Systemic data management for mathematical modelling of environmental problems

With the diversification of modelling activities encouraged by versatile modelling tools, handling their datasets has become a formidable problem. A further impetus stems from the emergence of the real‐time forecasting culture, transforming data embedded in computer programs of one‐off modelling activities of the 1970s‐1980s into dataset assets, an important feature of modelling since the 1990s, where modelling has emerged as a practice with a pivotal role to data transactions. The scope for data is now vast but in legacy data management practices datasets are fragmented, not transparent outside their native software systems, and normally “monolithic”. Emerging initiatives on published interfaces will make datasets transparent outside their native systems but will not solve the fragmentation and monolithic problems. These problems signify a lack of science base in data management and as such it is necessary to unravel inherent generic structures in data. This paper outlines root causes for these problems and presents a tentative solution referred to as “systemic data management”, which is capable of solving the above problems through the assemblage of packaged data. Categorisation is presented as a packaging methodology and the various sources contributing to the generic structure of data are outlined, e.g. modelling techniques, modelling problems, application areas and application problems. The opportunities offered by systemic data management include: promoting transparency among datasets of different software systems; exploiting inherent synergies within data; and treating data as assets with a long‐term view on reuse of these assets in an integrated capability.