• E-mail
• Print
• Comment
• Font Size
• Digg
• del.icio.us
• Discuss article

Development of Empirical, Geographically Specific Water Quality Criteria: a Conditional Probability Analysis Approach1

Posted on: Sunday, 23 October 2005, 03:01 CDT

By Paul, John F; McDonald, Michael E

ABSTRACT:

The need for scientifically defensible water quality standards for nonpoint source pollution control continues to be a pressing environmental issue. The probability of impact at differing levels of nonpoint source pollution was determined using the biological response of instream organisms empirically obtained from a statistical survey. A conditional probability analysis was used to calculate a biological threshold of impact as a function of the likelihood of exceeding a given value of pollution metric for a specified geographic area. Uncertainty and natural variability were inherently incorporated into the analysis through the use of data from a probabilistic survey. Data from wadable streams in the midAtlantic area of the U.S. were used to demonstrate the approach. Benthic macroinvertebrate community index values (EPT taxa richness) were used to identify impacted stream communities. Percent fines in substrate (silt/clay fraction, < 0.06 mm) were used as a surrogate indicator for sedimentation. Thresholds of impact due to sedimentation were identified by three different techniques, and were in the range of 12 to 15 percent fines. These values were consistent with existing literature from laboratory and field studies on the impact of sediments on aquatic life in freshwater streams. All results were different from values determined from current regulatory guidance. Finally, it was illustrated how these thresholds could be used to develop criterion for protection of aquatic life in streams. (KEY TERMS: sediment; wadable streams; benthic community condition; statistical analysis; aquatic ecosystems; standards.)

INTRODUCTION

A range of procedures are being used around the world for establishing criteria for the protection of water quality (Jimenez et al., 1999; Yin et al., 2003; Borja et al, 2004; Kamizoulis and Saliba 2004, Kay et al, 2004). In the United States, the U.S. Environmental Protection Agency (USEPA) is responsible for implementing the Clean Water Act (CWA) (Russo, 2002), which is the major national act for protecting water quality. The USEPA implements some aspects of the CWA by providing guidance for the control of pollutants through development of Water Quality Standards (WQS). These WQS serve as the foundation for pollution control and are a fundamental component of water quality management. They define the goals for a waterbody by designating its uses, setting criteria to protect those uses, and protecting water quality through antidegradation provisions. The criteria are developed for the protection of aquatic life as well as for human health.

Water quality criteria (WQC) for individual chemical pollutants (such as heavy metals and synthetic organic compounds) have been developed as national criteria (e.g., USEPA, 1994). These national criteria have been developed through laboratory bioassays, where exposure to a single pollutant can be maintained under controlled conditions (Hohreiter and Rigg, 2001; Rausina et al, 2002; Fisher and Burton 2003). As progress has been made in controlling these individual pollutants, a shift has occurred toward control of nonpoint source pollution (e.g., runoff, nutrients, and sedimentation). Consistent with this is the increased use of biological indicators to assess the condition of the environment (Niemi and McDonald, 2004). Recent developments suggest that adopting national criteria may not be sufficiently protective of the biota in various subregions (Perry and Vanderklein, 1996; USEPA, 200Oa), thus leading to a greater reliance on field generated data. The recent development of WQC for nutrients is an example of geographically specific criteria developed from field data (USEPA, 200Oa).

Sedimentation in streams is an example of a nonpoint source pollution problem (Spooner et al., 1991). Excessive sediment is a major cause of impairment in waterbodies across the country (USEPA, 2002). Development of water quality criteria for suspended and bedded sediments for the protection of aquatic life provides a challenge since the traditional approach using laboratory bioassays (dose-response studies) may not be applicable (Perry and Vanderklein, 1996).

A focus on the response of aquatic communities to sedimentation emphasizes protection of these communities from adverse effects of sedimentation and is consistent with the use of biological criteria. Establishing a criterion for sedimentation allows the source of excess sediments to be addressed for regulation or remediation of the problem. However, identification of the source of the sediments is not necessary for the development of criteria and is not discussed further.

In this paper, it is shown how a conditional probability analysis can be used with empirical, probabilistic monitoring data for aquatic resources to establish thresholds of impact for a stressor for a specified geographic area. Scientifically defensible thresholds are a necessary first step in establishing protective criteria by environmental managers. The approach is demonstrated by applying it to wadable streams in the midAtlantic region of the United States, and establishing thresholds for impact of sedimentation on the streams in this region. Finally, an illustration is presented on how these thresholds could be used to develop a criterion for protection of aquatic life in these streams from sedimentation.

Figure 1. Mid-Atlantic Region of the U.S. With the Wadable Stream Sampling Sites Used in This Study.

DATA SOURCES

The field data used in this paper are available through the USEPA Environmental Monitoring and Assessment Program (EMAP) web site (USEPA, 2004). These data were collected from the mid-Atlantic region streams in 1993 and 1994 and include 102 stream segments in first to third (Strahler) order wadable streams (Figure 1). These segments were selected for sampling using a spatially balanced probability design (Stevens, 1997; Stevens and Olsen, 1999). Inclusion probabilities for each sampled stream segment were determined using the sample sizes for each Strahler order and the total length of streams within each order in the region. Sampling locations within stream segments were chosen randomly. Quantitative data for stream macroinvertebrates, habitat, and water quality were collected at each site (for specifics see Lazorchak et al., 1998; Kaufman and Robinson, 1998; Herlihy et al., 2000; and Klemm et al., 2002). Sampling took place during a two-month sampling window each year from April through midJune.

Stream Macroinvertebrate Data

Stream benthic macroinvertebrates are a robust measure of stream condition, integrating temporal pollutant exposure. They are responsive to changes in in-stream sediment levels (Davis and Lathrop, 1992; Covich, 1999). Benthic stream community taxa in the orders Ephemeroptera (mayflies), Plecoptera (stoneflies), and Trichoptera (caddisflies) (collectively known as EPT) are considered reasonably sensitive indicator organisms since they exhibit a decrease in taxa richness with increased degradation of stream conditions (Loch et al., 1996; Barbour et al., 1999; Zweig and Rabeni, 2001; Klemm et al., 2002; Kaller and Hartman, 2004). The EPT taxa were used to identify impacted stream segments in the mid- Atlantic; when EPT taxa were less than 9 in the stream segments, the stream segments were considered to be impacted (Davis and Scott, 2000; Klemm et al., 2002).

Indicator for Sedimentation

Sediments, including suspended and bedded, can directly affect stream biota or indirectly affect stream biota through changes in habitat. For example, excessive suspended sediments in aquatic systems can cause increased turbidity and decreased light penetration. Altered light regimes can directly alter primary productivity and increase shading of submerged macrophytes (Canfield et al., 1985; Best et al., 2001). Excess fine sediments can fill in gaps between larger substrate particles, embedding the larger particles and eliminating interstitial spaces that would otherwise be used as habitat for reproduction, feeding, and cover for invertebrates and fish (Suttle et al., 2004). For example, bedded sediments in streams and rivers can cause the loss of spawning habitat for salmonids due to increased embeddedness (Young et al., 1991).

For the purpose of this paper, percent fines in the substrate is used as a surrogate indicator for sedimentation in streams. The percent fines (silt/clay fraction, less than 0.06 mm) is a direct measure of the smallest class of sediments and is strongly correlated with sediment embeddedness, a source of the most likely to be resuspended sediment, and an indirect measure of suspended sediment levels in the water column. Streams containing a larger fraction of fine sediment would be expected to have a benthic community at greater risk for impact (Zweig and Rabeni, 2001). Details of the protocols for sample collection and analysis for percent fines can be found in Lazorchak et al. (1998), Kaufmann et al. (1999), and Klemm et al. (2002). Percent fines is determined by visual examination at 11 equally spaced stream transects. The data used were restricted to stream segments with pools.

Reference Conditions

Reference con\ditions are expectations as to the condition of biological communities in the absence of any human disturbance (Plafkin et al., 1989; Gerritsen et al., 1994). These conditions provide an estimate of natural variability in biological condition and habitat quality that can be expected to occur. Few streams in the mid-Atlantic area are undisturbed. Reference conditions in the mid-Atlantic have been identified (Waite et al., 2000) using a set of selected chemical and habitat conditions from unimpacted or minimally impacted streams in the area. These chemical and habitat measures are used to identify the best available biological conditions in the area streams.

The chemical and habitat parameters used in the selection of reference conditions for wadable streams in the mid-Atlantic area (Waite et al., 2000) are: acid neutralizing capacity (> 50 eq/L), chloride (< 100 eq/l), sulfate (< 400 eq/l), total nitrogen (< 750 g/ 1), total phosphorous (< 20 g/1), and mean rapid bioassessment protocol (RBP) habitat score (> 15). The RBP habitat score encompasses the variety and quality of the substrate, channel morphology, bank structure, and riparian vegetation. It is measured on a scale of 1 to 20, where 1 is very poor habitat and 20 is excellent habitat (Barbour et al., 1999). Waite et al. (2000) define a stream segment as reference if all six of the chemical and habitat parameters are within the desired levels. Reference sites were identified that met these criteria for the extant data. Thus, the number and distribution of reference site conditions were not predetermined and were used strictly for comparative purposes.

METHODS

Conditional Probability Analysis (CPA)

The probability of observing a certain event y is denoted P(y). Data acquired with a probability survey design provide estimates of the probability of occurrence for a sampled variable. For example, consider a sampling frame that includes all stream segments in a state. If 75 percent of the sampled stream segments exhibit impacted benthic communities, then the likelihood of observing benthic impact in any of the stream segments in the state is 75 percent.

Functionally, to determine the probability of impact in the stream when some value of the pollution metric, x^sub C^, is exceeded, P (y = 1 | x > x^sub C^), a two-step procedure is used. Using the survey data, one identifies a subset of the sampled resource (e.g., stream segments) for which x > x^sub C^ (i.e., the stream segments are stratified based on the value of the pollution metric x); and from this subset of stream segments in which x > x^sub C^, one determines those segments in which the biological conditions are impacted. This is the subset of the sampled stream segments in which the pollution metric exceeds a specific value (x^sub C^), and which are also biologically impacted. The probability of the biology being impacted in stream segments when x > x^sub C^ over the entire range of observed x provides an empirical conditional probability curve. Confidence intervals (CIs) for this empirical curve can be estimated by bootstrap resampling (Manly, 1997). Bootstrapping assumes that the distribution of a population can be determined by resampling the original data. A bootstrap sample consists of drawing a sample of size N from the original data (of size N) with replacement, which is then used to calculate a bootstrap value for conditional probability, P (y = 1 | x > x^sub C^). One thousand samples were generated for the bootstrap distribution. The 90 percent and 95 percent CIs were determined from the empirical percentiles (Insightful Corp., 2001).

Identifying Thresholds of Impact

Threshold levels for pollutants that elicit different levels of biological impact in stream segments of a region need to be identified for eventual use in developing criteria. A threshold of impact was identified as a changepoint separating the empirical conditional probability curve into two parts, that part of the curve above the changepoint and that which is below it. For those samples that are above the changepoint, the probability of impact is different from what one would expect for the entire geographic area. A confounding factor in the identification of a changepoint is that these two groups created by the changepoint are not independent (i.e., the numbers used to create the points above the changepoint are a subset of the numbers used to create the points below the changepoint). Thus, a traditional t-test cannot be used in the determination of the changepoint since the data are not independent (Venables and Ripley, 1997). The identification of the changepoint was by using a weight-ofevidence approach with three different techniques. These techniques are: nonoverlapping confidence intervals, change in curvature of fitted curve, and nonparametric deviance reduction. Other possible techniques could be used to identify a changepoint. In this demonstration, specific values for factors and CIs were selected only as examples. Values used in an actual application of this approach would depend on the particular management requirements and objectives.

The use of nonoverlapping CIs to determine a changepoint involves determining when the lower CI of the empirical curve no longer overlaps the upper CI of the unconditional value (Cherry, 1996; Rahlfs, 1997; Cherry, 1998; Austin and Hux, 2002). This procedure is a conservative estimate for significant difference, since the CIs could overlap when the values are significantly different (Austin and Hux, 2002). The bootstrap percentile CIs based on a bootstrap distribution of 1,000 samples were used for this evaluation. The α-level for the nonoverlapping CI must be adjusted to account for the one-sided nature of this test, whereas the α-level for developing the CIs for the curves was based on a two-sided test (i.e., a factor of 2 in the α-level).

The deviance reduction point generally can be determined, but it may or may not be of biological significance. Uncertainty in the deviance reduction changepoint (90 percent and 95 percent CIs) is estimated from the empirical percentiles for the bootstrap distribution from resampling 1,000 times (Manly, 1997). An approximate χ^sup 2^ test was used to determine the significance of the changepoint. The test assumes that the deviance reduction divided by the scale parameter is approximately χ^sup 2^ distributed with 1 degree of freedom (Venables and Ripley 1997). A large deviance reduction will result in a small p-value, and the consequent rejection of the null hypothesis (H^sub 0^: no changepoint).

Biological Importance of Identified Thresholds

For use in criteria development, some level of biological importance needs to be associated with the threshold of impact value that is identified. The changepoint value determined by each technique must separate the samples so that the probability of impact for samples above the threshold would be different from what one would expect for the entire geographic area. A summary of literature values on the response of fish and benthic invertebrates at low reported levels of percent fines in the substrate (Newcombe and Jensen, 1996; Bash et al., 2001; Berry et al., 2003) was used to identify biological importance.

Statistical Analysis of Data

Figure 2. Cumulative Distribution Function for EPT Taxa Richness for All Stream Miles and for Stream Miles That Exhibit Reference Condition Characteristics.

The CCDF is the distribution for a subset of the total data, subsetted by (or conditioned on) a second variable [F(y | x)]. The reverse CCDF is similar to Equation (7), that is, 1-F(y | x). The reverse functions are consistent with the CPA results, which are expressed as a threshold (i.e., exceeding some value x^sub C^).

RESULTS

The CDF for EPT taxa richness is shown in Figure 2. Approximately 42 percent of the stream miles across the region were observed to have EPT taxa richness less than 9, indicating impacted benthic community conditions. Out of 100 stream segments that had valid values for the indicators used in this study, 16 met the reference condition requirements. In 91 percent of the reference condition stream miles, benthic communities were found that had EPT taxa equal to or greater than 9 (Figure 2). The EPT taxa richness generally declines as the percent fines increases (Figure 3, correlation coefficient, r, is -0.50). The fraction of EPT taxa richness variance explained using a linear regression with percent fines as the predictor is 0.25, suggesting that percent fines does appear to have a substantial effect on EPT taxa richness.

The reverse CDF and reverse CCDFs for percent fines in the substrate were expressed as a proportion of stream miles (Figure 4). The sampled stream segment values were weighted by inclusion probabilities to convert to stream miles. The distribution for impacted benthic communities is displaced to the right of that for benthic communities in good condition (Figure 4). The distribution for reference conditions is shifted to the left (towards lower percent fines) than that for unimpacted streams (Figure 4), since these are the best observed conditions.

The CPA approach suggests that when percent fines in the substrate is greater than 49 percent, there is a 100 percent probability that the benthic communities are impacted (Figure 5). All sites with percent fines in the substrate in excess of 49 percent had EPT taxa richness less than 9. As the percent fines approaches zero, there is a background level of impact on EPT taxa richness from all sources of stress in the region (mean = 42 percent, 95 percent confidence interval of 30 to 56 percent). Thus, irrespective of the level of percent fines in the substrate, approximately 42 percent of the stream miles in the region will likely exhibit an impact on EPT taxa richness. Therefore, to detect a significant signal due to percent fines in the substrate affecting the EPT ta\xa richness, the upper confidence limit on the estimate of the background impact (e.g., 56 percent, Figure 5) must not overlap with the lower confidence limit on the probability of benthic impact curve in Figure 5. The point at which this occurs is when the percent fines in the substrate is 14.8 percent (Figure 5). This is a threshold of impact, and is statistically distinguishable from background within this geographic area. The mean probability of observing impacted EPT taxa richness associated with this threshold is 67 percent.

Figure 3. Plot of EPT Taxa Richness Against Percent Fines in Substrate (silt/clay fraction, less than 0.06 mm). Horizontal line for EPT taxa richness = 9. Solid circles are stream segments that exhibit reference condition characteristics. Open circles are segments not satisfying reference condition characteristics.

The CPA identified threshold of 14.8 percent fines (from nonoverlapping CIs) would translate into approximately 47 percent of the total stream miles in the geographic area exceeding the threshold (from Figure 4). Similarly, only a small percentage of streams with reference condition characteristics (6 percent) or good benthic conditions (21 percent) would exceed the 14.8 percent fines threshold, but a much larger percentage of impacted streams (74 percent) would exceed it (from Figure 4). These values provide an estimate of the number of "false positives" for this value of a threshold for percent fines as the indicator of sedimentation. Because multiple stressors often impact stream communities, one cannot estimate the "false negatives." A community not stressed by the stressor of interest might be stressed in some other way.

Figure 4. Reverse Cumulative Distribution Function (CDF) for Percent Fines in the Substrate (silt/clay fraction, less than 0.06 mm) for Stream Miles Across Entire Area, and Reverse Conditional CDFs of Stream Miles for Impacted Benthic Conditions (EPT taxa richness less than 9), Unimpacted Benthic Conditions (EPT taxa richness equal to or greater than 9), and Reference Conditions. Horizontal lines are where the threshold of 14.8 percent fines intersects curves.

Figure 5. Probability of Observing EPT Taxa Richness Less Than 9 (benthic impact) in Mid-Atlantic Streams (open circles) if Specified Value of Percent Fines in the Substrate (silt/clay fraction, less than 0.06 mm) is Exceeded. Solid line is fit of Equation (4) (see Table 1). Dotted lines are 95 percent confidence intervals (CIs) from bootstrap estimation.

The coefficients from the nonlinear least squares regression for Equation (4) are given in Table 1, with the fitted curve shown in Figure 5. This technique also determined a threshold of impact of 14.8 percent fines in the substrate (Table 2). Using the nonparametric deviance reduction technique, a threshold of impact of 15.3 percent fines in the sediment, with p = 0.03, was identified. All three techniques for identifying a threshold of impact from the conditional probability analysis yielded consistent results (Table 2).

TABLE 1. Coefficients (mean value and confidence limits) From Nonlinear Least Squares Regression of Equation (4) for Percent Fines in Substrate Against Probability of Impacted Benthic Community (EPT taxa richness less than 9).

These three different techniques all determined thresholds of impact that separated the data such that a difference could be detected from what would be expected for the entire geographic area. For the first technique, the existence of nonoverlapping CIs provided the difference. For the second technique, the nonoverlap of CIs for the curvature parameters established the difference. In the third technique, the null hypothesis of no changepoint was rejected (p = 0.03).

TABLE 2. Summary of Thresholds of Impact for Percent Fines in Substract (silt/clay fraction, less than 0.06 mm) Identified by Conditional Probability Analysis (mean threshold and confidence intervals, when available) Using Three Techniques.

The literature supports a biological response of fishes to the thresholds for percent fines in the substrate: survival of salmonids have been shown to be negatively affected when fines exceed 10 to 20 percent (McNeil and Ahnell, 1964; Burns, 1970; Tappel and Bjornn, 1983; Chapman, 1988; Peterson et al. 1992; Argent and Flebbe, 1999). These studies for salmonids were all for a larger sediment size range (silt/clay/ sand) than was chosen for purposes of this demonstration. However, the silt/clay fraction is always less than or equal to the silt/clay/sand fraction of the same sample. These threshold values are consistent with the lower end of reported response levels in the literature (see Newcombe and Jensen, 1996).

DISCUSSION

Scientifically defensible numeric criteria are highly desirable for water quality protection programs responsible for preventing the impairment of aquatic systems. Historically, for single chemical pollutants, carefully controlled laboratory bioassays have been conducted. From the dose-response relationship derived from these bioassays, an appropriate criterion for the pollutant was developed that is protective of aquatic life. Unfortunately, this historical approach is not applicable for nonpoint source pollution. At low levels, nonpoint source materials may not be a pollutant, but may be necessary for the functioning of the aquatic systems (e.g., nutrients, sediments) and their levels may naturally fluctuate over a geographic area. Only when the levels become excessive (usually in conjunction with anthropogenic activity), do they become pollutants and require criteria development for their control. The development of scientifically defensible approaches for establishing thresholds, and eventually criteria, for nonpoint source pollution is a critical need for regulatory agencies (USEPA, 2003b). Any approach undertaken to develop nonpoint source criteria must take into account the natural variability of the pollution occurring across the geographic area of interest, and the impact of other stressors that are likely impacting the aquatic systems as well.

Use of the CPA can establish realistic thresholds for the impact on stream biotic condition by nonpoint source pollution. This approach was applied to establish a threshold of sediment impact on a susceptible biological community in wadable streams in the midAtlantic region of the U.S. The mid-Atlantic was selected because of the extensive amount of research and monitoring of streams in this region (e.g., Howard et al, 1999; USEPA, 200Ob), which provided the information base needed that would satisfy the conditions for application of CPA. The necessary conditions were: (1) monitored data must be collected based on a probability based sampling design; (2) there must be some metric that can quantify the pollution parameter of interest; (3) there must be a response metric sufficiently sensitive to respond to the extant levels of the pollution parameter of interest; (4) independent studies must be available that identify the characteristics of an impacted response metric; and (5) the pollution parameter must be capable of exerting a strong effect on the response metric.

The streams in the mid-Atlantic region met these criteria. Sufficient empirical data were available from a probability monitoring design for wadable streams in the region (see McDonald et al., 2004; data are available from http://www.epa.gov/emap/). The probabilistic sampling allows for statistically rigorous extrapolation from the sites sampled to the entire region of interest. Sedimentation was a major stressor in these streams (Howard et al., 1999; USEPA, 200Ob), and sufficient information was available on the percentage of fines in the substrate to allow its use as a surrogate for sedimentation (Klemm et al., 2002). Data had been collected on EPT taxa richness at these sites and related to stream condition (USEPA, 200Ob). Davis and Scott (2000) had determined a level of EPT taxa species richness (less than 9) below which mid-Atlantic highland streams were likely to be impacted and in relatively poor condition. Last, EPT taxa richness responded strongly to sedimentation (Figure 3).

It was decided not to develop thresholds for protecting against impact based on not exceeding a pollution metric value (i.e., P(y = 0 | x ≤ x^sub C^), where y = 0 represents unimpacted conditions). The approach of looking for a threshold which, if not exceeded, would indicate a high probability of encountering unimpacted conditions, is the approach taken for criteria developed from laboratory toxicological studies. What makes this appropriate for laboratory studies is the ability to control for all stressors other than the one for which criteria are being developed. These studies provide y = 0 (unimpacted) if x < x^sub C^. This approach does not work when dealing with actual field data, as one cannot control for all of the myriad stressors that are affecting the biological communities. The biological response observed reflects the cumulative response to all of the stressors. As the data for a specific stressor is analyzed, and as the magnitude of that stressor decreases, one would not expect a continual increase in the likelihood of unimpacted conditions, unless all of the stressors are strongly correlated. Reducing one stressor would still leave other stressors eliciting impact (in the case of 42 percent of the stream miles in Figure 2). Therefore, any calculation of P(y = 0 | x ≤ x^sub C^) would likely be confounded by other stressors (i.e., value for unconditional probability would not be zero). Thus, the thresholds identified with the conditional probability analysis (and based on the reverse CCDF) are thresholds of impact, above which the likelihood of impact is high. One is able to pull out the signal in the mixture of multiple stressors with the conditional probability analysis because the stressor chosen was strong enough to elicitan impact as the magnitude of the stressor increased. With the conditional probability analysis approach, one is protecting the aquatic resource against the likelihood of impact.

The EPT taxa richness metric was used to identify impacted stream communities. The taxa in the orders Ephemeroptera (mayflies), Plecoptera (stoneflies), and Trichoptera (caddisflies) respond similarly to sedimentation as estimated by percent fines in the substrate (Figure 3), with the probability of impact increasing as percent fines increased. Other functional biological groupings that responded similarly to EPT were benthic invertebrate scrapers and intolerant taxa richness, while noninsect benthic invertebrates, benthic invertebrate scavengers, and tolerant taxa richness responded with the probability of impact decreasing as percent fines increased. Conditional probability plots could have been generated using any of these groupings of the benthic invertebrate community, if a level of biological impact could be independently assigned (similar to EPT taxa richness less than 9).

With the CPA approach, traditional statistics cannot be used to ascertain the threshold of impact. Instead, a weight-of-evidence approach based on three separate techniques was used. However, this assumed that there was a consistency in the threshold levels identified using these disparate techniques. The CPA does provide relatively consistent thresholds of impact for the percent fines in the substrate, irrespective of which of the three techniques were applied (Table 2). These CPA thresholds contrasted markedly with the threshold values obtained with the two ad hoc approaches currently practiced for developing aquatic criteria based on monitoring data from sites across a geographic area (USEPA, 200Oa). The two techniques consist of setting thresholds with either the levels of stressor associated with streams in the 75th percentile of the reference stream miles sampled or the 25th percentile of all stream miles sampled. Using these approaches, the threshold for percent fines would be approximately 1.9 percent based on all stream miles and approximately 7.1 percent based on reference stream miles (Figure 4). These values are substantially lower than estimates from this study, and fall outside of the 95 percent confidence limits (Table 2). These values are also substantially lower than literature thresholds for percent fines.

While the literature supports the biological importance associated with the thresholds identified, the agreement of these threshold values with lower values from the literature does not validate the thresholds. However, it does gives credence to the conditional probability analysis approach for identifying realistic thresholds for use in development of criteria for protection of aquatic life. The CPA can provide environmental managers with an additional tool to evaluate the tradeoffs of setting different criteria. Using CPA, environmental managers can examine a given criterion and the tradeoffs: the likely number of stream miles that actually have good biological communities when the criterion level is exceeded and the number of streams that have impacted biology when the criterion level is not exceeded. This would allow the protection of the ecosystems to be more quantitative and explicit when being weighed in conjunction with economic considerations.

The CPA approach for threshold of impact can be combined with information on reference conditions and toxicological data to develop candidate values for water quality criteria. The steps in this process that provide the candidate values are listed below.

1. Acquire the survey data (probability based) that includes candidate pollutant for criterion development and biological response metrics.

2. Use available information on reference conditions (physical, chemical, and habitat metrics) to define impacted biological conditions in terms of biological metrics.

3. Conduct conditional probability analysis (probability of impact if value of candidate pollutant is exceeded).

4. Identify threshold of impact from conditional probability analysis results.

5. Evaluate identified threshold of impact against reference conditions, impacted conditions, and good conditions. Evaluate identified threshold for biological importance.

Nonetheless, additional work must be done to evaluate and validate the conditional probability analysis approach for identification of thresholds of impact. This could be accomplished by using other survey data from other geographic areas where the conditions for CPA are met. The approach should also be tested with other pollution and response parameters to confirm that this is a robust approach, which can be used for identifying realistic thresholds of impact.

SUMMARY

The conditional probability analysis approach can be used to develop realistic thresholds of impact for nonpoint source pollution on aquatic benthic communities in waterbodies across a region. However, this approach is predicated on the following conditions: (1) monitored data have been collected based on a probability based sampling design; (2) some metric must be available that can quantify the pollution parameter of interest; (3) a response metric sufficiently sensitive to respond to the extant levels of the pollution parameter of interest must be available; (4) independent studies must be available that identify the characteristics of an impacted response metric; and (5) the pollution parameter must be capable of exerting a strong effect on the response metric. In the example presented here, realistic thresholds of impact on EPT taxa richness due to sedimentation in midAtlantic wadable streams were identified. Threshold values from CPA were found to be in the range of 12 to 15 percent fines in the substrate, based on three different techniques for threshold identification. These threshold values were found to be consistent with existing literature from laboratory and field studies. These values were quite different from those determined with the current ad hoc practice, with the current practice appearing to produce overly prescriptive thresholds. Development of scientifically defensible thresholds are a necessary first step for managers in establishing protective criterion. However, thresholds determined with CPA, or other methods, should not be used exclusively to set water quality criteria, as other additional factors (e.g., designated uses, ecotoxicological data, economics) must be considered by managers when establishing criteria and standards.

Paul, John F. and Michael E. McDonald, 2005. Development of Empirical, Geographically Specific Water Quality Criteria: A Conditional Probability Analysis Approach. Journal of the American Water Resources Association (JAWRA) 41(5):1211-1223.

1 Paper No. 04095 of the Journal of the American Water Resources Association (JAWRA) (Copyright 2005). Discussions are open until April 1, 2006.

ACKNOWLEDGMENTS

Thanks to Brian Hill, Jim Wickham, Walter Berry, Jerry Pesch, John Van Sickle, Phil Kaufman, and anonymous reviewers for the critical and constructive reviews that they provided on various versions of this manuscript. Special thanks goes to Steve Hedtke for the encouragement to apply the conditional probability analysis approach to sediment criteria and to the EMAP-Surface Waters team for generating high quality data and making these data available to all. S-Plus software was used for statistical analyses and graphical displays. The research described in this paper was funded by the U.S. Environmental Protection Agency. This paper was not subjected to Agency review, and therefore does not necessarily reflect the views of the Agency. Mention of trade names or commercial products does not constitute endorsement or recommendation for use.

LITERATURE CITED

Argent, D.G. and PA. Flebbe, 1999. Fine Sediment Effects on Brook Trout Eggs in Laboratory Streams. Fisheries Research 39(3): 253- 262.

Austin, PC. and J.E. Hux, 2002. A Brief Note on Overlapping Confidence Intervals. Journal of Vascular Surgery 36(1):194-195.

Barbour, M.T., J. Gerritsen, B.D. Snyder, and J.B. Stribling, 1999. Rapid Bioassessment Protocols for Use in Streams and Wadeable Rivers: Periphyton, Benthic Macroinvertebrates and Fish (second Edition). EPA 841-B-99-002, U.S. Environmental Protection Agency, Office of Water, Washington, D.C.

Bash, J., C. Berman, and S. Bolton, 2001. Effects of Turbidity and Suspended Solids on Salmonids. University of Washington, Center for Streamside Studies. Available at http://depts.washington.edu/ cwws/Outreach/Publications/Salmon%20and%20Turbidi ty.pdf. Accessed in August 2005.

Berry, W, N. Rubinstein, B. Melzian, and B. Hill, 2003. The Biological Effects of Suspended and Bedded Sediment (SABS) in Aquatic Systems: A Review. U.S. Environmental Protection Agency, Office of Research and Development, Internal Report. Available at http://www.epa.gov/waterscience/criteria/sediment/ appendixl.pdf. Accessed in August 2005.

Best, E.PH., C.P. Buzzelli, S.M. Bartell, R.L. Wetzel, WA. Boyd, R.D. Doyle, and K.R. Campbell, 2001. Modeling Submersed Macrophyte Growth in Relation to Underwater Light Climate: Modeling Approaches and Application Potential. Hydrobiologia 444(1-3): 43-70.

Borja, A., V. Valencia, J. Franco, I. Muxika, J. Bald, M. J. Belzunce, and O. Solaun, 2004. The Water Framework Directive: Water Alone, or in Association With Sediment and Biota, in Determining Quality Standards? Marine Pollution Bulletin 49(l-2):8-ll.

Boward, D.M., P.F. Kazyak, S.A. Stranko, M.K. Kurd, and T.P. Prochaska, 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.

Burns, J.W., 1970. Spawning Bed Sedimentation Studies in Northern California Streams. California Fish and Game Quarterly 56(4):253- 270.

Canfield, D.E.G\., K.A. Langeland, S.B. Linda, and T.W. Haller, 1985. Relations Between Water Transparency and Maximum Depth of Macrophyte Colonization in Lakes. Journal of Aquatic Plant Management 23:25-28.

Chapman, D.W., 1988. Critical Review of Variables Used to Define Effects of Fines in Redds of Large Salmonids. Transactions of the American Fisheries Society 117(1):1-21.

Cherry, S., 1996. A Comparison of Confidence Interval Methods for Habitat Use-Availability Studies. Journal of Wildlife Management 60(3):653-658.

Cherry, S., 1998. Statistical Tests in Publications of the Wildlife Society. Wildlife Society Bulletin 26(4):947-953.

Covich, A.P., 1999. The Role of Benthic Invertebrate Species in Freshwater Ecosystems. Bioscience 49(2):119-127.

Davis, WR. and J. Scott, 2000. Mid-Atlantic Highlands Streams Assessment: Technical Support Document. EPA/903/B-00/004, Mid- Atlantic Integrated Assessment Program, Region 3, U.S. Environmental Protection Agency, Ft. Meade, Maryland.

Davis, WS. and J.E. Lathrop, 1992. Freshwater Benthic Macroinvertebrate Community Structure and Function. In: Sediment Classification Methods Compendium, EPA 823-R-92-006, pp. 8-1 to 8- 26. U.S. Environmental Protection Agency, Office of Water, Washington, D.C.

Fisher, D.J. and D.T. Burton, 2003. Comparison of Two U.S. Environmental Protection Agency Species Sensitivity Distribution Methods for Calculating Ecological Risk Criteria. Human and Ecological Risk Assessment 9(3):675-690.

Gerritsen, J., J. Green, and R. Preston, 1994. Establishment of Regional Reference Conditions for Stream Biological Assessment and Watershed Management. In: Proceedings, Watershed '93, A National Conference on Watershed Management. EPA/840/R-90/002, U.S. Environmental Protection Agency, Office of Water, Washington, D.C., pp. 797-801.

Herlihy, A.T., O.P. Larsen, S.G. Paulsen, N.S. Urquhat, and B.J. Rosenbaum, 2000. Designing a Spatially Balanced, Randomized Site Selection Process for Regional Stream Surveys: The EMAP Mid- Atlantic Pilot Study. Environmental Monitoring and Assessment 63(1):95-113.

Hogg, R.V. and J. Ledolter, 1992. Applied Statistics for Engineers and Physical Scientists. Macmillan Publishing Co., New York, New York.

Hohreiter, D.W, and O.K. Rigg, 2001. Derivation of Ambient Water Quality Criteria for Formaldehyde. Chemosphere 45(4-5):471486.

Insightful Corp., 2001. S-Plus 6 for Windows, Guide to Statistics, Volume 2. Insightful Corp., Seattle, Washington.

Jimenez, B., J. Ramos, and L. Quezada, 1999. Analysis of Water Quality Criteria in Mexico. Water Science and Technology 40(10):169- 175.

Kaller, M.D., and K.J. Hartman, 2004. Evidence of a Threshold Level of Fine Sediment Accumulation for Altering Benthic Macroinvertebrate Communities. Hydrobiologia 518:95-104.

Kamizoulis, G. and L. Saliba., 2004. Development of Coastal Recreational Water Quality Standards in the Mediterranean. Environment International 30(6):841-854.

Kaufmann, P.R., P. Levine, E.G. Robinson, C. seeliger, and D.V. Peck, 1999. Quantifying Physical Habitat in Wadeable Streams. EPA/ 620/R-99/003, U.S. Environmental Protection Agency, Washington, D.C.

Kaufmann, PR. and E.G. Robinson, 1998. Physical Habitat Characterization in Environmental Monitoring and Assessment Program - Surface Waters. In: Field Operations and Methods for Measuring the Ecological Condition of Wadeable Streams, J.M. Lazorchak, D.J. Klemm and D.V. Peck (Editors). EPA/620/R-94/004F, U.S. Environmental Protection Agency, Office of Research and Development, Washington, D.C., pp. 77-118.

Kay, D., J. Bartram, A. Pruss, N. Ashbolt, M.D. Wyer, J.M. Fleisher, L. Fewtrell, A. Rogers, and G. Rees, 2004. Derivation of Numerical Values for the World Health Organization Guidelines for Recreational Waters. Water Research 38(5):1296-1304.

Klemm, D.J., K.A. Blocksom, W.T. Thoeny, RA. FuIk, A.T. Herlihy, RR. Kaufmann, and S.M. Cormier, 2002. Methods Development and Use of Macroinvertebrates as Indicators of Ecological Conditions for Streams in the Mid-Atlantic Highlands Region. Environmental Monitoring and Assessment 78(2):169-212.

Lazorchak, J.M., D.J. Klemm, and D.V. Peck (Editors), 1998. Environmental Monitoring and Assessment Program-Surface Waters. Field Operations and Methods for Measuring the Ecological Condition of Wadeable Streams, EPA/620/R-94/004F, U.S. Environmental Protection Agency, Washington, D.C.

Loch, D.D., J.L. West, and D.G. Perlmutter, 1996. The Effect of Trout Farm Effluent on the Taxa Richness of Benthic Macroinvertebrates. Aquaculture 147(l-2):37-55.

Manly, B.F.J., 1997. Randomization, Bootstrap and Monte Carlo Methods in Biology. Chapman and Hall, New York, New York.

McDonald, M., R. Blair, D. Bolgrien, B. Brown, J. Dlugosz, S. Hale, S. Hedtke, D. Heggem, L. Jackson, K. Jones, B. Levinson, R. Linthurst, J. Messer, A. Olsen, J. Paul, S. Paulsen, J. Stoddard, K. Summers, and G. Veith, 2004. The Environmental Protection Agency's Environmental Monitoring and Assessment Program. In: Environmental Monitoring. G. B. Wiersma (Editor). CRC Press LLC, New York, New York, pp. 649-668.

McNeil, W.J. and W.H. Ahnell, 1964. Success of Pink Salmon Spawning Relative to Size of Spawning Bed Materials. U.S. Fish and Wildlife Service Special Scientific Report, Fisheries 469, Washington, D.C.

Newcombe, C.P. and J.O.T. Jensen, 1996. Channel Suspended Sediment and Fisheries: A Synthesis for Quantitative Assessment of Risk and Impact. North American Journal of Fisheries Management 16(4):693-727.

Niemi, G.L. and M.E. McDonald, 2004. Application of Ecological Indicators. Annual Review of Ecology, Evolution, and Systematics 35:89-111.

Perry, J. and E. Vanderklein, 1996.Water Quality: Management of a Natural Resource. Blackwell Science, Cambridge, Massachusetts.

Peterson, N.P., A. Hendry, and TP. Quinn, 1992. Assessment of Cumulative Effects on Salmonid Habitat: Some Suggested Parameters and Target Conditions. University of Washington, Center for Streamside Studies, TFW-F3-92-001, Seattle, Washington.

Plafkin, J.L, M.T. Barbour, K.D. Porter, S.K. Gross, and R.M. Hughes, 1989. Rapid Bioassessment Protocols for Use in Streams and Rivers: Benthic Macroinvertebrates and Fish. U.S. EPA Report EPA/ 440/4-89/001, Washington, D.C.

Qian, S.S., R.S. King, and C.J. Richardson, 2003. Two Statistical Methods for the Detection of Environmental Thresholds. Ecological Modelling 166(l-2):87-97.

Rahlfs, V.W., 1997. Understanding and Evaluating Clinical Trials [Letter]. Journal of the American Academy of Dermatology 37(5):803- 804.

Rausina, G.A., D.C.L. Wong, W Raymon Arnold, E.R. Mancini, and A.E. Steen, 2002. Toxicity of Methyl Tert-Butyl Ether to Marine Organisms: Ambient Water Quality Criteria Calculation. Chemosphere 47(5):525-534.

Russo, R.C., 2002. Development of Marine Water Quality Criteria for the USA. Marine Pollution Bulletin 45(1-12):84-91.

Spooner, J., L. Wyatt, S.W. Coffey, S.L. Brichford, J.A. Arnold, M.D. Smolen, G.D. Jennings, and J.A. Gale, 1991. Nonpoint Sources. Research Journal of the Water Pollution Control Federation 63(4):527- 536.

Stevens, D.L., Jr., 1997. Variable Density Grid-Based Sampling Designs for Continuous Spatial Populations. Environmetrics 8:167- 195.

Stevens, D.L., Jr. and A.R. Olsen, 1999. Spatially Restricted Surveys Over Time for Aquatic Resources. Journal of Agricultural, Biological, and Environmental Statistics 4(4):415-428.

Suttle, K.B., M.E. Power, J.M. Levine, and C. McNeely, 2004. How Fine Sediment in Riverbeds Impairs Growth and Survival of Juvenile Salmonids. Ecological Applications 14(4): 969-974.

Tappel, RD. and T.C. Bjornn, 1983. A New Method of Relating Size of Spawning Gravel to Salmonid Embryo Survival. North American Journal of Fisheries Management 3:123-135.

USEPA (U.S. Environmental Protection Agency), 1994. Water Quality Standards Handbook (second Edition). EPA/823/B-94/005, U.S. Environmental Protection Agency, Office of Water, Washington, D.C.

USEPA (U.S. Environmental Protection Agency), 200Oa. Nutrient Criteria Technical Guidance Manual: Rivers and Streams. EPA822-B-00- 002, U.S. Environmental Protection Agency, Office of Water, Washington, D.C.

USEPA (U.S. Environmental Protection Agency), 200Ob. MidAtlantic Highlands State of the Streams. EPA/903/R-00/005, U.S. Environmental Protection Agency, Region 3, Philadelphia, Pennsylvania.

USEPA (U.S. Environmental Protection Agency), 2002. National Water Quality Inventory: 2000 Report. EPA-841-R-02-001, U.S. Environmental Protection Agency, Office of Water, Washington, D.C.

USEPA (U.S. Environmental Protection Agency), 2003a. Developing Water Quality Criteria for Suspended and Bedded Sediments (SABS). Potential Approaches: A U.S. EPA Science Advisory Board Consultation. U.S. Environmental Protection Agency, Office of Water, Draft Report. Available at http://www.epa.gov/ waterscience/ criteria/sediment/s ab-discussion-paper.pdf. Accessed in August 2005.

USEPA (U.S. Environmental Protection Agency), 2003b. Setting Priorities to Strengthen the Foundation for Protecting and Restoring the Nation's Waters, Strategy for Water Quality Standards and Criteria. EPA-823-R-03-010, Office of Science and Technology, Office of Water, U.S. Environmental Protection Agency, Washington, D.C.

USEPA (U.S. Environmental Protection Agency), 2004. Environmental Monitoring and Assessment Program (EMAP): MidAtlantic Streams 1993- 96 Data Sets. Available at http://www. epa.gov/emap/html/datal/ surfwatr/data. Accessed in August 2005.

Venables, WN. and B.D. Ripley, 1997. Modern Applied Statistics With S-Plus (second Edition). Springer, New York, New York.

Waite, I.R., A.T. Herlihy, D.P. Larsen, and D.J. Klemm, 2000. Comparing Strengths of Geographic and Nongeographic Classifications of Stream Macroinvertebrates in the Mid-Atlantic Highlands, USA. Journal of the North American Benthological Society 19:429-441.

Yin, D., H. Jin, L. Yu, and S. Hu, 2003. Deriving Freshwater Quality Criteria for 2,4-Dichlorophenol for Protection of Aquatic Life in China. Environmental \Pollution 122(2):217-222.

Young, M.K., WA. Hubert, and T.A. Wesche, 1991. Selection of Measures of Substrate Composition to Estimate Survival to Emergence of Salmonids and to Detect Changes in Stream Substrates. North American Journal of Fisheries Management 11:339-346.

Zweig, L.D. and C.F. Rabeni, 2001. Biomonitoring for Deposited Sediment Using Benthic Invertebrates: A Test on 4 Missouri Streams. North American Benthological Society 20(4):643-657.

John F. Paul and Michael E. McDonald2

2 Respectively, Research Environmental Scientist and Director, Environmental Monitoring and Assessment Program, U.S. Environmental Protection Agency, Mail Drop 343-06, Research Triangle Park, North Carolina 27711 (E-Mail/Paul: Paul.john@epa.gov).

Copyright American Water Resources Association Oct 2005

Source: Journal of the American Water Resources Association

More News in this Category