Global Data Evaluation and Visualisation Laboratory

The requirement

Earth observing systems, both ground-based and satellite, are now providing vast amounts of data on the properties and evolution of the various components of the Earth system. The assimilation of these data into sophisticated numerical models is being undertaken by the Data Assimilation Research Centre, and by weather forecasting centres such as the European Centre for Medium Range Weather Forecasts. This process produces quality controlled, 4-D (space-time) global data sets of the atmosphere and ocean, with long-term plans to extend the application of the data assimilation methodology to other components of the Earth system, e.g. the land. The production of such global data sets is vital for the full exploitation of earth observation data in studies of the complex processes that govern the interactions of the Earth system and the consequent changes in the system, e.g. in the climate.

It is, however, a major intellectual challenge to understand the complexities of processes and interactions in the Earth system from the large arrays of gridded data produced by data assimilation. (The same comment applies to the output of global climate prediction models.) For example, the multi-faceted evolution of three-dimensional structures in the atmosphere are often very difficult to comprehend. It can take many years of painstaking inspection of time-sequences of two-dimensional (still) pictures to build up a mental picture of a particular phenomenon. Many well-respected scientists, with long research experience, acknowledge their great difficulty in doing this.

A much more effective way to investigate the evolution of the Earth system is through computer animations, run singly or simultaneously, in two or three dimensions. A central requirement is the ability to extract different fields easily and quickly from multi-decadal, global data sets and to run simultaneous animations of several related fields, portraying the evolution of the atmosphere over several decades in 10 minutes (say) of animation. The research and teaching potential of such a capability are enormous.

This proposal is therefore to establish a Global Data Exploration and Visualisation Laboratory (GDEVL), hosted by the DARC core unit at the University of Reading, with the following design criteria:

• The system, as first installed, must satisfy precise performance criteria (the example given above is one of them) without providing extra, expensive functionality that may not be utilised.
• The system must be readily upgradeable to avoid technological redundancy, and be capable of expansion as a permanent facility to meet the evolving needs of a growing community of Earth system scientists.
• The system must be well suited not only to the needs of basic research, but also to those of postgraduate training.
• DARC is a distributed collaboration involving several universities and institutes, and operating as a virtual centre. The system must cater for this structure, as well as for a distributed community of potential users who could derive benefit from the system as a nascent national facility.

The solution in outline

The proposed hardware system, together with advanced graphics software, meet the above design criteria. The system is based around a multi-terabyte raid array data store, a server and a high performance graphics workstation. Data are received by the server through its connection to Janet and are stored in its incoming format in the data store.

The server may also be used in conjunction with a multi-processor, to manipulate these data into a format more amenable to visualisation. For example, a series of model analyses files each comprising data on several meteorological fields at a single time might be converted into short time-series files, each containing a single field. These resulting time-series would also be stored on the data store.

Remote users can process data from the data store with the results being made available via a secure ftp server for eventual migration to their home institutes. A high performance graphics workstation is the final step in the chain, allowing subsets of data selected by the server to the presented visually to the scientist.