OR/14/042 The development of modelling platforms
Not all scientific questions requiring modelling need researchers to use an integrated modelling approach. We must be led by the research requirements not the technology. Currently in the UK we do not have an readily accessible infrastructure that is suitable for linking and running models in a “plug and play” style that is accessible and user friendly. The development of modelling platforms should therefore be concurrent with research being undertaken and the technical solutions to the research problems being solved should as much as possible be generic. Platforms have the ability to open up our models to a wider community, but need to be future proofed as much as possible in terms of models and data structures.
What do we mean by a platform and what does it do?
A basic framework is essentially an environment in which the user can explore data using a selection of data manipulation tools, models and visualisation methods.
A platform provides a structure which facilitates the integration of model components and environmental datasets in a way that can deliver the outputs required by users; including an objective assessment of the uncertainties associated with the output.
A platform, therefore, consists of an infrastructure consisting of hardware and software with a defined set of standards for interaction between code and datasets.
What do we need a platform to be able to do?
The platform needs to be flexible so that the software and hardware can be changed easily to answer different questions using the same platform. The scale of the questions that you are trying to answer will ultimately drive what hardware you need i.e. a laptop to solve a dredging problem or HPC to solve UK flooding problem.
No single platform would be able to answer all questions posed; a small number of platforms would provide a much smarter and cost effective approach.
The platform environment does not need to be too complex. We already have tools that if combined could become a platform.
It is essential within a linked modelling environment that we develop generic standards that are suitable for a number of communities that describe the models, data inputs and outputs, and the computing environment. Metadata must be created in order to provide information to non-specialist users of the platforms. This would have to be tailored to different communities (households/policy makers). Areas that would need to be addressed are:
- Different layers of metadata
- Metadata is essential if impacts are unexpected, automating the construction of a model-chain
- Platform needs to generate its own metadata
- Version data/library is important
- Liability and traceability are important (litigation drives things in the US)
- We need to be more rigorous with linked components
Scaling is also important, as different properties may emerge at different scales. For example, gridded data is available at different resolutions, but it may not be meaningful to use data generated at one resolution to drive models at a finer scale. These problems may be unexpected where multiple systems are interacting therefore:
- Platforms should have the tools to explore impacts
- Platforms could be used as a means of bridging scales
- Scale bridging needs to be done in an intelligent way
- Built-in checks are important to tell you when something is not scaled properly