Reducing the sampling frequency of groundwater monitoring wells

As part of a joint LLNL/SRTC project, a methodology for selecting sampling frequencies is evolving that introduces statistical thinking and cost effectiveness into the sampling schedule selection practices now commonly employed on environmental projects. Our current emphasis is on descriptive rather than inferential statistics. Environmental monitoring data are inherently messy, being plagued by such problems as extremely high variability and left-censoring. As a result, real data often fail to meet the assumptions required for the appropriate application of many statistical methods. Rather than abandon the quantitative approach in these cases, however, the methodology employs simple statistical techniques to bring a measure of objectivity and reproducibility to the process. The techniques are applied within the framework of decision logic, which inrerprets the numerical results from the standpoint of chemistry-related professional judgment and the regulatory context. This paper presents the methodology's basic concepts together with early implementation results, showing the estimated cost savings.