Geoscience Reference
In-Depth Information
Longley writes: '… GeoComputation has become integral to our understanding of spatial struc-
ture' (Longley, 1998b, p. 83). However, there is no reason to assume that only quantitative geog-
raphy and GIS will benefit; indeed, those non-quantitative areas of geography which are concepts
or theory rich but data poor may also have much to gain; see Openshaw and Openshaw (1997) and
Openshaw (1996, 1998a).
It is also argued that there will be no highly visible HPC revolution that suddenly sweeps all before
it. Instead the HPC revolution is silent and almost invisible. Most of the potential users probably still
read the wrong literature and attend the wrong conferences to notice what is going on. A faster PC
is merely the sharpest tip of a massive iceberg of HPC developments. Yet in those areas that need it
and where a computational paradigm may be helpful, then there is a way forward. If the current HPC
machines are too slow, then be patient; soon there will be much faster ones, but you need to start
developing the new approaches now and then safeguard your software investment by using portable
programming languages and conforming to emerging software standards. However, you do not need
access to the world's fastest HPC to start the process rolling. With modern parallel programming
tools, you can now write portable scalable codes that can be developed and proven to work on low-
end HPC platforms (e.g. workstation farms) before moving on to real-world large-scale applications.
Indeed you can even assemble your own workstation farms and test out your applications locally and
secure in the belief that if it works well on your workstation farm, it will probably do far better on a
real HPC machine. See Adnan et al. (2014) in this topic for more information on parallel computing.
It is an interesting thought that GC could act as an attractor for computationally minded scientists
from other fields. It is becoming apparent that the problems of using HPC are generic and disci-
pline independent. Cross-discipline research initiatives could be a useful way forward until critical
masses of users within specific disciplines appear. In a geographical context, the combination of
large amounts of data due to GIS, the availability of new AI and CI techniques and other types of
computer-intensive analysis and modelling technology and the increasing accessibility to HPC look
set to create a new style of computational geography that in the longer term will revolutionise many
aspects of the subject by creating new ways of doing nearly all kinds of geography. However, if this
is to happen, then we need to attract computationally experienced researchers from outside. GC has
a most critical and focal role to play in this process.
The essential challenge is to use HPC to extend and expand our abilities to model and anal-
yse all types of geographical systems and not merely those which are already quantitative and
computerised. It would be a dreadful waste if all they were used for was to make old legacy tools
run faster resulting in a kind of HPC-based revival of old-fashioned quantitative geography. The
opportunities are far broader than any backward looking view would suggest. In some areas,
almost instant benefits can be gained, for example, by switching to computationally intensive sta-
tistical methods to reduce reliance on untenable assumptions or to discover new information about
the behaviour of models. In other areas, whole new GC applications will emerge. In general, it is
likely that those with access to the biggest and fastest parallel hardware may well be best placed
to develop leadership in this new form of internationally competitive computational-based geo-
graphical science. As HPC continues to develop, it is likely that many subjects, not just geography,
will have to undergo a major change in how they operate as HPC is more widely recognised as a
paradigm in its own right.
In a world full of unimaginable data riches, maybe (just maybe) we can compute our way out of a
massive data swamp fenced in by scores of traditional restrictions and discover how best to do more
useful things with it. It is increasingly recognised that data are the raw materials of the informa-
tion age. They are extremely relevant to commerce and the functioning of society. New scientific
discoveries, new knowledge, new ideas and new insights into the behaviour of complex physical and
human systems will increasingly have to be created by a new kind of knowledge industry, some-
thing equivalent to a new knowledge manufacturing process. Maybe GC could become the geogra-
pher's own version of this knowledge processing industry of the new millennium. It will not be easy
because many of the systems being studied are non-linear, chaotic, noisy and extremely complex in
