In many instances the process of spatial analysis follows a number of well-defined (often iterative) stages: problem formulation; planning; data gathering; exploratory analysis; hypothesis formulation; modeling and testing; consultation and review; and ultimately final reporting and/or implementation of the findings. GIS and related software tools that perform analytical functions only address the middle sections of this process, and even then may be relatively “blunt” instruments.
Having first identified and formulated the problem to be addressed (often a significant task in itself), and developed an outline plan, the first task typically involves obtaining the data which are to be the subject of analysis. This immediately raises many questions that have an important bearing on subsequent stages: what assumptions have been invoked in order to represent the “real-world” and what are the implications of this for subsequent analysis? how complete are the data — spatially and temporally? how accurate are the data (spatially, temporally and in terms of measured attributes)? are all of the datasets compatible and consistent with one another — what sources are they drawn from and how do they compare in terms of scale, projection, accuracy, modeling, orientation, date of capture and attribute definition? are the data adequate to address the problem at hand? Can the available data be combined and processed with the resources available?
The second stage, once the data have been obtained and accepted as fit for purpose (and/or as the best available), is often exploratory. This may involve: simple mapping of the data, points, lines, regions, grids, surfaces, computation of rates, indices, densities, slopes, directional trends, level sets, classifications etc.; or more complex and dynamic exploration of the data, such as brushing and linking (see further, Section 5.2.1). One or more analytical techniques and tools may be utilized at this and subsequent stages.
The third stage will depend upon the objective of the analysis. In many instances presentation of the results of exploratory analysis in the form of commentary, maps, descriptive statistics and associated documents will complete the process. In others it will involve the development and testing of hypotheses about the patterns observed, and/or modeling of the data in order to undertake some predictive or optimization exercise. Frequently the result of this process is a series of possible outcomes (scenarios) which then need to be summarized and presented for final analysis and decision-making by stakeholders, interest groups, policy makers or entrepreneurs. The process may then iterate until an agreed or stable and robust flow is achieved, from problem specification to data selection, and thence to analysis and outcomes.
This kind of process can be formalized and may be implemented as a standard procedure in operational systems or as part of a planning process. Such procedures may involve relatively lengthy decision cycles (e.g. identifying the next location for a new warehouse as customer demand grows) or highly dynamic environments, for example controlling road traffic lights and routing to reflect the type and density of traffic in real time.
To a substantial degree spatial analysis can be seen as part of a decision support infrastructure, whether such decision-making reflects purely academic interest or commercial, governmental or community interests (see further Batty and Densham, 1996; Miller and Shaw, 2001, Chapter 8). Increasingly, to reflect this trend, GIS vendors have developed tools to facilitate such processes. This has been through the provision of Internet-enabled input, output and processing facilities, and through improved usability, speed, interactivity, spatial decision support systems (SDSS) and visualization techniques. Increasingly these have involved the use of formal decision-support systems, from simple cost-benefit analysis and Multi-Criteria Evaluation (MCE) techniques to the more sophisticated and formalized multi-criteria process such as Saaty’s Analytical Hierarchical Process (AHP) and his more recent Analytical Network Process (ANP) models ― see Saaty (1980, 1999), commercialized as the Decision Lens and the Superdecisions suites of software.
With these improvements in data and toolsets the range and experience of users has grown and will continue to develop. For this reason it is essential that the underlying principles are well-understood and not misused, accidentally or deliberately, leading to ill-informed or invalid decision-making. All providers and users of such systems have a responsibility to ensure that information generated is communicated in an unambiguous and unbiased manner, and to ensure that limitations and assumptions made are transparent and well-understood. These concerns are of particular importance and relevance to those in the print, internet and broadcast media, especially television, where visualization is so important and timeslots for presentation so brief.