One problem that I have encountered recently seems to arise out of some current operational definitions of spatial analysis. Some forms of what I would personally identify as spatial analysis are playing a major role in GIS applications today (e.g., transportation optimization models). Strangely enough, recent conversations I have had with some prominent geographers who identify themselves as “spatial analysts” reveal that they consider these activities to be “modeling” and not really spatial analysis. I conclude from these informal conversations that “real” spatial analysis is somehow viewed as deeply rooted in spatial statistics and that non statistical, analytic methods are considered to be “outside” of spatial analysis. My personal view is that both spatial statistics and analytic spatial models (both descriptive and optimizing forms) are simply complementary aspects of what should be a more comprehensive working, and generally accepted, definition of “spatial analysis.”
If we accept this broader definition of spatial analysis, then the question
becomes why are some aspects of it widely accepted in the GIS area while
others are not? The answer does not lie in a lack of participation by geographers
or a lack of knowledge of spatial analysis by GIS developers as Longley
and Batty (and others) have suggested.2 Rather, it may be found
in the understandable commercial orientation of the companies developing
GIS technology. The incorporation of spatial optimization models into the
GIS has resulted in massive cost savings by users of the technology that
has, in turn, led to demand for more sophisticated and easier to use tools
in this area. It has yet to be clearly shown that the incorporation of
other forms of spatial analysis would generate a similar level of utility
to institutions regularly dealing with complex, real–world spatial problems.
It is my belief that other forms of spatial analysis could make similar
contributions if (a) an effective attempt were made to establish their
clear utility within the context of large–scale, real–world problems, and
(b) their present, traditional view of spatial problems is replaced with
a broader and more realistic one. Let me comment on each of these in turn.
My first comment relates to the creation of a demand among the rapidly
increasing GIS community for the results that can be provided by the tools
in question (e.g., spatial statistics). The problem here is not unlike
the map projection problem that has dismayed the GIS community for years.
Basically the user’s question has been “Why bother with map projections?
Things work OK if I ignore them.” Even self–styled “GIS consultants” were
telling clients how much money they could save in data conversion of they
just forgot about all that stuff. It has now been generally demonstrated
that such an approach leads to expensive errors. We need to demonstrate
that the other components of spatial analysis (as contrasted to the highly
successful spatial optimization models) can make a real difference. It
will not be easy, and is unlikely to lead to many academic brownie points,
but it can be done.
The second point is one that I have brought up on several previous occasions. Because of the difficulties involved in both conceptual structures and computation, we have elected to deal with representations of the world around us that are far too limited. For example, many of our spatial views are one–dimensional in nature where the “space equals distance” assumption is so ingrained that it is never even mentioned. What about the other dimension of the two–dimensional space that we contend we deal with? Direction is generally ignored in geography but if we look around, we see that directional statistics play an important role in other disciplines such as ecology, oceanography, and geology.3 We have also generally neglected the role of time in our work in spatial analysis. In reality, what are called for are powerful spatial models and tools that work in a multi–dimensional space–time framework. It is my contention that melding existing spatial analysis approaches with GIS represents the single most important action that will permit significant developments to take place in this direction.
Operationally, I believe that the spatial analysis community urgently needs to follow the lead of the GIS community into an object–oriented view of how they carry out their work. Four years ago a colleague of mine, Dr. Randy Jackson, presented a detailed and cogent argument for the utility of object–orientation in Regional Science. He made this argument prior to the recent, major move of GIS developers into object–oriented development tools (e.g., ESRI’s Map Objects and Net Engine). This move has created a major change in GIS technology and it is clearly time for the spatial analysis community to think very seriously about Jackson’s seminal proposals.4
In conclusion, I feel that both spatial analysis and geographic information systems are on the brink of a major revolution that will lead to substantial increases in the scope and power of both areas. This will not be successful unless the two areas become much more highly integrated both conceptually and operationally. I would hope that the forthcoming meeting would lead to such an integration.
1. A brief position paper prepared in conjunction with the forthcoming 1998 Varenius Workshop on Status and Trends in Spatial Analysis.
2 Paul Longley and Michael Batty, 1996. Spatial Analysis: Modelling in a GIS Environment. Cambridge: GeoInformation International.
3 Marida, K.V., 1972. Statistics of Directional Data. New York: Academic Press is a classic reference and Gaile, Gary C., 1980, Directional Statistics. Norwich: GeoAbstracts summarized this area for geographers. Little interest has developed, perhaps due to a lack of relevant theory in human geography.
4 Jackson, Randall W., 1994. “Object–oriented modeling in Regional Science: an advocacy view,” Annals of Regional Science.
Additional study at SSRC Summer Institutes in Mathematics for the Social Sciences, Stanford University, 1955 and 1957
Association of American Geographers, Honors Award (1993) for: seminal work in quantitative techniques, transportation geography, computer modeling and simulation, and for pioneering research in geographic information systems.
Received Certificate of Honor (1990) from the Academy of Sciences of the People's Republic of China for: significant contributions to the development of geographic information systems in China
The teaching software system, OSU Map-for-the-PC, received the best academic software award (1990) from the Association of American Geographers.
Listed in Who’s Who in America, American Men and Women of Science, Who’s Who in Science and Engineering, etc.
“Brushing Spatial Flow Data Sets,” (with Lin Liu) in the Proceedings, 1997 Joint Statistical Meeting.
“Recent advances in the exploratory analysis of interregional flows in space and time,” (with Zaiyong Gou, Lin Liu and James Saunders), in Z. Kemp (ed.), Innovations in Geographic Information Systems. London: Taylor & Francis, 1997. [Keynote address given at the 4th National United Kingdom GIS Research Conference, Canterbury, England, April 1996.]
“Geodemographic analysis and GIS technology in college and university admissions planning,” (with Jim Herries), in the Proceedings of the Sixth International Conference on Business Demographics, 1996.
“Strategies for real–time spatial analysis using massively parallel SIMD computers: an application to urban traffic flow analysis,” (with Demin Xiong), International Journal of Geographical Information Systems, vol. 10, no. 6, pp. 769-789, 1996.
“Technical issues surrounding the integration of GIS with three–dimensional numerical models of spatial processes,” (with Phillip Chu and Kenneth Bedford) in Proceedings, GIS ‘95. Selected for reprinting in Heit, Parker and Shortreid (eds.), 1995. GIS Applications in Natural Resources 2.
“The potential methodological impact of geographic information systems
on the social sciences,” in Allen, Zubrow and Green (eds.), Interpreting
Space: GIS and Archaeology. London: Taylor & Francis,
Ltd., 1990.