The use of biological assessments using fish, macroinvertebrates and algae and the derivation of “biocriteria” as numeric goals stream and river health has advanced significantly over the past 10-20 years. Ohio and Maine pioneered the use of these techniques, but most states now use one or more aquatic assemblages to assess stream quality. Important uses of this data include accurate identification of aquatic life impairment and stressor identification. Accurate classification of natural features that can influence aquatic biota is important for deriving meaningful goals for waters. Early biological classification efforts used geographic-based classification schemes such as aquatic ecoregions in concert with important local factors such as catchment size; however there is now interest in whether classification systems can be refined.

Recognizing the importance of accurate classification for the best use of biological assessment and on which to basis watershed stressor identification, U.S. EPA funded, through its Science to Achieve Results (STAR) program, a series of grants in different regions of the U.S. These grants focused on innovative ways to refine classification of aquatic assemblages to support efficient monitoring strategies and to diagnose the causes of biological impairment. One characteristic of these STAR grants has been the use of multidisciplinary teams to address the range of important classification factors in watersheds. Ohio University’s effort is focused on the Western Allegheny Plateau ecoregion. We used Ohio’s extensive existing biological (fish and macroinvertebrates), habitat and chemical databases to explore the strength of geographically-dependent constructs (e.g., Level IV ecoregions) and geographically independent variables (e.g., habitat, water chemistry, land use, modeled hydrological variables, and geomorphic variables including a RGA). Our approach was to use exploratory statistics, MRPP tests, and Canonical Correspondence Analyses to understand whether these abiotic variables could be used to refine the classification framework for Ohio’s biocriteria. Beyond the historical variables collected during routine intensive water quality surveys, in 2005 and 2006 we added geomorphic variables including detailed physical survey data and at a subset of sites a Rapid Geomorphic Assessment (RGA). The goal of the RGA data was to provide information on channel stability, identified under a conceptual model based on understanding the process of channel evolution. This strategy implies that more insight into the processes that form physical habitat will explain more of the variation observed in aquatic assemblages. The addition of algal community data was designed to widen the realm of biological responses to include more information related to community production and ecological function.

Because of our a priori knowledge of the strength of stream size of fish and macroinvertebrate assemblage structure we divided the data into five stream size categories and tested the classification strength of these groups on fish and macroinvertebrate assemblages across several hundred historical sites in the WAP ecoregion of Ohio. We tested classification strength using Bray-Curtis similarity indices and Multi-Response Permutation Procedures (MRPP). This technique is designed to test the classification strength of geographic groups or constructs, i.e., do fish assemblages vary more within regions than between regions (null = no difference between regions). As expected, there was a statistically significant difference among stream size groups with the fish assemblages results more strongly classified than the macroinvertebrate assemblages. Based on these results, pairwise comparisons and sample size constraints we divided data into three stream size groupings for further analyses: small headwater streams (SH, < 5 sq mi), large headwater streams (HW, 5-20 sq mi) and wadeable streams (WD, > 20 sq mi).

We next tested the classification strength of level IV ecoregions (subecoregions) on fish and macroinvertebrate assemblage data at reference sites. Although we had multiple years of data for some reference sites, testing showed no difference between the use of latest data, most species rich samples or samples with the highest IBI or ICI scores. Thus depending on the intent of the test we used either latest data or most speciose samples. In addition to Level IV ecoregions we tested alternate geographic groupings including stream basins, groups assigned randomly, and the results of a cluster analysis that used the biological assemblages to establish groups (i.e., maximum separation).

There were statistical significant differences among all except the groups constructed by random assignment with the classification strength of the cluster groups greater than subecoregions and river basins which were generally similar in strength to one another. We subsequently conducted indicator species analyses that provided measures of species or taxa most responsible for significant classification categories. Indicator species analyses indicated that most differences were related to species replacements that would have little consequence for the biological indices (i.e., IBI or ICI) used to assess stream condition. We concluded that subecoregions or stream basin difference were not biologically significant enough to warrant alteration of index calibration or expectations (.e.g., biocriteria).

We then explored geographically independent abiotic variables at reference sites using Canonical Correspondence Analyses (CCA) analyses on a suite of abiotic variables (reduced using PCA from a larger set of variables). Initial results indicate that variables that reflect stream size, stream gradient and anthropogenic impacts (e.g., land use, conductivity) explain statistically significant variation in fish assemblages. Stream size is already a part of the IBI and ICI index calibration used in Ohio’s biocriteria and thus accounted for in biocriteria calculation and setting of biological expectations. Anthropogenic impacts as indicated by land use and dissolved materials variables (e.g. conductivity) of the magnitude observed in the reference data were also expected because Ohio’s reference site definition is least impacted, not pristine, and acknowledges a range of stressor on streams because of adjacent human land uses including agriculture. Ohio has dealt with this level of impact by establishing tiered aquatic life uses in Ohio that for these reference streams includes an “Exceptional Warmwater Habitat” use (EWH) for the best streams and a “Warmwater Habitat Use” for remaining sites. The calculation of biological endpoints (biocriteria) also excludes approximately 10-12% of reference sites as impaired under a realization that best management practices are not yet always widely developed or employed. Ohio’s monitoring approach builds in a re-sampling strategy to account for improvement in BMPs and water treatment. In fact a recent assessment of 15 years of reference site re-sampling in Ohio has documented improving reference site conditions, likely due to refined wastewater treatment in larger streams and rivers and better agricultural practices (no till and conservation tillage) in small streams.

Stream gradient, however, is not currently a source of variation incorporated in the calibration or application of the IBI. Cluster analyses combined with indicator species analyses identifies groups of streams that are identifiable as higher gradient waters, low gradient waters and a larger group of sites with both attributes. States that have more extensive low gradient regions have developed wetland stream IBI indices to deal with this source of biological variation. The generally non-regional distribution of low gradient streams in the WAP ecoregion precludes the need for a separate IBI and as such the assemblage differences are more modest. We propose that Ohio calibrate its expectations separately for its baseline (WWH) aquatic life use for low gradient streams in the WAP and potentially other ecoregions. A critical aspect of the process would be the ability to identify low gradient waters that are extensive enough to warrant this altered expectation and we expect that GIS-derived measures of floodplain width in long reaches of low gradient streams would be a useful a prior tool.

Chapter 2 - Identification of Key Stressors Limiting Aquatic Life in the WAP Ecoregion of Southeast Ohio

Our exploration of classification refinements for the WAP ecoregion of Ohio resulted in a relatively minor adjustment to current practice with the suggestion that streams with extensive reaches with very low gradient would be useful alternative classification strata. This portion of our studies deals with methods for refining watershed and regional approaches to identifying stressors responsible for aquatic life impairment. Identification of impaired aquatic communities is only the first step in managing biological condition of our Nation’s waters. A primary goal of the Clean Water Act (CWA) is the restoration of the biological integrity of the Nation’s waters and an essential component of that is to identify the stressors limiting aquatic life. Once this is accomplished, strategies must be developed for managing and reducing these stressors to restore and maintain biological integrity in surface waters.

Much of focus, through the CWA and section 303(d) has been on chemical stressors because of their obvious early impact from dischargers and runoff in urban and industrial areas. As technologies have been implemented that have reduced concentrations of discharge-related chemicals, other non-point chemicals, sediments and non-chemical stressors (habitat, sediments, flow alteration) have been recognized as the major limiting factors on aquatic assemblages. The goal of this chapter of the report is to demonstrate watershed-based statistical and analytical approaches for discriminating the effects of multiple stressors (chemistry, habitat, sediments and hydrology) on aquatic assemblages in southeast Ohio watersheds.

We used two scales of data for this study; ecoregion wide (WAP) data and data focused on two HUC-11 scale watersheds in this ecoregion, Leading Creek and the Shade River. Data for both of these scales included historical data on fish, macroinvertebrates, habitat, water chemistry, land use and hydrology but also included a more recent (2006-2007) subsample of data that include the aforementioned indicators plus data on geomorphology indicators, data on a rapid geomorphic assessment (RGA) tool and data on algal communities. Because of limitations to generating GIS coverages for all historical sites in the ecoregion we randomly selected approximately 160 sites from Ohio’s historical dataset of wadeable streams in the WAP ecoregion of Ohio to represent a gradient of stress as measured by biological indicators and stratified by the subecoregions to maintain a broad spatial coverage.

We used a modification of the WERF Integrated Impact Analysis (IIA) (Paulson et al. 2001) that relies on multiple statistical tools to identify the major stressors limiting fish and macroinvertebrate assemblages. We used Principal Components Analysis (PCA), Canonical Correspondence Analysis (CCA), Stepwise Multiple Regression (SMR), and Regression Tree Analyses (RTA) using PC-Ord v5.0 (PCA, CCA) and S-Plus 8.0 (SMR, RTA) statistical software versions of these tools. Three of these methods (PCA, SMR, RTA) use indices or metrics of these indices as response indicators and the CCA analyses uses the raw species data. Convergence on the same suite of significant stressor varies increases confidence that these variables have strong explanatory power in the WAP ecoregion. We also applied a standard weight of evidence approach to identify stressors in the Shade and Leading Creek Huc-11 watersheds.

As recommended by the IIA methodology we reduced the number of stressor variables by applying a series of PCA analyses by stressor category and then combinations of stressors to identify those stressors that explain the most variation in stressors data in this region by examining the loading estimates on each eigenvalue. The final list of stressor included QHEI substrate score, QHEI channel score, Forest Area Connectivity, Human Use Index, Percent Forest in Catchment, Percent Urban in Catchment, Percent Agriculture in Riparian, log(number of mine openings, log(population density), log(conductivity), log(min. pH), log(min. DO), and the log(total phosphorus). At this point, these variables were used in CCA analyses of fish and macroinvertebrate assemblages, and as independent variables in SMR and RTA analyses.

Although there were some subtle differences among statistical methods, all identify the widespread importance of habitat variables (e.g., channel and substrate conditions) and agricultural impacts (% agriculture in the riparian) and the local but important influence of what can be severe impacts from mining (e.g., pH) and urbanization (e.g., % urban land use). Of the various methods we applied we expect the regression tree analyses will provide the most understandable output for watershed scientists.

The site-specific weight of evidence approach used by Ohio EPA builds off of regional studies that link stressors to response variables, typically in a univariate fashion often at statewide or regional scales (Ohio EPA 1999). A reliance on regional studies linking stressors to responses will identify the key predominant stressor in a region, but often misses very local impacts that might be unique combinations of stressors not included a regional study because data is too sparse. A reliance on a site-by-site or even single watershed study can lack a broad enough stressor gradient to clearly identify important limiting factors. Too narrow a spatial focus can be a problem when trying to model a stressor-response relation for a TMDL, acid mine abatement effort, or other local watershed management effort. To aid in local stressor identification efforts we created a series of univariate ceiling and/or floor relationships between key stressors and fish and macroinvertebrate response variables (IBI, ICI, sensitive taxa metrics) for the Western Allegheny Plateau and paired this with results of the regional regression tree analyses to identify stressors in the Leading and Shade River watersheds. We propose these tools along with other tools that are being developed by relating individual taxa responses to stressor over suitable regional gradients can more accurately identify limiting stressors and drive management actions in watersheds.

Macroinvertebrate sampling in Little Raccoon Creek