New computer technologies, such as the GRID, seem likely to change the way that environmental models are constructed and used. The GRID is a new hardware and software initiative based on distributed high-performance parallel computers, linked by fast network connections that, to the user, should appear as a single machine. The concept is that the user should not have to worry about where the data necessary for a project are stored, nor where any computational tasks are run. To the user, the software (or 'middleware') should make the GRID appear as a desktop machine. The possibility of using GRID-scale computer networking to link together distributed database and computational engines means that it will become possible to couple together models of many more different environmental systems across disciplinary boundaries and across national administrative boundaries. In fact, this is already possible and is already happening on a limited basis, as demonstrated, for example, in the regional water resources models under construction in Denmark and in the national environmental management models being used in the Netherlands.
There is, however, a real question raised about how these types of interdisciplinary model might be best implemented. In the past, comprehensive modelling systems have been constructed as large complex computer programs. These programs were intended to be general, but have proven to be expensive to develop, difficult to maintain and difficult to apply because of their data demands and needs for parameter identification. With GRID computing technology it will be possible to continue in the same vein, but with more coupled processes and finer spatial and temporal resolutions for the predictions. It is not clear, however, whether this will result in a real improvement in model accuracy and use, because the problems inherent in the current generation of distributed environmental model do not necessarily easily go away with improvements in space and time resolutions of the component models.
There may be another approach, one that will be explored in this commentary. One of the features of having the possibility of these large-scale models is that everywhere is represented. We will have environmental models of everywhere. Once all places are represented within the flexible GRID-based system outlined in what follows below, the data may assume a greater importance than model structures as a means to refine the representation of each place within a learning framework. The result may be a new way of looking at environmental modelling, one that transcends the