An environmental consulting firm conducts RCRA groundwater detection monitoring at an active industrial hazardous waste landfill to determine whether down-gradient compliance wells exhibit statistically significant evidence of contamination above up-gradient background levels. Statistical hypothesis testing is structured as: Null hypothesis $H_0: \mu_{\text{down}} = \mu_{\text{background}}$ (no release), versus Alternative hypothesis $H_1: \mu_{\text{down}} > \mu_{\text{background}}$ (contamination present). Under US EPA Unified Statistical Guidance for Groundwater and NCEES principles, select the TWO true statements regarding this statistical testing program.
- AA Type I error (false positive) occurs if the statistical test concludes that groundwater contamination is present when the facility has not actually released any contaminants
- BA Type II error (false negative) occurs when a clean monitoring well is falsely declared contaminated, triggering unnecessary corrective action
- CWhen background monitoring data follows a normal distribution, an Upper Tolerance Limit (UTL = X̄ + k·s) or prediction limit can establish a compliant threshold above which a down-gradient value indicates statistical contamination
- DSetting an extremely small significance level (such as α = 0.0001) always maximizes the statistical power (1 - β) to detect real contaminant leaks early
Show the solution
Statistical analysis of groundwater monitoring data at waste facilities requires balancing environmental protection against false facility alarms under US EPA 40 CFR Part 264 Subpart F regulations. Step 1: Definition of statistical errors: A Type I error (probability $\alpha$) is a false positive—rejecting $H_0$ when $H_0$ is true (falsely alleging a leak when none occurred, Statement A). A Type II error (probability $\beta$) is a false negative—failing to reject $H_0$ when $H_1$ is true (failing to detect an actual leak; Statement B is incorrect because it confuses Type I and Type II errors). Step 2: Threshold estimation: In detection monitoring, an Upper Tolerance Limit (UTL) or Upper Prediction Limit (UPL) calculated from up-gradient background data as $\bar{X} + k \cdot s$ defines an upper boundary that contains a specified coverage percentage of the population with confidence $(1 - \alpha)$; values above this threshold trigger verification retesting or assessment monitoring (Statement C). Statement D is false because reducing $\alpha$ directly increases $\beta$, which decreases statistical power ($1 - \beta$) and causes small releases to be missed.