Class Details

The 1-day course, Best Practices for Efficient Sampling Designs, discusses sampling designs for contaminated soils that go beyond simple random or “gridded” grab-sample formats. More advanced designs can reduce sampling and analytical costs while simultaneously improving data quality and usability. This course is presented using common sense concepts (not statistical equations!). Project managers who attend will find they are more confident providing critical reviews of proposed sampling designs and communicating their data needs to their contractors.

The course will begin with a brief review of the difficulties posed by generating data from heterogeneous environmental media such as soils, sediments, and groundwater aquifers. Newer technologies and best practices that outperform older strategies are briefly described. Data sets from actual sites illustrate the pitfalls of older practices.

The first modules lay the groundwork to understand why common sampling designs can fall short:

• Misleading data and faulty decisions can be the consequences of contaminant & matrix heterogeneity if countermeasures are not taken.
• Collecting representative data requires reconciling the mismatch between the tiny spatial scale of sample collection/analysis and the much, much larger spatial scale of cleanup decisions.
• Representative data cannot be generated unless a project-wide conceptual site model (CSM) is used to reconcile these different scales when planning a sampling project.
• The CSM is the scientific (not necessarily statistical) hypothesis that is tested, modified, and refined until confident decision-making is possible.
• Reconciling the scale mismatch requires careful forethought to ensure samples are taken at the right time, in the right place, and in the right way, so that sampling density is increased in the right places, even if the overall number of samples is decreased.

Subsequent modules discuss how to improve the performance of traditional designs when they are used. Caveats on the uses and misuses of the Visual Sample Plan (VSP) software package will also be covered. The benefits in design performance and cost-savings offered by non-traditional designs will be demonstrated.

• Sampling designs are based on two different sampling goals: searching (such as searching for a source, hotspot, or contamination boundaries) and parameter estimation (such as calculation of a mean and UCL).
• Non-traditional designs to be covered include multi-increment sampling and compositing (these will be presented as 2 different sampling strategies) to estimate averages; and geoBayesian designs that can bound contaminated areas to a higher degree of confidence, but with far fewer samples, than traditional designs.
• Real-time adaptive/dynamic refinement of sampling and analytical quality is key for affordably achieving high quality data despite the complications of contaminant variability.

This course will be instructed by Deana Crumbling and Steve Dyment, Technology Innovation and Field Services Division, EPA Office of Superfund Remediation and Technology Innovation and Robert Johnson, Argonne National Laboratory.