Journal of Loss Prevention in The Process Industries, Vol.9, No.6, 401-411, 1996
CRAMD - A database for the validation of models used in chemical risk assessment
The objective of the project was to design and construct a database of experimental data for model validation and development applications. The experimental data were to cover source term and dispersion behaviour over flat and complex terrain and were to be collected from CEC funded research projects and other suitable sources of experimental data. The main results of the project are: A prototype database containing specially chosen data handling and processing facilities for model validation and development applications has been constructed. The database has been named CRAMD-Chemical Risk Assessment Modelling Database. The number of data sets which could be incorporated into the database were limited by project resource constraints. Representative examples containing different data formats and experimental measurement techniques were therefore selected in order to demonstrate the data handling and processing capabilities of the CRAMD prototype. Key design requirements were that the database should be readily accessible and easily extended to cope with new data formats and modelling applications. These were achieved by basing CRAMD on an object-oriented software platform with user access via the Windows-based World Wide Web (WWW) internet connection. Difficulties were experienced with data collection which exemplify the need for a central EC archiving facility in this work area. Examples were found of data being discarded, lost or poorly documented, There was also a poor response to formal requests to supply data for the database. It is suggested that future CEC funding for experimental and model validation investigations should be conditional on all measurement results which are used or acquired during the work be returned to the CEC in a fully documented, computer readable format for archiving. Clearly, CRAMD could provide a suitable receptacle for such data.