Geoscience Reference
In-Depth Information
The effectiveness of a variety of prophylactic and social interventions are considered against the
spread of the virus. Interestingly, Ferguson notes, in an accompanying narrative, the exceptional
computational demands of this simulation process - the processing required a computer with 20
times the memory of a typical PC and 'ten high-powered computers in parallel, but even then the
final runs took more than a month of computer time' (Ferguson, 2005).
Similar work has been undertaken by the EpiSims group at the Santa Fe Institute and later
the Virginia Biotech Institute (Eubank et al., 2004). This work, which was originally grounded in
transportation research (see Section 10.10), has also been extended to energy and other commodity
markets (Atkins et al., 2007), although typically limited to the urban and regional scale. At the other
extreme, Epstein has more recently attempted to trump the parochial aspirations of national scale
modelling with a global model of billions of agents (Epstein, 2009). Whether such an ambitious
project is capable of yielding valuable results is a matter of conjecture; nevertheless, it is clear that
the implied requirements for both data and computation are substantial, to say the least.
In our own work, we have majored on the link between MSMs and agent-based models (ABMs).
Since MSM is typically firmly grounded in empirical contexts with planning and policy objectives,
this also has the effect of scaling up ABM to a realistic, applied scale. In the work of Rene Jordan,
the adoption in the context of local housing market dynamics (Jordan et al., 2012) leads to a model
in the style of Schelling with a real-world perspective, which has also been attempted elsewhere
(Mahdavi et al., 2007). Hybrid models of MSM and ABM have also been developed in the con-
text of retail markets (Heppenstall et al., 2005), crime simulation (Malleson and Birkin, 2011) and
demographic forecasting (Wu and Birkin, 2012).
Although the movement from the aggregate to the individual is a natural and positive scientific
progression, it is one that has only been realised as new computational resources have been made
available and relies heavily on the availability of even more powerful machines in the future. Only
the most basic individual-level models, such as Thomas Schelling's early 1D attempts, can be cal-
culated by hand. However, to make use of the most powerful resources, a researcher requires a high
degree of programming skills and often membership to a physical institution such as a university.
Hence, e-Research, which provides a platform for nonexperts to use high-performance computers
and advanced models, is essential for the wider uptake of simulation and GC. The NeISS project,
discussed previously, is an example of such a provision.
10.9 MOVEMENT FROM THE DEDUCTIVE TO THE INDUCTIVE
In recent years, the movement to data-intensive research has become significant in the physical
sciences to the extent that Bell et al. (2009) feel ready to characterise this as a fourth paradigm in
scientific inquiry (following earlier paradigms of experimental science, deductive reasoning and
simulation). Such trends have not been lost on social scientists. In particular, the contribution of
Savage and Burrows (2007) is noteworthy in challenging the dependence on social surveys, which
are limited in both their scope and size - this argument is most specifically directed towards the dis-
cipline of sociology, but clearly, the considerations are more general. Stan Openshaw's earlier con-
tributions in this regard can be seen as typically forward-looking. For example, early applications
of the geographical analysis machine to the search for cancer clusters (Openshaw et al., 1988) are
a direct example of the data-driven approach to pattern detection. Rather than testing for abnormal
incidence of disease around candidate sites, the whole procedure is inverted so that huge numbers
of possible clusters are tested for significance and potential explanations are reserved for later. The
similarities to AstroGrid are marked, in which the sky is searched for potentially interesting phe-
nomena with no a priori view of where they might be found.
The new revolution in data-intensive science is supported by a plethora of new information
sources about the physical and social world. On the one hand, data which were formerly pro-
tected under commercial licences are now becoming much more freely available, for example, in
the United Kingdom as part of the Open Data initiative. The publication of an array of mapping
Search WWH ::




Custom Search