Spatial Statistics for Road Safety Modelling in Nairobi County
Publication
Using Poisson regression and Bayesian spatial models to identify high-risk road accident corridors.
This study applies spatial statistical methods to identify high-risk road accident corridors in Nairobi County, providing an evidence base for targeted safety interventions by the National Transport and Safety Authority (NTSA).
Methodology
The study uses road accident data from NTSA for the period 2016–2022, geocoded to 250m road segments across Nairobi's 17 sub-counties. Poisson regression models are fitted to accident counts with covariates including road type, speed limit, junction proximity, and pedestrian density.
Bayesian spatial models (CAR models implemented in R-INLA) are applied to account for spatial autocorrelation between adjacent road segments, producing smooth risk surface maps that identify clusters of elevated accident risk beyond what the raw data alone would suggest.
Key Findings
The highest-risk corridors identified are the Thika Superhighway (particularly the Githurai–Roysambu stretch), Mombasa Road between the Airport Roundabout and Mlolongo, and the Ngong Road–Dagoretti Corner area. Pedestrian-vehicle conflicts at unsignalised crossings account for 43% of fatalities in the study area.
Policy Implications
The study recommends targeted installation of pedestrian bridges at the 12 highest-risk crossing points identified by the spatial model, and speed enforcement camera deployment at 8 road segments with persistently elevated crash rates.