An evaluation of measures for quantifying map information

Abstract

A real-time map must not contain too much information. Therefore, we need measures of map information that could be guidelines for the selection of data layers and the real-time generalisation process. In this paper we evaluate measures of the amount of information and the distribution of information. The evaluation is performed by (1) defining measures, (2) implementing the measures, (3) computing the measures for some test maps, and finally (4) comparing the values of the measures with human judgement of the map information. For amount of information, we found that the measures number of objects, number of points and object line length had better correspondence with human judgement than object area. We also found that measures based on the size of Voronoi regions of objects (respectively points) can be used for identifying the distribution of information. The results are based on the testing of only building objects. Future work should extend the test, using all object types.

Introduction

A major issue in cartography is the usability of the map. For traditional paper maps this has been studied thoroughly. However, new technology has enabled new types of map usage such as interactive real-time applications using the web and mobile devices. These maps can be tailored for a specific purpose and even for a specific user (Reichenbacher, 2004, Gartner, 2004). This large freedom to tailor sets requirements on new analytical measures, or constraints, that describe the usability of the map.

This study is part of the project “The Swedish Planning Portal” (Planeringsportalen). By using this portal a user (employed at companies or governmental/local authorities) should be able to find planning information and especially geographic information related to physical planning. To enable the user to download and view the geographic information, web services–following the Open Geospatial Consortium (OGC) standards–will be set up. As for maps in general, it is important for a web-based map service like a planning portal that the information presented is as usable as possible. This means that is should be easy for the user to read and comprehend the maps.

To improve the usability of a map, cartographic generalisation is applied. In recent years, the generalisation research has tried to model the overall process of generalisation using constraints (Harrie and Weibel, 2007). A constraint can be seen as requirements that should be obtained in the generalisation process. The constraints can be classified into the following types (see Ruas and Plazanet, 1996, Weibel and Dutton, 1998, Harrie, 2003): position, topology, shape, structural, functional and readability (legibility). The five first types concern the representation, i.e. vital aspects of the map should not be lost in the generalisation process. The final type, readability constraints, concerns the ease with which the user can read the map.

There are two major types of readability constraints of a map. The first type concerns the visual perception. The map objects must be readable for a normal user. Robinson (Robinson, 1952, in MacEachren, 1995) suggested that cartographic objects should be designed considering human perception, using, for example, a definition of the smallest noticeable lettering size difference. For screen maps the paper map definitions can be rather coarse (Spiess, 1995), which is why specific definitions are needed.

The other type of readability constraints concerns the amount of map information. Even if the map objects, and features within the objects, are large enough the map reader cannot comprehend the map if it contains too much information (see Bjørke, 1996, Li and Huang, 2002). The amount of map information has an even greater importance in real-time maps than for traditional paper maps, as real-time maps should be read and understood relatively quickly. Therefore we should strive for establishing measures for the amount of information in a real-time map and let these measures act as constraints for selecting data layers and in the real-time generalisation process.

The aim of this study is to evaluate some measures of map readability that eventually should be used as constraints for the selection of data layers and in real-time generalisation. The paper is organised as follows. Section 2 includes a literature review of quantifying map information. In Section 3, we present our work. First, map readability is divided into the amount of information and the distribution of information; then, we propose some analytical measures for each category. These measures are evaluated in a case study. The paper ends with our conclusions.