Geoscience Victoria (GSV) is redeveloping its geological information systems. A key facet of this work is generating a robust logical data model based on geological, rather than software vendor-specific, principles (for example, there is no partitioning of data into spatial or aspatial domains). The model primarily informs the development of our system's persistence layer but also constrains the over-arching business systems and user interfaces, all of which may be implemented on a variety of platforms.
Initially, GSV adopted the North American Data Model (NADM-C1) as a conceptual model due to its unparalleled breadth of scope and strong geological focus. Using the Unified Modelling Language we turned the NADM-C1 into a provisional logical model, defining specific properties and additional behaviour required by GSV. Another constraint was introduced by our involvement in the development of the Geoscience Mark-up Language (GeoSciML). Initially, GSV were able to implement the GeologicUnit components of GeoSciML but later work modelling geological structures was undertaken before the GeoSciML modelling. Fortunately, both exercises had a common conceptual basis and personnel. Future modelling, addressing our field data, will use the recent OGC Observation and Measurements model to minimise our future modelling effort.
Adopting this approach does present challenges. Models for data exchange, such as GeoSciML, are a good foundation for this work, but care must be taken when adapting the model to define a persistence layer. Behaviour suitable during exchange may not be appropriate for long term storage or manipulation of data. Also, generic elements used during exchange may not define behaviour explicitly enough.
A strength of basing the modelling of the persistence layer on GeoSciML, which conforms to the ISO General Feature Model, is that it defines intrinsic behaviour in a system?such as whether entities have identity in your domain (essentially are they real and independent). However, care must be taken applying this subdivision into Features with identity and Objects. Although Features are typically considered to have a geographic location, it is not essential. For example, aspatial concepts such as orogenic episodes are Features with geological identity. GSV's work has extended a pre-existing business model where the equivalent of a Feature is strongly bound to its spatial representation. Implementing a geological data model in this essentially spatial framework means we are unable to explicitly specify true geological identity throughout the model.
While it cannot be expected to provide a complete definition of system behaviour, a well-designed, geology-based logical model is a valuable part of a complete system requirements specification. Developed with care it can be used over time to constrain the implementation of systems on a variety of platforms.