ARCHIVED REPORTS_XR0012101 - Laserfiche WebLink

Home Browse Search

selected for the site because it makes no assumptions about the underlying distribution of j the data (e.g., normal or log-normal distributions). i i Interpretation of Results As with any quantitative evaluation, the results of the Mann-Kendall test for trend should be considered within the context of an overall understanding of site-specific groundwater conditions. That is, the statistical calculation should not be treated as a "black box." Instead the data should also be evaluated after each sampling round within the context of the site conceptual model, to avoid misinterpretation or misuse of the results. As an example, it is common for the Mann-Kendall test to be overly influenced by the most recent analytical result. If that result is greater than all the previous results, than the Mann-Kendall statistic may indicate a statistically significant trend. Use of the Mann-Kendall Test at the Site: The Mann-Kendall statistical test was performed at all site monitoring wells for compounds that were detected in more than half of the groundwater sampling rounds. Because an assumed concentration of half the laboratory reporting limit is used in the test for non- detection results, the test is meaningless when half or more of the results for a compound are non-detect. Therefore, the test was not performed for analytes and wells where the analyte was not detected in more than half the analyses. Likewise, non-detection results for which no detection limit is known were not used in the analysis. For several wells in the immediate vicinity of the ozone sparging system (i.e., wells NP-1-22, NP-1-65, NP-2-22, NP-2-60, NP-2-116, U-3 and U-10), the Mann-Kendall test was run both j for all historic groundwater data and for the results obtained subsequent to the installation of the upgraded ozone sparging system. i The presence of trends in chemical concentration for each well evaluated are presented for the 80%, 90%and 95% confidence intervals (i.e., P=0.05, 0.1 and 0.2), and the coefficient of variation is presented to evaluate plume stability where no trend was found at these levels of confidence. Where no trend is apparent, a coefficient of variation below 1.0 is generally indicative of a stable plume. Where no trend is indicated for these confidence intervals, a general indication of the trend (if any exists) can be inferred for a lower degree of confidence by use of the S-statistic. Positive values of the S-statistic correspond to generally upward trends and negative values correspond to downward trends. The probability value, P, is used to determine the confidence level of the trend. A regulatory guidance document(Wisconsin Register, 2001) is included in this Appendix, as it provides a clear explanation of the statistical test, its significance and its limitations. More detailed information can be found in Statistical Methods for Environmental Pollution Monitoring(Gilbert, 1987).