Ecological Indicators 7 (2007) 751–767 This article is also available online at: www.elsevier.com/locate/ecolind Improving biological indicators to better assess the condition of streams M.T. Southerland a,*, G.M. Rogers a, M.J. Kline b, R.P. Morgan b, D.M. Boward c, P.F. Kazyak c, R.J. Klauda c, S.A. Stranko c a Versar, Inc., ESM Operations, 9200 Rumsey Road, Columbia, MD 21045, USA University of Maryland, Center for Environmental Science, Appalachian Laboratory, Frostburg, MD 21532, USA c Maryland Department of Natural Resources, 580 Taylor Avenue, Annapolis, MD 21401, USA b Received 2 November 2005; received in revised form 16 August 2006; accepted 18 August 2006 Abstract Biological indicators of stream condition are in use by water resource managers worldwide. The State of Maryland and many other organizations that use Indices of Biotic Integrity (IBIs) must determine when and how to refine their IBIs so that better stream condition information is provided. With completion of the second statewide round in 2004, the Maryland Biological Stream Survey (MBSS) had collected data from 2500 stream sites, more than doubling the number of sites that were available for the original IBI development. This larger dataset provided an opportunity for the MBSS to address the following shortcomings in the original IBIs: (1) substantial disturbance apparent in some reference sites, (2) fish IBIs could not be applied to very small streams, (3) natural variability within IBIs (based on regions) resulted in some stream types (e.g., coldwater and blackwater streams) receiving lower IBI scores and (4) one IBI was not able to discriminate degradation as desired (i.e., Coastal Plain fish IBI). Therefore, development of new fish and benthic macroinvertebrate IBIs was undertaken to achieve the goals of: (1) increased confidence that the reference conditions are minimally disturbed, (2) including more natural variation (such as stream size) across the geographic regions and stream types of Maryland and (3) increased sensitivity of IBIs by using more classes (strata) and different metric combinations. New fish IBIs were developed for four geographical and stream type strata: Coastal Plain, Eastern Piedmont, warmwater Highlands and coldwater Highlands streams; new benthic macroinvertebrate IBIs were developed for three geographical strata: Coastal Plain, Eastern Piedmont and Highlands streams. The addition of one new fish IBI and one new benthic macroinvertebrate IBI partitioned natural variability into more homogeneous strata. At the same time, smaller streams (i.e., those draining catchments <300 ac), which constituted a greater proportion of streams (25%) sampled in Round Two (2000–2004) than Round One (1995–1997), because of the finer map scale, were included in the reference conditions used to develop the new IBIs. The resulting new IBIs have high classification efficiencies of 83–96% and are well balanced between Type I and Type II errors. By scoring coldwater streams, smaller streams and to some extent blackwater streams higher, the new IBIs improve on the original IBIs. Overall, the new IBIs provide better assessments of * Corresponding author. Tel.: +1 410 740 6074; fax: +1 410 964 5156. E-mail address: southerlandmar@versar.com (M.T. Southerland). 1470-160X/$ – see front matter # 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.ecolind.2006.08.005 752 M.T. Southerland et al. / Ecological Indicators 7 (2007) 751–767 stream condition to support sound management decisions, without requiring substantial changes by cooperating stream assessment programs. # 2006 Elsevier Ltd. All rights reserved. Keywords: Indices of Biotic Integrity (IBI); Fish; Benthic macroinvertebrate; Coldwater streams; Reference condition 1. Introduction Biological indicators of stream condition are in use by water resource managers worldwide. The State of Maryland and many other organizations that use these indicators must determine when and how to refine them so that better stream condition information is provided. This paper describes analyses undertaken to improve the biological indicators used by the Maryland Biological Stream Survey (MBSS), a statewide monitoring program conducted by the Maryland Department of Natural Resources (DNR). The benefits of these changes are described as well as the costs to the MBSS and cooperating programs. The MBSS is a probability-based sampling program that can describe streams at varying spatial scales (Klauda et al., 1998). An objective of the MBSS is to assess the status and trends in biological integrity for all 9400 non-tidal stream miles (based on the U.S. Geological Survey 1:100,000 stream network) in Maryland. Therefore, the MBSS provides estimates of the biological condition of streams statewide using indicators based on references of biological integrity. Karr and Dudley (1981) used reference condition as the basis for their definition of biological integrity, i.e., ‘‘the ability of an aquatic ecosystem to support and maintain a balanced, integrated, adaptive community of organisms having a species composition, diversity and functional organization comparable to that of natural habitats in the region.’’ Multi-metric Indices of Biotic Integrity (IBIs), originally developed by Karr et al. (1986), are the most common indicators of stream condition in use today. Most IBIs develop their expectations for the structure and function of biological assemblages from reference sites. Originally, however, the variability in these reference sites was not explicitly modeled nor was the ability of the indicator to distinguish deviation from reference condition directly tested. Currently, it is standard practice to test the performance of IBIs by determining the percentage of reference sites and known degraded sites that are correctly classified. This was first done in Maryland by Weisberg et al. (1997) for the Chesapeake Bay Estuary. More recently, researchers have demonstrated the utility of empirically modeling reference condition from reference sites as exemplified in Bailey et al. (2004) ‘‘reference condition approach.’’ Thus, IBI development today involves the careful testing of the performance of individual metrics and their combinations as indicators that work best for the geographic regions and stream types of interest. The MBSS developed the first fish (Roth et al., 1998) and benthic macroinvertebrate (Stribling et al., 1998) IBIs for Maryland in 1998. Subsequently, Roth et al. (2000) refined the Maryland fish IBI and Southerland et al. (2004) developed a stream salamander IBI for Maryland. These original Maryland IBIs have performed well, helping Maryland DNR and other agencies better characterize and manage State waters, and have produced dozens of assessments and research findings (e.g., Vølstad et al., 2003b). At the same time, these IBIs had the following shortcomings: (1) substantial disturbance apparent in some reference sites, (2) IBIs could not be applied to smallest streams, (3) natural variability within IBIs (based on regions) resulted in some stream types (e.g., coldwater and blackwater streams) receiving lower IBI scores and (4) one IBI was not able to discriminate degradation as desired (i.e., Coastal Plain fish IBI). Specifically, either a more general IBI was applied to two classes of streams (e.g., both Highlands and Piedmont streams for the benthic macroinvertebrate IBI) or no IBI was calculated (e.g., streams draining catchments of less than 300 acres for the fish IBI). To better assess Maryland streams, IBIs that accurately characterize stream condition in more stream classes were needed. With completion of the second statewide round in 2004, the MBSS had collected data from approximately 2500 stream sites, M.T. Southerland et al. / Ecological Indicators 7 (2007) 751–767 more than doubling the number of sites that were available for the original IBI development. Therefore, development of new fish and benthic macroinvertebrate IBIs was undertaken with the following goals:  increase confidence that the reference conditions used are minimally disturbed by refining the criteria for selecting reference sites;  better capture the full range of natural variation in reference condition within Maryland by including more reference sites from unique geographic regions and stream types (e.g., small streams);  increase the sensitivity of IBIs for distinguishing human disturbance by segregating variation into more classes of reference condition. At the same time, development of the new IBIs had to take into account the following practical constraints:  fewer reference sites are available to characterize reference condition when a larger number of geographic or stream type classes are used;  IBIs developed for larger geographic or stream type classes may be less sensitive for distinguishing between reference condition and degraded condition (because they encompass greater natural variability, see Kilian, 2004). Complete details on this IBI development are provided in the Maryland DNR report available at www.dnr.state.md.us/streams (Southerland et al., 2005a). Future refinements to the IBIs will also be documented in this report. 2. Development of new IBIs With these objectives and constraints in mind, we undertook development of new fish and benthic macroinvertebrate IBIs for Maryland following the same steps used to develop the original MBSS IBIs:      develop the database; identify reference and degraded sites; determine the appropriate strata; test the candidate metrics; test and validate the indices. 753 2.1. MBSS database It is essential that the data used to develop IBIs (e.g., reference sites) are comparable to the data collected at the sites of concern (test sites). A virtue of the MBSS is that the same biological, chemical, physical habitat and land use data are collected for all sites used in stream assessment and indicator development. The MBSS is also ideal for the development of IBIs because the sampling protocols are rigorously applied through annual training sessions and a quality assurance program (Roth et al., 2005a). MBSS sites are selected using a probability-based design applied to all first- through fourth-order streams in Maryland based on a map scale of 1:100,000 (Roth et al., 2005b). Benthic macroinvertebrates are sampled in the spring and identified to genus or lowest practical taxon in the laboratory from 100organism subsamples. Fish are sampled in the summer using double-pass electrofishing of 75-m stream segments. Water chemistry and physical habitat data are collected from these same segments. Land use information is extracted from Maryland Office of Planning data for the catchments draining to each segment. For more details on MBSS methodologies see Roth et al. (2005b). As was done for the original IBIs and described in Roth et al. (2000), we developed an integrated dataset that included all site and landscape environmental variables linked to the biological data and their derived attributes, such as tolerance values and functional groups. The original IBIs were developed with data collected from 1994 to 1997 from a maximum of 1098 sites, divided into 732 calibration sites and 366 (33%) validation sites. The dataset for the new IBIs included all samples from 1994 to 2004, totaling 2508 sites with 353 (14%) reserved for validation. We believed this large number of sites provided us with enough reference sites to create reference conditions for additional classes of Maryland stream types. Small numbers of (or no) reference sites in a stream type (e.g., coldwater streams) prevent development of effective IBIs. 2.2. Better reference conditions Using reference sites that are minimally disturbed is perhaps the most important component of IBI development. If reference sites are only relatively less 754 M.T. Southerland et al. / Ecological Indicators 7 (2007) 751–767 disturbed than other sites (such sites are often referred to as least disturbed), assigning quality levels to IBI scores becomes problematic. Therefore, we reviewed the reference criteria used in the original MBSS IBIs (a site must meet all criteria to be designated as a reference site) to identify changes that would result in greater confidence that the new reference sites were minimally disturbed. We decided to retain the following reference criteria that we believe reflect levels at which individual stressors will probably not result in adverse effects to stream biota (Roth et al., 2000):  pH  6 or blackwater stream (pH < 6 and DOC  8 mg/l);  ANC  50 meq/l;  DO  4 ppm;  nitrate  300 meq/l (4.2 mg/l);  remoteness rating: optimal or suboptimal;  aesthetics rating: optimal or suboptimal;  in stream habitat rating: optimal or suboptimal;  no channelization;  no point source discharges. Because the remoteness variable was replaced with ‘‘distance to nearest road’’ in Round Two and the channel alteration variable was replaced with ‘‘channelization,’’ comparable replacement criteria were developed and applied to Round Two sites. Specifically, the surrogate ‘‘remoteness’’ variable was obtained by converting the distance to nearest road value p toffiffiffiaffiffiffiffi0–20 ffiffiffiffiffiffiffiffiffiffiffiscore ffiffiffiffiffiffiffiffiffiffiffiusing ffiffiffiffiffi the equation: ¼ 0:615 þ 0:733 meters from road (Paul et al., 2003). A regression of this new remoteness variable on the original variable yielded a reference criterion threshold of 70. For Round Two sites, the reference criterion of no channelization was indicated by a ‘‘no’’ value for the channelization variable. At the same time, we believed that the land use criteria were not strict enough to eliminate sites with adverse effects. Based on analysis of urban effects on stream condition (Vølstad et al., 2003b), the presence of original reference sites with relatively high levels of urban land (i.e., 5–20%) indicated that not all reference sites were minimally disturbed. Therefore, we changed the minimum allowable forested land use from 25% to 35% of the catchment area and the maximum allowable urban land use from 20% to 5% of the catchment area. In addition, studies have indicated that wider vegetated riparian buffers often ameliorate land use effects (see Naiman and Décamps, 1997; Lee et al., 2004), so the minimum allowable riparian buffer width was changed from 15 to 30 m. These changes in land use and riparian width thresholds resulted in a smaller proportion of stream sites meeting the reference site criteria. Using the original reference site criteria, 152 of the 1098 Round One sites (14%) were designated as reference sites. Using the new criteria, 196 of the total 2508 sites (8%) were designated as reference. Fig. 1 shows that the cumulative distribution of stream miles with equal or greater benthic macroinvertebrate IBI scores for Round Two sites meeting the new reference criteria is to the right (i.e., higher quality) than sites meeting the original reference criteria. This result is consistent with greater confidence that the sites are minimally disturbed and, since the total number of reference sites is greater than that used to develop the original IBIs, the characterization of reference condition should be more robust. We retained the criteria for degraded sites from the original IBI development procedures as follows (a site failing any one of the criteria is designated as a degraded site):  pH  5 and ANC 0 meq/l (except for blackwater streams, DOC  8 mg/l) (n = 23 sites);  DO  2 ppm (n = 20); Fig. 1. Cumulative distribution of stream miles with benthic macroinvertebrate IBI scores for: (1) MBSS sites sampled in 2000–2004 (solid line), (2) subset of 2000–2004 sites meeting original reference criteria (dashed line) and (3) subset of 2000–2004 sites meeting new reference criteria (dotted line). M.T. Southerland et al. / Ecological Indicators 7 (2007) 751–767  nitrate > 500 meq/l (7 mg/l) and DO < 3 ppm (n = 0);  in stream habitat rating poor and urban land use >50% of catchment area (n = 15);  in stream habitat rating poor and bank stability rating poor (n = 34);  in stream habitat rating poor and channel alteration rating poor (n = 69);  urban land use >50% of catchment area and riparian buffer width = 0 m (n = 48). A total of 170 of the 2508 sites (7%) were designated as degraded using these criteria. 2.3. Full range of natural variability While the original benthic macroinvertebrate IBI was applied to all sampled streams, the MBSS recognized that the reference sites did not adequately capture the natural variation of fish assemblages in small streams. This was in part due to the lower natural abundance of fish and fewer fish species in small streams, but also due to the small number of reference sites in these streams. Therefore, streams draining catchments of less than 300 ac (i.e., where the number of fish and fish species sampled were frequently less than 100 and 5, respectively) were not rated using the original fish IBI (Roth et al., 2000). This resulted in 98 (11%) of streams sampled from 1995 to 1997 being not rated for fish because of their small size (an additional 5% of sites were not rated because they were dry and therefore not sampleable in the summer). From 2000 to 2004, the MBSS sampled streams from a new 1:100,000-scale map that included a greater proportion of small streams than was sampled from 1995 to 1997. Specifically, while 11% of streams sampled in 1995–1997 drained <300 ac, 25% of streams sampled in 2000–2004 drained <300 ac. Only 5% of streams draining <300 ac had <100 fish sampled, so data limitation was not a justification for excluding all <300 ac streams. Therefore, we attempted to include these smaller streams in the development of the new fish IBI. We included all stream sizes in the analyses, creating a more representative but more variable reference condition; subsequently, we investigated partitioning this variability into separate small stream or coldwater stream type classes (see next section). 755 2.4. More classes of reference condition As described above, the 196 reference sites meeting the more restrictive reference criteria are representative of the best 8% of streams in Maryland. These reference sites were randomly selected within 84 primary sampling units (i.e., individual and combinations of Maryland 8-digit watersheds) and are distributed across all regions and stream types. This distribution allowed us to identify stream classes using empirical data, which is generally preferable to a priori classification (Hawkins et al., 2000). The goal of the classification step in indicator development is to partition the variability in reference condition into homogeneous regions or stream types that are best addressed with separate IBIs. The original fish IBIs were developed for three geographical strata: Coastal Plain, Eastern Piedmont (as defined by fish distributions ending at Great Falls) and Highlands (Fig. 2). The original benthic macroinvertebrate IBIs were developed for the Coastal Plain and non-Coastal Plain (Highlands and Eastern Piedmont combined). For the new IBIs, we performed cluster analyses as described in Roth et al. (2000) to identify groups of sites with similar biological assemblages (as represented by log-transformed percentages of species abundance). To ensure sufficient sample size, sites meeting the original reference criteria (i.e., 261 of the 2500 sites) were used in the analysis. Separate cluster analyses were done for fish and benthic macroinvertebrates. Another approach to identifying useful strata is to apply cluster analysis or other multivariate techniques to the metrics likely to be used in the IBIs (Angermeier et al., 2000). The cluster analyses in this study indicated that fish assemblages divided into four fairly distinct groups: Coastal Plain, Eastern Piedmont, small streams in the Highlands (draining <3000-ac catchments) and large streams in the Highlands (Fig. 2). Smaller clusters were also significantly different, but only these larger clusters could be associated with consistent abiotic variables (e.g., geographic boundaries or stream size). The differentiation among benthic macroinvertebrate assemblages was similar but less strong; comparisons of the benthic macroinvertebrate taxa list among the groups also indicated consistent differences. Distinct blackwater assemblages were not discernable in cluster analyses for either fish or benthic 756 M.T. Southerland et al. / Ecological Indicators 7 (2007) 751–767 Fig. 2. Map of ecological regions of Maryland and distribution of non-degraded sites with distinctly different fish assemblages. Each stratum represents sites that were found to be similar using cluster analysis. Thick lines denote the separation (from left to right) of the Highlands, Eastern Piedmont and Coastal Plain regions. Medium lines denote Maryland 8-digit watershed boundaries. Thin lines denote Maryland stream network at 1:100,000 scale. macroinvertebrates, nor were there enough blackwater reference sites to create a separate stratum. Nonetheless, other evidence, including analysis of sentinel blackwater sites (T. Prochaska, personal communication, 2005), indicates that there are differences that may justify not rating blackwater sites with the new fish and benthic macroinvertebrate IBIs. Using the original fish IBIs, the MBSS determined that many smaller streams meeting the reference criteria, and especially coldwater streams, were scoring lower than larger reference streams by approximately one-third. These erroneously low scores could lead to designating small streams as impaired when they are not. Therefore, we used both the segregated Highlands stream strata (small and large streams separately) and the combined Highlands stream stratum in subsequent indicator development steps. We also developed a coldwater streams stratum based on current and likely sustainable distributions of brook trout (M. Kline, personal communication, 2005) for use in indicator development. The coldwater stratum included all streams west of Evitts Creek in western Maryland. Isolated brook trout streams in the Catoctin Mountain area and parts of the Patapsco, Gunpowder and Susquehanna watersheds were not included in the coldwater streams stratum. The selection of each of these geographic or stream type strata has an ecological basis and potential for improving the performance of IBIs by reducing the variation in reference condition within each stratum. The number of reference sites occurring in each stratum is a practical limitation to IBI development. Bailey et al. (2004) recommends using 5–10 reference sites per class (stratum) as a minimum; experience of the MBSS indicates that 40 reference sites in each stratum is effective for developing IBIs. Even though more restrictive reference criteria were used to develop the new IBIs, the large dataset of 2500 sites still provided enough reference sites (approximately 40) for fish IBI development in each different stream type (Table 1). For the new benthic macroinvertebrate IBI, the coldwater stratum was not used because, unlike fish, benthic macroinvertebrate assemblages Table 1 Reference and degraded sites occurring in each geographic or stream type stratuma Coastal Plain Eastern Piedmont Warmwater Highlands Coldwater Highlands a Reference Degraded 52 43 53 48 82 40 35 13 Includes both calibration and validation data. M.T. Southerland et al. / Ecological Indicators 7 (2007) 751–767 are not typically depauperate in minimally disturbed coldwater streams. 2.5. Testing candidate metrics In developing the original fish and benthic macroinvertebrate IBIs, the MBSS compiled and tested more than 100 candidate metrics (Roth et al., 2000; Stribling et al., 1998). For the new IBIs, we retained all metrics that showed promise in the original analysis (i.e., all that had significantly different values for reference and degraded sites) and added selected new candidate metrics. The list of candidate metrics for the new fish IBI included 44 original metrics and the following new metrics:  Pirhalla (2004) habitat tolerance metrics;  log-transformed metrics that included sculpins (10 metrics including versions adjusted for catchment area);  observed/expected (O/E) for fish (Stranko et al., 2005). The list of candidate metrics for the new benthic IBI included 51 original metrics and the following new versions of original metrics calculated based on new benthic macroinvertebrate tolerance values (derived separately for urban and agricultural land use effects) calculated from the MBSS dataset (Bressler et al., 2004):     number and percentage of intolerants; percentage of tolerants; Hilsenhoff Biotic Index (HBI); Beck’s index. The log-transformed metrics were included because analysis indicated that the original fish IBIs might have been overly influenced by sculpin abundance. The other new metrics were not available when the original IBIs were developed. Each candidate fish metric and its predicted response to stress are described in Appendix B. Detailed results of testing each metric can be found in Southerland et al. (2005a). As was done for the original fish IBIs, metrics of fish abundance and species richness were tested within each stratum, both as raw values and values adjusted 757 for stream size (Roth et al., 2000). We tested all candidate metrics by comparing mean values and distributions between reference and degraded sites in each stratum, in combined strata and statewide. We also looked at including and excluding sites with no fish, sites draining <300 ac and sites with <60 benthic macroinvertebrates to evaluate these effects on the metrics. These different comparisons ensured that the usefulness of each metric for all possible IBIs was considered. The ability of fish metrics to discriminate between reference and degraded sites (i.e., the number of such sites correctly classified) was similar when sites draining <300 ac were included. Ann Roseberry Lincoln (personal communication, 2005) found no evidence of a bias in benthic IBIs resulting from small stream size. Only three reference sites had <60 benthic macroinvertebrates, so all sites were included in the metric testing. The first step in metric testing was to test for significant differences in: (1) the mean values between reference and degraded sites using the Mann–Whitney U-test and (2) the distributions of values using the Kolmogorov–Smirnov test. The next step was to score the metrics based on the distribution of values observed at reference sites within each stratum. In developing the original IBIs we scored each metric as 5, 3 or 1, depending on whether its value at a site approximates, deviates slightly from or deviates greatly from conditions at reference sites (Karr et al., 1986). In other IBI applications (e.g., Barbour et al., 1996; Fore et al., 1996; Lyons et al., 1996; Blocksom, 2003), a number of different methods have been used to establish scoring thresholds, based on varying subdivisions of observed values. For the new IBIs, we retained our discrete scoring approach so that direct comparisons with the original IBIs could be made. In our analysis, threshold values for each selected metric were established as approximately the 10th and 50th (median) percentile values for reference sites (see Fig. 3), and were established separately for each stratum. For each metric expected to decrease with degradation, values below the 10th percentile were scored as 1. Values between the 10th and 50th percentiles were scored as 3, as they fell short of median expected values for reference sites. Values above the 50th percentile were scored as 5. Scoring 758 M.T. Southerland et al. / Ecological Indicators 7 (2007) 751–767 Fig. 3. Schematic illustration of the process used to derive and interpret scores for the MBSS Indices of Biotic Integrity (IBIs). Scores are based on the distribution of reference sites, as depicted in the top figure. The bottom figure shows hypothetical reference sites in the context of other hypothetical sites, including those with known degradation. was reversed for metrics expected to increase with degradation (e.g., values below the 50th percentile were scored as 5, and values above the 90th percentile were scored as 1). In this method, both the upper and lower thresholds are independently derived from the distribution of reference site values. The 10th percentile threshold for designating scores of 1 represents our intent to identify values that are outside the natural expectation for reference sites. To test the discriminatory power of each candidate metric, we evaluated the degree of overlap between metric values at reference and degraded sites by examining the number of sites scoring above and below the lower threshold. A classification efficiency was calculated as the percent of reference sites with values scoring 3 plus degraded sites scoring <3, out of the total number of sites evaluated. Reference sites misclassified as degraded (score <3) and degraded sites misclassified as reference (score 3) make up the remainder of the sites. A high classification efficiency indicates a small amount of overlap between values for reference and degraded sites. In addition to overall classification efficiencies, classification efficiencies were also reported separately for reference and degraded sites. The term discrimination efficiency is often applied to the percentage of degraded sites alone that are correctly classified (Gerritsen et al., 2000). Most candidate metrics were significantly different between reference and degraded sites, and many had high classification efficiencies (i.e., exceeding 70%). Certain metrics in some strata exceeded 90%. Classification efficiencies were used as the primary means of selecting metrics for potential inclusion in the IBIs. Among similar metrics (e.g., number of species versus number of native species to describe species richness), the best performing metric (balanced across strata for core metrics) was used. The classification efficiencies of the fish abundance and richness metrics were very similar for both raw scores and scores adjusted for catchment area. We selected only adjusted metrics for inclusion in IBI testing because they make ecological sense and are consistent with the original MBSS IBIs. The lognormal metrics of sculpin abundance rarely had good classification efficiencies and were not selected; it is possible that sculpin absence at apparent reference sites is actually linked to current (or historical) degradation rather than unaccounted for natural differences. The observed/expected (O/E) metric for fish species could not be calculated for MBSS sites outside the values used to develop the model (i.e., later sample years), so the metric was not selected for IBI testing. In the future, refinement of the O/E models with more data may support its use as an independent indicator of stream condition. Some of the Pirhalla metrics performed adequately but were not better than traditional metrics, so they were not selected. The number of salamander species metric was also tested and had a high classification efficiency in the Coastal Plain and small stream Highlands, but was not selected because salamander sampling is not currently conducted at all MBSS sites. Some metrics with narrow thresholds, i.e., number of benthic species adjusted for M.T. Southerland et al. / Ecological Indicators 7 (2007) 751–767 catchment area, percent non-tolerant suckers (all suckers except white sucker), percent Tanytarsini, number of Ephemeroptera and number of Scrapers, are essentially presence/absence metrics (in some cases no scores of 3 were assigned). Three such metrics were included in the original benthic IBIs. We evaluated the effect of eliminating these ‘‘presence/ absence’’ metrics (or using the number of benthic species without adjustment) but determined that the original formulations performed better (i.e., had higher classification efficiencies). 2.6. IBI combinations and testing As with the original IBIs, we iteratively tested many combinations of metrics to develop the new fish and benthic macroinvertebrate IBIs. For each combination, an index was calculated as the mean of the metrics included, scaled from 1 to 5. Classification efficiencies of different metric combinations (indices) were calculated as above, separately for reference and degraded sites, and overall. Individual IBI combinations were done separately for each of the provisional strata. At first, the combinations of metrics for IBI testing were selected in a stepwise manner, starting with the best performing metric (i.e., highest classification efficiency). Additional metrics were added as long as they increased the overall classification efficiency of the index. In no stratum did the classification efficiency improve after a second metric was added. This is a result of the very high classification efficiencies achieved by individual metrics in each stratum. To ensure that the final IBIs were a more complete representation of the fish and benthic assemblages (as recommended by Karr et al., 1986 and done for the original MBSS IBIs), we selected a core of four metrics that performed well and represented different assemblage characteristics for each of the strata (in the coldwater Highlands stratum, only two of these core metrics were used). The core metrics for the new fish IBI were abundance per square meter; number of benthic species (adjusted for catchment area); percentage of tolerant fish; percentage of generalists, omnivores and invertivores. Only the abundance per square meter and number of benthic species (adjusted for catchment area) core metrics were included in the coldwater Highlands fish IBI. The core metrics for the 759 new benthic IBI were number of taxa; number of Ephemeroptera, Plecoptera and Trichoptera (EPT) taxa; number of Ephemeroptera taxa; percentage of benthic macroinvertebrates intolerant to urban stress (after Bressler et al., 2004). The core fish metrics represent four of the five assemblage components identified by Karr et al. (1986): species richness and composition, indicator species, trophic composition and fish abundance and condition. The reproductive function component was not represented in the core metrics as no reproductive metrics had high classification efficiencies. Subsequently, we attempted to improve the performance of the IBIs by adding other metrics to the core suite in the same stepwise fashion. Additional metrics were added until they no longer improved the classification efficiency of the index. The provisional small stream Highlands stratum had the lowest classification efficiency and the large stream Highlands stratum had so few degraded sites that its performance was suspect. At the same time, the coldwater stratum performed well and effectively captured most of the streams draining catchments <5000 acres, so the small stream Highlands stratum was abandoned and the remaining Highlands streams combined as a separate warmwater Highlands stratum (i.e., the remaining Highlands streams outside the geographic boundaries of the coldwater streams stratum). For the four final strata, two additional metrics were added to each new fish IBI, improving the calibration classification efficiency to at least 83%. For the final three new benthic macroinvertebrate IBIs, two to four additional metrics were added to the core suite, improving the calibration classification efficiency to at least 85%. 2.7. IBI validation As described above, we reserved 353 of all sites sampled from 1994 to 2004 for validation of the new MBSS IBIs. This number of sites is comparable to the number of validation sites used to develop the original MBSS IBIs. For the fish IBIs in the Highlands and Coastal Plain, the overall classification efficiencies of the validation sites were even higher than the calibration classification efficiencies (88%). The validation classification efficiency for the fish IBI in the Eastern Piedmont was 760 M.T. Southerland et al. / Ecological Indicators 7 (2007) 751–767 lower at 71%, but this validation is less reliable because only seven reference and degraded sites from this region were in the validation dataset by chance. The validation classification efficiencies for the benthic IBI were higher than for calibration in the Coastal Plain at 96%, somewhat lower in the Eastern Piedmont at 86% (but again there were only seven validation sites in this stratum) and comparable in the Highlands at 88%. These high classification efficiencies using only validation sites indicate that the performance of the IBIs was not derived from overfitting to the calibration dataset. Therefore, the IBIs are likely to be robust when applied to new data. 3. Comparison of original and new IBIs Using the indicator development process described above, we created new MBSS fish and benthic macroinvertebrate IBIs as shown in Tables 2 and 3. The new fish IBIs differ from the original IBIs in that they divide the original Highlands stratum into two strata, one for coldwater Highlands streams and one for the remaining warmwater Highlands streams. In addition, smaller streams (i.e., those draining <300 ac catchments) that were not included in the original fish IBI development have been included in the new IBIs; therefore the new IBIs can be applied to these smaller streams (25% of stream miles in 2000–2004). The new benthic IBIs differ from the original IBIs in that they divide the original non-Coastal Plain stratum into new Highlands and Eastern Piedmont strata. As with the original benthic and new fish IBIs, smaller (<300 ac) streams are included in the new benthic IBIs. The number and composition of metrics differ between the new and original IBIs. The following metrics from the original fish IBIs are included in the new fish IBIs for the same strata: number of benthic species (adjusted for catchment area); percent tolerants; percent generalists, omnivores and invertivores. The abundance per square meter metric that appeared in the original Coastal Plain and Eastern Piedmont fish IBIs is now in all four new fish IBIs. The new Coastal Plain fish IBI has only six metrics compared to the eight metrics in the original IBI, and the only new metric is the percent non-tolerant suckers (i.e., all suckers except white sucker). The new Eastern Table 2 New fish IBIs for Maryland by stratum and with metric scoring thresholds Thresholds 1 Coastal Plain Abundance per square meter Number of benthic species adjusted Percent tolerants Percent generalists, omnivores, invertivores Percent non-tolerant suckers (all suckers except white sucker) Percent abundance of dominant species Eastern Piedmont Abundance per square meter Number of benthic species adjusted Percent tolerants Percent generalists, omnivores, invertivores Biomass per square meter Percent lithophilic spawners Warmwater Highlands Abundance per square meter Number of benthic species adjusted Percent tolerants Percent generalists, omnivores, invertivores Percent insectivores Percent abundance of dominant species Coldwater Highlands Abundance per square meter Percent tolerants Percent brook trout Percent sculpins 3 5 <0.45 0 >97 100 0.72 0.22 68 92 0 2 >69 40 <0.25 <0.09 >68 100 1.25 0.26 45 80 <4.0 <32 8.6 61 <0.31 <0.11 >80 >96 0.65 0.25 39 61 <1 >89 33 38 2.24 0.81 0 0 0.88 0.22 0.14 0.44 Piedmont fish IBI has six metrics compared to the nine metrics in the original IBI and includes no new metrics. The new warmwater Highlands fish IBI has six metrics compared to the seven metrics in the original Highlands IBI, while the new coldwater Highlands fish IBI has only four metrics, including two new metrics appropriate to its stream type: percent brook trout and percent sculpins. The following metrics from the original benthic IBIs are included in the new benthic IBIs for the same strata: number of taxa and number of EPT taxa. The new Coastal Plain IBI also includes the percent Ephemeroptera and number of scraper taxa metrics M.T. Southerland et al. / Ecological Indicators 7 (2007) 751–767 Table 3 New benthic IBIs for Maryland by stratum and with metric scoring thresholds Thresholds 1 3 5 Coastal Plain Number of scraper taxa Number of EPT taxa Number of Ephemeroptera taxa Percent intolerant to urban Percent Ephemeroptera Number of Scraper taxa Percent climbers <14 <2 <1 <10 <0.8 <1 <0.9 22 5 2 28 11 2 8 Eastern Piedmont Number of taxa Number of EPT taxa Number of Ephemeroptera taxa Percent intolerant to urban Percent Chironomidae Percent clingers <15 <5 <2 <12 >63 <31 25 11 4 51 24 74 Combined Highlands Number of taxa Number of EPT taxa Number of Ephemeroptera taxa Percent intolerant to urban Percent Tanytarsini Percent scrapers Percent swimmers Percent Diptera <15 <8 <3 <38 <0.1 <3 <3 >50 24 14 5 80 4 13 18 26 from the original IBI, plus three new metrics: number of Ephemeroptera taxa, percent intolerant to urban stressors and percent climbers. The new benthic IBI for Eastern Piedmont includes number of Ephemeroptera taxa, and the new benthic IBI for Highlands includes number of Ephemeroptera taxa and percent Tanytarsini, both of which were included in the original non-Coastal Plain IBI. The Eastern Piedmont benthic IBI has six metrics and the Highlands IBI has eight metrics compared to the nine metrics in the original non-Coastal Plain IBI. The new Eastern Piedmont benthic IBI includes three new metrics and the Highlands IBI four new metrics. In addition to including different combinations of metrics, the new IBIs have different scoring thresholds. Because a new set of reference sites was used to develop the new fish and benthic IBIs, the metric values at the 10th and 50th percentiles of reference were different. For example, the degradation threshold (above which a score of 3 is given) for the abundance 761 per square meter metric changed from 0.42 (old) to 0.45 (new) in the Coastal Plain fish IBIs and from 0.56 to 0.25 in the Eastern Piedmont fish IBIs. For the Coastal Plain benthic macroinvertebrate IBIs, the degradation threshold for number of taxa changed from 11 to 14 and the threshold for the number of EPT taxa changed from 3 to 2. Larger changes occurred for some metrics and are attributable to changes in the reference condition resulting from stricter criteria, more small streams and chance. 3.1. Comparison of how original and new IBIs score reference condition As described above, the new MBSS IBIs were developed using a more restrictive set of reference sites (8% of all sites versus 14% of all sites for the original IBIs). Because stricter thresholds for land use and riparian disturbance were applied, we are more confident that the new IBI reference sites are minimally disturbed. At the same time, the reference sites for the new fish IBIs included smaller streams draining <300 ac that were not included in the original fish IBI. In addition, the sampling design for 2000–2004 on the 1:100,000-scale stream network resulted in more small streams being sampled. The distribution of the new reference sites included 38% that were <1000 ac, compared to 19% of the original reference sites. Including more small streams in the new reference condition means that more natural variability is included in the new IBIs, but that they are representative of smaller streams as well (i.e., they can now be rated). The mean score for all new reference sites was 3.7 using the original fish IBIs and 4.0 using the new fish IBIs. Mean reference site scores were 3.6 for the original benthic IBIs and 3.9 using the new benthic IBIs. For reference sites <1000 ac, the mean benthic IBI was 3.7 compared to 3.6 for larger streams and the fish IBI was 3.3 compared to 3.9. This indicates that smaller streams still scored somewhat lower on the fish IBI using the new fish IBI. 3.2. Comparison of how all streams score with original and new IBIs We conducted a direct comparison of the original and new MBSS IBIs (both fish and benthic macroinvertebrate) by applying them to the 2000–2004 762 M.T. Southerland et al. / Ecological Indicators 7 (2007) 751–767 MBSS dataset of 1367 sites. The statewide mean for the new fish IBIs was virtually unchanged with an original IBI of 2.91 and a new IBI of 2.93. The statewide mean of the new benthic IBIs was only 3% higher, increasing from 2.96 to 3.07. On a regional basis, the greatest difference in mean scores between original and new IBIs was an increase of 0.64 (16%) for the Coastal Plain benthic IBI and 0.32 (8%) for the Highlands fish IBIs. In the other regions, the mean benthic IBIs decreased 9% in the Highlands and 2% in the Eastern Piedmont, using the new IBIs. The mean fish IBIs decreased 5% in the Coastal Plain while staying the same in the Eastern Piedmont. On a county basis, 17 (one-third) of the 48 possible original and new mean IBI pairs (24 counties times both fish and benthic IBIs) changed by 0.5 units or more. The greatest increase was 1.14 for the benthic IBI in Caroline County and the greatest decrease was 0.58 for benthic IBI in Frederick County. Most of these changes were for the benthic IBIs in the nonCoastal Plain, which was separated into Highlands and Eastern Piedmont strata in the new benthic IBIs. The distributions of stream miles among the four MBSS condition classes (good, fair, poor, very poor) were also somewhat different between the original and new IBIs (Fig. 4). For both the new fish and benthic IBIs, the proportion of stream miles statewide changed by less than 10% in each condition class. The new Highlands fish IBI resulted in 16% more good streams and fewer fair and very poor streams. The increase in proportion of good streams is likely attributable to the appropriately higher scores for coldwater streams which have their own stratum in the new fish IBIs. Overall, the distribution of stream miles in each condition class was more even with the new IBIs than with the original IBIs. The influence of the new coldwater Highlands fish IBI is also evident when comparing it to the original Highlands fish IBI. The mean score for coldwater streams was 3.56 with the new fish IBI and 2.75 with the original fish IBI, an increase of 0.81 (20%). Although a separate blackwater stratum was not developed, the new Coastal Plain fish IBI rates blackwater streams 0.34 (8%) higher and the new Coastal Plain benthic IBI rates them 0.74 (18%) higher. This is likely due to the greater number of blackwater reference sites sampled in 2000–2004 and used to develop the new IBIs; only 24 blackwater Fig. 4. Percentage of stream miles in each condition class statewide for 2000–2004 sites scored with original (98) and new (05) MBSS IBIs. Fish and benthic IBIs are shown separately. reference sites were used to develop the original IBIs, while 64 were used for the new IBIs. Even though the blackwater stream type may still be scored lower than other types, the new IBIs better represent the expectation for natural blackwater streams. The different IBI scores that result from using the new IBIs rather than the old IBIs also affect the designations of watersheds as impaired according to Maryland’s biological criteria (MDE, 2005). These biological criteria are applied to Maryland 8-digit watersheds (or combined watershed Primary Sampling Units, PSUs) with 10 or more MBSS sample sites. Mean IBIs and one-sided 90% confidence interval values are calculated to give one of three ratings:  Does not meet criteria ( fails): If the mean and upper bound of the one-sided 90% confidence interval (upper) of either index (FIBI or BIBI) is less than 3.0, the 8-digit watershed (or PSU) is listed as failing to meet the proposed criteria.  Meets criteria ( passes): If the mean and lower bound of the one-sided 90% confidence interval M.T. Southerland et al. / Ecological Indicators 7 (2007) 751–767 (lower) of both indices (FIBI and BIBI) are greater than or equal to 3.0, the 8-digit watershed (or PSU) is listed as meeting the proposed criteria.  Inconclusive: All other cases are inconclusive. Applying the original MBSS IBIs to 2000–2004 data, 40 watersheds fail, 37 are inconclusive and 7 pass biological criteria; using the new MBSS IBIs, 31 watersheds fail, 41 are inconclusive and 12 pass. Overall, 22% fewer watersheds fail biological criteria with the new IBIs. The most frequent changes in the designation of individual watersheds are the 17 watersheds that failed with the original IBIs, but that are inconclusive with the new IBIs. In addition, among the 37 watersheds that were inconclusive with the original IBIs, 24 (65%) remain inconclusive with the new IBIs, while 5 pass and 8 fail. 4. Discussion As stated at the outset, refinements to biological indicators can improve the information on stream condition provided to water resource managers. Such changes also entail development costs and infrastructure costs to the host and cooperating programs. Therefore, indicator changes should only be undertaken with specific goals in mind. In Maryland, the development of new fish and benthic macroinvertebrate IBIs achieved the goals of: (1) increasing confidence that the reference conditions are minimally disturbed (less disturbed reference sites), (2) including more natural variation across the geographic regions and stream types of Maryland (including small streams among reference sites so that these streams could be rated) and (3) increasing sensitivity of IBIs with more classes (one additional stratum each for the fish IBI and benthic macroinvertebrate IBI) and different metric combinations (new IBIs with higher classification efficiencies). This was largely possible because of the large number of stream sites in the 1994–2004 MBSS dataset. At the same time, adoption of the new IBIs does entail costs to the MBSS and cooperating programs in the form of: (1) making assessments of stream condition using the original IBIs obsolete, (2) changing impairment decisions, (3) requiring staff and resources to implement the new IBIs and (4) 763 potentially affecting consistency with stream monitoring programs that choose to retain the original IBIs. Nonetheless, these new MBSS IBIs are generally consistent with the old IBIs; they remain transparent and understandable, and they provide clear thresholds of impairment for both the biointegrity and interim (fishable and swimmable) water quality goals. Using the new IBIs it is still possible to calculate joint estimates between sampling rounds and detect trends in stream condition. 4.1. Minimally disturbed reference condition The reference conditions used to develop the new MBSS IBIs represented only 8% of all stream sites in Maryland. They did not include original reference sites that were most likely to be affected by land use changes. For these reasons, we are more confident that the new IBIs are based on minimally disturbed reference conditions for Maryland streams. This is important so that degraded streams are not identified as meeting reference conditions. 4.2. IBIs that better predict degradation Given minimally disturbed reference conditions, the ability of IBIs to distinguish deviation from those reference conditions is based on how predictably IBI scores change with disturbance. This ability to predict deviation comes from: (1) choosing metrics that vary predictably and precisely with disturbance and (2) combining these metrics into an index that consistently changes with disturbance across the natural variation gradients encountered. We reduced the natural variation that each new IBI had to address by increasing the number of geographic or stream type classes, i.e., the number of new MBSS IBIs. We now have four rather than three fish IBIs and three rather than two benthic IBIs. In the case of the new fish IBIs, we increased the natural variation of reference condition by adding smaller streams <300 ac, but this did not adversely affect the performance of the new IBIs given the four strata. Within each stratum (i.e., new IBI), the combination of metrics changed from the old IBIs and in every case the ability of the IBI to distinguish reference from degraded sites (i.e., the classification efficiency) increased (Table 4). By convention, classification 764 M.T. Southerland et al. / Ecological Indicators 7 (2007) 751–767 Table 4 Comparison of classification efficiencies (CEs) between original and new MBSS IBIs Region Fish IBI Coastal Plain Eastern Piedmont Highlands Warmwater Highlands Coldwater Highlands Benthic IBI Coastal Plain Non-Coastal Plain Eastern Piedmont Highlands Original IBI calibration (validation) CE New IBI calibration (validation) CE Reference CE (%) Degraded CE (%) 74 (72) 90 (94) 86 (75) 85 (88) 96 (71) 89 95 80 97 83 (88) 85 85 78 87 (96) 89 83 93 (86) 91 (88) 94 93 92 88 87 (72) 88 (82) CEs are the percentage of reference and degraded sites that are correctly classified by each IBI. efficiencies above 80% are good and above 90% are excellent. In addition to these good-to-excellent overall IBI classification efficiencies, each new IBI was effective at correctly classifying both reference and degraded sites (Table 4). Misclassification of reference sites (saying they are degraded when they are not) is essentially a false positive or Type I error. Among the new fish IBIs, the classification efficiencies for reference sites ranged from 80% to 95%; among the new benthic IBIs, these classification efficiencies ranged from 89% to 94%. Misclassification of degraded sites (saying they are not degraded when they are) is essentially a false negative or Type II error. Among the new fish IBIs, the classification efficiencies for degraded sites ranged from 78% to 97%; among the new benthic IBIs, these classification efficiencies ranged from 83% to 92%. Low misclassification rates for both reference and degraded sites indicate that the new MBSS IBIs are a good balance between both types of error, i.e., not many degraded streams will be missed, nor will we be ‘‘crying wolf’’ about streams that are actually not degraded. This is important because indicators that make frequent errors or are biased in one direction are of little use for management decisions. 4.3. Applying the new MBSS IBIs The MBSS IBIs are central to water resource management in Maryland and have special implications for the designation of watersheds as impaired under Section 303d of the Clean Water Act (CWA). Therefore, it is critical that stream condition ratings be founded in ecological knowledge and solid science. The MBSS recognizes that there are no truly pristine streams left in Maryland; most have a history of human disturbance and all are affected by atmospheric deposition. Nonetheless, there are high quality streams in Maryland that can be accurately called minimally disturbed and equated with Biological Condition Gradient (BCG) level 2, ‘‘minimal changes in structure and function’’ (EPA, 2005). Adoption of the new IBIs provides us with more confidence that the reference conditions we are using to create IBIs and rate stream condition reflect BCG level 2, rather than BCG level 3, ‘‘evident changes in structure and minimal changes in function.’’ In addition to indicating when stream condition deviates from reference condition (i.e., is degraded), IBIs provide a means of determining the degree to which streams deviate or the ‘‘severity of failing’’ to meet the criterion (Bailey et al., 2004). The original MBSS IBIs used four ‘‘bands’’ of IBI scores to designate stream condition: 1.0–1.9 very poor, 2.0–2.9 poor, 3.0–3.9 fair and 4.0–5.0 good. This convention was retained for the new IBIs. Given the new reference conditions, these bands can be more confidently assigned to the biointegrity goal of CWA (good) and the interim goal of CWA (fair). The two additional bands (i.e., the poor and very poor classes of stream condition) are consistent with variability in stream condition relative to reference condition. Creation of more stream condition bands is not justified by the precision of the IBIs. The limits on M.T. Southerland et al. / Ecological Indicators 7 (2007) 751–767 IBI precision are to be expected, as IBIs balance sensitivity to degradation and incorporation of natural variability. While IBIs are founded in the concept of biological integrity, they are only a rough approximation of the ecological structure and function of stream resources. We argue that protection of biological diversity in its most expansive definition (CEQ, 1993; Noss and Cooperrider, 1994) cannot be achieved solely through the use of IBIs. Augmented or separate monitoring and assessment focused on rare species and habitats is needed to fully protect stream ecosystems (see Kazyak et al., 2005). 4.4. Continuity across the original and new IBIs We determined that the improvements in the performance of the new IBIs, especially the more accurate coldwater Highlands and Coastal Plain fish IBIs, and the ability to rate the large number of small streams with the fish IBI warranted adoption of the new IBIs. At the same time, the final construction of the new fish and benthic IBIs for the MBSS is very similar to the original MBSS IBIs. The basis in reference condition, the discrete 1–3–5 scoring and the four bands of stream condition were retained (an analysis showing that the use of continuous 0–100 scoring based on the range of values for all sites does not improve on discrete scoring is described in Southerland et al., 2005a). More elaborate modeling of reference condition (e.g., independent of geographic or stream type classification) was not incorporated. While new IBIs need to be calculated (using new metric combinations and thresholds), the IBI application process is unchanged. This is important so that cooperating programs (e.g., many Maryland counties) do not have to undertake infrastructure changes to continue their programs. As needed, the new MBSS IBIs can be calculated for stream sites sampled in the past to maintain continuity of the long-term MBSS dataset. It is also possible to convert IBI results between different sampling periods by using regressions between the original and new IBIs. In general the regression R2 are about 0.75; lower R2 occurs for the non-Coastal Plain benthic IBI where two new strata have been created and for the original Highlands fish IBI when compared to the new coldwater fish IBI, as expected. 765 Five of the metrics in the new MBSS benthic IBIs are shared by the benthic indices (Stream Condition Indices) of Virginia and West Virginia. The metric combinations in these indices performed adequately in Maryland but with lower classification efficiencies. Similarly, the new MBSS IBIs also share metrics with the Montgomery County IBIs. Comparability studies (Vølstad et al., 2003a; Southerland et al., 2005b) indicate that the indices for all these programs can be readily integrated. Acknowledgements We would like to thank the entire MBSS team, including field crews, laboratory staff and analysts, that sustain the program and provide the data needed to develop and apply the IBIs. Funding for this effort came primarily from the Maryland DNR, but was supplemented by U.S. EPA. Many other cooperators contributed valuable discussion. Useful comments on the manuscript were provided by Paul Angermeier and Greg Pond. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/ j.ecolind.2006.08.005. References Angermeier, P.L., Smogor, R.A., Stauffer, J.R., 2000. Regional framework and candidate metrics for assessing biotic integrity in Mid-Atlantic Highland streams. Trans. Am. Fisheries Soc. 129, 962–981. Bailey, R.C., Norris, R.H., Reynoldson, T.B., 2004. Bioassessment of Freshwater Ecosystems: Using the Reference Condition Approach. Springer Science/Business Media, Inc., New York, NY. Barbour, M.T., Gerritsen, J., Griffith, G.E., Frydenborg, R., McCarron, E., White, J.S., 1996. A framework for biological criteria for Florida streams using benthic macroinvertebrates. J. N. Am. Benthol. Soc. 15, 185–211. Blocksom, K., 2003. A performance comparison of metric scoring methods for a multimetric index for Mid-Atlantic Highlands streams. Environ. Manage. 31 (5), 670–682. Bressler, D.W., Paul, M.J., Stribling, J.B., 2004. Development of Tolerance Values for Benthic Macroinvertebrates in Maryland. 766 M.T. Southerland et al. / Ecological Indicators 7 (2007) 751–767 Tetra Tech, Inc., for Versar, Inc., and Maryland Department of Natural Resources. Council on Environmental Quality (CEQ), 1993. Incorporating Biodiversity Considerations Into Environmental Impact Analysis Under the National Environmental Policy Act. Council on Environmental Quality, Executive Office of the President, Washington, DC. U.S. Environmental Protection Agency (EPA), 2005. Use of Biological Information to Tier Designated Aquatic Life Uses in State and Tribal Water Quality Standards. Review Draft. U.S. Environmental Protection Agency, Washington, DC. Fore, L.S., Karr, J.R., Wisseman, R.W., 1996. Assessing invertebrate responses to human activities: evaluating alternative approaches. J. N. Am. Benthol. Soc. 15, 212–231. Gerritsen, J., Burton, J., Barbour, M.T., 2000. A Stream Condition Index for West Virginia Wadeable Streams. Report by Tetra Tech, Inc., to U.S. EPA Region 3, Environmental Services Division and U.S. EPA, Office of Science and Technology and Office of Water, Washington, DC. Hawkins, C.P., Norris, R.H., Gerritsen, J., Hughes, R.M., Jackson, S.K., Johnson, R.K., Stevenson, R.J., 2000. Evaluation of the use of landscape classifications for the prediction of freshwater biota: synthesis and recommendations. J. N. Am. Benthol. Soc. 19, 541–556. Karr, J.R., Dudley, D.R., 1981. Ecological perspective on water quality goals. Environ. Manage. 5, 55–68. Karr, J.R., Fausch, K.D., Angermeier, P.L., Yant, P.R., Schlosser, I.J., 1986. Assessment of Biological Integrity in Running Waters: A Method and its Rationale, vol. 5. Illinois Natural History Survey Special Publication, Champaign, IL. Kazyak, P.F., Kilian, J.V., Stranko, S.A., Hurd, M.K., Boward, D.M., Millard, C.J., Schenk, A., 2005. Maryland Biological Stream Survey 2000–2004. Stream and Riverine Biodiversity, vol. 9. Report to Monitoring and Non-Tidal Assessment Division, Maryland Department of Natural Resources. CBWPMANTA-EA-05-6. Kilian, J.V., 2004. Fish Assemblages in Non-tidal Streams of Maryland: The Influences of Assemblage Spatial Variability on the Index of Biotic Integrity. Master of Science Thesis. University of Maryland, Frostburg. Klauda, R., Kazyak, P., Stranko, S., Southerland, M., Roth, N., Chaillou, J., 1998. The Maryland biological stream survey: a state agency program to assess the impact of anthropogenic stresses on stream habitat quality and biota. Environ. Monit. Assess. 51, 299–316. Lee, P., Smyth, C., Boutin, S., 2004. Quantitative review of riparian buffer guidelines from Canada and the United States. J. Environ. Manage. 70, 165–180. Lyons, J., Wang, L., Simonson, T.D., 1996. Development and validation of an Index of Biotic Integrity for coldwater streams in Wisconsin. N. Am. J. Fisheries Manage. 16, 241–256. Maryland Department of Environment (MDE), 2005. Water Quality Standards. http://www.mde.state.md.us/ResearchCenter/Data/ WaterQualityStandards. Naiman, R.J., Décamps, H., 1997. The ecology of interfaces: riparian zones. Annu. Rev. Ecol. Syst. 28, 621–658. Noss, R.F., Cooperrider, A.Y., 1994. Saving Nature’s Legacy: Protecting and Restoring Biodiversity. Defenders of Wildlife. Island Press, Washington, DC. Paul, M.J., Stribling, J.B., Klauda, R., Kazyak, P., Southerland, M., Roth, N., 2003. A Physical Habitat Index for Freshwater Wadeable Streams in Maryland. Monitoring and Non-Tidal Assessment Division, Maryland Department of Natural Resources. CBWP-MANTA-EA-03-4. Pirhalla, D.E., 2004. Evaluating fish-habitat relationships for refining regional Indexes of Biotic Integrity: development of a tolerance index of habitat degradation for Maryland stream fishes. Trans. Am. Fisheries Soc. 133, 144–159. Roth, N., Erb, L., Rogers, G., Vølstad, J., 2005a. Maryland Biological Stream Survey 2000–2004. Quality Assurance Report, vol. XV. Report to Monitoring and Non-Tidal Assessment Division, Maryland Department of Natural Resources. Roth, N.E., Vølstad, J.H., Erb, L., Weber, E., Kazyak, P., Stranko, S., Boward, D., 2005b. Maryland Biological Stream Survey 2000–2004. Laboratory, Field, and Analytical Methods, vol. V. Report to Monitoring and Non-Tidal Assessment Division, Maryland Department of Natural Resources. CBWPMANTA-EA-05-3. Roth, N.E., Southerland, M.T., Chaillou, J.C., Kazyak, P.F., Stranko, S.A., 2000. Refinement and Validation of a Fish Index of Biotic Integrity for Maryland Streams. Report to Monitoring and NonTidal Assessment Division, Maryland Department of Natural Resources. CBWP-MANTA-EA-00-2. Roth, N., Southerland, M., Chaillou, J., Klauda, R., Kazyak, P., Stranko, S., Weisberg, S., Hall Jr., L., Morgan II, R., 1998. Maryland Biological Stream Survey: development of an Index of Biotic Integrity for fishes in Maryland streams. Environ. Monit. Assess. 51, 89–106. Southerland, M., Rogers, G., Kline, M., Morgan, R., Boward, D., Kazyak, P., Klauda, R., Stranko, S., 2005a. New Biological Indicators to Better Assess Maryland Streams. Prepared for Monitoring and Non-Tidal Assessment Division, Maryland Department of Natural Resources. Southerland, M., Vølstad, J., Erb, L., Weber, E., Rogers, G., 2005b. Proof of Concept for Integrating Bioassessment Results from Three State Probabilistic Monitoring Programs. Report to U.S. Environmental Protection Agency, Mid-Atlantic Integration Team, Fort Meade, MD. Southerland, M.T., Jung, R.E., Baxter, D.P., Chellman, I.C., Mercurio, G., Vølstad, J.H., 2004. Streamside salamanders as indicators of stream quality in Maryland. Appl. Herpetol. 2, 23–46. Stranko, S.A., Hurd, M.K., Klauda, R.J., 2005. Applying a large, statewide database to the assessment, stressor diagnosis, and restoration of stream fish communities. Environ. Monit. Assess. 23, 99–121. Stribling, J.B., Jessup, B.K., White, J.S., Boward, D., Hurd, M., 1998. Development of a Benthic Index of Biotic Integrity for Maryland Streams. Report to Monitoring and Non-Tidal Assessment Division, Maryland Department of Natural Resources. CBWP-MANTA-EA-98-3. Vølstad, J.H., Neerchal, N.K., Roth, N.E., Southerland, M.T., 2003a. Combining biological indicators of watershed condition from M.T. Southerland et al. / Ecological Indicators 7 (2007) 751–767 multiple sampling programs—a case study from Maryland, USA. Ecol. Indicators 3, 13–25. Vølstad, J.H., Roth, N.E., Mercurio, G., Southerland, M.T., Strebel, D.E., 2003b. Using environmental stressor information to predict the ecological status of Maryland non-tidal streams as 767 measured by biological indicators. Environ. Monit. Assess. 84, 219–242. Weisberg, S.B., Ranasinghe, J.A., Dauer, D.M., Schaffner, L.C., Diaz, R.J., Frithsen, J.B., 1997. An estuarine benthic Index of Biotic Integrity (B-IBI) for Chesapeake Bay. Estuaries 20, 149–158.