Integrating empirical and simulated big data in theory testing for urban systems

DENISE PUMAIN
JULIE FEN-CHONG
CLARA SCHMITT

Transcript

1. Insite, Venice, May 5-8 2014 1
2. Intra-urban and Inter-urban theory 2
3. Cities and Systems of cities as generalised ICT • Cities as places for innovation through individual and collective interaction • Cities connect through their networks a variety of complementary resources; they create new resources thus reduce local and conjonctural uncertainties • Innovation diffuses through circulation of information inside and between cities within the system of cities impulses for urban capital accumulation (gains from adaptation, attraction of new resources) impulses for urban capital accumulation (gains from adaptation, attraction of new resources) Feed back effect from the system: incitation to innovate because of interurban competition • Hierarchical process of innovation diffusion and substitution of activities scaling with city sizes • Path dependent process in socio-spatial evolution (accumulation of physical, human, symbolic capital) leading to increased levels of complexity in society and technology
4. Urban systems as multi-level networks Networks (Intertwined Individual (Intertwined Individual strategies) Rozenblat, 2008
5. Intra-urban evolution and observed big data 5
6. Urban systems as societal adaptators / creators MESO-LEVEL: the city • critical travel time : 1 hour (Zahavi) • low speed networks (x by 5 since 1800) • strong interactions (3/person/day) Density and price gradient (centre-periphery), Fractal spatial organisation, Functional zoning and social differentiation etc.
7. Urban systems as societal adaptators / creators Residents (Low gradient) Places of work (High gradient) Density and price gradient (centre-periphery) M. Wegener (High gradient) Traffic flows (High gradient)
8. • Urban theory: Observed big data and intra-urban evolution Centrality Urban Assets Accessibility multi-level (fractal) center-periphery patterns and discontinuities under gravitational and social constraints • Big data: real time information on intra-urban mobility knowledge about the relative « value » of all urban sites at a fine level of space-time resolution
9. The pervasive city Avaibility of digital footprints
10. • Ile-de-France Region • 12 000 square meters • Anonymous users • March-April 2009 • (2 weeks) Mobile phone footprints 10 • (2 weeks) French users Number of users per day 160 000 – 280 000 Number of footprints 2,3 – 5 millions
11. Cell area Cities boundaries Cell borders Imprecision issues Network imprecision (center/periphery gradient) Network station (x,y coordinates) Imprecise location Stop ? 12:00:00 ? 22:00:00 Temporal imprecision (uncertainty)
12. From individual footprints to the rythms of a city First steps : exploratory analysis Mobile phone footprints Visualization Spatio-temporal aggregation Second step : modeling and adding semantics Using network theory to identifiy spatial communities Adding semantic to trajectories
13. A center-periphery organisation of mobility 13 Flows of French users between 9 and 10 AM 31/03/2009 (Paris)
15. The rythms of the city Mean vectors of flows at 6 PM Mean vectors of flows at 9 AM
16. 16 Flows of roaming users between 9 and 10 AM 31/03/2009 (Paris)
18. Airport Charles de Gaulle Touristic hotspots Disneyland Paris center Flows of roaming users between 9 and 10 AM 31/03/2009 (Paris)
19. Theory construction and testing through simulation 19
20. Urban systems as societal adaptators / creators MACRO-LEVEL: system of cities • critical travel time : 1 day (E. Reclus) • high speed networks (x by 40 since pre-industrial) • weak interaction (less frequent) Hierarchy of sizes, Scaling between population size and number of activities, Functional geodiversity Etc.
21. Urban systems as societal adaptators / creators Population (Belhedi, 2004) Population Pace et al., 1990) Hierarchy of sizes Rank (France) (Tunisie) (Belhedi, 2004) Rank (Guerin-Pace et al., 1990)
22. Urban systems as societal adaptators / creators y = 0.0015x1.1589 R2 = 0.8851 1000 10000 100000 1000000 Financial activities y = 0.0031x1.1397 R2 = 0.9078 1000 10000 100000 1000000 Finance and Insurance France USA South Africa y = 0,0006x1,28 R2 = 0,78 10 100 1000 10000 100000 1000000 Finance, Insurance, Real Estate scaling between population size and number of activities Scaling parameter > 1: leading economic sectors (ex :FIRE: Finance, Insurance, Real Estate) 100 10000 100000 1000000 10000000 100000000 Population 100 10000 100000 1000000 10000000 100000000 Population 1 1000 10000 100000 1000000 10000000 Population β = 1.14 95% CL : 1.11-1.17 R² = 91 % β = 1.28 95% CL : 1.23-1.33 R² = 78% β = 1.16 95% CL : 1.13-1.19 R² = 89 %
23. Simulated big data and inter-urban competition • Urban theory: City size Innovation adoption Attractivity hierarchy of city sizes and functional diversity under gravitational and network constraints • Big data: exploration of possible variations through fluctuations; ranking of constraining factors; identification of path dependence effects and possible/plausible futures Theory construction and testing
24. A lack of empirical data… Data & knowledge Theories y = 0.0031x1.1397 R2 = 0.9078 100 1000 10000 100000 1000000 10000 100000 1000000 10000000 100000000 Population Finance and Insurance y = 0,0006x1,28 R2 = 0,78 1 10 100 1000 10000 100000 1000000 1000 10000 100000 1000000 10000000 Population Finance, Insurance, Real Estate => Regularities to explains…. City size Innovation adoption Attractivity Set of hypothesis on the generative mechanisms of socio-spatial processes ? adoption
25. A virtual laboratory to test theoritial hypothesis Simulation models Set of hypothesis on the generative mechanisms of socio-spatial processes Theories ? Simulation models Data & knowledge Simulation outputs
26. SIMPOP (1996) URBAN EVOLUTIONARY THEORY SIMPOP 2 (2006) The Simpop familly First generation SimpopLocal & SimpopNet (2012) MARIUS (2014) (2012) Second generation
27. Simulation models Data & Set of hypothesis on the generative mechanisms of socio-spatial processes Theories A virtual laboratory to test theoritial hypothesis ? Experiments Simulation models Data & knowldege Simulation outputs
28. 1 Demographic growth 2 Innovation impact The SimpopLocal model 4 Innovation acquistition By diffusion Pdiffusion = f(P i , P k , d ik ) Pi dik 3 Innovation acquistition by creation Pcreation = f(P i ) Pi Pk
29. 3 1 2 3 calibration objectives : Data & Knowledge SimpopLocal : model evaluation by calibration Evaluation of the quality of the model calibration [calibration error]
30. 3 1 2 Evaluation of the quality of the model calibration Simulation of SimpopLocal Evolutionnary algorithms SimpopLocal : model evaluation by calibration Grid computing
31. SimpopLocal : sufficient mechanisms ?
32. SimpopLocal : evaluation by Calibration Profils
33. Calibration error SimpopLocal : necessary mechanisms ? Innovation lifetime Calibration error Innovation diffusion probability Calibration error
34. Simulation models Data & Set of hypothesis on the generative mechanisms of socio-spatial processes Theories Simulation models and Theory construction ? Experiments Simulation models Data & knowldege Simulation outputs
35. Thank you for your attention! 35 http://geodivercity.parisgeo.cnrs.fr/blog/