SITE INFORMATION AND CORRESPONDENCE

Home Browse Search

Peter MacNicholl 11 April 2007 Dieldrin in Groundwater at Tracy Depot Page 2 exposure point concentrations using ProUCL 4.0, which properly takes into account the large number of samples in which dieldrin was below the detection limit. 2. Prior Agreement with Regulators: An e-mail is attached to this report, sent by Mr. Xuan Tran of USEPA Region 9, dated 30 January 2007, and addressed to Mr. Marshall Cloud of Defense Logistics Agency. This e-mail states that USEPA Region 9, San Francisco Bay Re- gional Water Quality Control Board, and DTSC will allow the Department of Defense to eliminate dieldrin as a chemical of potential concern in groundwater, provided an assessment of the highest concentration detected shows risks and hazard in the "risk management range" for domestic use of the groundwater. Mr. Tran goes on to state that risks or hazards higher than this might trigger re-opening the Record of Decision for the groundwater at this site. 3. Underestimation of Risk and Hazard: CmAX for dieldrin in groundwater was 0.19 µg/L, but the Department of Defense used 0.049 µg/L as their exposure point concentration. This latter value'is stated to be the 95% upper confidence limit on the mean concentration (95UCL), as estimated by ProUCL 3.0. Cancer risk and non-cancer hazard were estimated to be 5 E-5 and 0.41 using the 95UCL. These values would have been more than fourfold higher using CmAx, or—2 E-4 and—1.6. These latter results would have failed the criteria laid out in Mr. Tran's e-mail. Dieldrin should be retained as a chemical of potential concern in groundwater. 4. Exposure Point Concentration Using ProUCL 4.0: In case the 95UCL is needed for any future calculations or regulatory decisions, we re-calculated this value for the Department of Defense with several corrections: a. Duplicates: Table 5 contains 153 analyses for dieldrin in groundwater. We found that twelve of these results were either duplicates or confirmation samples. These are not in- dependent samples, so we culled the data set by using the higher of the two results in each case. An additional ten samples were analyzed twice, once with Method 8081 and again with Method 8081A. These are apparently independent samples, so we retained them in the analysis. b. Detection Limits: The Department of Defense used one-half the detection limit as the value for non-detects in ProUCL 3.0. Although Superfund guidance recommended this practice in 1989, it has been shown to be incorrect (see Helsel, 2005) and new guidance has been issued (USEPA, 2002). ProUCL 4.0 uses the non-parametric Kaplan-Meier method for these censored data. In Table 1 attached to this memorandum we show the culled data set up for analysis in ProUCL 4.0 (beta), a program which properly takes into account non-detected data and mul- tiple detection limits. This version of ProUCL will be available to the public within a few weeks. Table 2 shows the output for ProUCL 4 using the data in Table 1. The frequency of detection is 39 of 141 analyzed. The data for dieldrin in groundwater follow a lognormal dis- tribution (Table 2, center right). However, "reverse" Kaplan-Meier statistics are required to accommodate the many non-detects and multiple detection limits (Helsel, 2005). The rec- ommended 95UCL is 0.041 µg/L, shown at the bottom right of Table 2.