ArticlePDF Available

Abstract

Christopher Alexander, a British-American scholar, famously differentiated an old (natural) city from a new (planned) one in structure. The former resembles a “semilattice”, or a complex system encompassing many interconnected sub-systems. The latter is shaped in a graph-theoretical “tree”, which lacks the structural complexity as its sub-systems are compartmentalized into a single hierarchy. This structural distinction can explain, or perhaps determine “the patina of life” in old urban districts and the lack of such in new ones. Alexander's idea, although widely influential, remains contested for its lack of empirical support. Subsequent literature failed to distinguish the structural differences between the old and new cities in systematic ways, nor is his asserted structure-life relationship verified with rigor. This study aims to test Alexander’s urban structural theory under a comprehensive research framework. We translated his constructs and premises into a mathematically testable form. The structural qualities of an urban street network, conceived as “semilattice”, “complex network” and “living structure”, were measured using graph-topological indicators. Urban life was captured using a combination of Twitter activities, Point-Of-Interests, and walking trips, aggregated at the district level. The structure-life relationship was tested statistically, after controlling for urban form and socio-demographic confounders, including land use, density, block size, parks, income, age, and demographics. This research design was implemented in London, New York, Hong Kong, and Gdansk. Our results support Alexander's early works that an old urban district contains more “semilattice” than new ones. This quality can be captured by Meshedness Coefficient, a graph-network indicator for a semilattice-shaped street network and a strong predictor for urban life. The same cannot be observed for “complex network” with consistency, and we found no independent associations between “living structure” and life, contrary to existing literature. The study shed light on the hidden relationships between urban spatial structure and behaviors, in both the virtual and physical world. We uncovered the British-American predilection of Alexander’s theory, which is well-supported by observations in London and New York yet less so in Hong Kong or Gdansk, suggesting the need for a locally-sensitive approach. The analytical tools developed can be of value for planning research and practice.
Landscape and Urban Planning 226 (2022) 104469
0169-2046/© 2022 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Research Paper
A city is not a tree: a multi-city study on street network and urban life
Jianxiang Huang
a
,
b
,
*
, Yuming Cui
c
, Haoliang Chang
d
, Hanna Obracht-Prondzy´
nska
e
,
Dorota Kamrowska-Zaluska
f
, Lishuai Li
g
a
8/F Knowles Building, Department of Urban Planning and Design, The University of Hong Kong, Pokfulam Road, Hong Kong SAR, China
b
The University of Hong Kong Shenzhen Institute of Research and Innovation, 5/F, Key Laboratory Platform Building, Shenzhen Virtual University Park, No.6, Yuexing
2nd Rd, Nanshan, Shenzhen 518057, China
c
Department of Architecture, 7/F Knowles Building, The University of Hong Kong, Pokfulam Road, Hong Kong SAR, China
d
Department of Advanced Design and Systems Engineering, City University of Hong Kong, Hong Kong, China
e
Department of Spatial Management, University of Gdansk, ul. Jana Ba˙
zy´
nskiego 8, 80-309 Gda´
nsk, Poland
f
Department of Urban Design and Regional Planning, Faculty of Architecture, Gdansk University of Technology, 11/12 Gabriela Narutowicza Street, 80-233 Gdansk,
Poland
g
School of Data Science, City University of Hong Kong, 16-201, 16/F, Lau Ming Wai Academic Building, 83 Tat Chee Avenue, Hong Kong SAR, China
HIGHLIGHTS
Christopher Alexanders urban structural theory was tested empirically in four cities.
Urban life was measured using Twitter, POI, and walking behavior data.
Urban structure was measured using graph-analytics of street network data.
Semilattice-shaped networks, measured in Meshedness Coefcient, is conducive to life.
The same cannot be found for networks of Living Structure measured in Ht Index.
ARTICLE INFO
Keywords:
Urban new data
Urban planning
Street network
Urban life
Christopher Alexander
ABSTRACT
Christopher Alexander, a British-American scholar, famously differentiated an old (natural) city from a new
(planned) one in structure. The former resembles a semilattice, or a complex system encompassing many
interconnected sub-systems. The latter is shaped in a graph-theoretical tree, which lacks the structural
complexity as its sub-systems are compartmentalized into a single hierarchy. This structural distinction can
explain, or perhaps determine the patina of lifein old urban districts and the lack of such in new ones. Al-
exanders idea, although widely inuential, remains contested for its lack of empirical support. Subsequent
literature failed to distinguish the structural differences between the old and new cities in systematic ways, nor is
his asserted structure-life relationship veried with rigor. This study aims to test Alexanders urban structural
theory under a comprehensive research framework. We translated his constructs and premises into a mathe-
matically testable form. The structural qualities of an urban street network, conceived as semilattice, complex
networkand living structure, were measured using graph-topological indicators. Urban life was captured
using a combination of Twitter activities, Point-Of-Interests, and walking trips, aggregated at the district level.
The structure-life relationship was tested statistically, after controlling for urban form and socio-demographic
confounders, including land use, density, block size, parks, income, age, and demographics. This research
design was implemented in London, New York, Hong Kong, and Gdansk. Our results support Alexanders early
works that an old urban district contains more semilatticethan new ones. This quality can be captured by
Meshedness Coefcient, a graph-network indicator for a semilattice-shaped street network and a strong predictor
for urban life. The same cannot be observed for complex networkwith consistency, and we found no inde-
pendent associations between living structure and life, contrary to existing literature. The study shed light on
the hidden relationships between urban spatial structure and behaviors, in both the virtual and physical world.
We uncovered the British-American predilection of Alexanders theory, which is well-supported by observations
* Corresponding author at: 8/F Department of Urban Planning and Design, Knowles Building, The University of Hong Kong, Pokfulam Road, Hong Kong SAR, China
E-mail addresses: jxhuang@hku.hk (J. Huang), ymcui@connect.hku.hk (Y. Cui), hlchang4-c@my.cityu.edu.hk (H. Chang), hanna.obracht-prondzynska@ug.edu.pl
(H. Obracht-Prondzy´
nska), dzaluska@pg.edu.pl (D. Kamrowska-Zaluska), lishuai.li@cityu.edu.hk (L. Li).
Contents lists available at ScienceDirect
Landscape and Urban Planning
journal homepage: www.elsevier.com/locate/landurbplan
https://doi.org/10.1016/j.landurbplan.2022.104469
Received 17 August 2021; Received in revised form 4 March 2022; Accepted 8 May 2022
Landscape and Urban Planning 226 (2022) 104469
2
in London and New York yet less so in Hong Kong or Gdansk, suggesting the need for a locally-sensitive
approach. The analytical tools developed can be of value for planning research and practice.
1. Introduction
Christopher Alexander, a British-American scholar, formulated the
relationship between urban structure and life in an article A City Is Not a
Tree(Alexander, 1965). A natural (old) city differs from a planned
(new) one by structure: the former is shaped like a semilattice, a
mathematical term for a complex system encompassing large numbers of
interconnected sub-systems; whereas the latter is structured in a graph-
theoretic tree, compartmentalized into a single hierarchy. This struc-
tural difference explains why, or perhaps determines that a planned city
lacks the patina of lifeand it is entirely unsuccessful from a human
point of view(Alexander, 1965, p.1). He went on to assert a (lively) city
cannot be planned, at least not in the modernist manner, since they (the
planners) cannot encompass the complexity of a semilattice in any
convenient mental form (ibid. p.16). Alexanders structure-life theory
was revisited and updated continuously throughout his career, such as
Notes on the Synthesis of Form(Alexander, 1966), A Pattern Language
(Alexander, 1977), A New Theory of Urban Design(Alexander, 1987)
and more recently, The Nature of Order(Alexander, 2002), yet its key
message remained largely consistent.
Alexanders idea is of lasting inuences over the last half-a-century.
A City Is Not A Tree, along with A Pattern Languageand A New
Theory of Urban Design were listed among Cuthberts 40 classic urban
design texts (Cuthbert, 2007), widely cited, taught, and followed by two
generations of researchers and practitioners. Alexander was credited, in
part, of inspiring New Urbanism, an urban design movement advocating
for interconnected, compact network of streets over the fragmented cul-
de-sacs, or dead-ends in the suburbs (Katz, 1993; Park and Newman,
2017). His work is often compared with Walter Christallers Central
Place Theory of the quantity, the size and the location of human set-
tlements in the discipline of economic geography (Brush, 1966), and the
Deep Structure in Bill Hilliers Space Syntax Theory, referring to the
typological depth of urban spatial networks (Hillier, 1989; Davis, 2021).
Alexander is considered by Michael Batty, an urban spatial data scien-
tist, as path-breakingin revealing the complexity and diversity as the
essential features of urban live (Batty, 2006; 2015); he is also regarded
as one of the founders of the emerging science of complex networksby
urban science scholar Luis Bettencourt (2016, p.48).
The scientic rigor of Alexanders work remains contested. Stephen
Marshall (2012) criticized A City Is Not A Tree as rooted in analogies
and thought experiments, which have not been tested in systematic
ways, while subsequent researchers tend to simply cite Alexander in
outline without critical scrutiny. Gabriel and Quillien (2019) consider
Alexander and the architectural and urban planning community largely
failed to nd scientic explanations for his intuitions to be taken seri-
ously outside of the arts. Such failure, noted by Michael Mehaffy (2019),
owes much to the artist-architectsobsession in creating newness for its
own sake. Dovey and Pafka (2016) argued that the pseudo-scientic
tendencies of Alexander matters less, under the pretext that urban
design knowledge is broader and cannot be reduced to numbers. Alex-
anders structure-life relationship was rarely proven or falsied in
empirical literature. A search of scholarly work published in the wake of
Alexander yielded a trio of papers which tested his ideas partially
(Mohajeri, French, & Batty, 2013; Jiang and Ren, 2019; de Rijke et al.,
2020). Their ndings were insufcient to settle the disputes in literature.
Questions remained whether the asserted structural difference can be
observed consistently between the natural (old) and planned (new)
cities? Which Alexandrian structural qualities (and their mathematical
expressions) are predictive of urban life? And to what extent is Alex-
anders theory applicable across cultures and societies?
This study aims to test Alexanders urban structural theory under a
comprehensive research framework. Our objectives are to 1) test
whether there are systematic strucutral differences between old and new
cities or urban districts, 2) statistically examine the structure-life re-
lationships, 3) test the generalizability of his theory across geographical
and cultural context. A cross-sectional research design was adopted in
168 urban districts in four cities: London, New York, Hong Kong, and
Gdansk. The spatial structure of street networks in each district was
measured using a trio of graph-network indicators, operationalized ac-
cording to Alexanders three major constructs: semilattice, complex
network, and living structure. Urban life was measured using a
combination of social media data, Point-Of-Interests (POI), and walking
trips. The asserted structure-life relationships were tested using statis-
tical control for urban form and socio-demographic confounders.
2. Relevant works
Recent advancement in street network and urban data analytics
supports the revisiting of Alexander. His conception of urban structure
has inspired a large body of quantitative literature, covering the eld of
mathematics, transport, geography, and urban sciences. His construct of
urban life has also been captured by a wave of literature using urban
new data obstained from social media platforms, mobile devices, and
POIs, etc., although his key premises have rarely been tested empiri-
cally. Relevant literature is summarized in Fig. 1.
2.1. Urban structure and network analysis
Semilatticeand treewere proposed by Alexander (1965) to
describe the complexity of an urban system and the degree of over-
lapping of its sub-systems. At rst, he presented the semilattice-tree
dichotomy in set theory, which was later replaced with graph theory
by Harary and Rockey (1976), the two mathematicians advocating for
the graph-representation of urban structures using linksand nodes,
which has since become the mainstream approach. Semilatticecan be
measured by counting the number of loops in a network, and it can be
conveniently computed using Meshedness Coefcient, a graph-
theoretical indicator measuring the number of loops in a network as a
proportion of the maximum allowable number (Buhl et al., 2004). The
treequality, in contrast, implies the lack of structural complexity and
the absence of overlapping among sub-systems. It has been captured by
the Treeness Indicator, dened as the ratio between the length of street
segments not within a loop and those of the entire network (Xie and
Levinson, 2007).
Complex networkappeared in Notes on the Synthesis of For-
m(Alexander, 1966) and later in A Pattern Language(Alexander,
1977), referring to the convoluted, non-trivial features found in
real-world networks. It has since been operationalized into many
graph-topological indicators. Examples include Betweenness Centrality,
a measure representing the role of a node in connecting others in its
surroundings (Freeman, 1977), Closeness Centrality, the sum of the
lengths of the shortest paths from a nodeto the rest (Rodrigue,
Comtois and Slack, 2016), and many others such as Proximity and
Integration (Porta, Crucitti and Latora, 2006). Alternative indicators
were developed in transport literature to measure the complexity of a
transportation network, including the proportion of three-way /
four-way road intersections, cul-de-sacs (dead-end), or the average
block size(Kerr et al., 2007). To date, complex networkremains an
active domain of scientic research.
Living structurewas coined in The Nature of Orders(Alexander,
2002), referring to a structure containing overlapping details across
many levels of scale to form a coherent whole. Alexander regard the
J. Huang et al.
Landscape and Urban Planning 226 (2022) 104469
3
living structureas both a physical phenomenon and a philosophical
view fundamental to life and beauty. His listed 15 qualitative properties
from levels of scaleto contrast to describe a living structure, and
his list was substantiated by Nikos Salingaros (2018) using fractals, a
concept for objects of repetitive patterns at increasing and decreasing
scales. A mathematical classication of the living structurewas given
by Jiang and Yin (2014) based on the head/tail breaks, a method of
dividing elements of a network into the headand tailusing the
arithmetic mean of a designated data value. This division continues until
the distribution of headis no longer heavy-tailed. This method was
further developed into the H
t
Index, an indicator capturing the degree of
skewed distribution of a networks property, or far more small things
than large ones(Jiang and Ren, 2019).
2.2. Urban life
Urban life, synonymously referred to as vitality, vibrancy, or liven-
ess, is often viewed as an attribute of a place. As an outcome measure,
urban life has been captured by a wave of recent literature in the domain
of urban planning and geography. For example, Yue and Zhu (2019)
measured urban vitality in Wuhan, China using social network review
data, and they found vitality in association with density, mixed use, and
accessibility indicators such as Betweenness Centrality and Closeness
Centrality. Long and Huang (2017) studied POI and web-based review
data and concluded that urban block size is negatively associated with
urban commercial vitality. Alternatively, urban life has been measured
using mobile phone data (de Nadai et al. 2016) or social media activities
(Franca et al., 2015; Chen et al., 2019).
A more robust measure of urban life, which recently ourished in the
eld of urban studies, is to combine multiple evidence from both the
virtual and the real world. Huang et al. (2021), for instance, studied the
perception of city images using both geo-coded Instagram data and
questionnaire. Johnson et al. (2019) studied the perceived benets from
urban greenspace using both Twitter dataset and semi-structured in-
terviews. Wang et al., (2018) studied the perceived attractiveness of the
Olympic Forest Park in Beijing using both social media and survey data;
the above studies exposed the biases of the virtual world data and they
suggested a combined approach. In a recent paper, Fang et al. (2021)
constructed a multi-faceted measure of urban vitality using indicators of
concentration, accessibility, livability, and diversity. Interestingly, they
concluded that vitality associated positively with various street network
metrics such as connectivity and closeness, largely consistent with Al-
exanders idea yet without citing him.
2.3. Street network and travel behaviors
The impact of street network on travel behaviors have been studied
extensively in transport and health literature. Again, many overlapped
with Alexander in inputs (street network) or outcomes (walking trips)
without citing him. In general, researchers consider street network
characteristics such as connectivity, network density, and patterns,
contribute positively to walking, biking, and transit ridership (Marshall
and Garrick, 2010). In particular, walking is positively correlated with
land use diversity, intersection density, and the numbers of destinations
within walking distances, while public transit usage positively associ-
ated with the proximity to transit and street network patterns, with land
use diversity as a secondary factor (Ewing and Cervero, 2010). A well-
connected street network is found to promote walking, while poorly
connected ones promote driving (Oakes, Forsyth and Schmitz, 2007;
Berrigan, Pickle and Dill, 2010), although a large proportion of obser-
vations were made in the British-American context. In a global-scale
analysis, Barrington-Leigh and Millard-Ball (2019) cautioned that the
associations between street network attributes and travel behaviors may
not be universal, and a place-based approach is necessary.
2.4. Empirical testing
Alexanders premises and constructs have rarely been tested empri-
cally, despite repeated calls for doing so (Marshall, 2012; Gabriel and
Quillien, 2019; Mehaffy, 2019). A survey of academic publications
yielded a trio of recent publications which tested Alexander partially. de
Rijke et al. (2020) studied the graph-network properties of six cities and
concluded that old (natural) cities, notably Amsterdam, Rome, and
Geneva, contained more living structurethan new (planned) ones such
as Chandigarh, Levittown, and Brasilia. Their study did not, however,
measure urban life; nor was their choice of the six cities sufciently
justied to rule out possible selection biases. In another study, Mohajeri,
French and Batty (2013) analyzed the evolution of Londons street
network over two centuries. They found more loop-like structures in
historical London boroughs and more tree-likestructures in new ones,
hence partially supported Alexander, although they have not studied
urban life either. A third study by Jiang and Ren (2019) measured both
Fig. 1. A summary of literature related to the Alexanders theoretical constructs, measurement, and empirical testing, relative to the scope of this study.
J. Huang et al.
Landscape and Urban Planning 226 (2022) 104469
4
urban street networks and Twitter activities, a proxy for urban life.
They found that H
t
Index, a mathematical indicator for living struc-
ture, can predict the location of geo-coded tweets in the UK in support
of Alexander, although their analysis failed to statistically control vari-
ables such as density, land use, social and demographics, begging the
question of whether the H
t
Index is independently associated with
urban lifeor in interaction with confounders.
3. Research gaps
The following knowledge gaps remained concerning Alexanders
urban structural theory:
First, existing literature failed to distinguish the structural differ-
ences between the old and new cities in systematic ways. One can argue
that, by common sense, street networks are perhaps more fragmented in
newly built suburbs and more inter-connected in historical city centers.
But these evidence suffer from the lack of a counterfactual, that is,
whether such structural differences can be consistently observed be-
tween an old and a new city that are similar in every other aspects.
Second, his asserted structure-life relationships have not tested with
statistical rigor. Existing evidence fell short of determining whether the
structure-life relationships are independent of or in interaction with
urban form and socio-demographic confounders. Each of his construct
(i.e., semilattice, complex network, living structure) has been
substantiated and operationalized by his followers (Salingaros, 2005;
Jiang and Ren, 2019; Mehaffy, 2019), although without consensus over
which construct ‘behavein consistency with his premise.
Third, the generalizability of Alexanders theory across cultures and
societies remains unclear. His writings drew heavily upon British and
American cases, where he spent most of his academic career, while his
idea has been tacitly accepted as universal, exemplied by numerous
urban planning and design case studies, and by his own built work from
Japan to Mexico. We note with caution, however, that it is hard to
imagine how any of the above could inform the external validity of
Alexander in any meaningful way.
There is a need for a comprehensive, multi-city study to test Alex-
anders urban structural theory. Such a study will not only help resolve
the ongoing theoretical disputes over a classic text, but also enhance its
relevance for contemporary cities and planning practices. Evidence
collected from multiple cities can shed light on the external validity of
Alexanders theory across cultures and societies.
4. Methods
We translated Alexanders theory into a mathematically testable
form. His key constructs were measured using a variety of performance
and graph-network indicators. We used statistical models to test the life-
structure relationships and the structural differences between old and
new cities or districts. The above research design was adopted in four
cities: London, New York, Hong Kong and Gdansk, where empircal
datasets of behaviors and street networks have been compiled and
analyzed.
4.1. Empirical translation of Alexander
A testable form of Alexanders theoretical premise is provised in Eq.
(1), in which urban life is expressed as a function of urban structure,
urban form and socio-demographic conditions.
L=f(S,U,E,D,e)(1)
where L stands for the patina of life, measured in Twitter activities,
POI, and walking trips; S represents urban structural indicators, oper-
ationalized following Alexanders constructs of Semilattice-Tree,
Complex Network, and Living Structure; U represents confounding
urban form conditions, such as density, land use mixture, block size, and
parks. E & D represent the socio-demographic characteristics; e is the
unmodelled error term.
A conceptual framework is then derived according to the above
(Fig. 2), with blue arrows specifying the hypothesized relationships
between urban life and urban structure.
4.2. Measuring urban life
We used three data sources, namly Twitter, POI, and walking trips, to
capture the patina of lifeor the multi-faceted human-built environ-
ment interactions.
Twitter is a popular social media platform which allows users to
communicate in concise, instant messages about life events (Lau, Collier
and Baldwin, 2012). It is also popular among researchers as a data
source to capture the intensity of human activities, since Twitter data
(tweets) are accessible from the public domain by default. We used
Twitter API, a computer program interface (Morstatter et al., 2013) to
exhaustively retrieve tweets in each city within a predened bounding
box, marked by the longitudinal and latitudinal coordinates of the
lower-left and upper-right corners (see Fig. 4). The retrieved tweets
contain text, user ID, time stamp, and, for a proportion of users who
enabled the location sharing by default, the location of the mobile de-
vice from which the tweet was sent. The range of accuracy for GPS-based
smartphone positioning is supposedly <5 m under an open sky, or up to
30 m in a density city (PNT, 2008), sufcient for our study purposes. The
original Twitter database were cleaned by removing repetitive tweets
and those from fake accounts. Tweets containing GPS coordinates from
mobile devices were geo-coded using ArcMap 10.4 software. For each
district, the density of geo-coded tweets (Tden ) can therefore be
computed by dividing the number of geo-coded tweets by its area.
Point-Of-Interest (POI) refers to an umbrella category of destina-
tions, from tourist attractions to local community amenities such as
restaurants, library, community centers. It is often used as a proxy for
human activities, since its presence depend on and generate pedestrian
trafc (Zukin, 2010; Yue & Zhu, 2019). We used the density of POI per
square kilometer (Pden) to measure urban life, computed by dividing the
POI count by the area of the district. The POI databases in the four cities
were acquired via OpenStreetMap (OSM), an open access digital plat-
form enriched by millions of contributors (OSM, 2019). The database
covers a range of destinations from landmarks, e.g., the Empire State
Building in Manhattan, New York to corner shops in Londons West End.
Walking trip reects the intensity of people walking from place to
placein Alexanders remarks (Alexander, 1965, p.14), and it is another
accepted proxy for urban life in research literature (Sung, Lee and
Cheon, 2015; Yue and Zhu, 2019). We computed the area-weighted
density of walking trips (Wden) according to Eq. (2) below.
Wden =Dwden*100*Wperc *Wn(2)
where Dwden is the dwelling density (people / km
2
); Wperc is the per-
centage of trips carried out on foot, and Wn is the number daily trips per
person, all acquired from ofcial travel demand survey databases.
4.3. Measuring urban structure
Following the graph-theoretical approach specied by Steadman
(2004), we represented the structure of a real city using a mathematical
graph consisted of nodes, i.e. a street intersection or a dead-end, and
links, the street segments. By street, we refer to a public road acces-
sible to both vehicular and pedestrian trafc, therefore excluding pri-
vate roads, hiking trails, or bicycle tracks, etc. The street network
database for each district in the four cites was prepared in the following
procedure. First, road network databases were acquired from ofcial
sources (Section 3.5); Redundant road segments and trafc islands were
removed; fragmented road polylines were joined together. We removed
the pseudo node, a type of node connected by only two adjacent links;
J. Huang et al.
Landscape and Urban Planning 226 (2022) 104469
5
we then re-joined the remaining two links together using the Unsplit
command in ArcMap 10.4 software. Other loosely connected street
segments were also checked and xed. The above procedure was
implemented using the sDNA software (Cooper, 2013) and automated in
Python. For quality assurance, the results were independently checked
against Google Street Map by two trained researchers. The prepared
database was then used to compute network indicators below.
The semilatticequality was measured using Meshedness Coef-
cient (
α
), a graph indicator to describe the overlaps and redundancy in a
network, following the equations developed by Buhl et al. (2004) and
later by Rodrigue et al. (2016).
α
is a normalized value between 0 and 1,
with a larger value suggesting a semilattice, while a smaller
α
implies a
tree(Fig. 3). A mathematical expression of
α
is provided in Equation
(3), where n and k are the number of nodes and links in the network,
which were computed using the Network Dataset Functionin ArcMap
10.4 software; u and v were the current and the maximum possible
number of loops in the network, which were expressed using the Euler
characteristic (Spanier, 1981) as kn+1 and 2n 5 respectively.
α
=u
v=kn+1
2n5(3)
The complex network quality of a street network was measured
using node-based betweenness metrics. We used the Average Between-
ness Centrality (Bt), dened as the arithmetic mean of Betweenness
Centrality (Bti)of a total of n nodes across a network, and it is expressed
in Eq. (4). Bt has been used to measure the interactive relationships
within a network (Ding et al., 2015). A high Bt value suggests a complex,
highly connected network with intense trafc ow, and possibly intense
urban life. For walking trips, we adopted an analysis radius of 600m in
accordance with literature (Calthorpe, 2004). The above computation
was implemented using ArcMap 10.4 and sDNA software.
Bt =n
i=1Bti
n=n
i=1Σx=y=z
6ij(i)
6ij
n(4)
The living structurequality were measured based on the presence
of fractals across various levels of scale in a street network. We adopted
the H
t
Index (Ht), a mathematical indicator developed by Jiang and Yin
(2014), which classies the degree of heavy-tailed distribution of node-
based betweenness in a street network. Ht was computed following the
approach outlined in the same work: all nodes in a given street network
were rst ranked according to their betweenness values (Bti), from large
to small; the list was then divided in two using the mean Bti value as a
threshold. The head listcontains nodes with above-average values,
while the rest belonged to the tail list. At each scale, if the length ratio
between the head and tail list was smaller than 0.4, this meant there
were far more small things than large ones, and the head/tail break
calculation would repeat using the head list at the next scale, or it would
stop otherwise. Ht was the number of times the head/tail break calcu-
lation had been repeated. We automated the above workow using a
script written in Python.
4.4. Confounders and statistical analysis
Human activities tend to interact with confounding urban form and
socio-demographic variables, as suggested by literature at length (Ja-
cobs, 1961; De Nadai et al., 2016; Ye, Li, & Liu, 2018; Lu, 2019). These
confounders have been calculated in the following steps and controlled
in statistical analysis.
Land use mixture (M) was computed using Shannons entropy (Eq.
(5)) based on ofcial land use databases. M is denoted as a function of
the area of land use category i as a proportion of the total area of the
district (Pi) and the number of land use categories (n). In this study n =
4, including residential, commercial, manufacturing, and park land.
M= − n
i=1
Pi*lnPi
ln(n)(5)
Dwelling density (Dwden) was computed using the total population
divided by the area of the district. Block size was measured using the
density of street network intersections (Bksize) in each city district. Bksize
was calculated using the number of street intersections divided by the
area of the district, the higher its value, the smaller the average block
size in the district. Park has been associated with urban vitality, espe-
cially walking behaviors in a number of studies (Ball et al., 2001; Sung,
Lee and Cheon, 2015; De Nadai et al., 2016). Park coverage (Pkperc) was
calculated as the percentage of the park land in the district.
Socio-demographic confounders include Mean Household Size
Fig. 2. A conceptual framework for testing Alexanders theory on urban structure and urban life.
Fig. 3. (a) A semilatticenetwork with redundant connections and loops (
α
=
0.24). (b) A treenetwork without loops (
α
=0.00).
J. Huang et al.
Landscape and Urban Planning 226 (2022) 104469
6
(Hsize) measured by the number of persons in each household, Median
Household Income ($hh ) per annum (converted to US dollars), Home
Ownership Percentage (Operc ), the percentage of households owning
homes, Demographic Groups, the percentage of population in the top-
two demographic groups (D1, D2), depicted by ethnicity in London and
New York, or linguistically in Hong Kong and Gdansk, and Age Groups
measured by the percentages of the elderly (Aold) and the young (Ayoung)
population. The relationships between urban structure and urban life
were tested using a multivariate Ordinary Least Squares (OLS) regres-
sion model shown in Eq. (6).
L=β0+β1S+β2M+β3Dwden +β4Bksize +β5Bdage +β6Bdstdv
+β7Pkperc +β8Hsize +β9$hh +β10Operc +β11 D1
+β12D2+β13 Aold +β14 Ayoung +
μ
(6)
where the dependent variable L represented urban life indicators (Tden ,
Pden, Wden); S was the street network indicators (
α
, Bt,Ht); M, Dwden ,
Bksize, Bdage , Bdstdv, Pkperc, were confounding urban form conditions;
Hsize, $hh , Operc D1, D2, Aold and Ayoung were the socio-demographic con-
founders; β0 β14 were the intercept and regression coefcients;
μ
was
the unmodelled error term. Assumptions of the OLS regression were
checked robustly. The data distribution of each variable above was
checked against a normal distribution curve. Multicollinearity diagnosis
for each independent variable was performed using the Variance Ina-
tion Factor (VIF) (Thompson et al., 2017). We also checked the spatial
autocorrelation of the OLS regression using Morans I (Moran, 1950),
implemented using the spatial regression package sg162 (Pisati, 2001)
and STATA MP/16.1 software.
The above OLS regression was performed for each city and all four
together. The structural distinctions between old and new districts were
tested on semilattice (
α
), complex network (Bt) and living
structure(Ht), using t-test.
4.5. Study locations, data and sources
Our research design was implemented in New York, London, Hong
Kong, and Gdansk, four metropolises with relatively developed street
network, data accessibility, and digital infrastructure. All ranked high in
smartphone penetration and social media usage (Go-Globe, 2015). The
choice of the four cities allows us to test the generalizability of Alex-
ander: London and New York frequently appear in Alexanders writings
and both are typical of what German geographer Burkhard Hofmeister
(1970) referred to as Anglo-American cities. Hong Kong and Gdansk
were never mentioned by Alexander, although Hong Kong can be
marginally related to the Anglosphere due to its colonial history.
The unit of analysis was the urban district, although named differ-
ently in each city. We divide the total of 168 urban districts in four cities
(excluded two sparsely inhabited ones in Hong Kong) into the oldand
newby local plans(Fig. 4)
Greater London (London) (5130N 05W), the largest metropol-
itan area in the UK, covered an area of 1,596 km
2
and accommodated a
population of 8.2 million. London consisted of 33 Local Authority Dis-
tricts (LAD) designated by the UK Parliament (1963), each measuring
some 50 km
2
in land area and 220,000 residents on average. The nat-
uraland plannedcities in London were distinguished according to the
1944 Greater London Plan (ACO, 1946), which had transformed London
into . a large number of communities, each sharply separated from all
adjacent communitiesThe structure is a tree(Alexander, 1965, p.6).
Fig. 4. Map of (a) Greater London, (b) the City of New York, (c) Hong Kong SAR, and (d) the City of Gdansk with the coordinates of the geographical bounding boxes
for Twitter data-mining and borders of urban districts. The old districts are differentiated from the new ones according to local planning documents.
J. Huang et al.
Landscape and Urban Planning 226 (2022) 104469
7
Road centerlines and travel mode choice data were obtained from
Transport for London (TfL, 2019); land use and open space databases
were obtained from the Ofce of National Statistics (ONS, 2010); de-
mographic and economic information, including population density,
household income, home ownership, ethnicity, etc., were acquired
through the Greater London Authority (GLA, 2018).
The City of New York (New York) (4042N 7400W) was the
largest city in the USA. It covered an area of 777 km
2
and it was home to
a population of 8.4 million. New York consisted of 59 Community Dis-
tricts, each measuring 13 km
2
in land area and hosting 140,000 people
on average. New Yorks historical districts, desginated by the Regional
Plan of New York and Its Environs in 1929 (CRPNYIE, 1929), were
located mostly in Manhattan and Brooklyn, while newer districts were in
Bronx, Queens, and Staten Island (Fig. 4 (a)). Road centerline data were
obtained from ofcial sources (NYC OpenData, 2017). Travel data were
collected from the citys 2019 report of Citywide Mobility Survey (DOT,
2019). Land use, and open space information were acquired from the
Primary Land Use Tax Lot Output (PLUTO) data (DCP, 2020).
Household income data were acquired from the 2018 American Com-
munity Survey (US Census Bureau, 2020); demographic data were ac-
quired from New York City Government Department of City Planning
(DCP, 2018).
Hong Kong Special Administrative Region (Hong Kong) (2218N
11412E) was a vibrant metropolis in Southern China, covering a land
area of 1,106 km
2
and home to 7.1 million residents (HKCSD, 2017).
Hong Kong consisted of 44 Secondary Planning Unit (SPU) (HKPD,
2017), each accomodates some 170,000 people and covers 22 km
2
of
land on average. The two sparsely inhabited SPUs (# 81 & 91 of remote
islands) were excluded from analysis. The old districts are located
mainly on the Hong Kong Island and Kowloon Peninsula (Loo and Chow,
2008), while newer districts consist of new towns in the New Territories.
The two sparsely inhabited SPUs, #81 and #91, were atypical of Hong
Kongs urban conditions and they were dropped from analysis.
The City of Gdansk (Gdansk) (5422N 1838E) was a large city in
Northern Poland, covering an area of 263 km
2
and home to a population
of 470,800 (COGD, 2020). The city consisted of 34 administrative
neighborhoods, each measuring 6.7 km
2
in land area and 12,800 in
population on average. Gdansks historical urban core dated back to the
Middle Ages (Lorens, Kamrowska-Załuska & Kostrzewska, 2014), while
the new growth started after the WWII. The old districts were
demarcated using the Protected Historical Neighborhoods designated by
the municipality (GDA, 2018). Databases of road centerlines, de-
mographics, land use, parks, and walking behaviors were obtained from
ofcial sources (NCGCD, 2020; GDA, 2016).
5. Results & discussion
We analyzed datasets from the four cities. Results are discussed
relative to Alexanders theoretical premises, constructs, generalizability
and practical implications.
5.1. Data characteristics
The indicators for urban life, structure, urban form and socio-
demographic confounders are summarized in Table 1. We acquired
151 million tweets within the bounding boxes covering the four cities
between May 2016 and December 2018. Some 20 million tweets con-
tained mobile GPS coordinates. After extensive cleaning, 2.5 million can
be geo-coded within the boundaries of the 168 districts. We retrieved
227,805 geo-coded POIs from OSM in the 168 districts.
New York had the highest Twitter, POI, and walking trip densities,
whilst Gdansks were the lowest. Structure-wise, New York had the
highest
α
at 0.38 and London had the lowest
α
of 0.16. London and
Gdansk had the highest and lowest values in Bt and Ht. Hong Kong had
the highest dwelling density of 21,000/km
2
, contrary to the lowest of
3,790/km
2
in Gdansk. Income was the highest in New York and lowest
in Gdansk. Percentage-wise, Gdansk had the most elderly population
(16% >65), while New York had the largest share of young people (22%
<18). Home ownership was higher in Hong Kong and Gdansk (>=50%)
and lower elsewhere.
5.2. Distinguishing old districts from new ones
We found more semilatticeand complex networkqualities in old
urban districts than in new ones. Fig. 5 summarizes
α
, Bt, and H
t
by old
and new districts in four cities. The mean
α
and Bt were consistently
higher in old districts than in new ones (Fig. 5 (a) and (b)). A t-test of the
two rejected the null hypotheses at 95% condence level, suggesting
both semilattice and complex network are viable constructs, and
α
Table 1
The mean value and the range of indicators for urban life, urban structure, urban form, and socio-demographic confounders for the 168 urban districts in four cities.
London New York Hong Kong Gdansk
Observations 33 59 42 34
Urban Life Indicators
Density of Geo-Coded Tweets, #/km
2
2,363(6330,961) 3,260(2961,415) 2,435(139,220) 86(11,380)
POI Density, #/km
2
160(23960) 1,393(2354,214) 125(1786) 90(6353)
Density of Walking Trips, 10
6
#/d.km
2
0.68(0.111.83) 3.32(0.05 9.23) 0.87(0.005.19) 0.36(0.011.42)
Urban Structure Indicators
Meshedness Coefcient (
α
) 0.16(0.100.23) 0.38(0.230.56) 0.17(0.070.28) 0.19(0.120.30)
Average Betweenness (Bt) 1,678(8853,685) 700(4041,473) 1,014(114190) 521(911,619)
Head/Tail Index (Ht) 8.00 (511) 7.20(59) 6.11(38) 5.56(48)
Urban Form Confounders
Land Use Mixture (M), ratio 0.84(0.530.96) 0.52(0.190.84) 0.62(0.030.88) 0.62(0.300.85)
Dwelling Density, 10
3
/ km
2
7.38(2.2116.10) 17.38(2.3942.49) 21.07(0.05106.05) 3.79(0.0917.95)
Street Intersection Density, #/km
2
128(59250) 112(63226) 107(1538) 42(5127)
Park Coverage, % 32(560) 7(130) 30(183) 61(1288)
Socio-Demographic Confounders
Average Household Size, ppl./hh. 2.4(1.63.0) 2.7(1.73.4) 3.0(2.34.2) n.a.
Median HH. Income, 10
3
US$/yr. 55.90(40.5889.70) 66.58(25.73148.15) 57.26(20.23201.67) 16.79(14.6518.15)
Home Ownership Percentage, % 22(738) 27(471) 53(2881) 50(2664)
Primary Demographic Group
1
, % 59(2783) 32(185) 79(4192) 58(1100)
Secondary Demographic Group
2
,% 9(225) 30(675) 9(148) 28(098)
Perc. of Old (>65), % 12(618) 12(622) 16(928) 22(835)
Perc. of Young
3
, % 20(1127) 22(734) 11(718) 16(1126)
1
White for New York and London, Cantonese-speaking group for Hong Kong, and English-language Twitter users for Gdansk.
2
Hispanic and Asian for New York and London, English-speaking group for Hong Kong, Polish language Twitter users for Gdansk.
3
<18 for New York and Gdansk, <15 for London, and <14 for Hong Kong.
J. Huang et al.
Landscape and Urban Planning 226 (2022) 104469
8
and Bt are viable indicators to capture the structural differences between
the old and new. The same cannot be observed on the living structure
qualities. The differences between the old and new districts measured in
Ht were inconsistent (Fig. 5 (c)). Old districts in Gdansk, London, New
York had lower Ht than new ones, contrary to literature, although such
differences in London and New York were statistically insignicant.
Hong Kong was the only normality, where Ht was higher in its old dis-
tricts than in new ones (p =0.00). Our ndings contradicted de Rijke
et al. (2020) who reported higher Ht in old cities (Amsterdam, Rome,
Geneva) than in new ones (Chandigarh, Levittown, and Brasilia),
although our observation was at the district level (intra-city) vs. de
Rijkes at the city level (inter-city). Nevertheless, this inconsistency
raises questions on either the construct of living structureor the
effectiveness of Ht in capturing the structural differences between old
and new cities.
A possible explanation of why old districts contained more semi-
latticelies in the road geometric standards, which were adopted in the
UK and the US concurrently with Alexanders observations. An example
is illustrated in Fig. 6, in which
α
was computed for each of the four-
stage evolution of the predominant street patterns outlined by Stephen
Marshall (2005).
α
was high in the pre-industrial Altstadt (
α
=0.26)
and the Bilateralof the 1920 s (
α
=0.35). It reduced in the later stage
of Conjoint (
α
=0.22) and further in Distributory(
α
=0.01). The
transition took place around the 1950s, when A City is Not a Treewas
conceived. The transition was driven, in Marshalls view, largely by the
road geometric standards in the UK, the US, and later elsewhere (ITE,
1984; HA, 1999; AASHTO, 2018). These standards purportedly played a
role in reducing the semilatticequalities in newly planned districts by
rejecting the traditional, well-connected street network as dysfunc-
tionaland unplanned, while promoting the cul-de-sacs and three-way
intersections (Marshall, 2005).
5.3. Urban structure and urban life
The predictive powers of street network indicators on urban life were
tested using correlational statistics. Table 2 shows nine bivariate
Fig. 5. Testing the structural differences between old and new districts, measured by (a) Meshedness Coefcient, (b) Average Betweenness, and (c) H
t
Indexin the
168 urban districts in four cities.
Fig. 6. The evolution trajectory of street network outlined by Stephen Marshall (2005) with declining
α
, suggesting a shift from semilatticetowards tree
over time.
J. Huang et al.
Landscape and Urban Planning 226 (2022) 104469
9
regression models, each regressing an urban life indicator (Twitter, POI,
walking trips) on a street network indicator (
α
, Bt, Ht).
A semilattice-shaped street network is a strong predictor for urban
life in the physical world. In model 13,
α
contributed positively and
statistically signicantly to Twitter density, walking trip density and
POI density (p =0.000). Particularly, model 2 and 3 had high R
2
at
0.712 and 0.570, suggesting that
α
was more predictive for POI and
walking, and both are proxies for life in the physical world. In con-
trary, the R
2
in model 1 was much lower at 0.184, suggesting that
α
was less predictive of Twitter activities.
The complex networkquality is a strong predictor for urban life of
the virtual world. Bt contributed positively to Twitter density (p =
0.000) and the regression model explained 36% of the variations in
Twitter activities, which was a proxy of urban life in the virtual
world. In comparison, the R
2
values for model 5 & 6 were lower
(0.055 & 0.019), suggesting that complex networkquality were
less effective in explaining life in the physical world.
The living structure quality is an inconsistent predictor for urban
life. Ht correlated positively with Twitter and POI density (p =
0.000) in model 7 and 8. The result in Model 7 agree with Jiang and
Yin (2019) that Ht can predict the location of tweets. However, the
R
2
values for model 7 and 8 were much lower than those of Model 2,
3, and 4, suggesting that Ht was perhaps a weaker predictor for urban
life than
α
and Bt. Also, we found no correlations between Ht and
walking trips in model 9 (p =0.157).
The semilattice indicator
α
contributed independently to urban life,
which is a strong support for Alexander. Table 3 summarized the results
of three multivariate models of regressing urban life indicators (i.e.,
Twitter, POI, walking trips) on
α
, while holding urban form and socio-
demographic confounders constant.
α
was found to have associated
positively with Twitter density (p =0.001), POI density (p =0.000), and
walking trip (p =0.000) in all three models with controls.
α
out-
competed the dwelling density in Model 1 and it also outcompeted
mixed use and street intersection density in Model 3, so that these urban
form confounders became insignicant with
α
in the models. We found
no spatial autocorrelation in the above regression models in four cities
(Morans I =-0.424, p =1.329), which justied the use of OLS regres-
sion. A multi-collinearity check yielded high VIF values (>10) for var-
iables of building age, household size and home ownership (Bdage , Bdstdv,
Hsize,Operc ), hence they were dropped from analysis to ensure the sta-
bility of regression results.
We found that a new (planned) district, on average, is less active on
twitter compared with old ones, as it is shown in the negative coefcient
of the New District dummy variable in Model 1 of Table 3 (p =0.003).
There is more in a new (planned) district that is punitive to Twitter
users, aside from the street network, mixed use, density, and block size
already in control, that our model did not capture. This is an important
nding since the lack of vitality is a common pitfall for new town
development, and placed-based social media activities is increasingly
viewed as a planning success. The semilatticestructure contributed
positively to urban life in both old and new districts, with or without
controlling the New District dummy, suggesting that such structure-life
relationship is consistent.
Additional tests were conducted on the predictive power of Bt and Ht
on urban life, after controlling the same confounders (Table 4). Bt
associated positively with Twitter activities (p =0.000), but not with
Table 2
Regressing Twitter and POI density (logarithmic) on Semilattice, Complex
Networkand Living Structureindicators for the 168 urban districts in four
cities.
Regression
Coefcient
(p-value)
R
2
Urban Life in
Virtual World
Urban Life in Physical world
Twitter Density
(Logarithmic)
POI Density
(Logarithmic)
Walking Trip
Density
Semilattice-
Tree(
α
)
Model 18.593
(0.000)
***
R
2
=0.184
Model 212.833
(0.000)
***
R
2
=0.712
Model
313.081
(0.000)
***
R
2
=0.570
Complex
Network(Bt)
Model 41.932
(0.000)
***
R
2
=0.364
Model 55.738
(0.002)
**
R
2
=0.055
Model 60.387
(0.073)
R
2
=0.019
Living
Structure(Ht)
Model 70.667
(0.000)
***
R
2
=0.139
Model 80.437
(0.000)
***
R
2
=0.104
Model 9 0.157
(0.190)
R
2
=0.010
* p <0.05 **p <0.01 ***p <0.001.
Table 3
Regressing urban life indicators on
α
, controlling for urban form and socio-demographic confounders.
Model 1
Twitter Density
Model 2
POI Density
Model 3
Walking Trip Density
Dependent Variable Coef. (p-value) Coef. (p-value) Coef. (p-value)
Meshedness Coefcient (
α
) 5.514 (0.001)*** 10.593 (0.000)*** 7.778 (0.000)***
Urban Form Confounders
Land Use Mixture, ratio 3.698 (0.000)*** 1.923 (0.000)*** 0.482 (0.327)
Dwelling Density, 10
6
/km
2
2.684 (0.781) 10.342 (0.047)* 60.219 (0.000)***
Street Intersection Density, #/km
2
0.006 (0.002)** 0.004 (0.000)*** 0.002 (0.147)
Percentage of Park, % 0.007 (0.284) 0.005 (0.170) 0.006 (0.305)
New District (New =1, Old =0) 0.839 (0.003)** 0.031 (0.832) 0.325 (0.152)
Economic Attributes
Median Household Income, 10
6
US$/yr. 10.883 (0.016)* 5.885 (0.015)* 8.0857 (0.030)*
1st Demographic group, %
1
0.023 (0.003)** 0.008 (0.048)* 0.001 (0.859)
2nd Demographic group, %
2
0.024 (0.001)** 0.002 (0.955) 0.009 (0.126)
Perc. of Old (>65), % 0.097 (0.000)*** 0.002 (0.903) 0.0052 (0.815)
Perc. of Young, %
3
0.111 (0.000)*** 0.038 (0.016)* 0.022 (0.368)
Intercept 6.370 (0.000)*** 0.460 (0.492) 1.976 (0.056)
Number of obs. 168 168 168
R-squared 0.714 0.857 0.741
* p <0.05 **p <0.01 ***p <0.001.
1
White for New York and London, Cantonese-speaking group for Hong Kong, and Polish-speaking group for Gdansk.
2
Black and Asian for New York and London, English-speaking group for Hong Kong, and English-speaking group for Gdansk.
3
<18 for New York and Gdansk, <15 for London, <14 for Hong Kong.
J. Huang et al.
Landscape and Urban Planning 226 (2022) 104469
10
POI (p =0.112) nor walking (p =0.165), suggesting it can predict life in
the virtual world but not in the real world. We found no statistically
signicant associations between Ht and all three urban life indicators
(Model 79), contrary to earlier results in Table 2 where Ht associated
positively with Twitter and POI density without controls. Our ndings
suggest that such association is not robuest, the Ht effect may disappear
after controlling for urban form and socio-economic confounders. The
inconsistency raised doubt on the reliability of living structureand its
mathematical expression (Ht) in explaining urban life.
5.4. Generalizability
Alexanders theory can better explain observations in London and
New York, less so in Hong Kong, a former British colony, and much less
so in Gdansk. The results of city-by-city multivariate regressions,
including the regression coefcient for
α
and R
2
values, are summarized
in Table 5.
α
contributed positively (p-values within the range of 0.000
0.034) to Twitter, POI, and walking trips in London and New York
(Model 16), suggesting observations from the two cities strongly sup-
port Alexander. In Hong Kong,
α
correlated positively with Twitter (p =
0.000) and POI (p =0.001), but not with walking trips (Model 79). In
Gdansk (Model 1012),
α
was predictive of only walking trips (p =
0.008), not Twitter (p =0.54) nor POI (p =0.133), suggesting evidence
from the Polish city did not t Alexanders theory well.
5.5. Discussion
This study marked a rst attempt to test Alexanders urban structural
theory under a comprehensive framework. Overall, our results sup-
ported his early works outlined in A City Is Not A Tree(Alexander,
1965) and Notes on the Synthesis of Form(Alexander, 1966). An old
urban district contains overlapping structures in resemblance to a
semilatticeand complex network, more so than new districts. This
structural difference can explain variations in urban life, independent of
other well-known confounders such as mixed use, density, and block
size, etc. A well-connected, complex city encourages human movement,
and a by-product to this is increased social interactions, captured in both
Twitter and POI datasets. In this sense, Alexander is a pioneer in
observing, documenting, and theorizing the fragmentation of street
network and its socio-behavioral consequences.
Our ndings diverged from his later work (Alexander, 2002) and
follow up literature. We found no supporting evidence for Ht as an
effective measure of living structure. Ht in historical urban cores were
not higher than those of new districts, contrary to de Rijke et al., (2020).
We found the relationships between Ht and the location geo-coded tweet
were not robust, which disappeared after controlling urban form and
socio-economic confounders. More evidence is needed on Ht and Alex-
anders construct of living structure.
We uncovered a possible British-American predilection of Alexan-
ders theory, which is well-supported by observations from London and
New York, less so in Hong Kong and Gdansk. We speculate that Alex-
anders observations weighed heavily upon British and American cities
in the 1950s and 60s. Hong Kongs new towns, however, began late after
the 1970s, when the city resisted car-oriented suburban sprawl and
embraced high-density, transit-oriented development (Cervero and
Murakami, 2009). Gdansk adhered to the gridiron street pattern
throughout its post-war development, its suburbanization began only
after Polands transition to the market economy in the 1990s (Mazur-
kiewicz and Zupanˇ
ciˇ
c, 2015). Evidence from both Hong Kong and
Gdansk might not have reached Alexander in time, if at all, for him to
adjust the theory. A locally-sensitive approach is therefore needed to
anchor case study research and planning practices, while Alexanders
theory should be interpreted with caution outside of the British-
American context.
The analytical tools developed can be of value for urban planning.
α
and Bt, as exemplied in this study, can serve as performance indicators
concerning urban life as a goal. Fig. 7 and Fig. 8 provide two examples,
in which
α
and Bt are plotted against POI and Twitter density.
α
is a
strong predictor for POI density, and Bt for geo-coded tweets. Urban
districts with the highest
α
and Bt, such as the Midtown Business District
in Manhattan, the Srodmie´
scie (Main Town) in Gdansk, Mong Kok and
Central in Hong Kong, are also among the most lively in both the real
and virtual world.
Lessons can be drawn on street network planning. Apparently, our
evidence suggest that a compact, well-connected street network pro-
motes urban life in both old and new districts, and it should therefore be
encouraged in practice. Cul-de-streets, three-way intersections are to be
discouraged since they reduce the link-to-node ratio and
α
. Further, our
ndings suggest that perhaps a semilattice structure can be planned,
exemplied by some of the liveliest districts in Fig. 9. The gridiron street
patterns in Midtown Business District in Manhattan was a ‘planned
outcome from the 19th century (Bridges, 1811), which rivaled the even
older districts such as the Battery Park City and Greenwich Village in
both
α
and life. This may serve as a counter example to Alexanders view
that a lively city can never be planned; it can and had been done before,
although not in the manner of the modernist urban planning against
which he rebelled.
The research framework we adopted is reductionist in nature, which
is an important limitation. It would be infeasible, nor to our intention, to
capture the richness of Alexanders theory in its entirety. The indicators
Table 4
Regressing urban life indicators on complex network and living structure
indicator for the 168 urban districts in four cities, controlling urban form and
socio-economic confounders.
Independent
Variable
Dependent Variable (Urban Life Indicators)
Twitter Density
(Logarithmic)
POI Density
(Logarithmic)
Walking Trip
Density
Semilattice(
α
) Model 1
5.514(p =0.001)***
R
2
=0.714
Model 2
10.593(p =0.000)
***
R
2
=0.857
Model 3
7.778(p =
0.000)***
R
2
=0.741
Complex
Network(Bt)
Model 4
1.398(0.000)***
R
2
=0.753
Model 5
0.288(0.112)
R
2
=0.737
Model 60.307
(0.165)
R
2
=0.670
Living Structure
(Ht)
Model 70.149
(0.153)
R
2
=0.661
Model 80.082
(0.256)
R
2
=0.735
Model 9
0.151
(0.087)
R
2
=0.695
* p <0.05 **p <0.01 ***p <0.001.
Table 5
Regressing urban life indicators on
α
for the 168 urban districts in four cities,
controlling for urban form and socio-demographic confounders.
Dependent Variables
Independent
Variable
Twitter Density
(Logarithmic)
POI Density
(Logarithmic)
Walking Trip
Density
α
in London Model 119.028
(0.002)
**
R
2
=0.941
Model 29.909
(0.004)
**
R
2
=0.948
Model 37.604
(0.001)
***
R
2
=0.941
α
in New York Model 46.498
(0.034)
**
R
2
=0.758
Model 54.044
(0.000)
***
R
2
=0.869
Model 62.368
(0.007)
**
R
2
=0.990
α
in Hong Kong Model 728.276
(0.000)
***
R
2
=0.858
Model 812.852
(0.001)
***
R
2
=0.896
Model 96.428
(0.067)
R
2
=0.854
α
in Gdansk Model 103.310
(0.542)
R
2
=0.501
Model 113.731
(0.133)
R
2
=0.762
Model 121.069
(0.008)
**
R
2
=0.967
* p <0.05 ** p <0.01 *** p <0.001.
J. Huang et al.
Landscape and Urban Planning 226 (2022) 104469
11
Fig. 7. Scatterplot of urban life in the physical world (POI density) and
α
for the 168 urban districts in four cities.
Fig. 8. Scatterplot of urban life in the virtual world (Twitter density) and
Bt for the 168 urban districts in four cities.
Fig. 9. Examples of urban neighborhoods resembling semilatticeand high levels of urban life.
J. Huang et al.
Landscape and Urban Planning 226 (2022) 104469
12
of urban lifewere intermediate measures, and can only be so, since the
patina of lifeis an unmeasurable outcome. We have also left out other
important concepts such as beauty. The two-dimentional graph-rep-
resentation of urban structure might be awed, which cannot gauge the
social, economic, and cultural connectivity among various urban ele-
ments and their contributions to life.
Our study is also limited in several technical aspects. The Twitter
database is vulnerable to sampling biases, i.e., over-sampling the young,
technology-savvy, and male, while under-representing the elderlies,
low-incomers, or minorities as reported by literature (Rost et al., 2013;
Wang et al., 2016). Similarly, POI and travel demand survey data may
also suffer from the missing data bias. These drawbacks may be
compensated, to a certain extent, by combining multiple data sources
and cross-checking the results with each other, as demonstrated in our
analysis.
As a postscript, we do not expect our ndings to be used to validate
or falsify Alexanders ideas; those are based on his observations made six
decades ago. Our societies have since been profoundly transformed, so is
the way people live, work and socialize with each other. Alexander
might not have guessed that digital technologies lend powerful mean-
ings to monitor, analyse, and understand cities and people today with
precision and granularity. The observations presented in this study are
and should be of relevance at present, not in retrospect. Going forward,
Alexanders work should continue to inspire, except that the eld has
now converged with multiple burgeoning domains of scientic research
(Oakes, Forsyth, & Schmitz, 2007; Berrigan, Pickle, & Dill, 2010; Ewing
& Cervero, 2010; Fang, He & Wang, 2021). The research landscape has
been immersed in new theories, new data and new tools of growing
sophistications. To strengthen the scientic rigor of our discipline, one
might as well do what we did in this paper, to bring powerfully-narrated
yet rarely-tested texts into a system of well-supported hypotheses. We,
the community of researchers and professionals, bear responsibility to
continuously rene and rebuild urban planning theory, using the best
data and analytical tools now at our disposal.
6. Conclusion
This study marks a rst attempt to test Alexanders urban structural
theory under a comprehensive research framework. Our evidence cor-
roborates Alexanders early work that a semilattice-shaped street
network is conducive to life. The intra-city structural distinctions be-
tween old and new districts can be found in consistency across all four
cities. In contrast, we found inconsistent correlations between urban life
and the Ht Index, suggesting a need to further investigate Alexanders
living structure construct and the use of Ht Index as its performance
indicator. Alexanders theory can better explain evidence in London and
New York than in Hong Kong and Gdansk. A locally-sensitive approach
is therefore needed to anchor case study research and planning prac-
tices. Our study contributes new evidence to research literature on street
network, travel behaviors and subsequent socio-behavioral impact. The
analytical protocols are of value for urban planners and researchers in
the four cities and elsewhere. The practical message is that the Mesh-
edness Coefcient and the Node-Based Average Betweenness can serve
as performance indicators in both new town and urban renewal de-
velopments, concerning urban life as a goal.
Acknowledgements
The work was supported in part by the National Natural Science
Foundation of China (No 51978594), the Hong Kong Research Grants
Council Theme-Based Research Scheme under Grant T32-101/15-R, the
General Research Fund under Project 11209717 and the SEED grant
from the Faculty of Architecture, the University of Hong Kong.
References
AASHTO (2018) A Policy on Geometric Design of Highways and Streets. 6th edn, American
Association of State Highway and Transportation Ofcials. 6th edn. Washington DC:
AASHTO. Available at: www.transportation.org.
ACO. (1946). Reviewed work: Greater London Plan, 1944 by Patrick Abercrombie. The
Geographical Journal, 108(1/3), 9394. https://doi.org/10.2307/1789338
Alexander, C. (1965). A city is not a tree. Architectural Forum, 122(1), 5862.
Alexander, C. (1966). Notes on synthesis of form (second pri). Cambridge, MA, USA:
Harvard University Press.
Alexander, C. (1987). A new theory of urban design. New York: Oxford University Press.
Alexander, C. (2002). The nature of order: An essay on the art of building and the nature of
the universe, Book 1 The phenomenon of life (1st ed.). Center for Environmental
Structure.
Alexander, C. (1977). A Pattern Language: Towns, Buildings, Construction. New York:
Oxford University Press.
Ball, K., et al. (2001). Perceived environmental aesthetics and convenience and company
are associated with walking for exercise among Australian adults. Preventive
Medicine, 33(5), 434440. https://doi.org/10.1006/pmed.2001.0912
Barrington-Leigh, C., & Millard-Ball, A. (2019). A global assessment of street-network
sprawl. PLoS ONE, 14(11). https://doi.org/10.1371/journal.pone.0223078
Batty, M. (2006). Hierarchy in cities and city systems, in Hierarchy in Natural and Social
Sciences. doi: 10.1007/1-4020-4127-6_7.
Batty, M. (2015). ‘Alexanders challenge: Beyond hierarchy in city systems and systems
of cities. A City Is Not a Tree (50th Anniversary ed.). Portland, OR, USA: Sustasis
Press.
Berrigan, D., Pickle, L. W., & Dill, J. (2010). Associations between street connectivity and
active transportation. International Journal of Health Geographics, 9, 20. https://doi.
org/10.1186/1476-072X-9-20
Bettencourt, L. (2016). The complexity of cities and the problem of urban design. In
M. W. Mehaffy (Ed.), A city is not a tree (50th Anniversary ed., pp. 4761). Portland,
OR, USA: Sustasis Press.
Bridges, W. (1811) Map of the city of new york and island of manhattan with explanatory
remarks and references. New York, USA.
Brush, J. E. (1966) ‘Walter Christaller. Central Places in Southern Germany. Translated
by Carlisle W. Baskin. Pp. 230. Englewood Cliffs, N.J.: Prentice-Hall, 1966. $9.95,
The ANNALS of the American Academy of Political and Social Science, 368(1), pp.
187187. doi: 10.1177/000271626636800132.
Buhl, J., et al. (2004). Efciency and robustness in ant networks of galleries. European
Physical Journal B, 42(1), 123129. https://doi.org/10.1140/epjb/e2004-00364-9
Calthorpe, P. (2004). The next American metropolis, The sustainable urban development
reader, p. 73.
Cervero, R. and Murakami, J. (2009). Rail and property development in Hong Kong:
Experiences and extensions, Urban Studies, 46(10), pp. 20192043. doi: 10.1177/
0042098009339431.
Chen, T., et al. (2019). Identifying urban spatial structure and urban vibrancy in highly
dense cities using georeferenced social media data. Habitat International, 89, Article
102005. https://doi.org/10.1016/j.habitatint.2019.102005
Cogd. (2020). Database of census records by department of population register. Gdansk,
Poland: City Ofce of Gdansk. Available at: https://gdansk.stat.gov.pl/en/
publications/population/demographic-situation-of-pomorskie-voivodship-in-
2020,2,4.html.
Cooper, C. (2013). Detailed measure descriptions of sDNA. UK: Cardiff.
CRPNYIE (1929). Regional Plan of New York and Its Environs Volume I: The Graphic
Regional Plan, Regional Plan Association. Philadelphia: Regional Plan of New York and
Its Environs. Available at: https://rpa.org/work/reports/regional-plan-of-new-york-
and-its-environs (Accessed: 1 January 2021).
Cuthbert, A. R. (2007). Urban design: Requiem for an era Review and critique of the
last 50 years. Urban Design International, 12(4), 177223. https://doi.org/10.1057/
palgrave.udi.9000200
Davis, H. (2021). Christopher Alexander and Bill Hillier: Overlaps and Divergences. United
Kingdom: UCL. Available at: Https://vimeo.com/506091135.
DCP (2018). Census Demographics at the NYC City Council district (CNCLD) level. Available
at: https://data.cityofnewyork.us/City-Government/Census-Demographics-at-the-
NYC-City-Council-distri/ye4r-qpmp (Accessed: 1 May 2020).
DCP (2020). PLUTO and MapPLUTO. Available at: https://www1.nyc.gov/site/planning/
data-maps/open-data/dwn-pluto-mappluto.page (Accessed: 30 May 2020).
Ding, R., et al. (2015). Complex network theory applied to the growth of Kuala Lumpurs
public urban rail transit network. PLoS ONE, 10(10). https://doi.org/10.1371/
journal.pone.0139961
DOT (2019) 2019 Citywide Mobility Survey Results. New York. Available at: https://
www1.nyc.gov/html/dot/downloads/pdf/nycdot-citywide-mobility-survey-report-
2019.pdf.
Dovey, K., & Pafka, E. (2016). The science of urban design? Urban Design International, 21
(1), 110. https://doi.org/10.1057/udi.2015.28
Ewing, R., & Cervero, R. (2010). Travel and the built environment: A meta-analysis.
Journal of the American Planning Association, 76(3), 265294.
Fang, C., He, S., & Wang, L. (2021). Spatial characterization of urban vitality and the
association with various street network metrics from the multi-scalar perspective.
Frontiers in Public Health, 9, Article 677910. https://doi.org/10.3389/
fpubh.2021.677910
Franca, U., et al. (2015). Visualizing the ‘heartbeat’” of a city with tweets. Complexity,
21(6), 280287. https://doi.org/10.1002/cplx
Freeman, Linton (1977). A Set of Measures of Centrality Based on Betweenness.
Sociometry, 40(1), 3541. https://doi.org/10.2307/3033543
J. Huang et al.
Landscape and Urban Planning 226 (2022) 104469
13
Gabriel, R., & Quillien, J. (2019). A search for beauty/A struggle with complexity:
Christopher Alexander. Urban Science. https://doi.org/10.3390/urbansci3020064
GDA (2016). Gdansk Research on mobility 2016 (Gda´
nskie Badanie Ruchu). Gdansk,
Poland. Available at: https://www.brg.gda.pl/attachments/article/282/Raport-III.
pdf.
GDA (2018). Study of conditions and directions of spatial development in the City of
Gda´
nsk adopted by the City Council on 23 April 2018 (Studium uwarunkowa´
n I
kierunk´
ow rozwoju przestrzennego miasta Gda´
nska). Gdansk, Poland. Available at:
https://www.brg.gda.pl/planowanie-przestrzenne/studium-uwarunkowan-i-
kierunkow-zagospodarowania-przestrzennego.
GLA (2018). London Borough Proles. Available at: https://londondatastore-upload.s3.
amazonaws.com/instant-atlas/borough-proles/atlas.html (Accessed: 18 June
2020).
Go-Globe (2015). Social Media Use in Hong Kong - Statics and Trends, Go-Globe. Available
at: http://www.go-globe.hk/blog/social-media-hong-kong/ (Accessed: 10
September 2016).
HA (1999). Design Manual for Roads and Bridges, Standardsforhighways.Co.Uk.
Birmingham.
Harary, F., & Rockey, J. (1976). A city is not a semilattice either. Environment and
Planning A, 8(4), 375384.
Hillier, B. (1989). The architecture of the urban object. Ekistics, 56(334335), 521.
HKCSD (2017). The 2016 Hong Kong Population By-census. Available at: https://www.
bycensus2016.gov.hk/en/bc-dp.html (Accessed: 1 September 2020).
HKPD (2017). Tertiary Planning Units. Available at: http://www.census2011.gov.hk/en/
tertiary-planning-units.html (Accessed: 28 December 2017).
Hofmeister, B. (1970). Anglo-Americas great cities: Major characteristics, recent trends,
regional variations. Geoforum, 1(3), 1729.
Huang, J. et al. (2021). The image of the City on social media: A comparative study using
Big Dataand Small Datamethods in the Tri-City Region in Poland, Landscape and
Urban Planning, 206, p. 103977. doi: 10.1016/j.landurbplan.2020.103977.
ITE. (1984). Recommended guidelines for subdivision streets: A recommended practice of the
Institute of Transportation Engineers. Washington DC: Institute of Transportation
Engineers.
Jacobs, J. (1961). The death and life of great American Cities. New York: Random House.
Jiang, B., & Ren, Z. (2019). Geographic space as a living structure for predicting human
activities using big data. International Journal of Geographical Information Science, 4
(33), 764779. https://doi.org/10.1080/13658816.2018.1427754
Jiang, B., & Yin, J. (2014). Ht-index for quantifying the fractal or scaling structure of
geographic features. Annals of the Association of American Geographers, 104(3),
530540. https://doi.org/10.1080/00045608.2013.834239
Johnson, M. L., et al. (2019). Mapping urban park cultural ecosystem services: A
comparison of twitter and semi-structured interview methods. Sustainability
(Switzerland), 11(21), 6137. https://doi.org/10.3390/su11216137
Katz, P. (1993). The new urbanism: Toward an architecture of community. New York:
McGraw-Hill Education.
Kerr, J., et al. (2007). Urban form correlates of pedestrian travel in youth: Differences by
gender, race-ethnicity and household attributes. Transportation Research Part D:
Transport and Environment, 12(3), 177182. https://doi.org/10.1016/j.
trd.2007.01.006
Lau, J., Collier, N., & Baldwin, T. (2012). On-line trend analysis with topic models:
#Twitter trends detection topic model online. Available at: International Conference
on Computational Linguistics (COLING), 15191534 Https://www.aclweb.org/ant
hology/C/C12/C12-1093.pdf.
Long, Y. and Huang, C. (2017).‘oes block size matter? The impact of urban design on
economic vitality for Chinese cities, Environment and Planning B: Urban Analytics and
City Science. doi: https://doi.org/10.1177/2399808317715640.
Loo, B. P. Y., & Chow, A. S. Y. (2008). ‘Changing urban form in Hong Kong: What are the
challenges on sustainable transportation? International Journal of Sustainable
Transportation Taylor & Francis, 2(3), 177193. https://doi.org/10.1080/
15568310701517331
Lorens, P., Kamrowska-Załuska, D., & Kostrzewska, M. (2014). Urban transformations of
gdansk bay metropolitan area. Urban Design, 130, 3033.
Lu, Y. (2019). Using Google Street View to investigate the association between street
greenery and physical activity. Landscape and Urban Planning. https://doi.org/
10.1016/j.landurbplan.2018.08.029
Marshall, S. (2005). Streets and patterns (1st ed.). New York: Spon Press.
Marshall, S. (2012). Science, pseudo-science and urban design. Urban Design
International. Nature Publishing Group, 17(4), 257271. https://doi.org/10.1057/
udi.2012.22
Marshall, W. E., & Garrick, N. W. (2010). Effect of street network design on walking and
biking. Transportation Research Record, 2198, 103115. https://doi.org/10.3141/
2198-12
Mazurkiewicz, W. M., & Zupanˇ
ciˇ
c, T. (2015). Sensitivity of public places in Gdansk
Osowa district. Journal AR Architecture, 1623.
Mehaffy, M. W. (2019). Assessing Alexanders later contributions to a science of cities.
Urban Science, 3(2), 59. https://doi.org/10.3390/urbansci3020059
Mohajeri, N., French, J. R., & Batty, M. (2013). Evolution and entropy in the organization
of urban street patterns. Annals of GIS, 19(1), 116.
Moran, P. A. (1950). Notes on continuous stochastic phenomena. Biometrika, 37(12),
1723. https://doi.org/10.1093/biomet/37.1-2.17
Morstatter, F., et al. (2013). Is the sample good enough? Comparing Data from twitters
streaming API with twitters rehose. In International Conference on WEblogs and
Social Media (pp. 400408). AAAI, 10.1007/978-3-319-05579-4_10.
De Nadai, M., et al. (2016). The death and life of great Italian cities: A mobile phone data
perspective. Proceedings of 26th International ACM Conference on World Wide Web
(WWW), 10.1145/2872427.2883084.
NCGCD (2020). Database of topographic objects (Baza Danych Obiekt´
ow Topogracznych
BDOT10k). Available at: https://www.geoportal.gov.pl/ dane/baza-danych-
obiektow-topogracznych-bdot (Accessed: 27 July 2021).
NYC OpenData (2017). NYC Street Centerline (CSCL). Available at: https://data.
cityofnewyork.us/City-Government/NYC-Street-Centerline-CSCL-/exjm-f27b
(Accessed: 1 May 2021).
Oakes, J. M., Forsyth, A., & Schmitz, K. H. (2007). The effects of neighborhood density
and street connectivity on walking behavior: The Twin Cities walking study.
Epidemiologic Perspectives & Innovations, 4(1), 16. https://doi.org/10.1186/1742-
5573-4-16
ONS (2010) Land use statistics (Generalised Land Use Database), Ofce for National
Statistics. Available at: https://data.gov.uk/dataset/4413c03c-762a-47ad-a865-
6c1eee77fe6c/land-use-statistics-generalised-land-use-database (Accessed: 18 June
2020).
OpenStreetMap (2019).
Park, Y., & Newman, G. D. (2017). A framework for place-making using Alexanders
patterns. Urban Design International, 22(4), 349362. https://doi.org/10.1057/
s41289-017-0040-1
Pisati, M. (2001). sg162: Tools for spatial data analysis. Stata Technical Bulletin, 60,
2137.
PNT (2008) Global positioning system standard positioning service performance standard.
Washington DC. Available at: https://www.gps.gov/technical/ps/2008-SPS-
performance-standard.pdf.
Porta, S., Crucitti, P., & Latora, V. (2006). The network analysis of urban streets: A primal
approach. Environment and Planning B: Planning and Design. https://doi.org/10.1068/
b32045
de Rijke, C. A., et al. (2020). Living structure as an empirical measurement of city
morphology. ISPRS International Journal of Geo-Information, 9(11), 677. https://doi.
org/10.3390/ijgi9110677
Rodrigue, J. P., Comtois, C. and Slack, B. (2016) The geography of transport systems, The
Geography of Transport Systems. London: Taylor & Francis Ltd. doi: 10.4324/
9781315618159.
Rost, M. et al. (2013). Representation and Communication: Challenges in Interpreting
Large Social Media Datasets. In Proceedings of the 2013 Conference on Computer
Supported Cooperative Work. San Antonio, TX, pp. 357362. doi: 10.1145/
2441776.2441817.
Salingaros, N. (2018). Fractals and Christopher Alexanders Fifteen Fundamental
Properties. Conscious Cities Anthology, 2018(1). https://doi.org/10.33797/cca18.04
Salingaros, N. A. (2005). Principles of urban structure. Amsterdam, Netherlands: Techne
Press.
Spanier, E. H. (1981) Algebraic topology, The Mathematical Gazette. New York: Springer.
doi: https://doi.org/10.1007/978-1-4684-9322-1.
Steadman, P. (2004). Editorial: Developments in space syntax. Environment and Planning
B: Planning and Design, 31(4), 483486.
Sung, H., Lee, S., & Cheon, S. (2015). Operationalizing Jane Jacobss urban design
theory: Empirical verication from the great city of Seoul, Korea. Journal of Planning
Education and Research, 35(2), 117130. https://doi.org/10.1177/
0739456X14568021
TfL (2019) Transport of London Road Network Dataset. Available at: https://data.gov.uk/
dataset/101434ec-c34d-486e-89dc-8d0714629e5b/t-road-network%0A (Accessed:
1 May 2021).
Thompson, C. G., et al. (2017). Extracting the variance ination factor and other
multicollinearity diagnostics from typical regression results. Basic and Applied Social
Psychology, 39(2), 8190. https://doi.org/10.1080/01973533.2016.1277529
UK Parliament (1963) London Government Act 1963. United Kingdom. Available at:
https://www.legislation.gov.uk/ukpga/1963/33/contents.
US Census Bureau (2020) American Community Survey 1-Year Estimates, Table S1903
(2005-2018). Available at: https://data.census.gov/ (Accessed: 1 May 2020).
Wang, G. et al. (2016) Will check-in for badges: Understanding bias and misbehavior
on location-based social networks, in Proceedings of the 10th International Conference
on Web and Social Media, ICWSM 2016.
Wang, Z., et al. (2018). Comparing social media data and survey data in assessing the
attractiveness of Beijing Olympic Forest Park. Sustainability (Switzerland), 10(2), 382.
https://doi.org/10.3390/su10020382
Xie, F., & Levinson, D. (2007). Measuring the structure of road networks. Geographical
Analysis, 39, 336356. https://doi.org/10.1111/j.1538-4632.2007.00707.x
Ye, Y., Li, D., & Liu, X. (2018). How block density and typology affect urban vitality: An
exploratory analysis in Shenzhen, China. Urban Geography, 39(4), 631652. https://
doi.org/10.1080/02723638.2017.1381536
Yue, H., & Zhu, X. (2019). Exploring the relationship between urban vitality and street
centrality based on social network review data in Wuhan, China. Sustainability
(Switzerland). https://doi.org/10.3390/su11164356
Zukin, S. (2010) Naked city: The death and life of authentic urban places. New York: Oxford
University Press. doi: 10.17323/1726-3247-2018-1-62-91.
J. Huang et al.
... Huang et al. tested Alexander's urban structural theory under a comprehensive research framework utilizing a combination of Twitter activities, Points-Of-Interest, and walking trips, and found no independent associations between "living structure" and life, contrary to existing literature. They also suggested the need for a locally-sensitive approach in future studies [40]. ...
Article
Full-text available
Following the rapid development and urbanization in China over recent decades, sustainable renovation of urban residences has become an important issue. This study aimed to develop an architectural teaching program integrating the study of local climate, vernacular buildings, climate-responsive design strategies, and simulation tools. A local house in a high-density neighborhood in Guangzhou, China, in a hot-humid climate area was selected for renovation in this teaching program. Investigations of the urban neighborhood development, the construction and climate-responsive characteristics of the vernacular houses, long-term thermal environment characteristics, and sustainable design strategies were conducted before the design project began. The guidelines of Active House combining passive strategies and active technologies were incorporated into the concept design. The students’ works represented their understanding of the characteristics of a hot-humid climate, responsive strategies for the local buildings, preliminary methodologies of micro-climate analysis, and technologies supporting sustainable building design. Thus, this program provided a valuable approach to the appropriate pedagogy for a research-integrated design studio within the context of sustainable architectural education development.
Article
Full-text available
Jane Jacobs (1961) theorized that four urban form conditions, namely, mixed use, short block, aged buildings, and density, are indispensable for the “exuberant diversity” and conducive to, or perhaps even determinant of, the success of a city district. Jacobs’ theory has been used widely as a reference point in case study research and policy and design prescriptions. We found five studies that attempted to test it more formally, using various performance indicators such as mobile phone activities, walking, crime and mortality. Their findings were inconsistent and unable to settle theoretical controversies. Questions remained as what performance indicators are most strongly associated with her urban form conditions? Are these conditions independently associated with desired outcomes or in combination and what are the interaction effects? Our study aimed to test Jacobs’ theory that urban form conditions contribute to the vitality and success of city districts. Jacobs’ urban form conditions were measured using GIS data for each of Hong Kong’s Tertiary Planning Unit. Performance outcomes were gauged using a combination of “new urban data”, comprising Twitter activities, sentiment tones, and Point-Of-Interest (POI), and “traditional data”, comprising walking commute, employment, and mortality. Urban context, income and demographic indicators were used as controls in fitting spatial regression models to predict measures of performances based on urban conditions. Results showed that Jacobs’ urban form conditions contribute positively to “vitality” indicators such as the density of tweets, walking trips, and POI, but not with “failure and success” indicators such as expressed sentiment on Twitter, employment, or mortality. Out findings suggest that her theory largely hold for Hong Kong, except that dwelling density should be substituted by building density, while tall buildings associated positively with desirable outcomes, contrary to Jacobs’ observation in the American context. More generally, we demonstrate how new urban data can be used to evaluate classic planning theories at scale.
Article
Full-text available
In the context of rapid urbanization in developing countries, the spatial organization of cities has been progressively restructured over the past decades. However, little has been done to understand how the physical expansion affected the reorganization of socioeconomic spaces in cities. This study explores the association between various street network metrics and urban vitality and how it changes across different scales using geographic big data through a case study of Wuhan, China. Urban vitality is characterized by four components: concentration, accessibility, livability, and diversity. The new technique of spatial design network analysis (sDNA) is employed to characterize street network metrics, including connectivity, closeness, betweenness, severance, and efficiency, with 16 localized network variables. Furthermore, the stratified spatial heterogeneity between street network metrics at multiple scales and the four components of urban vitality is investigated using the Geodetector tool. First, concentration, accessibility, and diversity decline with distance from the urban center, whereas livability has a fluctuating upward trend with distance from the urban core. Second, the correlation between street network characteristics and urban vitality is sensitive to different spatial scales. Third, connectivity explains the largest amount of the variance in urban vitality (over 40%), while both betweenness and closeness explain roughly 28% of urban vitality. Efficiency and severance contribute 22 and 10% to the spatial heterogeneity of urban vitality, respectively. The study sheds light on the mechanisms between street configurations and urban vitality from the multi-scalar perspective. Some implications are provided for the improvement of the streets' urban vitality.
Article
Full-text available
“The Image of the City” by Kevin Lynch is a landmark planning theory of lasting influence; its scientific rigor and relevance in the digital age were in dispute. The rise of social media and other digital technologies offers new opportunities to study the perception of urban environments. Questions remain as to whether social media analytics can provide a reliable measure of perceived city images? If yes, what implication does it hold for urban planners? This paper describes a study on the perception of city images using a combination of “big data” and “small data” methods in the Tri-City Region in Poland. The aims were to 1) test the hypothesis whether social media analytics can elicit Lynchian elements of city image in consistency with conventional methods, and 2) develop and evaluate social media-based indicators of Imageability for planning practice. Geo-tagged images and texts were collected from Instagram and Twitter, two popular social media platforms in Poland. Text-Mining, Image Processing, Clustering Analysis, Kernel Density Estimation, and Sentiment Analysis were used. Results were compared with benchmarks constructed from official GIS database, questionnaire responses and sketch maps. “District”, “landmark”, and “path” identified on social media were in good agreements with benchmarks, less so for “edge” and “node”. Two social media-based indicators have influenced the perception of a place: Instagramability, the frequency of a place captured on Instagram, was linked to its perception as an architectural landmark and tourist attraction, while Twitterability, the frequency of a place mentioned on Twitter by name, was linked to its perceived niceness and relevance to everyday life of communities. Methods developed in this study have theoretical and practical implications for urban planners.
Article
Full-text available
Human actions and interactions are shaped in part by our direct environment. The studies of Christopher Alexander show that objects and structures can inhibit natural properties and characteristics; this is measured in living structure. He also found that we have better connection and feeling with more natural structures, as they more closely resemble ourselves. These theories are applied in this study to analyze and compare the urban morphology within different cities. The main aim of the study is to measure the living structure in cities. By identifying the living structure within cities, comparisons can be made between different types of cities, artificial and historical, and an estimation of what kind of effect this has on our wellbeing can be made. To do this, natural cities and natural streets are identified following a bottom-up data-driven methodology based on the underlying structures present in OpenStreetMap (OSM) road data. The naturally defined city edges (natural cities) based on intersection density and naturally occurring connected roads (natural streets) based on good continuity between road segments in the road data are extracted and then analyzed together. Thereafter, historical cities are compared with artificial cities to investigate the differences in living structure; it is found that historical cities generally consist of far more living structure than artificial cities. This research finds that the current usage of concrete, steel, and glass combined with very fast development speeds is detrimental to the living structure within cities. Newer city developments should be performed in symbiosis with older city structures as a whole, and the structure of the development should inhibit scaling as well as the buildings themselves.
Article
Full-text available
Understanding the benefits received from urban greenspace is critical for planning and decision-making. The benefits of parks can be challenging to measure and evaluate, which calls for the development of novel methods. Crowdsourced data from social media can provide a platform for measuring and understanding social values. However, such methods can have drawbacks, including representation bias, undirected content, and a lack of demographic data. We compare the amount and distribution of park benefits elicited from (1) tweets on Twitter about Prospect Park, Brooklyn, New York (n = 451) with park benefits derived from (2) broad (n = 288) and (3) directed (n = 39) questions on two semi-structured interview protocols for park users within Prospect Park. We applied combined deductive and inductive coding to all three datasets, drawing from the Millennium Ecosystem Assessment’s (MEA) cultural ecosystem services (CES) framework. All three methods elicited an overlapping set of CES, but only the Twitter dataset captured all 10 MEA-defined CES. All methods elicited social relations and recreation as commonly occurring, but only the directed question interview protocol was able to widely elicit spiritual values. We conclude this paper with a discussion of tradeoffs and triangulation opportunities when using Twitter data to measure CES and other urban park benefits.
Article
Full-text available
Disconnected urban street networks, which we call “street-network sprawl,” are strongly associated with increased vehicle travel, energy use and CO2 emissions, as shown by previous research in Europe and North America. In this paper, we provide the first systematic and globally commensurable measures of street-network sprawl based on graph-theoretic and geographic concepts. Using data on all 46 million km of mapped streets worldwide, we compute these measures for the entire Earth at the highest possible resolution. We generate a summary scalar measure for street-network sprawl, the Street-Network Disconnectedness index (SNDi), as well as a data-driven multidimensional classification that identifies eight empirical street-network types that span the spectrum of connectivity, from gridiron to dendritic (tree-like) and circuitous networks. Our qualitative validation shows that both the scalar and multidimensional measures are meaningfully comparable within and across countries, and successfully capture varied dimensions of walkability and urban development. We further show that in select high-income countries, our measures explain cross-sectional variation in household transportation decisions, and a one standard-deviation increase in SNDi is associated with an extra 0.25 standard deviations in cars owned per household. We aggregate our measures to the scale of countries, cities, and smaller geographies and describe patterns in street-network sprawl around the world. Latin America, Japan, South Korea, much of Europe, and North Africa stand out for their low levels of street-network sprawl, while the highest levels are found in south-east Asia, the United States, and the British Isles. Our calculations provide the foundation for future work to understand urban processes, predict future pathways of transportation energy consumption and emissions, and identify effective policy responses.
Article
Full-text available
This study investigates the association between urban vitality and street centrality in Wuhan, China. Urban vitality was measured with social network review data. Street centrality was evaluated in terms of closeness, straightness, and betweenness in walking and driving mode. We constructed a square mesh to convert datasets of street centrality (segments) and urban vitality (points) into one unit for analysis. Geospatial visualization, a chi-square test, and correlation analysis were first employed to obtain an initial understanding of the spatial coupling relationship between urban vitality and street centrality. Then spatial regression models were applied to evaluate the significances and directions of the influences of street centrality on urban vitality. A geographical detector technique was utilized to further evaluate the magnitudes of these influences. The results suggest that street centrality plays an important role in shaping the spatial organization of urban vitality, and various street centralities presented marked differences in their association with urban vitality. More specifically, when considering street centrality in walking mode, betweenness affected urban vitality the most, followed by closeness and straightness. When considering street centrality in driving mode, straightness had the greatest influence on urban vitality, followed by closeness and betweenness.
Article
Full-text available
Beauty. Christopher Alexander’s prolific journey in building, writing, and teaching was fueled by a relentless search for Beauty and its meaning. While all around him the world was intent on figuring out how to simplify, Alexander came to embrace complexity as the only path to his goal. The Beauty and life of that which he encountered and appreciated—an Indian village, a city, a subway network, an old Turkish carpet, or a campus—lay in its well-ordered complexity. As a designer and maker he found that simplicity came from choosing—at every step—the simplest way to add the necessary complexity. The failure of so much of our modern world, in Alexander’s eyes, was oversimplification, wantonly bulldozing context, misunderstanding the relationships of part and whole, ignoring the required role of time in the shaping of shapes, and ultimately dismissing, like Esau, our birthright of Value in favor of a lentil pottage of mere Fact. Ever elusive, Beauty demands of her suitors a constant return of attention to see what might be newly revealed, and Alexander duly returned again and again in pursuit of the mystery. In this essay—essentially biographical and descriptive of one man’s endeavors—we examine the full arc of his work from dissertation to most recent memoir. We don’t shy away from his failures, and we don’t simplify his journey. We leave work done by other scholars for another day. We reach no conclusion, rather, we invite readers to reflect on what Alexander’s lifelong effort suggests to them about their own path, their own sense of aesthetics and order, innate cognitive shortfalls, and professional blind alleys.
Book
As cities have gentrified, educated urbanites have come to prize what they regard as "authentic" urban life: aging buildings, art galleries, small boutiques, upscale food markets, neighborhood old-timers, funky ethnic restaurants, and old, family-owned shops. These signify a place's authenticity, in contrast to the bland standardization of the suburbs and exurbs. But as Sharon Zukin shows in Naked City, the rapid and pervasive demand for authenticity--evident in escalating real estate prices, expensive stores, and closely monitored urban streetscapes--has helped drive out the very people who first lent a neighborhood its authentic aura: immigrants, the working class, and artists. Zukin traces this economic and social evolution in six archetypal New York areas--Williamsburg, Harlem, the East Village, Union Square, Red Hook, and the city's community gardens--and travels to both the city's first IKEA store and the World Trade Center site. She shows that for followers of Jane Jacobs, this transformation is a perversion of what was supposed to happen. Indeed, Naked City is a sobering update of Jacobs' legendary 1961 book, The Death and Life of Great American Cities. Like Jacobs, Zukin looks at what gives neighborhoods a sense of place, but argues that over time, the emphasis on neighborhood distinctiveness has become a tool of economic elites to drive up real estate values and effectively force out the neighborhood "characters" that Jacobs so evocatively idealized.
Article
Tracking human activities with social media (online social networks) and point of interest data to understand human dynamic distribution, behavior, and high-density urban environments is gaining importance in the domain of urban studies. Recently, social media data have proven to be a rich source of information, providing a novel way to derive urban spatial structures and their impact on the quality of life. Yet, integration of this wisdom in urban planning and policymaking has not been comprehensively investigated in high-density cities such as Hong Kong as it relates to spatial configurations. This study aims to investigate spatial structures and analyze social media data to apprise urban planning with knowledge of human activities. This study also seeks to introduce an exploratory analysis to develop a greater understanding of the interaction between social activities and urban space. The results show that function layout defines urban spatial structure and determines human social activities. The research provides insights regarding a better interpretation of the knowledge of social activities, underlying how well social activities reflect the corresponding urban spatial structure and gaining a detailed understanding of their respective variation in activities.