ARCHIVED REPORTS XR0008945 - Laserfiche WebLink

Home Browse Search

Plan view reaps for each of these three sites are included in Appendix 1 These maps show relevant site features and a set of time-series plots of historical groundwater concentration data for each monitor well At each of the three sites, soil cores were collected below the water table at three locations, nominally two within the plume and one background sample Soil sampling locations are indicated on the individual site maps included in Appendix 1 The soil samples were subjected to laboratory microcosm analysis under a variety of - environmental conditions These tests were intended to help confirm the presence and activity of microbes capable of degrading MTBE and TBA, as well as to provide an indication of a possible pathway of TBA formation from MtBE Additional discussion of these tests and test results are included in Section 5 of this report 3.1. Statistical analysis of groundwater monitoring data Statistical analysis of the time-dependent groundwater TBA and MTBE concentration data has been completed This analysis provides interpretation of increasing, steady, or decreasing trends in concentration, as well as confidence intervals of the trend A summary of the statistical tests used in the data evaluation follows Time (t) versus concentration (c) data for a single chemical (either TBA or MTBE) and single well, is examined in each analysis • Only detected data, and data without qualified data codes are included in the trend evaluation Non-detected data and data with qualified data codes are neglected in that they have higher levels of associated error than the detected, unqualified data • Simple substitution was not used (that is substituting a non-detect level or a fraction of non-detect level, as commonly done in risk assessment analysis) as the detection levels in this data set are generally variable and can be higher than other detected data in the same data set This approach for non-detects is reasonable for a trend analysis • A minimum of four detected data samples per well per chemical has been set as a criterion before inferring trends based on the statistical calculations More samples will yield narrower confidence limits The analysis is confined to a set of non-parametric statistical tests As a general point, non-parametric tests can be more indicative of a central tendency in the data, regression tests (which we have not applied) can be more significantly influenced by infrequent extreme values In analysis for each well • The data is transformed as x = t, y =1n(c) Non-parametric tests do not depend on the data distribution (by definition) This transformation yields a first-order estimate (in concentration) of a rate parameter • A Mann-Kendall statistic, S (Kendall, 1970, Gfbbons and Coleman, 2001), is calculated, and an associated probability is estimated • The Mann-Kendall statistic as applied to trend, evaluates a total count in change of concentration (increasing, +1, equal, 0, decreasing, -1) between all sequential permutations of data pairs (yi - y,)with xj >x, , in the N data points (x, y) 3