Using FME to Manipulate Various Sources of Data for Traffic Collision Spatio-temporal Correlation...

Using FME to Manipulate Various Sources of Data for Traffic Collision Spatio-temporal Correlation Analysis

Ying (Ariel) Luo, M.Sc., E.I.T.Traffic Safety Spatial AnalystCity of Edmonton, Office of Traffic Safety

City of Edmonton Office of Traffic Safety

• Established in 2006, first municipal office of traffic safety

in North America

• Five “E”s traffic safety strategy: Engineering, Enforcement,

Education, Evaluation and Engagement

• Focusing on traffic safety engineering, automated

enforcement and speed management, data analytics and

road user behavior studies

OTS Ultimate Goal: OTS Ultimate Goal:

00Fatal and Injury Collisions in EdmontonFatal and Injury Collisions in Edmonton

Outline

Project background Network-based traffic safety spatial analysis

Objectives Challenges Spatial Data Extract-Transform and Load (ETL)

using FME

Project background

Ran-off-road is the most severe collision cause in the City of Edmonton

Project background

Project background

Network Kernel Density Estimation

Identify hot spots along the roadway network

Density of seg. s (#collisions/km)

Summation of segments within search bandwidth

Search bandwidth

distance between collision i to end node of segment s

Network-based spatial analysis

Collisions are network constrained events!

City of Edmonton ROR Collisions Hot Spot Map (Grid-based)

Any distance related analysis will need network distance, instead of Euclidean distance

Network Kernel Density Estimation

Identify hot spots along the roadway network

Density of seg. s (#collisions/km)

Summation of segments within search bandwidth

Search bandwidth

distance between collision i to end node of segment s

Step 1: Build Network Topology

Segmentation: custom transformers Build network connectivity: some of the polylines

are drawn in opposite direction/ two-way segments are drawn in single line

Step 2: snap collision points to unit segments

Challenge- Snap to the correct direction

Step 3: Estimate density for each unit segment

For each unit segment, find

collisions within search bandwidth;

Calculate K function for individual

collision to the analysis segment;

Get segment kernel density.

Over 14,000Over 14,000

Over 6,000Over 6,000

City of Edmonton ROR Collisions Hot Spot Map (2009-2014,network-based)

(Grid-based hot spot map)

Location specific treatment

Next step- spatial correlation analysis

Spatial components

Alcohol outlet locations

Land use pattern

Infrastructure information (e.g. traffic control)

Enforcement activities

Socio-ecomonic factors

Non-spatial components

Weather, geometry, traffic volume etc.

Spatial Correlation Example (qualitative visualization)

Weekend Late-night ROR Density vs. Alcohol Outlets Density

Thank You!

Questions?

For more information: Ying (Ariel) Luo: ariel.luo@edmonton.ca City of Edmonton, Office of Traffic Safety

http://www.edmonton.ca/transportation/traffic-safety.aspx