Objective

The purpose of this research is to develop a variety of methods, techniques, and approaches for the analysis of spatial and time-space data and models, utilizing the ability of geographic information systems (GISs) to store, select, manipulate, explore , and display georeferenced data.

Problems of human health, social deprivation, global environmental change, industrial and economic development, and a host of other problems demand that we make sense of what is happening in the world around us. The term "spatial analysis" encompasses a wide range of techniques for analyzing, visualizing, simplifying, and theorizing about geographic data. Methods of spatial analysis can be as simple as taking measurements from a map or as sophisticated as the most abstract forms of mathematical statistics. At the same time, we are being flooded by the benefits of new technologies for Earth observation. New remote-sensing satellites are providing unprecedented amounts of data on aspects of the Earth environment, and new sources of demographic, social, and economic data are becoming available at finer spatial detail. Yet our ability to "drink from the firehose," extract meaning, and make useful decisions has not kept pace. We can no longer rely on the human eye and brain alone but must augment their powers through the development of improved techniques for sifting through data to find patterns and outliers, for creating more effective visualizations of data, for testing theories and hypotheses, and for making decisions.

To remain at the cutting edge of GIS technology, analytic and computational methods must be devised that allow for solutions to problems conditioned by GIS data models and the nature of spatial and space-time inquiries. New forms of statistical analysis are needed to assess the relationships between variables in a variety of spatial contexts. New theories must be devised to frame our understanding of relationships between variables at new levels of resolution and dimension. What is the relationship, for example, between moisture and plant growth when our reference is a square kilometer of earth space? How do we assess the clustering of cases of malaria when our environmental data are recorded in little rectangles one meter across?

Spatial data must be treated differently from other types of data. Stronger relationships exist within and among variables that are near to one another. Because the size and configuration of spatial units varies dramatically, we find relationships within and among variables that result from the nature of the spatial units as much as from the nature of the variables under study. Standing in the way of confirmatory spatial data analysis, including modeling, are questions related to spatial scale, spatial association, spatial heterogeneity, boundaries, and incomplete data. Without reasonable responses to these problems, the usefulness of GISs as analytical tools in a sophisticated research environment will come into question. Through the use of GIS s, highly visual methods of spatial analysis that previously were prohibitively expensive and computationally intensive have become accessible at reasonable costs.

The UCGIS Approach

Spatial analysis is the bridge that links fundamental data models to GIS technology, with the result that applications are enhanced and research findings are broadened and deepened. The University Consortium for Geographic Information Science (UCGIS ) emphasizes those research areas that integrate a variety of these activities. The GIS framework includes both the georeferenced data and the tools for data manipulation. The linkages to applications allow spatial analysts to inform applied practitioners of new, more profitable ways to conduct research, and, in like manner, practitioners are able to develop new analytic approaches useful to particular applied fields in the social, physical, and environmental sciences. UCGIS calls on spatial analysts from both the physical and the human sciences to assist in the development of spatial statistics, geostatistics, spatial econometrics, structural and time-space modeling, mathematics, and computational algorithms that can take advantage of the flexibility, capacity, and speed of GISs. Those who are well-schooled in theory, empiricism, data collection, data manipulation, programming, and computer technology will be in the best position to make advances in this field, but practitioners such as geographers, epidemiologists, ecologists, climatologists, regional scientists, landscape architects, and environmentalists can add much to the development of GIS-related research.

Importance to National Research Needs

For the United States to remain on the cutting edge of GIS technology, we must foster the development of appropriate analytical techniques in a variety of rapidly changing fields. By engaging in fundamental research in spatial analysis, we can achieve a better understanding of spatial scale, spatial association, spatial heterogeneity, spatial movement, and bounding effects, and we can develop more appropriate tools for modeling continuous and discrete data. We will improve our handling of very large spatial data sets (e.g., disaggregated census data, remotely-sensed data at a global scale), and we will discover the appropriate GIS tools for pattern recognition, data generalization, edge detection, and fuzzy pattern analysis. In the United States, enormous quantities of data are now available to help solve local and regional problems. We must devote energy to exploit this availability.

Benefits

The research topics outlined in the following section point to the priorities the scientific community must support as we move to the 21st Century. Better techniques of spatial analysis, coupled with GISs, will have applications that span a vast range of societal concerns:

• Disease distribution: The study of the transmission of infectious diseases such as dengue, malaria, and AIDS would benefit from the placement of disease incidence in a spatial-temporal ecological framework made possible by a GIS.

• Traffic management and land use planning: Real-time traffic analysis in a GIS framework will aid in the development of highway infrastructures, traffic and travel demand management, and land-use planning.

• Environmental problems: GISs can assist in the analysis of data extracted from models of water, air, and other types of environmental variables. Problems of fire control, species diversity, hydrology and flood control, hazard mitigation, and park usage are ideally suited for analysis with a GIS framework.

 
• Landscape characterization and measurement: A compelling problem of those using remotely sensed data for analyzing such things as land cover and land use is the classification of high-resolution data. If image analysis is brought into a GIS analytical framework, various classification schemes can be tested and used in the analysis of land cover data.

• Social, cultural, and economic analyses: Economists and other social scientists will have the opportunity to use block and county data to test theories by means of spatial econometric analyses. The development of the use of these data sets in a GIS framework will increase our understanding of all sorts of social processes, including patterns of employment and unemployment, crime, economic growth, and population change.

• Physical processes: The analysis of hydrologic and climatologic processes under varying physiographic regimes in a GIS framework will enable researchers to pinpoint trends (e.g., global change), identify anomalous events, and further applied research in these fields.

• Improving the accessibility and equity of opportunities and services: GIS can accommodate more sensitive configuration of economic activities and public sector services. GIS capabilities for handling spatial data allow researchers to develop detailed representations and analyses of the spatial distribution of disadvantaged populations and their access to opportunities and services. GIS-based techniques for solving sophisticated and realistic location and distribution problems can allow these systems to be configured to maximize accessibility and equity.

Future technologies must be not only spatial but also spatial-temporal. They must address certain key questions: How do we handle large spatial data sets (e.g., disaggregated census data or remotely-sensed data at a global scale)? What techniques can account for the ways that spatial data influence the type of analysis employed (e.g., scale and aggregation effects)? What generic GIS tools [e.g., Openshaw's (1994) pattern-spotters and testers, data simplifiers, edge detectors, and fuzzy pattern analyzers] are appropriate for spatial analysis? Long term (5 to 10 years)

• Develop methods for handling massive spatial data sets: As georeferenced data sets become larger, we must develop methods to use, pare, classify, and manipulate the rich information inherent in very large spatial data bases.

• Analyze spatial and space-time data: We must extend exploratory methods of analyzing spatial data to include space-time data so that we can develop models that better represent reality.

• Develop confirmatory (significance) procedures: We must develop statistical (Bonferroni-type) procedures that recognize the dependence of georeferenced data and allow researchers to engage in testing hypotheses.

• Extend use of the variogram and kriging: These useful geostatistical procedures are just now being incorporated into GIS frameworks. Research must continue in this area to take full advantage of GISs as research tools, including the use of these procedures in broader analyses and in models of continuous spatial phenomena.

• Analyze the impact of scale and the development of scale-independent procedures: Perhaps most fundamental to preparing GISs for spatial analytical work are the modules and algorithms that evaluate the effects that changes in scale have on research results. The development of scale correctives and scale independent methods is a compelling research need. Researchers have great interest in the degree to which interzonal analysis is affected by the configuration and regionalization of spatial data units.

• Study global versus local effects: Global analysis of massive data sets has proved superficial and inadequate. We need procedures and tests to use window, kernel, individual, and other local measures to find and equate the characteristics of nonstationary spatial data.

• Develop procedures to identify essential or extreme observations: We must develop procedures that can identify in a spatial setting key observations, groups of observations (clusters), or hot spots that draw attention to the anomalous regions.

• Develop computationally intensive procedures: We need to interface GISs with tools that take advantage of great increases in the capabilities of computational platforms. Computationally intensive tools can allow more effective use of large data sets, more sophisticated and extensive simulations of complex spatial phenomena, and the solution of complex location and distribution problems. Potential techniques include neural nets, fuzzy sets, wavelets, microsimulation, artificial intelligence, natural language processing of textual information, artificial life, real-time data analysis, numeric optimization techniques and massively parallel algorithms. These new techniques should be evaluated in a GIS environment to determine their usefulness in practical applications.

• Use econometric modeling in a GIS environment: Spatial econometrics is a new and burgeoning field. It is important to attempt to link the sophisticated procedures of the econometrician with the functionality and flexibility of GISs. In addition, we need to find appropriate estimators and testing devices for heterogeneous, nonuniform georeferenced data.

• Develop spatial interaction models in a GIS framework: Perhaps one of the most used model types among spatial analysts is the spatial interaction model. Developing these models in a GIS environment will provide marketing, transportation, and human interaction specialists with greater analytical power than is currently available. Techniques for visualizing spatial interaction flows in a sophisticated manner are also needed.

• Accommodate operations research: The field of operations research would benefit greatly from the functionality and power of GIS to manipulate data, especially in solving problems such as routing, location allocation, coverage, and other optimization. Most existing techniques require simplistic representations of spatial objects and relationships for the sake of tractability. We need GIS-based solutions that can manipulate spatial entities and recognize the complex spatial relationships that occur between these entities.

Anselin, L., 1996. SpaceStat, Version 1.80. Morgantown, WV: Regional Research Institute, West Virginia University.

Anselin, L., and A. Getis, 1992. Spatial statistical analysis and geographic information systems. Annals of Regional Science 26:1933.

Cressie, N., 1991. Statistics for Spatial Data. New York: Wiley.

Fischer, M. F., and P. Nijkamp, editors, 1993. Geographic Information Systems, Spatial Modeling and Policy Evaluation. Berlin: Springer Verlag.

Fotheringham, S., and P. Rogerson, editors, 1994. Spatial Analysis and GIS. London: Taylor and Francis.

Kaluzny, S. P., S. C. Vega, T. P. Cardoso, and A. A. Shelly, 1996. S+SpatialStats: User's Manual, Version 1.0. Seattle: MathSoft, Inc.

Miller, H. J., 1996. GIS and geometric representation in facility location problems. International Journal of Geographic Information Systems 10(7):791-816.

Openshaw, S., 1994. Two exploratory space-time attribute pattern analyzers relevant to GIS. In S. Fotheringham and P. Rogerson, editors, Spatial Analysis and GIS. London: Taylor and Francis.

Openshaw, S., 1996. Parallel simulated annealing and genetic algorithms for reengineering zoning systems. Geographical Systems 3:201-220.