An important function of GIS when coupled with environmental modeling is postprocessing and visualization. The postprocessing has already been
discussed in previous chapters. Here we will give some consideration to the visualization and communication of spatial information within the general context of this book. A good piece of modeling, with uncertainty reduced to a minimum, will fall flat if the results cannot be adequately communicated. This comes down to good cartographic skills with quantitative and qualitative data. It must also be recognized that not everybody can instantly recognize and feel comfortable with plan representations of multidimensional phenomena (Keates, 1982) and, therefore, the means of communication and its effectiveness need to be carefully considered. It is often a case of having to supplement maps with other representations, such as tables, graphs, and images. The Web is providing new tools in this regard.
The principle of communication through maps can be summed up as: “How do I say what to whom, and is it effective?” (Kraak, 2001). This concept
that underpins cartographic design which aims to create maps that can be understood effectively by users. The “what” aspect concerns the information content to be presented on a map—the data model. Such content needs to be analyzed and determined in relation to the objective of presenting such content. The aspect of “how” is the means by which the information is represented cartographically. In achieving the creation of a good map, the basic concept of the cartographic theory developed by Bertin (1967) can be regarded as key guidelines (see below). Another aspect is how a user reads the map, which leads to “whom”: the user. Users should always be considered in the process of creating maps with either an emphasis toward a particular group or more toward individual users, their expected background, likely level of understanding of the concepts being mapped, etc.
The “effective” aspect demonstrates the amount of information extracted and understood by users. The information intended to be communicated
through a map and the information retrieved by a user will rarely achieve an exact match, which can be viewed as the different levels of effectiveness. There can be information loss or, as in Chapter 8, Figure 8.6, levels of uncertainty in use can arise through confusion, ambiguity, or misrepresentation of the information to be imparted. In deciding an appropriate cartographic design, it is important then to analyze the characteristics of the information that is to be visualized and understood. Bertin (1967) has distinguished six visual variables that are important in thematic map construction and relate mostly to the symbology used to represent features. The variables are size (e.g., thickness of lines, size of points), shape (e.g., circles versus triangles, dashed lines), orientation (say, of labeling), color (scheme, range), value (e.g., the degree of contrast over the color range), and texture (e.g., smooth versus coarse patterning). These all need careful consideration and even some experimentation and testing on colleagues for legibility.
Tufte (1983) has further laid down some principles of graphical excellence, which can be well applied to thematic mapping. The cornerstone of graphical excellence is interesting data well presented; in other words, there must be something worth communicating. A well-designed presenta tion becomes a matter of substance, statistics, and design; it consists of complex ideas communicated with clarity, precision, and efficiency. What this means is that the graphics (map or other visualization) should give the viewer “the greatest number of ideas, in the shortest time, with the least ink in the smallest space” (Tufte, 1983). Above all, it requires telling the truth about the data. In thematic maps of quantitative data, the main challenge in telling that truth often boils down to the number of class intervals, the identification of class boundaries, and what symbology to use for each class.
Most GIS packages provide default approaches to help the user through this task, usually giving a choice of equal interval, equal area, natural breaks, standard deviations and quantiles, and inevitably in shades of red. One problem is the overuse of a particular default by individuals or within an organization without much thought given to the nature of the data or the nature of the message, results in throw-away graphics. The issues of good thematic map design are still being debated in the search for good representations that provide more objectively comparable maps (e.g., Brewer and Pickle, 2002). One key here is the normalization of data prior to map creation.
Unfortunately, unless the data are normally distributed (which very often they are not), this can lead to bias. Recently an alternative to the z-score, robust normalization, has been introduced (Brimicombe 1999b, 2000b). The process of robust normalization can be visualized in Figure 10.4 where there are initially two sets of data with contrasting distributions. The first stage of normalization is the centering of the boxplots through subtraction of the median. Sibley (1987) suggests division by the interquartile range to give standardized boxplots, but this gives values that can be difficult to interpret in terms of core and extreme values. Instead, robust normalization uses an asymmetric division in the form: This ensures that all data are transformed to have a median of zero and an interquartile range of [-1, 1]. Extreme values can be taken as values of RN < –3 or RN > 3 (equivalent to 2 standard deviations of a normal distribution). This also leads to an intuitive set of class intervals, which make maps more objectively comparable. Figure 10.5 gives a robust normalized mapping of the pipe burst case study. The eye can quickly pick out the one location where there is both a high extreme occurrence of cases as represented by density and a high extreme occurrence of risk. That would be the first area to prioritize for remedial action.
MacEachren and Kraak (1997) have identified different purposes for map visualizations depending on the intended audience, the degree of interaction with the viewer, and the degree of unknowns still to be resolved. Thus, maps can be used to explore data and issues, analyze them, provide synthesis, and for presentation. Further advances in data exploration useful to GIS and environmental modeling are interaction with the graphics through the use of “brushing” techniques with linked, alternative views of the data (Dykes, 1997). We are also now seeing the incorporation of images, panoramas, and video alongside maps in the exploration of real world phenomena (Dykes, 2000). Through the Web, we have seen the growth of clickable maps in a multimedia setting that facilitate pathways to data and information. This underscores the next issue we need to consider.