GROUNDWATER STATISTICAL METHOD CERTIFICATION WALTER SCOTT JR. ENERGY CENTER COAL COMBUSTION RESIDUAL MONOFILL POTTAWATTAMIE COUNTY, IOWA

Transcription

1

2 GROUNDWATER STATISTICAL METHOD CERTIFICATION WALTER SCOTT JR. ENERGY CENTER COAL COMBUSTION RESIDUAL MONOFILL POTTAWATTAMIE COUNTY, IOWA 1.0 INTRODUCTION Terracon Project No October 15, 2017 The Monofill is an existing coal combustion residual (CCR) landfill that receives materials from the Walter Scott Jr. Energy Center (WSEC) facility. Both the WSEC facility and the Monofill are located in Pottawattamie County, Iowa. The Monofill is permitted under the IDNR Operating Permit No. 78-SDP-26-06P issued May 2, 2007, with subsequent amendments. The site was developed as a Monofill in 2007 (Cell 1) and began receiving CCR in September The Monofill was constructed with a composite liner system including a 2-foot compacted clay liner and 60-mil high density polyethylene (HDPE) plastic liner. Since the site was developed as a Monofill, additional cells have been added. Monofill Cells 2, 3S, and 3N (2008), Cell 4 (2010), Cell 5 (2011), and Cell 6 (2012) were constructed and receiving CCR materials prior to October 19, Construction of Cell 7 commenced prior to October 19, 2015, and was completed in Cell 7 was approved for operation to receive CCR materials in On April 17, 2015, the United States Environmental Protection Agency (USEPA) issued the final version of the CCR rule for regulation and management of CCR materials at coal-fired units under subpart D of the Resource Conservation and Recovery Act (RCRA) [USNARA, April 2015]. The Federal CCR rule (40 CFR, Part 257) became effective on October 19, 2015 and applies to the Monofill. As required by the Federal CCR rule 40 CFR (b), prior to October 17, 2017, the owner or operator of the CCR unit must develop the groundwater sampling and analysis program to include selection and certification of the statistical procedures to be used for evaluating groundwater monitoring data as required by section The Federal CCR rule also requires that a narrative description of the statistical method(s) selected to evaluate the groundwater monitoring data be included in the certification. The statistical method(s) selected to evaluate the groundwater monitoring data at the Monofill is described in Section STATISTICAL METHOD The goal of the statistical analysis of the groundwater analytical data is to assist in the assessment of potential CCR impact in groundwater adjoining the Monofill. Following each Responsive Resourceful Reliable 1

3 compliance monitoring event, the groundwater analytical data measured in groundwater samples collected from site wells will be evaluated by the statistical methods noted in this section. The statistical methods will be in compliance with 40 CFR and the Statistical Analysis of Groundwater Monitoring Data at RCRA Facilities, Unified Guidance [EPA, March 2009]. The statistical methods will evaluate if a statistically significant increase (SSI) of contaminant concentrations has occurred in groundwater samples when comparing a compliance monitoring sample analytical result with background (baseline) groundwater analytical results. The SANITAS TM For Groundwater Statistical Software will be used to conduct the statistical analysis of groundwater analytical results for the site. However, if, during the period of the groundwater monitoring program at the Monofill, a comparable statistical software program is or may become available, WSEC reserves the right to change software packages. If, during the period of the groundwater monitoring program at the Monofill it becomes apparent a change in the statistical analysis method(s) is warranted, WSEC reserves the right to use any other statistical analysis method allowed by the Federal CCR rule (40 CFR, Part 257). If the statistical analysis method(s) are updated, this certification will be revised to reflect the change in statistical method. Trend analysis conducted using the Sen s Slope/Mann-Kendall method and an intra-well prediction interval have been chosen as the statistical methods to evaluate analytical data collected from site wells. These methods will be utilized to determine if a SSI of monitored analytes has occurred in groundwater adjoining the Monofill. An intra-well analysis is proposed to be conducted at the site to remove the influence of potential spatial variation in the hydrogeology that underlies the site. The utilization of an intra-well analysis instead of an inter-well analysis removes the potential of a false positive SSI as a result of hydrogeologic variation in natural, background monitored analytes. A prediction interval (limit) is selected to compare the compliance analytical results with the background analytical population. The prediction interval incorporates the variability of the background analytical population and establishes an upper limit which provides for a clear interpretation of when background levels have been exceeded and a SSI has occurred. A retesting method can be incorporated with the prediction interval method to aid in verifying if a SSI has occurred or disconfirm it and avoid unnecessary false positives which could trigger assessment monitoring or corrective action activities at the Monofill. Finally, the Sen s Slope/Mann-Kendall method will assist in identifying increasing trends of analyte concentrations in the groundwater at upgradient and downgradient monitoring points relative to the site. The method will also indicate the magnitude of the trend. This method is utilized in conjunction with the intra-well prediction interval method in evaluating if analytical data is stationary over time. The method is being incorporated into this certification based on recommendation in the Statistical Analysis of Groundwater Monitoring Data at RCRA Responsive Resourceful Reliable 2

4 Facilities: Unified Guidance document [EPA, March, 2009] which states that trend testing should be used in detection monitoring programs and for intra-well analysis methods. The proposed statistical methods comply, as appropriate, with the performance standards of section 40 CFR (g). The following subsections describe the statistical method procedures that will be utilized to analyze groundwater analytical results for the site. 2.1 Intra-Well Prediction Intervals The prediction interval is a statistical interval used to compare a single observation to a group of observations. The prediction interval is calculated to include observations from the same population with a specified confidence. In groundwater monitoring, a prediction interval approach may be used to make comparisons between background and compliance well data. The interval is constructed to contain all future observations with stated confidence. If any future observation exceeds this interval, this is statistically significant evidence that the observation is not representative of the background group. Parametric prediction intervals are the first choice when performing prediction interval statistics. The parametric alternative is constructed assuming the background data have a normal or transformed-normal distribution. During parametric prediction interval analysis, the mean and the standard deviation are calculated for the raw or transformed background data. The number of comparison observations is specified to be included in the interval. Once the interval has been calculated, at each sampling period, the mean of the compliance well observations is obtained. This mean is compared to see if it falls within the interval. If less than 15 percent of the background observations are non-detects, the non-detect values will be replaced with one half of the practical quantitation limit (PQL) prior to running the analysis [EPA, April 1989]. If more than 15 percent but less than 50 percent of the background data are less than the method detection limit (MDL), the data s sample mean and standard deviation are adjusted according to the Cohen Adjustment method. However, when the background data are not transformed-normal or contain between 50 and 90 percent observations below the MDL, SANITAS TM For Groundwater automatically constructs a non-parametric prediction interval. During non-parametric analysis, the highest value from the background data is used to set the upper limit of the prediction interval. If more than 90 percent of the background data are less than the MDL, a Poisson distributionbased prediction interval will be computed. The Poisson distribution is a probability distribution modeled for rare events. The Poisson probability of a detectable observation is rare unless there is an impact. The sum of the Poisson counts across background samples is computed by adding the number of parts per billion (ppb) across all observations for the background well. Prior to any calculation, non-detects are set to one-half the PQL and all trace values are evaluated as an average of the PQL. To test the upper prediction limit, the Poisson count of the sum of the next k observations from the well is compared to the 99 Responsive Resourceful Reliable 3

5 percent upper Poisson prediction limit. If this sum exceeds the prediction limit, there is significant evidence of an impact. 2.2 Sen s Slope Estimator/Mann-Kendall Test When used in conjunction with one another, the Mann-Kendall Test for temporal trend and the Sen s Slope Estimator are two types of evaluation monitoring statistics useful in determining the significance of an apparent trend and to estimate the magnitude of that trend. Prior to performing intra-well control chart or prediction interval statistics, the Sen s Slope Estimator/Mann-Kendall Tests are conducted to determine whether a significant upward trend in the data is present. The Mann-Kendall Test is non-parametric, meaning that it does not depend on an assumption of a particular underlying distribution. The test uses only the relative magnitude of data rather than actual values. Values reported by the lab as below the MDL are assigned values equal to one half the PQL. For wells having less than 41 data points, an Exact test is performed. This version of the Mann-Kendall assigns a positive or negative score based on the differences between the data points. The Mann-Kendall Statistic is then computed, which is the number of positive differences minus the number of negative differences. If the absolute value of the Mann-Kendall Statistic exceeds the absolute value of the critical value then a trend is significant at a 95 percent Confidence Level (two-tailed). Positive and negative values of Mann-Kendall Statistics and critical values respectively relate to increasing and decreasing trends. The plots contain the slope (units per year) and concentrations over time, the Mann-Kendall Statistic, the critical value, and an indication if the trend is or is not significant at the 95 percent confidence level for a two-tailed test. An important consideration in any graphical presentation is whether the data is significantly influenced by seasonal changes. If this is the case, then the data should be adjusted for seasonal influences. In order to make such a determination, there should exist at least eight and preferably sixteen observations for each parameter. However, seasonal influences will likely be first suspected from visual observation of the data graphs discussed above. 2.3 DETECTION VERIFICATION PROCEDURE Verification procedures will be conducted in accordance to Chapter 19 Prediction Limit Strategies with Retesting, Section 19.1 Retesting Strategies, pg 19-1 to 19-3 of the Statistical Analysis of Groundwater Monitoring Data at RCRA Facilities, Unified Guidance [EPA, March 2009]. Verification re-sampling is an integral part of the statistical methodology. Without verification re-sampling, much larger statistical limits would be required to achieve site-wide false positive rates of 5% or less. Furthermore, the resulting false negative rate would be greatly increased. The following procedure will be conducted for each compound determined to be initially above Responsive Resourceful Reliable 4

6 its statistical limit. Only compounds that initially exceed their statistical limit will be sampled for verification purposes. If one or more of the parameters are detected above their statistical limit, a verification resample will be collected within 90 days of the parameter(s) being detected above its statistical limit. A SSI will be recorded if verification of an elevated parameter is confirmed in a concentration greater than the control/prediction limit for the discrete verification re-sample. If the re-sampling program confirms that the initial sample represented a laboratory or sampling-induced outlier, the initial sample will be flagged as an outlier and the verification sample value will be used to eliminate bias from the prediction interval, which compares the most recent data points to calculated limits. The original sample value will be maintained in the database as an outlier. 3.0 REFERENCES EPA, April Statistical Analysis of Ground-Water Monitoring Data at RCRA Facilities: Interim Final Guidance. Washington DC: USEPA, Office of Solid Waste. EPA, March Statistical Analysis of Groundwater Monitoring Data at RCRA Facilities: Unified Guidance. Washington DC: USEPA, Office of Resource Conservation and Recovery. U.S. National Archives and Records Administration [USNARA], April CFR Parts 257 and 261: Hazardous and Solid Waste Management System; Disposal of Coal Combustion Residuals From Electric Utilities. Responsive Resourceful Reliable 5