GIS and Spatial Analysis in Urban and Regional

Research
Derek Bond
Abstract
The early optimism surrounding GIS has now faded. Few GIS implementations have
lived up to expectation. The most successful have been those at the operational level
where their main role has been to replace paper maps with digital ones. However, as the
cost of GIS implementation falls and spatial data becomes more readily available there
is a need to consider the future of GIS and spatial analysis. This is particularly important
in the field of urban and regional research especially where considerable investment in
GIS has already been made. This paper looks at the major issues that face regional and
urban research in the next stage of development of GIS and spatial analysis. The main
conclusion of the paper is that regional and urban statistics offices and researchers
are well placed to benefit from future developments in GIS and spatial analysis.

Derek Bond
School of Business, Retail and Financial Services
University of Ulster at Coleraine
Cromore Rd.
Coleraine , Co. Derry
United Kingdom
BT52 1SA
e-mail: [email protected]
C&R ISSN 1568-167X

Realising the Dream

Modern statistical techniques and theories have, generally, been developed to assist in
the reduction of large amounts of primary and secondary data into seemingly useful and
manageable information. Until recently this has meant that statistics have been primarily
concerned with data reduction. That is, trying to produce summary figures and tests which
describe something about the underlying data generation processes.
With the increasing performance and the falling costs of information and communication
technology (ICT) this reliance on mainly parametric based data reduction techniques
has been challenged. It is becoming increasing cost-effective for end-users to handle the
large-scale datasets themselves. One of the main areas where this trend is most
pronounced is in the handling and analysis of spatially referenced data using
Geographic Information Systems (GIS). However, many of the original dreams and
expectations of what GIS could offer do not seem to have been realised. In particular
their role as an essential tool for policy and decision-makers (see for example
Department of Environment (1987)) do not seem to have been met. Reviews of the
current topics in the GIS literature together with the research findings of studies on the
implementation of systems (for example Hendriks (1998) and Campbell and Masser
(1995)) show that on the whole GIS is being used mainly in support of operational level
tasks. In many cases they do little more than replacing paper maps and records with
digital equivalence. The use of GIS and spatial analysis at the strategic and policy
related levels are currently limited. Various perspectives on this have been put forward
covering aspects such as availability of tools to management issues (for example
Worrall and Bond (1996) argue that the main issues are organisational, Hendriks
(2000) similarly takes a 'learning organisation' perspective.)
Spatial analysis and GIS requires large investment in staff and time to provide adequate
data in formats that can readily be used for detailed spatial data analysis. Many regional
and urban statistics offices and research institutes have committed considerable
resources to this task. It is therefore important that the returns on such investments
are realised. In this paper and the other papers in this volume of Cities and Regions
various issues associated with spatial analysis and GIS in urban and regional
research will be critically examined. The aim is to show that while there might be a
concern with the general direction that GIS and spatial analysis research are moving
those concerned with urban and regional research are attempting to address issues
related to ensuring that the investment in spatial data and related software is realised.
This is done by considering research in the various aspects of spatial analysis and
GIS. These range from the availability of spatial data, the software needed to handle
the data to methods for handling the data.
SPATIAL DATA
Fundamental to the use of GIS and spatial analysis is the availability of spatial data.
This data is normally of two types: topographic and attribute. In the early days of GIS
there was a lack of readily available topographic data and it was a large resource
commitment to produce it. Now digital topographic data is readily available both in
raster and vector formats, though issues of copyright and access still exist. While
topographic data is now readily available few national spatial infrastructures exist

(c.f. Masser (1999) for discussion of the issues surrounding the development of national
spatial infrastructures). The availability of topographic data has reduced the overheads
associated with developing a GIS and has led to an increase in demand for more
spatially dis-aggregated and referenced attribute data to use with such systems. This
demand and a lack of national and international spatial data infrastructures have various
implications for those involved in regional and urban research and analysis. The most
common problems found in the literature are those of attaching consistent spatial
references to attribute data and of the requirement for fine detail small area statistics
which do not compromise on confidentiality. Caught in the middle of these problems
are Urban and Regional Statistics Offices. Most of these offices are both data users and
providers. While involved in some primary data collection, often as a by-product of the
administrative processes of their administrations, they often add value to data
collected by National Statistical Institutes (NSIs) and other bodies. Often these
secondary data sets do not contain spatial references that are consistent with the
topographic data or other attribute data used. The issue of attaching spatial identifiers is
a complex one (for example, see Visvalingam (1991) for discussion) and is linked to that
of statistical confidentiality.
Statistical confidentiality is often quoted as a major issue hindering the development of
GIS. The problem is that in the ideal world a GIS would contain the basic records of any
attribute dataset with consistent spatial references attached. In most countries this is
impossible as statistical disclosure rules do not permit the identification of individuals or
single organisations (see for example Marsh et al. (1992) for discussion on confidentiality
surrounding the 1991 United Kingdom's Census of Population and EuroStat (1992)
for general discussion of various issues surrounding statistical confidentiality). The issues
of statistical confidentiality and spatial tagging are addressed in this volume by
Wendy Treadwell.
While spatial tagging and statistical confidentiality are problems that need to be
addressed for the successful implementation of GIS in an organisation GIS also
provides tools with which to consider these issues. For example, GIS software is often
used to assist in the construction of statistically 'sound' spatial reporting regions.
However, as Coombes (1998) concluded, while discussing the construction of
'Travel To Work Areas' for England and Wales: '... the methods and software which is
needed for defining new sets of areas -..- are still too specialised to be found in many
GIS packages'.
The problems facing urban and regional statistics offices are two faced. They are often
trying to combine data from various data sources to provide high quality insights
into the functioning of their areas whilst being seen as the first port of call by
researchers and other end users looking for detailed secondary data on the area.
Much innovative research has been done by these offices to integrate spatial data
sets and provide consistent data and information to their end users (see, for example
Tammilehto-Luode and Backer (1999) for a description of a Scandinavian exercise
to provide small area statistics using GIS).

SHARING SPATIAL DATA

In an increasingly globalised world most socio-economic activities do not stop at international borders so there is a need to share data across international boundaries.
However, the problems of sharing data are many and range from the issues of
compatibility of formats and comparability of coverage to issues of commercial
exploitation. At the urban and regional level many of these issues are more acute
than at the national level. For example when comparing small area data minor
definitional differences can result in major problems.
In Worrall and Bond (1996) it was argued that the commercial or quasi-commercial
nature of much of the attribute data required for the successful development of GIS is a
major handicap in the success of major GIS implementations. This can be a particular
problem in small urban and regional statistics offices that are mainly concerned with
adding value to secondary data. The secondary data used is often at a resolution and
detail that is of use in the commercial sector for such things as marketing, utilities
management etc. and is main source of non-public revenue for the public body
collecting the data. With the increasing requirement in many countries that their public
sectors move to cost recovery models this has a serious implication for urban and
regional statistics offices.
Despite such problems many regional and urban statistics offices have managed not only
to obtain various spatially referenced data sets but also to share the data with other
offices on an international scale allowing for comparative statistics to be produced.
(see for example City of Helsinki (1999) and City of Nuremberg (1995) for
examples of how city offices have tried to produce comparable statistics).
ACCESSING SPATIAL DATA
Advances in information and communication technologies have led both to an
increased demand for spatially reference data and also to its availability, subject to
the commercial and confidentiality problems discussed above. Whilst being part of
the catalyst for this increased demand for data GIS has become to be seen as part of
the solution to the problem of developing effective access to large amount of
spatially referenced data. These very large spatial datasets are becoming available
through such outlets as digital libraries (Smith (1996)), data archives (Lievesley
(1992)) and data warehouses (Jarke et al. (2000)). Users are faced with the problem of
finding and correctly using relevant data from such sources. This can be an
intimidating task.
In recent years there has been a growth in interest in the use of metadata to help search
such databases. The problem is that there are no agreed standards for metadata
though in the social science community attempts are being made to standardise on the
'Dublin Core' (Lee-Smiltzer (2000)). Various projects, such as NESSTAR/FASTER
(www.faster-data.org) in the European academic community, are attempting to provide
user-friendly interfaces to attribute data using metadata-based interfaces. The problem
facing researchers and other users of the data sets is lack of detailed knowledge of the
information that can be derived from the data contained in these complex databases.

The visual element of GIS has always played a large part in its popular appeal and research
has shown (Smelcer & Carmel (1997)) that, provided the task is relatively simple, a map can
provide an ideal starting point for information discovery. This has led to growth in GIS based
interactive information seeking software. The aim of this software is to replace traditional textbased information retrieval strategies with ones based on visualisation and knowledge
discovery in databases (Fabrikant (2000)). Knowledge Discovery in Databases (KDD) has
been defined as 'the non-trivial process of identifying valid, novel, potentially useful, and
ultimately understandable patterns in data' (Fayyad et al. (1996) p.6). KDD is closely linked
with data mining and uses the computational power of modern IT to explore data sets rather
than relying on data reduction techniques. A simple example of the use of visualisation and
KDD is the Casweb Interface to the United Kingdom's 1991 Census of Population
(http://jimay.mcc.ac.uk/casweb/). In this volume the paper by Hans Voss et al. looks at, amongst
other things, the issues surrounding the implementation of map interfaces to complex data and
the role of KDD.
SPATIAL DATA ANALYSIS
It is often argued (c.f. Csillag and Kabos (1999)) that developments in methods of spatial
analysis have failed to keep up with those in GIS thus hindering its uptake. While there may
be some truth in this most of the discussion is couched in the traditional Fisher-Neyman
paradigm and emphasis is put on the need for parametric estimation and model building rather
than on KDD and Geographic Visualisations (GVis) (c.f. Maceachren et al. (1999)). The
development in the use of these techniques has come from two distinct fields. GVis has been
developed by mathematical geographers whereas KDD has grown from work on non-spatial
data mining in mathematics.
Traditional spatial analysis has also developed and adopted a more Exploratory Data Analysis
(EDA) approach. The paper in this volume by Peter Brown et al. is a good example, using
geodemographic profiling, of the current policy interface between GIS and spatial analysis.
The problem with approaches such as this is the reliance on the results of one data reduction
exercise. GVis and KDD allow the user to more flexibility explore the larger complex
datasets. The paper by Hans Voss et al. in this volume looks at the issues of visualisation and
KDD in spatial datasets and the paper by Maribel Santos and Luis Amarel gives an example
of how spatial reasoning might be applied in KDD.
GIS AND SPATIAL ANALYSIS SOFTWARE
To many end users the talk of GVis and KDD in spatial datasets seem a distant dream as
traditional GIS software, other than simple raster based mapping packages have had the
reputation of being extremely expensive (Pinals (1998)). However, as the papers by Patrick
Gerland et al. and Dang Van Due in this volume show there is now a wealth of low cost and
free software available for end-users. Similarly the paper by Maribel Santos and Luis Amarel
shows how KDD might be done on spatial datasets using standard software and the paper by
Natali Andrienko et al describes ready available visualisation and KDD software.
In the first of these papers Patrick Gerland et al. describe the GIS software developed by the
United Nations for population research which includes software for capturing, storing,
retrieving and analysing spatial data - all of which is readily available on the web
(http://www.undo.org/popin/softproj/index.htm) and are windows based for ease of use. In the
second of these papers Dang Van Duc outlines the web based MapOnline software which has
developed out of the UN's PopMap software. The paper by Maribel Santos and Luis Amarel
gives an example of how the SPSS add in 'Clementine' might be used for spatial KDD.
Clementine is a ready available datelining programme and again has an easy to use window
interface. The last of these papers by Natali Andrienko et al. discusses the use of the Java
based Descartes software (http://allanon.gmd.de/and/java/iris/). This software like MapOnline
is a web based spatial analysis tool allowing for KDD.

These four papers show that:

to use GIS and spatial analysis effectively there is no longer the need to invest heavily in
expensive cartographic based GIS software;
large amounts of data do not have to be stored locally; and
easy to use interfaces for even complex operations such as GVi's and KDD are now
becoming readily available
SUMMARY
This paper began by considering the generally rather pessimistic view of the cost
effectiveness of investment in GIS and spatial analysis. Through examples the paper has
shown that the 'state of health' of GIS and spatial analysis in urban and regional research is
good. Indeed from considering the information contained in the papers in this volume one
should conclude that the, often considerable, investment by city and regional statistical offices
in compiling and maintaining large-scale spatially referenced datasets, places them and
associated researchers in an extremely strong position to benefit from the current
developments in the area.

References
Campbell, H and I. Masser (1995): GIS and Organisations. London, Bristol: P.A. Taylor and
Francis.
City of Helsinki (1999): Comparative City Statistics 1999 City of Helsinki Urban Facts,
Yliopistopaino, Helsinki.
City of Nuremburg (1995) Facts and Figures in Comparison 1994 City of Nuremburg,
International Relation Office, Nuremburg.
Csillag F., and S. Kabos (1999): 'Convergence between GIS and spatial statistics: What?
Where? When?' Bulletin of the International Statistical Institute, Vol. 58, book 1, pp. 249252.
Department of the Environment (1987): Handling Geographic Information: report of the
Committee of Enquiry chaired by Lord Chorley. London.: H.M.S.O.
Eurostat (1992): International Seminar on Statistical Confidentiality: Proceedings,
Eurostat/ISI.
Fabrikant S.I. (2000): 'Spatialized Browsing in Large Data Archives' Transactions in GIS Vol. 4,
pp. 65-78.
Fayyad, U., G. Piatetsky-Shapiro and P. Smyth (1996): 'From Data Mining to Knowledge
Discovery in Databases: An Overview'. In Fayyad, U., G. Piatetsky-Shapiro, P. Smyth and R.
Uthurusamy (eds.) Advances in Knowledge Discovery and data Mining,. Menlo Park, CA.:
AAAI Press/MIT Press.
Garcia-Molina, H., J. Widom, J. Wiener, W. Labio, B. Lent and Y. Zhuge (1995)
Hendrik P.H.J. (1998): 'Information Strategies for Geographic Information Systems'
International Journal of Geographic Information Science, Vol. 12, pp. 621-639.
Hendrik P.H.J. (2000): 'An Organisational Learning perspective on GIS' International
Journal of Geographic Information Science, Vol. 14, pp. 373-396.
Lee-Smeltzer, K.H. (2000): 'Finding the Needle: controlled vocabularies, resource discovery
and Dublin Core' Library Collections Acquisitions and Technical Services, Vol. 24, pp. 205215.
Lievesley, D. (1992): 'The Role of the ESRC's Data Archive in the Dissemination of Data for
Secondary Analysis' Journal of the Market Research Society, Vol. 35, pp. 267-278.
Jarke M, C. Quix, D. Calvanese,M. Lenzerini, E. Franconi, S. Ligoudistianos, P. Vassiliadis,
and Y. Vassiliou (2000) 'Concept based design of data warehouses: The DWQ
demonstrators' Sigmod Record, Vol. 29, pp. 591-591.
Maceachren, A.M., M. Wachowicz, R. Edsall, D. Haug and R. Masters (1999): 'Constructing
knowledge from multivariate Data: integrating geographical visualisation with knowledge
discovery in database methods'
International Journal of Geographic Information Science, Vol. 13, pp. 311-334.
Pinals, D. (1998): 'Choosing A Mapping Software package that suits your company's needs,
Business Geographics July 98.
Smelcer, J. and E. Carmel (1997): 'The effectiveness of different representations for
managerial Problem Solving: Comparing Maps and tables' Decision Science,
Vol. 28, pp. 391-420.
Smith, T. (1996): 'Alexandria digital library Communications of the ACM,
Vol. 38, pp. 61-62.
Worrall, L. and D. Bond (1996): 'Geographical Information Systems, Spatial Analysis and
Public Policy' International Statistical Review, Vol. 65, pp. 365 -379.