[RF/Closure Report ] [Next Section]
The site lies along the south/southwest bank of Salado Creek. The Glen Rose Limestone Formation bedrock underlies the soil, at a depth of approximately four feet. Based on the RFI sampling results, closure of SWMU B‑11 required excavation and disposal of the waste material present at the site along with confirmation sampling to ensure all waste residue was removed. In addition, statistical analysis was conducted on RFI bedrock results to support RRS1 closure.
As described in Section 1.4.3, a small number of limestone samples collected from soil borings drilled during the RFI contained metals concentrations slightly exceeding CSSA background levels for the Glen Rose Formation. However, 30 TAC §335.553(d)(2) allows use of statistical comparison using the 95 percent confidence limit of the mean concentration of the contaminant as a representative value for the site. If all of the samples across the site are used to calculate an upper confidence limit (UCL), and the UCL is less than the established background level, then the site can be closed under RRS1.
To calculate the UCL, the data must be normally or log-normally distributed. To test the distribution of the data, the Shapiro-Wilk test of normality is used (if sample sizes are less than or equal to‑50). The Shapiro-Wilk test is included in the U.S. Environmental Protection Agency (EPA) software, ProUCL, used for the UCL calculations (EPA 2003; also located at: http://www.epa.gov/esd/tsc/install.htm). If the distribution is normal, the UCLs are calculated on the raw data. If the distribution is not normal, then the data are log transformed, and the Shapiro-Wilk test of normality applied to the transformed data. If the data were log-normally distributed, the UCLs were calculated based on the transformed data. If the Shapiro-Wilk test statistic (W) exceeds the critical value of the statistic, then that distribution is considered normal or log-normal, according to the distribution of the data.
If the data are neither normally nor log-normally distributed, then non-parametric UCL calculations may be made, with certain limitations. Non-parametric UCL calculations are considered distribution-free, that is, the computations do not make specific assumptions about the shape of the underlying distribution of concentrations, except that the samples are representative of the underlying distribution. For small sample sizes, a “Jackknife” procedure is commonly employed, which is a robust sampling technique where repeated samples are drawn from a given set of observations by omitting each observation in turn, yielding n data sets of size n-1. An estimate of the desired unknown parameter (e.g., sample mean) is then computed for each sample (EPA 2002), and because of the re-sampling, the “Jackknife” will approximate a normal distribution. These procedures, however, are sensitive to highly skewed data and/or data with a large standard deviation.
Using the soil boring data that was collected as part of the B‑11 RFI Report, concentrations of several metals (barium, chromium, nickel, and zinc) slightly exceeded CSSA background values for bedrock, and one VOC (dichlorodifluoromethane) slightly exceeded the RL. These analytes were statistically evaluated, and all statistical evaluation tables are provided in Appendix A. A total of 10 bedrock samples that were collected as part of the B‑11 RFI Report were used for the statistical analyses. For each metal or VOC, the distribution of the data was tested for normality as described above. For this data set, the Shapiro-Wilk critical value W0.05, 9 d.f. = 0.842 (d.f. is the degrees of freedom, generally the total number of samples minus one).
Data flags were treated in the following way for data analysis. If the data were flagged with an “F”, “J”, or “B” flag, the concentrations were retained in the data set. If the data were flagged with a “U” flag, the concentration was entered as ˝ the value of the sample quantitation limit (SQL).
2.1.1 Barium
Barium exceeded background levels at only one location, B11‑SB03 (4.5‑5.0 ft.), with a concentration of 43.41 mg/kg. The data are log-normally distributed (W = 0.858). The UCL calculated by ProUCL (EPA 2003) is 18.07 mg/kg barium. This UCL exceeds the background bedrock value for barium (10 mg/kg).
At the sample location where barium exceeded the background bedrock value and the calculated UCL value, this sample was at the interface of soil and bedrock. It is likely that during the sampling, some soil was introduced to the sample. If that was the case, the background soil concentration for barium is 186 mg/kg, and the UCL calculated above is well below the soil background concentration.
2.1.2 Chromium
Chromium exceeded background levels at only one location, B11‑SB03 (4.5‑5.0 ft.), with a concentration of 8.2 mg/kg. The data are normally distributed (W = 0.932). The calculated UCL is 5.42 mg/kg. The UCL for chromium does not exceed the background bedrock concentration (8.1 mg/kg). Because the calculated UCL for chromium does not exceed the bedrock background concentration, there is no need for further testing and SWMU B‑11 meets the requirements for RRS1 closure.
2.1.3 Nickel
Nickel exceeded background levels at only one location, B11‑SB03 (4.5‑5.0 ft.), with a concentration of 8.14 mg/kg. The data are normally distributed (W = 0.941). The calculated UCL is 5.47 mg/kg. Because the UCL for nickel does not exceed the background bedrock concentration (6.8 mg/kg), there is no need for further testing and SWMU B‑11 meets the requirements for RRS1 closure.
2.1.4 Zinc
Zinc exceeded background levels at only one location, B11‑SB03 (4.5‑5.0 ft.), with a concentration of 15.33 mg/kg. The data are log-normally distributed (W = 0.924). The calculated UCL is 7.69 mg/kg. Because the UCL for zinc does not exceed the background bedrock concentration (11.3 mg/kg), there is no need for further testing and SWMU B‑11 meets the requirements for RRS1 closure.
2.1.5 Dichlorodifluoromethane
Dichlorodifluoromethane exceeded background levels at one location, B11‑SB01 (9.5‑10.0 ft.), at a concentration of 0.0106 mg/kg. The data are not normally distributed or log-normally distributed, and therefore the UCL calculation will be based on a non-parametric “Jackknife” procedure The calculated UCL is 0.003576 mg/kg. There is no bedrock background concentration for dichlorodifluoromethane, but the laboratory RL is 0.005 mg/kg. Because the UCL for dichlorodifluoromethane is less than the laboratory RL, there is not need for further testing and B‑11 meets the requirements for RRS1 closure.
2.1.6 Chloroform
Chloroform was detected above RLs in three samples, but all results were flagged with a “B” because the compound was also detected in the associated blank sample. Because this analyte was associated with sampling or laboratory contamination, its results were not statistically evaluated.
2.1.7 Methylene Chloride
Methylene chloride exceeded background levels at all of the sites sampled. However, there was no known use of methylene chloride at CSSA, and none of the waste encountered during excavation activities indicated possible disposal of VOCs at the site. This chemical is a common component of vinyl gloves and other laboratory/sampling equipment, and its detection is often associated with laboratory/sampling contamination. Given the very low concentrations detected, the absence of any other VOCs, and the consistent low concentration range (0.0110–0.0417 mg/kg) in the samples, laboratory/sampling contamination is strongly suspected, and a statistical evaluation of the results was not conducted.
Although the geophysical survey conducted in April 1995 did not indicate the presence of subsurface anomalies and previous sampling results indicated the UCLs for reported Chemicals of Concern (COCs) to be less than background concentrations, the waste present on the soil surface and the disturbed ground surface indicated that buried debris was potentially present at SWMU B‑11. To address contaminated soils and waste material at the site, excavation activities were conducted from November 5, 2003 to November 14, 2003. During this period, 3,078 yds3 of waste and waste residue were excavated from the site. This volume included all waste at the site, as well as contaminated soils identified during RFI sampling activities. All excavation activities were conducted by Eagle Construction and Environmental Services of San Antonio, Texas, under the supervision of a Parsons construction manager. Waste material that was removed with the soil included concrete rubble, roofing debris, wire, and burnt wood. No ordnance-related materials were identified. The excavated area was approximately 450 feet long (northwest to southeast), and ranged from about 25 feet wide on the southeastern end to about 75 feet wide on the northwestern end. Along the Salado Creek side of the excavation, the excavation depth was very shallow, roughly 2 feet deep. On the opposite side of the excavation, into the southern bank of the creek, the excavation was approximately 8 feet deep.
Based on the location and nature of the waste disposed at SWMU B-11, it is deduced that this material was placed along the creek bank at some time in the past to shore it up and to route surface water flow to the east. An unnamed tributary flows into Salado Creek directly north of SWMU B-11, and a storage building is located approximately 250 feet to the south-southeast of the site. The bulk of the material waste material excavated at the site was located just to the south of this confluence. Today, during heavy precipitation events, flow in Salado Creek often jumps its current banks and the corner of the storage building sometimes floods.
Excavated material was stockpiled onsite during disposal characterization. Disposal of the waste was conducted from December 9, 2003 to December 11, 2003, under waste profile CG-25591, C-2. Manifests are provided in Appendix B. Toxicity characteristic leaching procedure (TCLP) results from the stockpiled waste indicated the material met non-hazardous Class 2 criteria, so the waste was disposed of at Covel Gardens Landfill in San Antonio, Texas. After excavation activities were complete, the site was graded so that the site surface was smooth. The following photographs show excavation activities at the site.
Confirmation sampling was performed on November 25, 2003 to verify that RRS1 criteria had been met. After the waste and waste residue were excavated, samples were collected on the sidewalls and bottom of the excavation pit. A total of eight sidewall samples and five bottom samples were collected in the excavation area, at a rate of approximately one per 100 linear feet. At the southeastern end of the excavation area, an excavation sidewall did not exist (waste was excavated leaving a level ground surface). Only excavation bottom
samples were collected in this area. All sample locations from both the February 2000 and November 2003 sampling events and the area excavated are shown on Figure B11‑9.
Each closure sample was analyzed for arsenic (by method SW7060A), barium, chromium, nickel, and zinc (method SW6010B), and cadmium (method SW7131A). The results were compared to the background soil concentrations. The analytical suite used for the closure samples was determined by the metals that exceeded CSSA background criteria as reported in the SWMU B‑11 RFI Report. The confirmation sampling results are presented in Table B11-2. For the confirmation sampling, all of the metals were below CSSA’s established background levels, with one exception in a duplicate sample. Cadmium in the duplicate of one sidewall sample, B11‑SW02, slightly exceeded the background soil concentration of 3.0 mg/kg. However, cadmium was present at a concentration below the background level in the parent sample.
Therefore, cadmium was statistically evaluated using the samples collected as part of the closure effort. The most conservative testing would include the highest value of cadmium (3.88 mg/kg) in the analyses, even though this sample is a field duplicate. For the cadmium data collected as part of the closure effort, if the field duplicate sample is included in the analysis (instead of the original sample), the data are not normally or log-normally distributed (Wcritical = 0.866, Wtest = 0.398) and a UCL cannot be determined for one of these underlying distributions. The data are moderately skewed, with σ in the interval between 0.5 and 1.0 (note: σ = 0.999, σ is the standard deviation of the data). Therefore, the non-parametric UCL can be calculated with the “Jackknife” procedure. The UCL calculated in this way is 1.06 mg/kg. The background soils value for cadmium is 3.0 mg/kg, and therefore, the calculated UCL is less than the background soils value. Results of this statistical analysis are presented in Appendix A.