Data Validation – Navigating the Divide Between Usable and Unusable Data

Data Quality, Management, and Review
Oral Presentation

Prepared by R. Bass, K. Garvin
SynTerra, 148 River Street, Suite 220, Greenville, South Carolina, 29609, United States


Contact Information: rbass@synterracorp.com; 864-421-9999


ABSTRACT

Environmental analytical data used for monitoring, compliance, and decision-making purposes must be reliable and legally defensible. Third-party data validation, typically performed by the consultant, ensures data quality and helps the user identify whether or not data are usable for project objectives.

Data quality assurance begins in the field and continues as samples are analyzed in the laboratory and results are assessed in the office. In the field, quality control (QC) samples such as field blanks, equipment blanks, trip blanks, and field duplicates are collected. The results of the field QC samples are examined during data validation and used to qualify sample results, if necessary, to indicate possible biases/unreliability. A systematic review of the laboratory report is also performed during data validation to evaluate holding times, sample preservation, laboratory blank contamination, spike recovery results, instrument calibration (accuracy), and laboratory duplicate samples (precision). It is important for the data user to understand laboratory assigned qualifiers and how they impact their data, as well as how they differ from the data validation assigned qualifiers.

Even as our industry improves and streamlines data validation, a combination of manual and automated processes is recommended. The degree to which data validation is performed is determined by the project’s data quality objectives and the degree to which uncertainty is introduced in the data validation process. Ultimately, a better understanding of the data validation process from field collection to site assessment and reporting will help subject matter experts and data analysts navigate the divide between usable and unusable data.