Objective

This research priority calls attention to the multidisciplinary issues related to spatial as well as spatio-temporal scale. The objective is to clarify and assess what scale and scale effects are, and through recognizing the fundamental existence of scale effects, determine how one can realistically approach and mitigate the scale effects. Primary issues therefore center on gaining a better understanding of how scale affects human perception; how to effectively and efficiently measure and characterize scale; how to use scale information in judging the fitness of data for a particular use; how to automate scale change and simultaneously represent data at multiple scales; how scale and change in scale affect information content, analysis, and conclusions about patterns and processes.

Background

Scale is not a new issue, nor is concern restricted to geographic information scientists. Scale variations have long been known to constrain the detail with which information can be observed, represented, analyzed, and communicated. Changing the scale of data without first understanding the effects of such action can result in the representation of processes or patterns that are different from those intended. For example, research has shown that reducing the resolution of a raster land cover map (going to larger cells) can increase the dominance of the contiguous classes, but decrease the amount of small and scattered classes (like wetlands in some locations) in the representation (Turner et al. 1989). The spatial scaling problem presents one of the major impediments, both conceptually and methodologically, to advancing all sciences that use geographic information. Likewise, temporal scaling, a separate but related issue, is not well understood and thus difficult to formalize. In an information era, a massive amount of geographic data are collected from various sources, often at different scales. Before these data can be integrated for problem solving, fundamental issues must be addressed.

Recent work on the scaling behavior of various phenomena and processes (including research in global change) has shown that many processes do not scale linearly. The implication is that in order to characterize a pattern or process at a scale other than the scale of observation, some knowledge of how that pattern or process changes with scale is needed so that the scaling process can be adjusted accordingly. Attempts to describe scaling behavior by fractals or self-affine models, which mathematically relate complexity and scale, have proven ineffective because the properties of many geographic phenomena are not strictly repeated across multiple spatial or temporal scales. Multifractals have shown some promise for characterizing the scaling behavior of some phenomena, but it is more likely that fractals will offer only a partial model. Alternative models are needed to understand the impacts that changes in scale have on the information content of databases. Examining the sensitivity of process and analytical models to scale will help scientists validate hypotheses, which in turn will improve geographic theory-building.

Despite a longstanding recognition of the implications of scale on geographic inference and decision making, many questions remain unanswered. The transition from analog (i.e., maps) to digital representations of geographic information forces users of those data to formally deal with these conceptual, technical, and analytical questions in new ways. It is easy to demonstrate by isolated example that scale poses constraints and limitations on geographic information, spatial analysis, and models of the real world. The challenge is to articulate the conditions under which scale-imposed constraints are systematic and to develop geographic models that compensate or standardize scale-based variation. Mishandling or misunderstanding scale can bias inference and reasoning and ultimately affect decision-making processes. New types of analyses, for example the Geographical Analysis Machine (GAM) proposed by Openshaw et al. (1987), may offer methods that are less sensitive to scale than traditional quantitative techniques.

The widespread adoption of geographic information systems (GISs) contributes to the scale problem, but it may offer solutions as well. GISs facilitate data integration regardless of scale differences. This is a problem whenever we try to use coarse aggregate data (like statewide or countywide data), and especially when we try to compare those data with less coarse or disaggregate data (like information by census tract or by individual survey). The capability to process and present geographic information "up" and "down" local-, regional-, and global-scale ranges has been advocated as a solution to understanding the global systems of both natural (e.g., global climate change) and societal (e.g., global economy) processes and the relationships between the two. Fundamental scale questions will benefit from coordinated research efforts among geographic information scientists with various interests and domain experts. Information systems of the future can sensitize users to the implications of scale dependence and provide scale management tools once we develop alternative models of scale behavior, an improved qualitative understanding of the effects of scale, novel methods for describing the scale of data, and intelligent automation methods for changing scale.

Ongoing research

Issues of scale affect nearly every GIS application and involve questions of scale cognition, the scale or range of scales at which phenomena can be easily recognized, optimal digital representations, technology and methodology of data observation, generalization, and information communication. These are very different types of questions. Effective research in the area of scale will require interdisciplinary efforts of geographers, spatial and/or geostatisticians, cartographers, remote sensing specialists, domain experts, cognitive scientists, and computer scientists. Research on scale is under way in geography (Hudson 1992), remote sensing (Quattrochi and Goodchild 1997), cartography (Buttenfield and McMaster 1991), spatial statistics (Wong and Amrhein 1996), hydrology (Sivapalan and Kalma 1995), and ecology (Ehleringer and Field 1993) among other areas. Scale research in many institutes, agencies, and in the private sector began in an ad hoc fashion. Motivated both by practical needs as well as theoretical development, recent attention is focused on formalizing the study of scale, on developing theory, and on exploring robust methods for information representation, analysis, and communication across multiple scales.

Research Priorities

It has become clear that global and regional processes have implications for local places and that individual and local decisions collectively have global and regional implications. Therefore, scientific information about global and regional patterns and processes must be understood on a local level and vice versa. As the policy-making and scientific communities come to terms with these relationships, systematic understanding about spatial and temporal variations in scale gain importance. Geographical information plays an ever larger role as we move to an increasingly automated information economy. Our understanding of scale and the management of data at various scales must keep pace. Ultimately data and information must inform and must produce better decisions.

• Definitions of scale concepts. There has been much confusion and misuse of the term "scale." Lam and Quattrochi (1992) set a basis for clarifying the term. More thorough work will help clarify all connotations of scale in a multidisciplinary context. These connotations are inherently related. A good understanding of the conceptual relationship between these connotations should benefit the geographic information users in conjunction with other standardization efforts in geographic information use.

   

• Systematized bases for scale-related decision making. Basic research on the effects of scale on information content and loss will yield practical information for the many users of geographical information. We need to ask, for example, if we lose a significant portion of our explanatory power when we represent global population trends by country as compared with representations at the level of primary divisions within countries. In an increasing array of management and policy settings, decisions about the appropriate scales of analysis are made every day. Identification of critical scales and of scale-invariant data sets or modeling procedures can make those decisions explicit and better informed.

   

• Practical guidance on data integration and use. Often, data at suboptimal or disparate scales are the best available. Using imperfect data that are available is in many cases preferable to using no data at all, but there are implications for the validity of results. For the user community, creation of knowledge about scale provides principles to improve a data set's fitness for use and guidelines by which to discount model results when necessary. For example, we already know that analysis using aggregate data cannot be used to impute the behavior of individuals (called the ecological fallacy). A better understanding of the problem of conflation (the procedure of merging the positions of corresponding features in different data layers) will enable the fusing of data sets, produced at different scales or produced at the same scale, from different sources. Conflation currently poses a serious impediment in the map overlay process, a critical component of GIS. A better understanding of conflation is also necessary for the integration of data sets produced by different agencies. 

   

• New methods for quantifying and compensating for the effects of scale in statistical and process models. Scientists and land managers apply a variety of analytical tools to answer geographic questions. Many existing methods do not allow the user to adjust or compensate for the effects of the data scale on the analysis results. Methods that are sensitive to scale will allow the inclusion of scale correctives, much like the correctives that compensate for inflation of the significance of statistical relationship in the presence of spatial autocorrelation.

• Intelligent automated generalization methods. GIS tools can be expanded to provide users with methods for intelligently changing the scale of their data. Basic research is still needed to understand how scale changes are perceived, and this in turn can inform interface design. Intelligent generalization will permit the encoding of raw observations in digital form to derive more responsive, application-specific archived representations.

• Improved understanding of cognitive issues of scale. Many scale questions involve human cognition (i.e., how humans perceive the world and information representing it). This issue is explicit especially during human-computer interaction and must be dealt with technically during interface development. It ultimately affects a chain of decisions. Basic research lays a foundation for answering many conceptual and technical questions about the proper use of spatial and temporal scales in geographic information processing.

 Potential projects:

• Develop more consistent definitions of scale and its associated attributes that can be "uniformly" understood or perceived across multiple disciplines. The challenge here is how to make the "standardized" lexicon be adopted and utilized multidisciplinarily.

• Encourage the development of a scale analysis module to be included in major environmental analytical and measurement methods, so that sensitivity analysis, robustness testing, and the effects of scale on the analytical findings can be assessed.

• It is important to point out that many existing scale studies rely heavily on the re-sampling methods to generate multi-scale data for analysis, and as such, the findings on the scale effects are not really due to the scale effects, but rather they are an artifact attributable to the use of different re-sampling methods (Weigel, 1996). The effects of re-sampling methods on the scale studies are an aspect requiring further research.

• More work need to be done to identify and develop efficient spatial/geostatistical techniques for assessing and characterizing the scale effects. This include refinement of existing techniques, such as fractals, spatial autocorrelation, Shannon index, geographical variance, and local variance, as well as exploration of new techniques such as wavelet analysis, local fractals, and multifractals (Hou, 1998). The indices derived from these techniques, if sufficiently discriminatory and information-rich, could be included as part of the metadata.

• Identify the ranges of scale over which an encoded attribute classification scheme is valid. A project of this type will help build a better understanding of the linkages between the scale of a spatial representation and the appropriate corresponding attribute detail. For example, land cover data represented as 30-meter pixels should include more detail (i.e., the number and types of classes, a hierarchical classification scheme to aid the aggregation of classes) in the land cover classification than similar data represented as 1-kilometer pixels. Studies of these relationships in specific settings and for specific data sets (e.g., land cover classification) will be performed with the aim of building theoretical bases for understanding these relationships more generally.

   

• Identify the optimal scales of analysis for common data sets, applications, and needs, and critical scales at which the content or structure of phenomena change suddenly (e.g., in vegetation or population). This work will involve the application of sensitivity analyses and spatial statistical tools for describing those sensitivities. Ultimately the goal is to allow users of geographic data to determine the appropriate scales prior to data collection or analysis.

• Develop methods for cost-benefit analysis comparing the use of pregeneralized data (e.g., soil polygons) versus automated generalization of raw data (e.g., soil data from collection points). "Cost" could refer to data collection costs and/or to computational cycles; "benefits" can similarly serve as a metaphor for data validity; each will depend on the needs of the particular application. As automated generalization methods become readily available (many are now in commercial GIS packages), users may wish to access data in the rawest form possible as opposed to using data generalized for a specific purpose. Research of this type would have implications for data collection and archival agencies as well as users.

• Develop improved methods for data generalization. A well researched, yet still unsolved, problem associated with spatial generalization is the creation of multiple versions of databases. Research in cartographic generalization has taken several directions including algorithmic design and testing, the design of models and conceptual frameworks, the application of expert systems, and the modeling of cartographic features. Thus far, most of the work in generalization has focused on what is termed "cartographic" generalization, which involves the graphical considerations associated with scale change. A second, less researched, area is in "model" generalization, by which generalization operators (simplification, smoothing, aggregation, agglomeration, and others) are applied to an original digital landscape model (DLM) in order to create secondary representations of the database, called digital cartographic models (DCM). These terms, DLM and DCM, are taken from the European cartographic literature, where a significant amount of work has been completed in this area.

• Develop improved methods for incorporating knowledge about scale into metadata. We need to model descriptive data about data sets (i.e., metadata) to assess the consequences of downloading data at a finer resolution than is needed for a particular GI S application. The question is to determine if the choice of appropriate scale can be made on the basis of metadata alone. With the expansion of global computer networks and the use of those networks for geographic data transmission, efficient modes of communicating data content are needed. We need to determine which modes are most appropriate for representing scale and to evaluate their relative effectiveness. The provision of information about scale aids in the education of users on how to assess appropriate scales for a given application. Scale is clearly a fundamental component of any metadata report.

• Design and develop a multi-scale database. We still lack the database management tools and associated functions needed to effectively and efficiently store multi-scale data, perform multi-scale analysis, and intelligently change scale. This project should be aimed at the development of a software product that would be useful in GIS applications.

 References

 Buttenfield, B. P., and R. B. McMaster (editors), 1991. Map Generalization: Making Rules for Knowledge Representation. New York: Longmont Scientific and Technical.

 Ehleringer, J. R., and C. B. Field (editors), 1993. Scaling Physiological Processes, Leaf to Globe. New York: Academic Press, Inc.

 Hou, R.-R., 1998. A Local-Level Approach in Detecting Scale Effects on Landscape Indices. M. S. Thesis, Louisiana State University.

 Hudson, J., 1992. Scale in space and time. In R. F. Abler, M. G. Markus, and J. M. Olson (editors), Geography's Inner Worlds: Pervasive Themes in Contemporary American Geography. New Brunswick, NJ: Rutgers University Press, pp. 280-300.

 Lam, N., and D. A. Quattrochi, 1992. On the issues of scale, resolution, and fractal analysis in the mapping sciences. The Professional Geographer 44:88-98.

 Openshaw, S., M. Charlatan, C. Wymer, and A. Craft, 1987. A Mark 1 Geographic Analysis Machine for the automated analysis of point data sets. International Journal of Geographical Information Systems, 1(4):335-358.

 Quattrochi, D. A., and M. F. Goodchild (editors), 1997. Scaling in Remote Sensing and GIS. Boca Raton, FL: CRC/Lewis Publishers, Inc.

 Sivapalan, M., and J. D. Kalma, 1995. Scale problems in hydrology: Contributions of the Robertson Workshop. Hydrological Processes 9(3/4):243-250.

 Turner, M. G., R. V. O'Neill, R. H. Gardner, and B. T. Milne, 1989. Effects of changing spatial scale on the analysis of landscape pattern. Landscape Ecology 3:153-162.

 Weigel, S. J., 1996. Scale, Resolution and Resampling: Representation and Analysis of Remotely Sensed Landscapes Across Scale in Geographic Information Systems. Ph.D. Dissertation, Louisiana State University.

 Wong, D., and C. Amrhein (editors), 1996. The Modifiable Areal Unit Problem. Special issue of Geographical Systems 3:2-3.

 Back to Research Priorities Revised White Papers