Terra Mentis has extensive knowledge and experience with the statistical evaluation of data including: the evaluation of data ranking to determine how appropriate the data might be to actually conduct a risk assessment, the housing of large data sets on databases and the statistical reduction of data to provide summaries for risk assessment. We have access to statisticians with Expert Witnesses on Superfund environmental projects for the private clients, as reviewers of SAPs and Risk Assessments for State Agencies, and as risk assessors/statisticians on RCRA and Superfund sites with soil, sediment, groundwater, surface water, and indoor air contamination. We have the broad experience across media and sites to know where the weak points typically are in sampling plans and interpretation. In particular, background issues are always complicated for any solid media and especially for indoor air. Many simple statistical methods of comparison are inadequate given the typical size of most background sample populations and the concern about whether the selected background population is representative of the site. New methods, such as interval ANOVA, show promise of alleviating some of the past statistical difficulties in this area. Whenever background is expected to be an issue, specific DQOs should be developed or evaluated to define how this issue will be addressed.
Another important step when dealing with data is to ensure that confidence in the data has been calculated. Subsequent to this step, the actual assessment of sampling adequacy is very dependent on the Data Quality Objectives and factors such as potential reuse or redevelopment of the site and will be reviewed in these contexts. For areas where no human populations are present or anticipated and the goal is to characterize groundwater or soil as sources to groundwater or air, characterization of sources to a certainty level of 75% may be adequate. Where sensitive sub-populations such as children or the elderly may be exposed or there are compounds that bioaccumulate, both a definition of sources and extent require more certainty. The number of samples required will depend on the statistical certainty needed for the population of concern. Certainties of 95-99% can be achieved, but the cost increases exponentially based on "power curves" that require very small sampling grids.