Transport Accessibility Index Project

Composite Transport Accessibility Index — Münster

A composite transportation-accessibility index for a 1,331-cell grid over Münster, Germany, combining street-network centrality, pedestrian network density, public transport access, and service frequency into a single spatial index — with spatial-autocorrelation (LISA) diagnostics on the result.
R (sf, spdep, tmap) Python (GeoPandas) libpysal / esda folium
Methodology

From five raw indicators to one index

The original analysis was built in R; a Python (GeoPandas/libpysal/esda) reproduction was added specifically to generate the interactive web map below — the same pipeline, run twice, in two ecosystems.

  1. Join five indicator layers — betweenness centrality, closeness centrality, pedestrian network density, public transport access, and service frequency — onto a 1,331-cell spatial grid (EPSG:25832) by grid ID.
  2. Build Queen contiguity spatial weights; apply a spatial lag to betweenness, closeness, and frequency (not to pedestrian/transit access, matching the original specification).
  3. Z-standardize all five (lagged) indicators.
  4. Composite index = mean of the five z-scores, then min-max normalized to [0, 1].
  5. Local Moran's I (LISA) on the normalized index — High-High / Low-Low / High-Low / Low-High clusters at p ≤ 0.05.
Grid cells1,331 Indicators combined5 Complete-data cells509 (38%) WeightsQueen contiguity
Histogram of the composite accessibility index across grid cells with complete data
Distribution of the composite index across the 509 grid cells with complete indicator coverage
Open interactive accessibility + LISA map

Known caveat (inherited from the original analysis): grid cells missing pedestrian-network or public-transport data (roughly 30–60% of cells, likely outside mapped network coverage) get a NaN composite index, filled to 0 before mapping/LISA — i.e. "no data" is treated as "worst accessibility." Worth revisiting (exclude vs. impute vs. flag) if this index is used beyond exploratory mapping.

Methods & Tools

Stack

IndicatorsBetweenness & closeness centrality, pedestrian network density, public transport access, service frequency
Spatial WeightsQueen contiguity (spdep / libpysal)
Index ConstructionZ-standardization, composite averaging, min-max normalization
DiagnosticsLocal Moran's I / LISA cluster classification
R Stacksf, dplyr, spdep, tmap
Python StackGeoPandas, libpysal, esda, folium