Characterizing street hierarchies through network analysis and large-scale taxi traffic flow: A case study of Wuhan, China

Liang Huang, Xinyan Zhu, Xinyue Ye, Wei Guo, Jiye Wang

Environment and Planning B: Planning and Design 0(0)-21, 2015

1. Research Questions

• What is the shape of the street hierarchy distribution based on network analysis and traffic flow with different times?

2. Theoretical background

• “Detecting street hierarchy helps understand the urban spatial structure and can shed light on the mechanism of urban development”.

3. Research methodology & Data

• Wuhan city data

3.1 Centrality measures

– “‘degree’ is connectivity that specifies the number of nodes that directly link a given node in a graph”.

– “‘betweenness’ is the number of times a node acts as a bridge along the shortest path between two other nodes, which reflects the intermediary location of a node”.

Where v_i, v_j, and v_k is distinct vertices, _jk(i) is the number of shortest path that pass through v_i, and _jk is number of shortest paths from v_j and v_k.

– UCINET software.

3.2 Power law distributions. (scale-free distributions)

– “When the probability of frequency of a particular value of some quantity varies inversely as a power of that value, the quantity is said to follow a power law (Newman, 2005)”.

– “The universality of the power law is that there are many small events and few large events”.

– “The straight line on the log-log plot is the signature of a power law”.

– ‘Pareto distribution’ consists of horizontal axis and P(X>x) on the vertical one.

3.3 Traffic flow distributions

– Wuhan city taxicabs data, GPS data (date, time, identification, longitude, latitude, velocity, driving direction (heading), and service status).

4. Results

Below Figure 2, and Figure7`s graph follow power laws. Also, as below Table 1, The connectivity and betweenness of Wuhan`s both natural and named streets are inclined to Top 20% rank. Even, bottom 80% rank has under average connectivity and betweenness. So, Author says that top 1% rank forms aorta of the street network of Wuhan.

As below Table 3, daily flow distribution among the streets are also inclined to Top20% rank. And also, graph of flow distribution follows power laws. As below Figure 12, Author says that there are obvious hour-to-hour changes in the percentage of hourly flow distribution. So he expects that these lead to traffic congestion, and argues that the street network of Wuhan needs to be improved.

(Connectivity & Betweenness & Daily flow 상위 20%에 집중됨–> Traffic congestion)

Size, connectivity, and tipping in spatial networks: Theory and empirics

Yuri Mansury, J.k. Shin

Computer, Environment and Urban System 54 (2015) 428-437

1. Research Questions & Hypothesis

• Is city`s connectivity related to population size? And how do these things have relationship?
• Is the overall size distribution of settlements related to city`s connectivity? And how do these things have relationship?

2. Introduction & Theoretical background

• The clustering of social contacts in cities suggests an intense pattern of interconnectedness. Also, density promotes the type of face-to-face contacts that create new ideas (Lobo & Strumsky, 2008) and facilitate the transmission of new knowledge (Henderson, 2007).
• It appears that large cities facilitate human interactions, and these in turn spur productivity (Granovetter, 2005).
• It appears that size and connectivity are positively correlated because a larger community provides more opportunities to form new linkages and maintain old relationships.
• Spatial constraints are relevant in social networks where distance matters, such as friendships and collaboration networks (Emmerich, Bunde, Havlin, Li, & Li, 2013).
• Nodes in most spatial models are assumed to be either randomly or uniformly distributed in steady sates (Eechkhout, 2004; Giesen, Zimmermann, & Suedekum, 2010). Thus, Simulating the connectivity impact of size requires a spatial network in which settlers interact on an uneven landscape.

3. Methodology & Data & Results

3.1 The network model (based on ‘Gravity Model’)

They assume that a pair of nodes can be joined by at most one link and that all existing links have equal weight. Interactions between the two cites(I and J) are measured like,

Where K_ij is a binary indicator = 1 if a link exists between node I and node J and = 0 otherwise.

Where <k>_Iⁱⁿ and <k>_I^out are the average number of intra-city and inter-city links, respectively, and m_I is city I`s population size. Then, K_Iⁱⁿ and K_I^out are determined by Gravity approach (Haynes & Fotheringham, 1984; Sen & Smith, 1995; Blumenfeld-Lieberthal & Portugali, 2012; Masucci, Serras, Johansson, & Batty, 2013).

Where d_IJ is the distance between two cities and C_x is a constant term (Gravity model).

Intra-city interactions also can be obtained from Gravity model. The log-log plot of connectivity against population size m converges to a straight line like

Where the power exponent can be viewed as the elasticity of intra city interactions and C₀ is constant term in intra-city and is smaller than or similar to 1 in inter-city.

These are consolidated by empirical support like below.

–>At this graph, Intra-city connectivity is more influenced by population size than inter-city connectivity. Also, Gravity forces are stronger in the smaller system.

(intra-city가 인구수의 영향 ↑, Smaller system에서 Gravity model 영향↑)

3.2 A model of a geometric network

– Geography matters in the real world for two reasons. First, cities are unevenly distributed across space, and their populations are clustered in urban agglomerations. Second, large cities tend to cluster with other cities while distance between cites increases.

They initialize the ABM with a random distribution of 100,00 agents across a two-dimensional 1000 X 1000 plane as considering the dispersion effect of moving costs and the centripetal force of increasing returns like below

And also, data is obtained like below

–>At this table, as urban growing, the number of cites decreases, average connectivity increases, and average external factor is reduced because interact with same-city neighbors vis-à-vis acquaintances increases.

(Urban Growing–>도시수↓, average connectivity↑, average external factor↓)

• Tipping point

– Micro-foundation tipping model can help them identify the influences of social processes on the spatial context.

– Tipping point: The time that incompatible views on social issues are completely converted to consensus through conversation.

Tipping point is expressed by formulation

Where y_i^t denotes agent i`s trait at time t, the trait retention rate, the diffusion rate.

Also, below data is obtained

–>, which shows that exponential distribution leads to a homogenous society, power laws leads to a pluralistic society and links are remained because agent`s trait convergences to a consensus in exponential distribution, and otherwise in power laws.

(exponential size distribution –>homogenous society, power law distribution –>pluralistic society)

–>Lastly, they randomly swap the locations of settlements, which weakens the local network in the largest city because swapping makes the minority trait extinct. So the city can be homogeneous society.

(location of settlement를 임의로 바꿈–>minority trait 없어짐–>local network ↓)

Tighe & Ganning (2016)

20141288 BAEK HAEIN

1. 논문 제목, 저자, 저널명
   Do Shrinking Cities Allow Redevelopment Without Displacement? An Analysis of Affordability Based on Housing and Transportation Costs for Redeveloping, Declining, and Stable Neighborhoods, J.Rosie Tighe & Joanna P. Ganning, Housing Policy Debate, 2016

   http://www.tandfonline.com/doi/abs/10.1080/10511482.2015.1085426

2. 연구질문
   As you can see in the title of the article, we are going to think about the affordability based not only on housing but also transportation costs for redeveloping, declining, and stable neighborhoods.
3. 이론적 배경
   “In weak market cities, excess real estate supply prior to downtown redevelopment is able to absorb new demand without displacing arts-based employment, an industry traditionally vulnerable to gentrification.” (p. 3)
   “Redevelopment often includes employment growth, which can reduce or eliminate commuting costs in surrounding neighborhoods.” (p. 3)
4. 연구가설
   -“We Hypothesize that such costs may be offset by improvements in access to public transportation, cycling, and walking options.” (p. 2)
   -Analyze sample of 83 shrinking cities’ block groups : “Whether neighborhood-level redevelopment in shrinking cities may support, rather than diminish, affordability when it is assessed as the cots of housing and transportation combined, rather than housing alone.” (p. 3)
5. 연구 방법론
   -Data : 2000 block-level population data to the 2010 block group boundaries, S. Census Bureau’s Longitudinal Employer-Household Dynamics
   LAI data
   -Unit of Analysis : Block Group level, Neighborhood level
   -Time : 2000-2010
   -Case
   High-High : high redevelopment index values and are surrounded by other neighborhoods with high values
   High-low : high index values that are surrounded by neighborhoods with low values
   Low-high : declining neighborhoods surrounded by redeveloping ones
   Low-low : declining neighborhoods surrounded by declining neighborhoods
   Non-significant block group
   -분석모델 : LAI (Location Affordability Index) model : estimate the total affordability for different income levels for every block group in the US, taking into account both housing and transportation costs.
6. 주요연구결과
   -Lower-income households pay a higher percentage of income.
   -Redevelopment neighborhoods housing has less affordable.
   -“regardless of income group, housing is least affordable in redeveloped neighborhoods, compared with stable or declining neighborhoods.” (p. 7)
   -Transportation costs : affordable in redevelopment neighborhoods.
7. Policy Implications
   Housing
   -HCV(Housing Choice Voucher) : Assistance to households earning as little as 30% of AMI(Area Median Income). High demand for affordable housing.
   Transportation
   -Improving walkability, safety, and infrastructure to support bike-oriented travel.
   -Public funding : bike/ped planning (project) ; facilitate access of areas holds potential.

1. 논문 제목, 저자, 저널명
   The Geography of the Recent Housing Crisis : The Role of Urban Form, Hongwei Dong & J. Andrew Hansz, Housing Policy Debate, 2016 (Vol.26, No.1, 151-171)

   http://www.tandfonline.com/doi/abs/10.1080/10511482.2015.1038575

2. 연구질문
   -“How the recent housing crisis is spatially distributed, and whether and how it is related to the geography of urban form, neighborhood composition, state foreclosure policies, and some other socioeconomic variables within and across American metropolitan areas.” (p. 150)
   -”How the patterning of the housing recession is related to urban form, neighborhood, composition, and other spatial and socioeconomic variables.” (p.157)
3. 이론적 배경
   The Spatial Dimensions of the Housing Crisis
   -High concentration neighborhoods those who are minorities and low-income households are more likely locating inner cities, which means more vulnerable to the crisis.
   -“Some middle-class neighborhoods in suburban and exurban areas were also deeply affected by the crisis.” (p.152)
   The Potential Role of Urban Form in the Crisis
   -Urban Form : Variation of housing prices in local markets in a housing crisis by offering different types of neighborhoods. Affecting neighborhood resilience in an economic crisis by influencing accessibility and household expenditure on transportation.
   -“Speculation that new urbanist neighborhoods are undersupplied in American cities because of policy constraints and exclusive zoning.” (p.153)
4. 연구가설
   -They made some variables to represent the land use, transportation, and housing characteristics
   -When they study about spatial variation across the metropolitan areas’ for the housing recession depth, most of the fast-growing metropolitan areas’ housing price declines in the recent recession. However, for the housing recession duration, it’s quite different with the depth. Areas with the fewer price declines, they experience longer housing recessions. But, some areas such as Southwest and Florida experienced deeper and longer housing recessions, and some experienced shallow and shorter housing recessions.
   -About within metropolitan areas, they divided the neighborhoods in two subgroups by size: more than 2.5 million people to the large area, less than 2.5 million people to the medium-sized. The reason why they did this is to test whether the landscape of the housing recession exhibited different patterns in different size of metropolitan areas. They supposed that with higher density, higher levels of mixed use, and lower levels of auto dependency, housing recessions were shallower and shorter since neighborhoods with these features are likely to provide better work and non-work accessibilities and more non-auto transportation options.
5. 연구 방법론
   -Data : CoreLogic, Census, ACS
   -Unit of Analysis : Zip code (Neighborhoods) + State, County
   -Time : 2005-2013
   -분석모델 : Hierarchical Linear Modeling
   -Variables : land-use density, mixed use, housing stock age, and auto dependency. By including land-use density and mixed use, we can know the built environment as well as work and non-work opportunities in neighborhoods. One significant thing is age of housing stock which can predict the housing foreclosures. Last thing, auto-dependency, can measure the extent to which households in a neighborhood are dependent on auto for their daily travels.
   -To account for association between housing recession and neighborhood composition, they developed median household income and the shares of African Americans and Hispanics in each neighborhood, so we can know whether the housing recession was felt more intensely by certain population segments.
   -Proportion of high-cost home purchase loans from 2004 to 2006 : effect high-risk lending.”
   -Distance to the city center : relationship between housing recession depth /duration
   -McGranahan natural amenity index : natural amenities on housing prices. A higher value, better natural amenities
6. 주요연구결과
   Housing Recession Depth
   Depth ∝ 1/(land-use densities) ∝ Auto dependency ∝ Housing stock age
   Housing Recession Duration
   Duration ∝ 1/lower densities ∝ Mixed use ∝ auto dependency.
   “The submodels by population size, however indicate that the conclusion holds for large metropolitan areas only.” (p.165) When other conditions are equal, “higher income neighborhoods with higher proportions of Hispanics were associated with shorter recessions.” (p.166)