Special study for the statistical evaluation of groundwater data trends. Final report

摘要

Analysis of trends over time in the concentrations of chemicals in groundwater at Uranium Mill Tailings Remedial Action (UMTRA) Project sites can provide valuable information for monitoring the performance of disposal cells and the effectiveness of groundwater restoration activities. Random variation in data may obscure real trends or may produce the illusion of a trend where none exists, so statistical methods are needed to reliably detect and estimate trends. Trend analysis includes both trend detection and estimation. Trend detection uses statistical hypothesis testing and provides a yes or no answer regarding the existence of a trend. Hypothesis tests try to reach a balance between false negative and false positive conclusions. To quantify the magnitude of a trend, estimation is required. This report presents the statistical concepts that are necessary for understanding trend analysis. The types of patterns most likely to occur in UMTRA data sets are emphasized. Two general approaches to analyzing data for trends are proposed and recommendations are given to assist UMTRA Project staff in selecting an appropriate method for their site data. Trend analysis is much more difficult when data contain values less than the reported laboratory detection limit. The complications that arise are explained. This report also discusses the impact of data collection procedures on statistical trend methods and offers recommendations to improve the efficiency of the methods and reduce sampling costs. Guidance for determining how many sampling rounds might be needed by statistical methods to detect trends of various magnitudes is presented. This information could be useful in planning site monitoring activities.