Uneven Intraurban Growth in Chinese Cities: A Study of Nanjing Yehua Dennis Wei Department of Geography and Institute of Public and International Affairs University of Utah Jun Luo Department of Geography, Geology and Planning Missouri State University
Outline 1. Introduction 2. Study area and growth patterns 3. Data and Methodology 4. Logistic GWR model 5. Spatial variations of urban growth 6. Conclusion
1. Introduction 1.1 Research on urban growth in China Two broadly defined groups: Institutional/political economy perspectives Process, mechanisms, theories growth machines development/entrepreneur states globalization, globalizing cities … Markusen: evidences, methodology…
Neoclassical/modeling approaches Land use/land cover change Location factors, growth determinants Statistics, GIS/RS, landscape metrics… Positivism, theory?
1.2 Modeling urban growth Statistical models: global models underlying forces 1.3 Urban growth Local, non-stationary process over the space Same set of factors have different influences on different areas of a city Context-sensitive theory?
Theories: Regional Development Industrial agglomeration (RS), remaking the Wenzhou model (EG) Methodology: GIS local analysis, LISA, ESDA, GWR, spatial regression… Regional development (PiRS) Urban growth/structure (EPB) 1.4 Objective
1) Local analysis/perspectives Explore spatially varying relationships between urban land expansion and influential factors Modeling: Logistic geographically weighted regression (GWR), a local regression technique 2) Socio-economic factors
2. Study area and Growth Patterns 2.1 Nanjing: coastal, Yangtze Delta From 1988 to 2000 Population: 4.88 million to 5.45 million Built-up area: 392 km 2 to 512 km 2 Study area: the majority of built-up areas, km 2
Population density 2000
Urban growth in Nanjing:
3.1 Data Census data Landsat TM imageries: 1988 and 2000 Image processing Classification: built-up, agriculture, forest and water body GIS: transportation, plan scheme, topographic and land use survey 3. Data and Methodology
3.2 Land use data sampling Sampling: combined systematic and random scheme Systematic sampling: extract regularly spaced points with 300m interval Extract all 1332 points with non-urban to urban land use conversion Randomly select 1350 points without land use conversion 2682 land use sample points
3.3 Variables inputs Dependent variable: Probability of non-urban to urban land conversion Explanatory variables: Proximity factors: proximity to economic nodes Neighborhood factors
Variables TypeDescriptions Dependent variable ChangeProbContinuousProbability of land use conversion Explanatory variable Proximity Dis2HwyContinuousDistance to inter-city highway Dis2LardContinuousDistance to local artery roads Dis2RailContinuousDistance to railways Dis2YRiverContinuousDistance to Yangtze River Dis2YBridContinuousDistance to Yangtze bridge Dis2MCenContinuousDistance to major city centers Dis2MNCenContinuousDistance to suburban centers Dis2InducContinuousDistance to industrial centers Neighborhood AgriDenContinuousDensity of agriculture land BuiltDenContinuousDensity of built-up land WaterDenContinuousDensity of water body ForeDenContinuousDensity of forest land
Agriculture Land Water body Forest land
4.1 Global logistic regression model 4. Logistic GWR model Findings: All explanatory variables are significant road infrastructure development local roads: more important than highways Land use constraints: forest, water City centers more important than subcenters
Explanatory variables BS.E.t valueExp(B) Constant Dis2Hwy Dis2Lard Dis2Rail Dis2YRiver Dis2YBrid Dis2MCen Dis2MNCen Dis2Induc AgriDen BuiltDen WaterDen ForeDen Sample size Log likelihood PCP70.1%
Weighting scheme: Fixed kernel vs Adaptive kernel N=138, Chosen by minimizing an AIC score 4.2 Logistic GWR model
4.3 Model comparison Global logistic modelLogistic GWR PCP70.1%85.6% RSS Moran’s I of residuals Significance test for spatial variability All parameters with p-value below 0.01 Significant spatial variability
MinMaxMeanStd.D%Positive%Negative Dis2Hwy Dis2Lard Dis2Rail Dis2YRiver Dis2YBrid Dis2MCen Dis2MNCen Dis2Induc AgriDen BuildDen WaterDen ForeDen Summary statistics for GWR parameters estimates
5. Spatial variations of urban growth pattern Parameters vary across space: local process All the variables except for Dis2Lard and ForeDen have both positive and negative parameter values Dis2Lard: significant all over the city (-) Other parameters have certain parts in the study area where they are non-significant Use inverse distance weighted (IDW) interpolation to generate parameter and t-statistic surfaces (30×30m)
GWR parameter surfaces: Roads: more negative effective in the north
GWR parameter t-statistic surfaces
GWR parameter surfaces: Centers: more effective in the north Influence of major centers: compact city Suburban centers: weak, local influence
GWR parameter t-statistic surfaces
GWR parameter surfaces: Neighborhood: varied effectiveness
GWR parameter t-statistic surfaces
Urban growth probabilities
6. Conclusions Findings: 1. Logistic GWR can significantly improve the global logistic regression for urban growth modeling: 2. Effects of determining factors have significant spatial variation 3. Interpretation of spatial process should be careful with spatial context; need for local analysis
Limitations: 1) Data: socio-economic variables Discussion: 1) The nature of theory: Theoretical statements 2) Local analysis vs. generalization 3) Representativeness, sampling bias Thank You and Questions?