Curtin (2013) calls on the Geographic Information Science (GISc) community to seize the opportunities surrounding network analysis in geographic information systems (GIS). If GISc researchers and GIS developers can sufficiently integrate networks into existing theoretical frameworks, construct robust methods and design compatible softwares, they could exert a strong geographically-minded influence on the expansion of network analyses in a wide variety of other disciplines.
Networks define fundamental and distinct data structures in GISc, that have not always been served in past GIS implementations. Historically, both non-topological and topological data models in GIS have been inefficient for performing network analyses, with constraining factors leading to repetitions and inconsistencies within the structure. Consequently, data models are required that explicitly treat the description, measurement and analysis of topologically invariant properties of networks (i.e. properties that are not deformed by cartographic transformations), such as connections between transport hubs or links in a social network.
The paper demonstrates that networks are pervasive in their everyday use for navigation of physical and social space. Linear referencing is applied as an underlying location datum, as opposed to a geographic or relative coordinate system, to signify distance along a path. Common metrics for distance between two geographic locations are often calculated by optimally traversing a network.
I think that in order for GIScientists to exert the kind of influence, as envisioned by Curtin, over future GIS network analysis research and its applications, they will need to embrace and address the computational challenges associated with current geographic data models. While they are well-positioned to do so, the ambiguity in ownership implied by the existence of this paper suggests that concurrently evolving fields should not be discounted.
-slumley