QA/QC protocols and documentation for ocean data – a case for standardization?

Since the earliest days of ocean sciences, the acquisition of observed data called for quality assurance and quality control procedures that had to be developed to ensure consistency of the collected information. The main goal is to exclude any grossly flawed data from the collected data stream either by preparing the used instrumentation in the best possible way and by eliminating any obviously wrong data within the data stream. However, only if one follows an agreed upon scheme or protocol and associated parametrization, it can be ensured that the information on the region of interest can be judged as traceable and repeatable. The issue that comes into the picture is that different observing configurations and environments will call for dedicated QA/QC procedures. This is what can be seen today, i.e., that every observing network like Ocean Networks Canada [1], OceanSITES [2], ARGO [3], etc. is following its own best practices. While these QA/QC practices are well documented at a high, procedural level, their individual application to e.g., time series may vary with respect to the selection and the sequence of individual tests. Further, the individual parametrization of these tests such as thresholds used or supporting data allows a large degree of freedom. As a result of these procedures, in most cases a verbose assessment of the data quality (good, bad) is determined and stored together with the raw data. However, the procedures used, the selection, sequence and parameterization of the tests remains undocumented, which makes verification impossible. For now, it is difficult to foresee whether a common standard for QA/QC can be developed and what any type of standard shall encompass. A new avenue maybe be opened by stripping down the QA/QC process into the essential functional steps taking into consideration the actual information flow. In analogy to the ISO 19115 series of standards one should consider to model QC processes and workflow models with UML models along with a common vocabulary or ontology on QC tests and their parameters. This would allow enough flexibility to integrate different workflows as necessary for the different observing scenarios and at the same time would allow for a rapid implementation of new or additional functional blocks into the basic QA/QC model. Based on this model we propose to develop a unified QA/QC documentation standard as well as a linked data-based replacement of the now used simple quality flags. The suggested method shall be exemplified by using the QA/QC routines developed as part of IOOS/QARTOD [4]. QARTOD has established authoritative QA/QC procedures for the US IOOS core variables which then enter the related manuals. The scope of QARTOD however, goes well beyond US observing efforts and includes the coordination with similar efforts of the Global Ocean Observing System (GOOS) and transatlantic programs like the EU funded project AtlantOS [5].