Network distance, as the name suggests, is the distance (or cost, time etc.) involved in traversing a network. In most instances the term refers to the lowest value distance measure between pairs or lists of points (generally network nodes or points upstream and/or downstream in hydrological applications). Typically the network is a logical collection of linked polylines, joined at network nodes (or vertices), and possessing a network topology (sometimes called a geometry). Network links and nodes may have additional explicit attributes, such as type of link, weights applied to links and nodes, rules governing behavior of flows (e.g. link directionality, permitted turns, barriers) etc., and intrinsic attributes (e.g. position and segment lengths). When network distances are computed by a GIS the procedure takes into account the various explicit and intrinsic attributes of the network and the rules applied for the problem selected. Thus network distance will vary in meaning according to the problem at hand. A variety of hybrid concepts exist, which may be modeled within some GIS packages, but generally only in a limited manner. These include multi-modal networks (thus seeking routes that may combine road, rail and bus for example), and mixed network/cost distance situations, in which sources or targets may not lie on the network, but for which sections of optimal routes (or network building programme) need to utilize existing network infrastructures.
Most GIS packages have not historically included extensive networking tools as part of their standard offering, but may provide limited functionality in this area (e.g. ArcGIS Utility Network Analyst and Linear Referencing tools). This situation is now changing: Manifold does include a range of network analysis tools via its Business Tools extension; ArcGIS V9 now ships with the new Network Analyst option as part of the main distribution kit; TNTMips now includes a network analysis module with a wide range of facilities; and the open source GIS, GRASS, includes a range of network analysis tools in its core modules for vector processing, with functions such as v.net.path generating shortest paths and outputting associated attribute data such as distances and times in linked attribute tables. A rather different tool, SANET, is specifically designed to carry out certain types of spatial analysis on a network (e.g. statistical analysis), and is provided as an ArcGIS add-in. TransCAD and Cube also differ from most GIS products as they are essentially “GIS for transportation” (GIS-T) offerings. TransCAD is built on a general purpose GIS engine, Maptitude and Cube is built on ArcGIS. Figure 4‑70 illustrates the use of network distance within the TransCAD shortest path facility, utilizing least distance (heavy line shown in blue) and least time (heavy line shown in green) with a single intermediate node. Such algorithms are widely used in web-based GIS providing routing and traffic advice (see for example, Google maps, http://www.mapquest.com). Output from these systems is in a variety of forms, including tabulation of section-by-section network distance/time and cumulative distance and/or time.
Figure 4‑70 Shortest and least time paths
Depending on the GIS package, for large networks with potentially complex routing it is often advisable to compute link lengths as an explicit field or table and work with this data rather than expecting the GIS to re-compute such information dynamically from an intrinsic field. For more details see Chapter 7.