Carpool Insight

Overview

Carpool Insight identifies business clusters in the Oslo/Akershus region where public transit is inadequate compared to driving, making them ideal targets for carpool initiatives.

The analysis uses a sequential filtering approach:

1. Phase 1: Find all business clusters (destination-first mapping)
2. Phase 2: Assess public transit quality for each cluster
3. Phase 3: Analyze residential catchment areas for top targets

1 Business Clustering

The first phase identifies geographic clusters of businesses in the Oslo/Akershus region.

Data Filtering

• Geography: Oslo (0301) and Akershus municipalities (32xx series)
• Employee count: 20-200 employees per business unit

Entity Type Handling

Norwegian businesses register two types of entities:

• Underenheter (sub-units): Physical operating locations (stores, offices, factories) for companies with multiple sites. These have beliggenhetsadresse (location address) - the actual work location.
• Hovedenheter (parent entities): Legal/registered company records. May only have forretningsadresse (business address) which isn't necessarily an operational site.

To accurately map work locations and avoid duplicates:

• Include all underenheter - actual physical locations
• Include hovedenheter without sub-units - single-location companies
• Exclude hovedenheter that have sub-units - prevents duplicates

Geocoding

Business addresses are geocoded using the Kartverket API. Results are cached locally to avoid redundant API calls.

Spatial Clustering (DBSCAN)

Businesses are grouped into clusters using the DBSCAN algorithm:

Super-Clusters

Nearby clusters are grouped into super-clusters representing shared commuter sheds. People traveling to the same general area can share corridors regardless of specific employer.

Output

Each cluster contains:

• Centroid point (geographic center)
• Boundary polygon (convex hull of member businesses)
• Total employee count and business count
• Auto-generated name based on street/location
• super_cluster_id and super_cluster_name for grouping

2 PT Disadvantage Assessment

Each cluster is scored on how poorly served it is by public transit. Higher score = worse PT = better carpool target.

Four Metrics

Metric	Weight	What is measured
Competitiveness	50%	PT time / driving time ratio
Access	30%	Meters to nearest frequent stop
Transfer	10%	Average number of transfers (12 origins)
Frequency	10%	Departures/hour (4-hour window)

Reference time: Tuesday 07:30

Combined PT Disadvantage Score

Target Selection Threshold

Only clusters with PT disadvantage >= 40 are shown in the Top 100 Targets list. This threshold filters out well-served downtown areas that would otherwise rank highly due to employee count alone.

Carpool Potential Ranking

Clusters are ranked by carpool potential, which balances PT disadvantage with cluster size:

Using logarithm reduces the impact of employee count differences, making PT disadvantage the primary ranking factor while still giving weight to cluster size.

3 Catchment Analysis

For super-clusters with high PT disadvantage, the system analyzes residential catchment areas to understand where potential carpoolers live.

Four Views

Isochrone Generation

Driving isochrones are fetched from the Mapbox Isochrone API for each super-cluster centroid:

• 15 minutes: Immediate catchment (darkest blue)
• 30 minutes: Primary catchment
• 45 minutes: Extended catchment (lightest blue)

Isochrones are pre-computed and cached to avoid API calls on each page load.

Wedge Corridor Analysis

The 30-minute isochrone is divided into directional wedges (N, NE, E, SE, S, SW, W, NW) to identify which roads carry the most potential carpoolers:

1. Divide isochrone into 8 compass direction wedges
2. Intersect each wedge with SSB population grid data
3. Sum residential population within each wedge
4. Identify the primary road in each direction (E6, E18, Rv4, etc.)

Catchment Output

For each super-cluster, the Catchment tab displays:

• Total catchment population within 30-minute drive
• Population bars showing breakdown by road/direction
• Percentage of total population for each corridor

Clicking a catchment item shows the isochrone rings on the map and highlights only the member cluster markers (hiding others for clarity).

P Parking Data Layer

An independent parking data layer provides context on parking availability across the region. Parking data is displayed as a standalone map layer and sidebar tab — it does not affect PT disadvantage scoring or carpool potential rankings.

Why Parking Matters

Parking availability is a key contextual factor for carpool analysis:

• Ample free parking: People already drive — high carpool potential
• Expensive/limited parking: People already use PT — lower carpool potential
• Park & Ride (innfartsparkering): Facilities that both compete with and complement carpooling

Data Source

Parking data comes from Statens vegvesens Parkeringsregister, an open API providing all registered public parking areas in Norway. Areas are filtered to the Oslo/Akershus bounding box.

Parking Categories

Category	Description
Innfartsparkering (P+R)	Park & Ride facilities near transit stations
Parkeringshus	Multi-story parking garages
Avgrenset område	Dedicated surface parking lots
Langs kjørebane	On-street parking

Map Display

Parking areas are shown as a GeoJSON layer with clustering at low zoom levels. Markers are color-coded by type, with Park & Ride facilities given a larger, distinct marker. A toggle button (P) in the map controls allows showing/hiding the parking layer independently of the active sidebar tab.

Parkering Tab

The sidebar tab displays a summary of all parking areas grouped by category, with P+R facilities listed first (most relevant for carpool analysis). Each entry shows name, capacity, and type. Clicking an item flies the map to that location.

Capacity details include: paid spaces, free spaces, EV charging spaces, and disabled-accessible spaces.

Data Sources