network based kernel density estimation for cycling facilities optimal location applied to ljubljana

24
NLB / 28.06.11 / p.1 NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana Nicolas Lachance-Bernard 1 , Timothée Produit 1 , Biba Tominc 2 , Matej Nikšič 2 , Barbara Goličnik 2 1 Geographic Information Systems Laboratory, Ecole polytechnique fédérale de Lausanne 2 Urban Planning Institute of the Republic of Slovenia, Ljubjlana The International Conference on Computational Science and its Applications Cities, Technologies and Planning, June 2011, Santander, Spain

Upload: geographical-analysis-urban-modeling-spatial-statistics

Post on 06-Jul-2015

356 views

Category:

Technology


1 download

DESCRIPTION

Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to LjubljanaNicolas Lachance-Bernard, Timothée Produit - Ecole polytechnique fédérale de LausanneBiba Tominc, Matej Niksic, Barbara Golicnik Marusic - Urban Planning Institute of the Republic of Slovenia

TRANSCRIPT

Page 1: Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

NLB / 28.06.11 / p.1NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana

Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

Nicolas Lachance-Bernard1, Timothée Produit1, Biba Tominc2, Matej Nikšič2, Barbara Goličnik2

1 Geographic Information Systems Laboratory, Ecole polytechnique fédérale de Lausanne

2 Urban Planning Institute of the Republic of Slovenia, Ljubjlana

The International Conference on Computational Science and its Applications – Cities, Technologies and Planning,June 2011, Santander, Spain

Page 2: Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

NLB / 28.06.11 / p.2NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana

Plan

• Introduction

• Conceptual background

• Methodology

• Ljubljana case study

Page 3: Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

NLB / 28.06.11 / p.3NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana

Plan

• Introduction

– Cycling and Urban Planning

– Challenges and Needs for Optimal Location of Cycling Facilities

• Conceptual background

• Methodology

• Ljubljana case study

Page 4: Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

NLB / 28.06.11 / p.4NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana

Cycling and Urban Planning

• Cycling?

– Promoted as one of the most appropriate ways of urban mobility

– Environmentally friendly, require less space, impacts on health

• Planning?

– Importance of cycling facilities provision for cycling development

• Germany: 12,911km (1976) 31,236km (1996)

• The Netherlands: 9,282km (1978) 18,948km (1996)

– Stated preference surveys: Facilities discontinuities, route attributes

• Goal?

– Cycling facilities: Right places (O-D), right corridors (Flux)

Page 5: Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

NLB / 28.06.11 / p.5NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana

Optimal Location of Cycling Facilities

• Opportunities

– GPS: Portable, lightweight, unobtrusive and low-cost

– Planners: Insights of current and future behaviors (monitoring)

• Past studies

– Aultman et al. 1997 – Bicycle commuter routes and GIS

– Dill and Gliebe 2003 – Bicycle and facilities in USA

– Jensen et al. 2010 – Speed and paths of shared bicycle in Lyon

– Menghini et al. 2010 – Route choice of cyclists in Zurich

– Winters et al. 2011 – Motivators and deterrents of bicycling

Page 6: Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

NLB / 28.06.11 / p.6NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana

Optimal Location of Cycling Facilities

• Challenges and Needs

– GPS tracking visual presentation: data volume

– Direct usage of GPS data in the planning practice: lack of methods

– GVI: free enriched geographic data sources (i.e. OSM)

Page 7: Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

NLB / 28.06.11 / p.7NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana

Plan

• Introduction

• Conceptual background

– Examples of Current GPS Tracking Projects

– Ljubljana Investigation Background

– Kerned Density Estimation (KDE)

– Network Based Kernel Density Estimation (NetKDE)

• Methodology

• Ljubljana case study

Page 8: Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

NLB / 28.06.11 / p.8NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana

Examples of Current GPS Tracking Projects

• San Francisco (USA) – Smart phones

– Weekly prize draw

– “Developing” facilities instead of “building” them

• Copenhagen (Denmark) – Web-based GIS portal

– 3,000 trips mapped by citizen VISUM model

– COWI A/S GPS tracking: before / after facilities improvements

• Barcelona (Spain) – Qualitative / Quantitative

– Bici_N project rent-a-cycles video/audio

– Data transfert from station to central DB for further analysis

Page 9: Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

NLB / 28.06.11 / p.9NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana

Ljubljana Investigation Background

• Stated preferences (2008)

– Web-based portal Geae+

• Cyclist description, trip information

• Digitalization of trip

• GPS track transfert from enabled device

– Low-Tech: Paper over map drawing

• Revealed preferences (2010)

– GPS tracking device

• User friendly, low-cost, accurate

• Data transfert by technicians

• Broader investigation

Page 10: Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

NLB / 28.06.11 / p.10NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana

KDE vs. NetKDE

• Kernel Density Estimator (KDE*)

– Operates in euclidean space

– Weights events by their radial distances from grid centroid

• Network Based Kernel Density Estimator (NetKDE*)

– Operates in a network constrained space

– Weights events by the distance from grid centroid measured along this network

* Density estimation function + Epanechnikow kernel function

NetKDE and KDE (2009-2011) by Timothée Produit, Nicolas Lachance-Bernard, Loic Gasser, Dr. Stephane Joost, Prof. Sergio Porta, Emanuele Strano

Page 11: Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

NLB / 28.06.11 / p.11NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana

KDE vs. NetKDE

KDE NetKDE

Page 12: Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

NLB / 28.06.11 / p.12NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana

KDE vs. NetKDE

KDE NetKDE

Page 13: Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

NLB / 28.06.11 / p.13NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana

Plan

• Introduction

• Conceptual background

• Methodology

– GPS Tracking

– Network and Grids

– Low Resolution KDE, High Resolution NetKDE

• Ljubljana case study

Page 14: Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

NLB / 28.06.11 / p.14NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana

GPS Tracking

• Device

– Sport tracker QSTARZ BT-Q1300s

– 62 x 38 x 7 mm, 10m accuracy

– One button (On/Off), mini USB port

– KML, GPX, CVS

– Tracking: 5 seconds, 15h autonomy

• Data

– CSV SHP (WGS84) Merge Projection (UTM33N) [Manifold]

Page 15: Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

NLB / 28.06.11 / p.15NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana

Network and Grids

• Open Street Map Network

– Source: Cloudmade website

– SHP (WGS84) 10km GPS Buffer Projection (UTM33N) Places digitalization Highway deleted[Manifold]

– Topology (0.5m connecting/merging) + attributes cleaning[ESRI ArcGIS model builder]

• Grids

– 100m: Low resolution multi-bandwidths KDE

– 20m: High resolution specific-bandwidths NetKDE[IDRISI]

Page 16: Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

NLB / 28.06.11 / p.16NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana

Plan

• Introduction

• Conceptual background

• Methodology

• Ljubljana case study

– Resources, Data and Calculations

– Low Resolution Grid KDE Results

– High Resolution Grid NetKDE Results

– Discussion

Page 17: Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

NLB / 28.06.11 / p.17NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana

Resources

• Software / Hardware

– Postgres/PostGIS/Python/QuantumGIS

– Windows XP 64

– Intel® Core™2 Quad CPU Q950 @ 3.GHz 7.83GB of RAM

Page 18: Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

NLB / 28.06.11 / p.18NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana

Data and Calculations

• Low resolution KDE 100m 425km2

13,630 segments, 42,342 gridpoints, 442,260 GPS pointsKDE bandwidths

[200m, 2500m] 24 X 100m steps (2-3h)

• High NetKDE/KDE 20m 20km2

8,114 segments, 314,250 gridpoints, 423,748 GPS pointsNetKDE bandwidths

60m (17h), 100m (19h), 200m (24h), 400m (27h)

KDE bandwidths[40m, 100m] 7 X 10m steps[200m, 1000m] 9 X 100m steps (total 18h)

Page 19: Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

NLB / 28.06.11 / p.19NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana

KDE results

100m grid

Bandwidths:

A)300m

B)500m

C)1000m

D)2000m

*Deciles distribution

Page 20: Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

NLB / 28.06.11 / p.20NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana

KDE results

20m grid

Bandwidths:

A)60m

B)100m

C)200m

D)400m

*Deciles distribution

Page 21: Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

NLB / 28.06.11 / p.21NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana

NetKDE

results

20m grid

Bandwidths:

A)60m

B)100m

C)200m

D)400m

*Deciles distribution

Page 22: Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

NLB / 28.06.11 / p.22NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana

Discussion

• NetKDE 20m (Visual analytics)

– 3:1 ratio - Shows flux corridors (a)

– 5:1 ratio - Smoothscorridors only (b)

– 10:1 ratio - Highlights axis and intersections (c)

– 20:1 ratio - Shows cyclist’smain area presence and main axis

Page 23: Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

NLB / 28.06.11 / p.23NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana

Discussion

• Research under rapid evolution…

– 3rd algorithm: Calculation optimization 90-95% (10h network-indexing, 5 min. for each steps)

– Current work on Barcelona, Ljubljana, Geneva, Glasgow, Baghdad

– Professional uses: Architects, Planners, Criminologs, Biologists

• Actual projects…

– Spatio-temporal and statistical analysis

– Fuzzy-map comparison (time, model, resolution, bandwidth)

– Testing Adapted Landscape metrics

– Testing HPC for calculation and subsequent analysis

– Prototyping the integration of NetKDE, KDE, MCA, … into SDI

Page 24: Network Based Kernel Density Estimation for Cycling Facilities Optimal Location Applied to Ljubljana

NLB / 28.06.11 / p.24NetKDE for Cycling Facilities Optimal Location Applied to Ljubljana

Thank you!