 21-27.

Karr, J.R., Chu, E.W., 1999. Restoring Life in Running Waters. Island Press, Washington, D.C.

Karr, J.R., Fausch, K.D., Angemier, P.L., Yant, P.R., Schlosser, I.J., 1986. Assessing biological integrity of running waters: a method and its rationale. Ill. Nat. Hist. Sur. 5, 1-28.

Klauda, R., Kazyak, P., Stranko, S., Southerland, M., Roth, N., Chaillou, J., 1998. Maryland stream biological survey: a state agency program to assess the impact of anthropogenic stress on stream habitat quality and biota. Environ. Monitor. Assess. 51, 299-316.

Klein, R., 1979. Urbanization and stream quality impairment. Water Resour. Bull. 15, 948-963.

MacRae, C., DeAndrea, M., 1999. Assessing the impact of urbanization on channel morphology. 2nd International Conference on Natural Channel Systems. Niagara Falls, Ontario.

May, C.R., Horner, J., Karr, B., Mar, B.W., Welch, E., 1997. Effects of urbanization on small streams in the Puget Sound lowland ecoregion. Water. Protect. Tech. 2, 483-494.

Miltner, R., White, D., Yoder, C.O., 2004. The biotic integrity of streams in urban and suburbanizing landscapes. Landscape Urban Plann. 69, 87-100.

Morley, S.A., Karr, J.R., 2002. Assessing and restoring the health of urban streams in the Puget Sound basin. Conserv. Biol. 16, 1498-1509.

Morse, C.C., Huryn, A.D., Cronan, C., 2003. Impervious surface area as a predictor of the effects of urbanization on stream insect communities in Maine, U.S.A. Environ. Monit. Assess. 89, 95-127.

Ohio EPA, 1987b. Biological Criteria for the Protection of Aquatic Life. Volume II: Users Manual for Field Assessment of Ohio Surface Waters. Division of Water Quality Monitoring and Asssessment, Surface Water Section. Columbus, Ohio.

Ohio EPA, 1989b. Addendum to Biological Criteria for the Protection of Aquatic Life. Volume II: Users Manual for Biological Field Assessment of Ohio Surface Waters. Division of Water Quality Monitoring and Assessment, Surface Water Section. Columbus, Ohio.

Rankin, E.T., 1989. The Qualitative Habitat Evaluation Index (QHEI), Rationale, Methods, and Application. Ohio EPA, Division of Water Quality Planning and Assessment, Ecological Assessment Section. Columbus, Ohio.

Rankin, E.T., 1995. The use of habitat assessments in water resource management programs, in: Davis, W., Simon, T. (Eds.), Biological Assessment and Criteria: Tools for Water Resource Planning and Decision Making. Lewis Publishers, Boca Raton, Florida, pp. 181-208.

Richards, C., Host, G., 1994. Examining land use influences on stream habitats and macroinvertebrates: a GIS approach. Water Resour. Bull. 30, 729-738.

Roth, N.E., Allan, J.D., Erickson, D.L., 1996. Landscape influences on stream biotic integrity assessed at multiple spatial scales. Landscape Ecol. 11, 141-156.

Roy, A.H., Rosemond, A.D., Paul, M.J., Leigh, D.S., Wallace, J.B., 2003. Stream macroinvertebrate response to catchment urbanisation (Georgia, U.S.A.). Freshwat. Biol. 48, 329-346.

Schaal, G.M., Motsch, B.R., 1997. State of Ohio land cover, 1994. Database derived from Landsat TM. Ohio Department of Natural Resources, Division of Real Estate and Land Management, GIS and Remote Sensing Services, Columbus, OH.

Schlosser, I.J., 1982. Fish community structure and function along 2 habitat gradients in a headwater stream. Ecol. Monogr. 52, 395-414.

Schuler, T.R., 1994. The importance of imperviousness. Watershed Protect. Tech. 1, 100-111.

Steedman, R.J., 1988. Modification and assessment of an index of biotic integrity to quantify stream quality in southern Ontario. Can. J. Fish. Aquat. Sci. 45, 492-500.

Trautman, M.B., 1981. The Fishes of Ohio. Ohio State University Press, Columbus, Ohio.

Wang, L., Lyons, J.L., Kanehl, P., 2003. Impacts of urbanization on stream habitat and fish across multiple scales. Environ. Manage. 28, 255-266.

Wang, L., Lyons, J.L., Kanehl, P., Gatti, R., 1997. Influences of watershed land use on habitat quality and biotic integrity in Wisconsin streams. 22, 6-12.

Weaver, L.A., Garman, G.C., 1994. Urbanization of a watershed and historical changes in a stream fish assemblage. Trans. Amer. Fish. Soc. 123, 162-172.

White, G.W., Totten, S.M., 1982. Glacial Geology of Northeastern Ohio. Bulletin 68. Ohio Division of Natural Resources, Columbus, OH.

Yoder, C.O., Rankin, E.T., 1995. The role of biological criteria in water quality monitoring, assessment, and regulation. Ohio EPA Technical Report. Ohio EPA, Monitoring and Assessment Section, Division of Surface Water. MAS/1995-1-3. Columbus, Ohio.

Yoder, C.O., Miltner, R., White, D., 1999. Assessing the status of aquatic life designated uses in urban and suburban watersheds, in: Kirschner, R. (Eds.), National Conference on Retrofit Opportunities for Water Resource Protection in Urban Environments. pp. 16-28.

Yoder, C.O., Miltner, R., White, D., 2000. Using biological criteria to assess and classify urban streams and develop improved landscape indicators. National Conference on Tools for Urban Water Resource Management and Protection. Chicago, Illinois.

|Table 1. List and summary statistics for variables used in the current analysis. Also listed are shortened variable names used in subsequent |

|tables. |

| | | |Sub-Catchment Scale |500 m Radius Local Zone |

|Variable |Abbreviated Name |N |Mean ( S.E.M. |Min - Max |Mean ( S.E.M. |Min - Max |

|

|Biological/Habitat Variables | | | | | | |

|IBI | |227 |29.48 ( .6 |12.00 – 50.00 | | |

|

|MIWB | |227 |5.54 ( .12 |0 – 9.02 | | |

|IBI Sub-metrics | |227 | | | | |

|1. Number of native species | | |10.89 ( .29 |0 – 24 | | |

|

|2. Number of darter species | | |1.30 ( .10 |0 – 6 | | |

|

|3. Number of headwater species | | |.97 ( .06 |0 – 6 | | |

|

|4. Number of cyprinid species | | |4.04 ( .14 |0 – 9 | | |

|

|5. Number of sensitive species | | |.23 ( .04 |0 – 3 | | |

|

|6. % of tolerant species | | |59.70 ( 1.76 |0 – 100 | | |

|

|7. % omnivores | | |26.55 ( 1.37 |0 – 100 | | |

|

|8. % insectivores | | |27.21 ( 1.55 |0 – 92.31 | | |

|

|9. % pioneer species | | |33.04 ( 1.48 |0 – 100 | | |

|

|10. Number of individuals | | |731.89 ( 56.86 |39 - 6523 | | |

|11. % simple lithophiles | | |30.76 ( 1.27 |0 – 82.98 | | |

|12. % of individuals with | | |1.00 ( .14 |0 – 15.43 | | |

|deformities, eroded fins, lesions, or| | | | | | |

|tumors | | | | | | |

|QHEI | |165 |61.66 ( .89 |25.00 – 86.50 | | |

|Land Use/Demographic Variables | | | | | | |

|% Urban |Urban |227 |16.12 ( 1.51 |0-100 |18.42 ( 1.60 |0 – 99.23 |

|% Open Grassland / Parkland / |Grass |227 |37.75 ( 1.91 |0-100 |23.84 ( 1.42 |0 – 94.45 |

|Agricultural Fields | | | | | | |

|% Shrub / Scrub |Shrub |227 |8.73 ( 1.08 |0-86.13 |3.67 ( .35 |0 – 28.33 |

|% Non-forested Wetland |Wetland |227 |9.00 ( .96 |0-100 |4.93 ( .52 |0 – 58.15 |

|% Open Water |Water |227 |.43 ( .12 |0-18.81 |.11 ( .05 |0 – 10.16 |

|% Forested |Forest |227 |27.87 ( 1.95 |0-100 |43.98 ( 1.69 |0 – 100 |

|% Barren |Barren |227 |.07 ( .02 |0-3.16 |.05 ( .03 |0 – 6.06 |

|Population Density (per km2) |Pop |227 |5130.20 ( 1157.28 |0 – 133772.57 |326.73 ( 29.14 |0 – 2320.92 |

|Housing Density (per km2) |House |227 |2329.72 ( 544.61 |0 – 71217.05 |39.92 ( 4.31 |0 - 378 |

|Catchment Area (km2) |Area |227 |5.06 ( 1.12 |.10 – 84.50 | | |

|Distance from Terminus (km) |Distance |227 |7.97 ( .69 |0-46.67 | | |

|Table 2. Pearson product moment correlations for variables aggregated at sub-catchment scale. * .05>P>.01, **.01>P>.001, *** P ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download