Introduction

Different Ways of Thinking about Street

Networks and Spatial Analysis

Bin Jiang1, Atsuyuki Okabe2

1Department of Technology and Built Environment, University of Gävle, Gävle, Sweden, 2School ofCultural and Creative Studies, Aoyama Gakuin University, Tokyo, Japan

Street networks, as one of the oldest transport infrastructures in the world, play a significant rolein modernization, sustainable development, and human daily activities in both ancient andmodern times. Although street networks have been well studied in a variety of engineering andscientific disciplines, including transportation science, geography, urban planning, economics,and even physics, our understanding of street networks in terms of their structure and dynamicsremains limited, especially when dealing with such real-world problems as traffic jams, pollution,and human evacuations for disaster management. Thanks to the rapid development of geographicinformation science and its related technologies, abundant street network data have been col-lected in an attempt to better understanding these networks’ behavior and human activitiesconstrained by these networks. For example, OpenStreetMap has assembled hundreds of thou-sands of gigabytes of data for streets and for other related geographic objects, such as publictransportation, building footprints, and points of interest. Given this context, we predict that, inthe near future, increasing amounts of research will be published regarding the underlyingstructure and dynamics of street networks. This special issue (SI) comprises five of the bestarticles from 19 articles submitted to and presented at the International Cartographic Associationworkshop on street networks and transport (https://sites.google.com/site/icaworkshop2013/).Due to time constraints, we were unable to include a number of other high-quality articles, but weare confident that these other articles soon will be added to the literature elsewhere.

One goal for this SI is to promote different ways of thinking about understanding streetnetworks, and of conducting spatial analysis. Network spatial analysis involves a set of statisticaland computational methods developed by Okabe and Sugihara (2012), as demonstrated in Shiodeand Shiode (2014), for analyzing events occurring along networks. Network spatial analysisclearly differs from conventional spatial analysis, which assumes a continuous Euclidean spacerather than space constrained to networks. Current network analysis in geographic informationsystems (GIS) essentially is geometry oriented; hence, some research issues related to underlyingstructure are hard to address. In this regard, the topological representation to be introduced in thefollowing text has enabled us to uncover the underlying scaling pattern of street networks. Fourof the SI articles (Gil 2014; Lerman, Rofè, and Omer 2014; Mohajeri and Gudmundsson 2014;

Correspondence: Bin Jiang, Department of Technology and Built Environment, University of Gävle, SE-80176 Gävle, Swedene-mail: bin.jiang@hig.se

Submitted: June 20, 2014. Revised version accepted: July 15, 2014.

Geographical Analysis (2014) 46, 341–344

doi: 10.1111/gean.12060© 2014 The Ohio State University

341

Wei and Yao 2014) have adopted or are well connected to the topological representationperspective, which is powerful for understanding street hierarchies or geographic forms andprocesses in general.

Conventionally, geographic space is considered as a continuous Euclidian space that isdivided or subdivided into different areas, which authorities often define and delineate adminis-tratively and legally. Data collected about geographic space are assigned to individual areas andtherefore are assumed to be homogenous in each area. Good reasons existed for this assumptionduring the small-data era, when data were not as rich as they are in the present big-data era. Forexample, for the purpose of protecting privacy, data are deliberately aggregated rather thananalyzed at an individual level; data often are estimated statistically or roughly approximated,rather than accurately observed, as is the case with a global positioning system. This situation haschanged dramatically in recent decades due to advances in geospatial technologies. Subse-quently, large amounts of geospatial data at an individual level (rather than an aggregate level)have been collected for spatial analysis and spatial decision making related to events such astraffic jams, traffic accidents, and street crime. Network spatial analysis provides a powerfulanalytical means with which to gain insights into such data; for more details, one can refer toOkabe and Sugihara (2012). Network spatial analysis is not limited to street networks and can beapplied to other networks, such as those for rivers, water pipelines, and power grids. Accompa-nied by the SANET software tool (http://sanet.csis.u-tokyo.ac.jp/), network spatial analysis isdefinitely a new addition to standard network analysis in GIS.

Although current network analysis in GIS is based on the notion of a graph in which streetjunctions and segments are represented by nodes and links, respectively, it is essentially geomet-ric in the sense of locations of the junctions, distances along the segments, and/or directionsbetween the segments or the junctions. This geometric representation of street networks, whichis rooted and commonly seen in the geography literature (Haggett and Chorley 1969), is veryhelpful for dealing with such real-world problems as routing, closest facilities, and serviceareas—this outcome is because the key issue for these problems is distance. However, ageometry-oriented network is of little help in addressing issues regarding the underlying structureof streets (note that, by streets, we mean entire named streets rather than street segments). Forexample, what is the average degree of street connectivity for a city? How does street connec-tivity differ from one city to another? How many intermediate streets must one pass to travelbetween streets A to B? To address any of these questions, we need a topological representationor a graph in which entire named streets are represented by nodes, while street intersections orjunctions are viewed as links. Such a graph, rooted in space syntax (Hillier and Hanson 1984) andrelated software tools (e.g., http://fromto.hig.se/~bjg/axwoman/), is purely topological, becauseit does not involve geometric properties such as locations, distances, and directions (Jiang andClaramunt 2004). Many critics are skeptical of this type of topological representation, arguingthat the geometric representation contains more metric information and therefore is better thanthe topological one (e.g., Ratti 2004). This is a misperception, because having more informationdoes not ensure the rendering of new insights. Rather, having more information at a more detailedresolution can prevent us from understanding the bigger picture, for example, the scaling patternof far more small things than large ones.

A topological representation is much more interesting and informative than a geometric onein terms of understanding an underlying structure. Only a very limited range of connectivityexists for either junctions or segments, whereas a very wide range of connectivity exists forindividual streets. In other words, only a few kinds of junctions or segments, but many kinds of

Geographical Analysis

342

streets, are possible. More importantly, the few kinds of possible segments are more or lesssimilar, and can be characterized by an average and a limited variance; the many kinds of possiblestreets, in contrast, lack an average and a limited variance for characterizing their distribution ofconnectivity—this is known as scale free, or scaling. To be more specific and for junctions inparticular, the smallest number of connections is 3, whereas the largest number could be around10; the ratio between these two extremes is between 3 and 4. However, the picture is quitedifferent for individual streets: the smallest and the largest number of connections, respectively,are one and dozens, yielding a ratio of the two numbers that can far exceed 4. A large ratio hereindicates the presence of far more less-connected than well-connected streets. In more generalterms, far more small things than large things exist, a universal scaling pattern observed in manynatural and societal phenomena (Bak 1996). A geometric representation is of little help foruncovering a scaling pattern. Moreover, the power of a topological representation lies exactly inits lack of geometric or metric information. In this regard, the London underground map adoptedthe same principle, being geometrically distorted and topologically retained, yet more informa-tive in terms of station-to-station connections. As a reminder, a major purpose of formulating andrefining models is to get a simplified counterpart (or abstraction) of reality in order to obtain newinsights into that reality. Although reality itself contains far more information than a model of it,reality can never be identical to such a model.

A scaling pattern, or the notion of far more small things than large ones, differs radically fromthat of more small things than large ones. The former indicates a nonlinear relationship—that is,small causes large effects and large causes small effects—whereas the latter is a linear relation-ship—that is, small causes small effects and large causes large effects. Street networks, or builtenvironments in general, bear this scaling property and therefore are nonlinear complex systems,which cannot be simply understood by Newtonian physics (Jiang 2014). In this context, chaostheory and complexity science provide a series of modeling tools for better understandingnonlinearity and complexity, such as fractal geometry, complex networks, and agent-basedmodeling (Mitchell 2011). The scaling pattern of far more small things than large ones recursmultiple times, rather than just once. In other words, a few intermediate scales exist between thesmallest and the largest, and all of the scales form a scaling hierarchy. This hierarchy can be derivedby applying head/tail breaks—a new classification scheme for data with a heavy-tailed distribution(Jiang 2013). The resulting number of classes or hierarchical levels is called the ht-index (Jiang andYin 2014), which can be used to characterize complexity of geographic features or fractals ingeneral. The larger an ht-index value, the more complex is the fractal it measures.

We hope readers of this SI will keep the foregoing commentary in mind as they read thesubsequent articles. Its perspective should harbor different ways of thinking about understandingstreet networks, and of conducting spatial analysis. For now, we would like to take this oppor-tunity to thank all the participants at the workshop for their active participation and discussions,the articles’ authors for contributing their excellent work, and the many reviewers (who will beformally acknowledged separately by the journal) for providing constructive feedback in a timelyfashion. We also would like to thank Dr. Xiaobai (Angela) Yao and Dr. Itzhak Benenson for theirassistance in making the workshop successful and fun, and Dr. Daniel Griffith and Dr. YongwanChun, the journal editorial team, for their trust and support in editing the SI.

References

Bak, P. (1996). How Nature Works: The Science of Self-Organized Criticality. New York:Springer-Verlag.

Bin Jiang and Atsuyuki Okabe Street networks

343

Gil, J. (2014). “Analyzing the Configuration of Multimodal Urban Networks.” Geographical Analysis46(4), 368–391.

Haggett, P., and R. Chorley. (1969). Network Analysis in Geography. London: Edward Arnold.Hillier, B., and J. Hanson. (1984). The Social Logic of Space. Cambridge: Cambridge University Press.Jiang, B. (2013). “Head/Tail Breaks: A New Classification Scheme for Data with a Heavy-Tailed Distri-

bution.” Professional Geographer 65(3), 482–94.Jiang, B. (2014). “Geospatial Analysis Requires a Different Way of Thinking: The Problem of Spatial

Heterogeneity.” GeoJournal 46(4), 341–344. doi: 10.1007/s10708-014-9537-y.Jiang, B., and C. Claramunt. (2004). “Topological Analysis of Urban Street Networks.” Environment and

Planning B: Planning and Design 31, 151–62.Jiang, B., and J. Yin. (2014). “Ht-Index for Quantifying the Fractal or Scaling Structure of Geographic

Features.” Annals of the Association of American Geographers 104(3), 530–41.Lerman, Y., Y. Rofè, and I. Omer. (2014). “Using Space Syntax to Model Pedestrian Movement in

Urban Transportation Planning.” Geographical Analysis 46(4), 392–410.Mitchell, M. (2011). Complexity: A Guided Tour. London: Oxford University Press.Mohajeri, N., and A. Gudmundsson. (2014). “The Evolution and Complexity of Urban Street Networks.”

Geographical Analysis 46(4), 345–367.Okabe, A., and K. Sugihara. (2012). Spatial Analysis along Networks: Statistical and Computational

Methods. Chichester, UK: Wiley.Ratti, C. (2004). “Space Syntax: Some Inconsistencies.” Environment and Planning B: Planning and

Design 31, 487–99.Shiode, S., and N. Shiode. (2014). “Microscale Prediction of Near-Future Crime Concentrations with

Street-Level Geosurveillance.” Geographical Analysis 46(4), 435–455.Wei, X., and X. Yao. (2014). “The Random Walk Value for Ranking Spatial Characteristics in Road Net-

works.” Geographical Analysis 46(4), 411–434.

Geographical Analysis

344