Network-Based Spatial Analysis
Also known as: network spatial analysis, network-constrained spatial analysis, spatial network analysis, NBSA
Network-based spatial analysis (NBSA) analyzes the distribution and interaction of spatial phenomena constrained to a network structure — such as roads, railways, or rivers — using network distance rather than straight-line (Euclidean) distance. It is the appropriate framework whenever movement, proximity, or risk is governed by the underlying network topology rather than open space.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use NBSA whenever the phenomenon of interest is constrained to a network: road traffic incidents, crime on street segments, pedestrian flows, pipeline failures, or species distribution along river networks. It is essential when Euclidean proximity is a poor proxy for actual accessibility or risk. Do not use it when events genuinely occur across an open planar surface (e.g., spread of an airborne pollutant across a field) — standard planar spatial analysis is more appropriate in that case. It also requires a reliable, topologically correct network dataset; poorly digitized networks undermine the analysis.
Strengths & limitations
- Measures distance and proximity in a way that matches how movement or risk actually operates along infrastructure.
- Avoids the false positives of planar hotspot analysis where Euclidean clusters span impassable barriers.
- Network KDE produces density estimates that respect connectivity, giving more realistic exposure or risk surfaces.
- Directly usable for resource allocation on networks (optimal placement of emergency services, sensors, or stops).
- Well-supported in GIS software (ArcGIS Network Analyst, QGIS, SANET toolbox, R spdep/sfnetworks).
- Requires a clean, topologically valid network dataset — gaps, dangling edges, or wrong directionality corrupt all downstream results.
- Computationally intensive for large networks because all pairwise shortest paths must be computed or approximated.
- Network KDE bandwidth selection lacks universal rules; results are sensitive to the chosen bandwidth h.
- Cannot easily incorporate off-network phenomena (e.g., a pedestrian cutting across a park).
- Statistical inference (significance tests) typically relies on Monte Carlo simulation, which can be slow.
Frequently asked
How is network-based spatial analysis different from standard GIS buffer analysis?
Buffer analysis draws circular zones using Euclidean distance, ignoring barriers and the actual path people or vehicles must travel. NBSA uses shortest-path network distance, so a location 200 m away by road may be closer in network terms than a location 150 m away across a river. This matters whenever movement is constrained to the network.
What software can run network-based spatial analysis?
SANET (free toolbox for ArcGIS, also available standalone), ArcGIS Network Analyst (commercial), the R packages sfnetworks and spNetwork, and the Python library OSMnx combined with NetworkX all support NBSA workflows. QGIS has basic network analysis tools via the QNEAT3 plugin.
What network data quality is needed before analysis?
The network must be topologically consistent: edges must connect at shared nodes, no dangles or gaps, correct directionality for one-way links, and projected in a planar coordinate system with meaningful length units. OpenStreetMap data often needs cleaning with tools like osmnx.simplify_graph or PostGIS topology validation before use.
How do I choose the bandwidth for network kernel density estimation?
Common choices are 100–500 m for pedestrian-scale phenomena and 500–2000 m for vehicle-scale phenomena. Cross-validation or likelihood-based methods (as implemented in the R spNetwork package) give data-driven bandwidth selection. Always report sensitivity of conclusions to bandwidth choice.
Can I apply Moran's I to network-constrained data?
Standard Moran's I uses a planar spatial weights matrix and is not appropriate for network data. A network-adapted weights matrix — where neighbors are defined by adjacency along the network and weights by inverse network distance — should be constructed instead, and significance tested by Monte Carlo permutation on the network.
Sources
- Okabe, A., Satoh, T., Furuta, T., Sugihara, K., & Okano, K. (2006). Generalized network Voronoi diagrams: Concepts, computational methods, and applications. International Journal of Geographical Information Science, 22(9), 965–994. DOI: 10.1080/13658810701587891 ↗
- Okabe, A., & Sugihara, K. (2012). Spatial Analysis Along Networks: Statistical and Computational Methods. Wiley. ISBN: 978-0470770818
How to cite this page
ScholarGate. (2026, June 3). Network-Based Spatial Analysis. ScholarGate. https://scholargate.app/en/spatial-analysis/network-based-spatial-analysis
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Geographically Weighted RegressionSpatial analysis↔ compare
- Hot Spot AnalysisSpatial analysis↔ compare
- Spatial AutocorrelationSpatial analysis↔ compare