Quality control flags (QC flags) are an important and powerful element of the quality assessment process in environmental science. They enable data providers and managers to attach simple data quality qualifying markers to data records. These markers can then be used by data users to quickly filter the data based on previous quality assessments and according to their own particular requirement. QC flags are widely used in the oceanographic community, and schemes have multiplied over the years, with individual programmes, data centres or group of researchers adopting, extending or creating schemes to best address their specific needs.Our aim here is to build upon the work started in the frame of the European programme SeaDataNet by Lesley Rickards, Reiner Schlitzer and colleagues. Taking into consideration existing QC flag schemes used for discrete water bottle samples but also other types of data, we need to: 1) discuss and document the pros and cons of the existing schemes; 2) recommend a mapping between existing schemes; and 3) recommend a strategy for the future. |
The file "wod09readme.pdf" (also avilable at ftp://ftp.nodc.noaa.gov/pub/WOD09/DOC/wod09readme.pdf) is the latest draft of the NODC World Ocean database 2009 documentation. Page 39 of this document describes the QC procedures and Table 12 (Page 47) illustrates all of the available QC flags. The latest versions of this and other documents can be found at http://www.nodc.noaa.gov/OC5/WOD09/docwod09.html.
The document "ODV4_QualityFlagSets.pdf" by Reiner Schlitzer summarizes all Quality Flag (QF) schemes (14 in total) that Ocean Data View (ODV) can handle and knows about. There is one page per scheme, containing the flag values and descriptions, as well as the mappings (adopted by ODV) to the other schemes. There are clickable links to the relevant documentation and to the pages of the other schemes.
The document "DefinitionQualityFlags.pdf" has the quality control flag definitions as used in the NODC World Ocean Database (see also the wod09readme.pdf document)
Comment from Reiner Schlitzer: “The different schemes range from simple (ODV) to extensive (BODC, PANGAEA, SMHI), and mappings are sometimes difficult. Quite a number of schemes contain flags that have nothing to do with quality but rather with the processing history of a value (e.g., “interpolated value”). Such values could still be good or bad, and usage of the particular flag basically leave the value without a quality assessment.”
Nominal data users often use QC flags determined or based on different QC schemes. For instance, the NODC World Ocean Database (WOD) uses a qc flag scheme that involves qc flags using numerical and subjective tests. A user may or may not use these qc flags. WOD also preserves when possible, the originator's numerical qc flags. The user then has a choice of qc flags. Other qc schemes elsewhere may use similar ideas.
One desirable outcome of this workshop is to discuss and hopefully recommend a set of QC flags that are based on operationally defined qc tests (e.g., a t, S, pressure data failed a density inversion test, a data failed an unrealistic large vertical gradient, data failed an unrealistic basin-scale data range, etc). In this way one can assign specific qc flags to data value that users can all at least understand. This will in principle reduce ambiguity in the meaning of some qc flag name convention.
There are different sets of QC flags. They have been developed in national data centers and within various national and international projects and each of them is justified by their authors. Some of them are based on expert's decisions, the other are based on the results of automatic evaluation (σ criteria, spite detection, climate range, etc.) and expert's decisions. It might be the best solution for this workshop is to make an overview of the existing schemes of QC flags and to recommend a simple one. I believe that more important outcome is to demonstrate the existing methods and tools for QC.
ODV4_QualityFlagSets.pdf updated with correct schema names and mapping of QC flags used by ODV (developed by Prof. Dr. Reiner Schlitzer, AWI).