Data’s quality assurance
The main objective of ORACLE is to put at the disposal of scientists and operational, data collected during the period of observation. These data require the installation of a quality procedure. This work falls under Irstea policy of quality, established on the whole of its Observation sites, and based on two networks, “Measures and Instrumentation” and “Database”. This quality relies on 1- the metrological aspect (establishment of rigorous protocols for device installation, systematization of the procedures, rigorous and regular maintenance), 2- data processing (frequent data repatriation, checking of coherence of the series, checking of coherence of variables and comparison between parameters) and 3- on the question of the data reconstitution. The objective is to bring in work a number of provisions at the managerial and technical level to lead to reliable results and to be able to prove this reliability (concept of confidence). This work is based in particular on recommendations for quality in research, such as the AFNOR document, FDX50-551.
1- Maintenance and metrology of ORACLE device:
The first stage concerning measurement quality was to set up a database of the ORACLE metrological park with the software installation of maintenance of measurement equipment(SPLIT4®). The technical data of each device, their precise localization on ORACLE, the date of each calibration and the events of maintenance are taken into account by the software. This allows us a better traceability of the equipment and its maintenance, a control of nonconformities and an improvement of the performances of measurement devices.
2- Data processing and 3- reconstitution of data :
Rainfall and Limnimetric records are automated and directly remotely transmitted on an internal database, every hour. Other data are repatriated at regular frequency on an internal database (every two months on average). Limnimetric data are stored after validation on the HYDRO bank, managed by the MEEDDAT.
All data are validated by the research engineer, manager of the Observatory. Semi-automated routines were set up to allow a faster validation of the 34,000 monthly data of ORACLE, but also a certain repeatability of the validation. The use of statistical tools (e.g., average, quartiles…) makes it possible I) to carry out an inter-comparison on data set over the full period of observation and II) to check series coherence. Tools and methods of validation, but also data reconstitution, are based on the experiment and expertise of Irstea on data management, developed for nearly 50 years. In each validated file the validated data and the reconstituted data are labelled.