Mapping the Spatial Distribution of Urban Traffic Congestion

Mapping the Spatial Distribution of Urban Traffic Congestion

A geospatial analysis framework for quantifying and visualizing intra-urban traffic congestion patterns across European cities using Google Distance Matrix API and spatial analytics

The Challenge

Urban traffic congestion represents one of the most persistent challenges facing modern cities, with profound impacts on economic productivity, environmental sustainability, and quality of life. While traditional congestion metrics provide city-wide or corridor-level assessments, they fail to capture the fine-grained spatial heterogeneity of congestion within urban areas. Policy makers and urban planners require detailed spatial intelligence to identify congestion hotspots, understand their distribution patterns, and design targeted interventions.

Existing congestion studies predominantly rely on aggregated indicators or limited sensor coverage, providing insufficient spatial granularity for neighborhood-scale analysis. The fundamental research questions driving this investigation were: How does traffic congestion vary spatially within European cities? Can we develop a systematic methodology to map intra-urban congestion patterns? What insights emerge from comparing congestion distributions across different urban contexts?

This research addresses these gaps by developing a comprehensive geospatial framework for measuring, analyzing, and visualizing traffic congestion at unprecedented spatial resolution across four European cities: Lisbon, Göteborg, Amsterdam, and Glasgow.

Research Methodology

Spatial Grid Framework

We implemented a systematic grid-based approach to partition each urban area into standardized spatial units. For each city, we:

  1. Defined urban study boundaries using official European urban morphological zones
  2. Generated systematic grid cells (1km² for Lisbon using ETRS89 LAEA projection; varying sizes for other cities)
  3. Computed grid centroids as representative points for origin-destination analysis
  4. Created comprehensive OD matrices with all possible grid-to-grid combinations
Example of spatial grid overlay on urban area

Figure 1: Systematic grid-based spatial partitioning methodology applied to urban study area

Travel Time Data Collection

We employed the Google Distance Matrix API to collect empirical travel time data under varying traffic conditions:

  • Traffic Scenarios: Three distinct traffic models were queried:

    • Pessimistic scenario: Morning rush hour (8:30 AM on Wednesday, March 15, 2018) - captures peak congestion
    • Best guess scenario: Current traffic conditions - represents typical conditions
    • Optimistic scenario: Free-flow conditions (4:00 AM on Sunday, July 15, 2018) - minimal traffic baseline

  • OD Matrix Queries: For each grid-to-grid pair, we collected:

    • Network distance (meters)
    • Travel duration without traffic (seconds)
    • Travel duration in traffic (seconds)
    • Implied average speed

Congestion Index Calculation

We computed the Travel Time Index (TTI) as our primary congestion metric:

TTI = tcongested / tfree-flow

Where:

  • tcongested = travel time during peak congestion (pessimistic scenario)
  • tfree-flow = travel time under optimal conditions (optimistic scenario)

A TTI of 1.5 indicates that a trip takes 50% longer during congestion compared to free-flow conditions.

Below is the step by step method.

Methodology followed in the research

Figure 2: Spatial distribution of Travel Time Index (TTI) showing congestion hotspots and patterns across urban areas

Technical Architecture

      ┌─────────────────────────────────────────────────────────────┐
      │                    DATA PIPELINE                            │
      ├─────────────────────────────────────────────────────────────┤
      │                                                             │
      │  Spatial Grid       Google Distance      Congestion         │
      │  Generation    -->  Matrix API      -->  Index              │
      │  (R/rgdal)          Queries              Calculation        │
      │                     (XML parsing)        (TTI, speed)       │
      │                                                             │
      └─────────────────────────────────────────────────────────────┘

      ┌─────────────────────────────────────────────────────────────┐
      │                  ANALYTICAL WORKFLOW                        │
      ├─────────────────────────────────────────────────────────────┤
      │                                                             │
      │  1. load_n_centroid()    - Create spatial grid centroids    │
      │  2. create_mat_od()      - Generate OD matrix               │
      │  3. data_build()         - Query Google API (3 scenarios)   │
      │  4. clean_1st()          - Data cleaning & speed calc       │
      │  5. aggregation()        - Spatial aggregation by origin    │ 
      │  6. aggrergate_n_merge() - Merge with spatial geometries    │
      │  7. end_of_aggregation() - Compute final TTI metrics        │
      │                                                             │
      └─────────────────────────────────────────────────────────────┘

Key Implementation Components:

  • R Scripts: Congestion_Index.R - Core analytical functions
  • Execution Pipeline: Excecution.R - City-specific processing workflows
  • Visualization: graphs.R, tables.R - Publication-quality outputs
  • Geospatial Libraries: rgdal, sp, sf, mapview, leaflet
  • Data Processing: dplyr, tidyr for data manipulation

Results & Findings

Spatial Congestion Patterns

Our analysis revealed distinct spatial heterogeneity in traffic congestion across all four cities:

Travel Time Index spatial distribution across study cities

Figure 3: Spatial distribution of Travel Time Index (TTI) showing congestion hotspots and patterns across urban areas

Key Spatial Findings:

  1. City Centers: Consistently exhibited highest TTI values (>1.5), indicating severe congestion during peak periods
  2. Suburban Areas: Generally showed lower congestion (TTI 1.1-1.3) with better free-flow conditions
  3. Radial Patterns: Congestion intensified along major arterial corridors connecting periphery to center
  4. Hotspot Clustering: Specific neighborhoods showed persistent congestion regardless of traffic scenario

Comparative City Analysis

Comparative congestion metrics across the four cities

Figure 4: Comparative analysis of congestion intensity and spatial extent across Lisbon, Göteborg, Amsterdam, and Glasgow

Inter-City Comparisons:

  • Lisbon: Exhibited highest average TTI with extensive congestion coverage
  • Amsterdam: Moderate congestion but highly concentrated in canal district
  • Göteborg: Lower overall congestion with localized bottlenecks
  • Glasgow: Variable patterns with corridor-specific congestion

Temporal Variability

Analysis of the three traffic scenarios revealed:

Δt = tpessimistic - toptimistic

Average Time Differences (minutes):

  • Short trips (<5km): +8-12 minutes during congestion
  • Medium trips (5-15km): +15-25 minutes during congestion
  • Long trips (>15km): +30-45 minutes during congestion
Distribution of congestion-induced travel time delays

Figure 5: Distribution of additional travel time during peak congestion compared to free-flow conditions

Speed Reduction Analysis

Average travel speeds showed significant reductions during congested periods:

  • Free-flow conditions: 40-65 km/h (depending on road network)
  • Peak congestion: 15-30 km/h
  • Speed reduction: 50-70% during rush hours

Technical Innovation

Scalable Geospatial Framework

We developed a modular R-based analytical pipeline that:

  • Handles large-scale OD matrices: Efficiently processes tens of thousands of origin-destination pairs
  • Integrates multiple APIs: Google Distance Matrix (legacy) with TomTom migration path
  • Supports multiple cities: Parameterized functions adapt to different coordinate systems and grid sizes
  • Produces publication-ready outputs: Automated generation of maps, charts, and statistical summaries

Data Quality & Validation

  • Error handling: Robust try-catch blocks for API failures
  • Data cleaning: Removal of invalid routes and zero-distance pairs
  • Outlier detection: Statistical filtering of anomalous travel times
  • Cross-validation: Comparison with ground-truth traffic data where available

Visualization Pipeline

The project implements sophisticated visualization capabilities:

# Automated map generation for all metrics
export_maps_n_graphs(city_data, 
                     city_name = 'Lisbon',
                     city_best, 
                     city_worst)

# Interactive leaflet maps with custom color schemes
mapview(lisbon, zcol = "tti")

Generates:

  • Choropleth maps: TTI, travel time differences, speed reductions
  • Distribution plots: Histograms and density curves for congestion metrics
  • Scatter visualizations: Distance vs. travel time relationships
  • Comparative charts: Multi-city statistical comparisons

Research Impact

Academic Contributions

This research demonstrates a replicable methodology for fine-grained urban congestion analysis that can be applied to any city with road network data. The spatial grid approach enables:

  • Neighborhood-level insights: Beyond traditional corridor or city-wide metrics
  • Comparative urban analytics: Standardized framework for cross-city studies
  • Policy-relevant evidence: Spatially explicit data for targeted interventions

Applications for Urban Planning

Transportation Planners can use these methods to:

  • Identify specific zones requiring congestion mitigation
  • Evaluate the spatial extent of infrastructure improvements
  • Prioritize public transit investments in high-congestion areas
  • Assess equity implications of congestion burdens across neighborhoods

Policy Makers gain:

  • Evidence-based justification for congestion pricing zones
  • Spatial targeting for intelligent transportation systems
  • Data-driven evaluation of urban development impacts on traffic

Methodological Advances

  • Open-source implementation: All code available for reproduction and extension
  • API-agnostic design: Adaptable to different routing service providers (Google, TomTom, HERE, etc.)
  • Modular architecture: Individual components can be modified or replaced
  • Scalability: Framework tested on cities ranging from 500,000 to 3+ million inhabitants

Real-World Application: Latin America Transportation Investment Analysis

The methodology developed in this research was successfully adapted and deployed for strategic transportation investment analysis in collaboration with the Inter-American Development Bank (IDB). This application demonstrated the framework’s versatility beyond urban congestion mapping:

Strategic Adaptation for Regional Analysis:

  • Scale Transition: From city-grid analysis to country/regional-level strategic planning
  • Point Selection: Instead of systematic grid centroids, strategic logistic points were selected across national transportation networks
  • Investment Focus: Analysis targeted potential transportation infrastructure investments with regional economic impact
  • Multi-Modal Integration: Expanded beyond road networks to include ports, airports, and major freight corridors

Methodological Extensions:

  • Strategic Node Identification: Key logistics hubs, border crossings, economic centers, and infrastructure bottlenecks
  • Regional Connectivity Analysis: Inter-city and cross-border transportation efficiency assessment
  • Investment Prioritization: Travel time improvements translated into economic development potential
  • Policy Integration: Alignment with national transportation master plans and regional development strategies

This application validates the framework’s adaptability from neighborhood-scale urban analysis to national-scale strategic planning, demonstrating its value for both detailed city planning and macro-economic transportation investment decisions.

Technology Stack:

  • R (4.x) - Statistical computing and geospatial analysis
  • Key Libraries:
    • Spatial: rgdal, sp, sf, maptools, raster
    • Visualization: ggplot2, mapview, leaflet, tmap
    • Data: dplyr, tidyr, tidyverse
    • API Integration: XML, RCurl

Data Sources:

  • Google Distance Matrix API (historical data collection)
  • TomTom Routing API (updated implementation)
  • European Urban Morphological Zones (EEA)
  • National grid systems (ETRS89 LAEA for Portugal, local projections for others)

Future Directions

Real-Time Congestion Monitoring

Extension to continuous monitoring systems using:

  • Streaming API integration
  • Temporal trend analysis (hourly, daily, seasonal patterns)
  • Predictive modeling for congestion forecasting

Multi-Modal Analysis

Expand beyond private vehicles to include:

  • Public transit travel times
  • Bicycle routing and infrastructure
  • Pedestrian accessibility metrics
  • Comparative mode choice implications

Environmental Integration

Link congestion patterns with:

  • Air quality sensor data
  • Noise pollution measurements
  • Carbon emissions estimates
  • Urban heat island effects

Research Approach: Geospatial analytics • API-driven data collection • Reproducible science Technologies: R • Spatial analysis • Interactive visualization • Statistical modeling Project Status: Research outputs published | Methodology documented | Code available

Interested in urban analytics, transportation research, or geospatial data science? This project demonstrates how open-source tools and API integration can produce high-quality research insights for evidence-based urban planning.