Investigating the link between the built environment and the incidence of pedestrian crashes in Cape Town, South Africa

Nteziyaremye, Pascal (2018-12)

Thesis (PhD)--Stellenbosch University, 2018.

Thesis

ENGLISH ABSTRACT: Pedestrians are the most vulnerable road users in the road environment, particularly in the developing world. To gain a better understanding of pedestrian crash causation, the built environment has been given much attention in the international traffic safety research. However, research of this nature is still scarce in the developing world, including South Africa. This study investigates the link between the built environment and the incidence of pedestrian crashes. The study used pedestrian crash data collected in Cape Town over a 3-year period between 2012 and 2014. The research method involved screening, geocoding and supplementing poor quality secondary data on pedestrian crashes. Moreover, the study applies a variety of analytical methods including univariate, bivariate, geospatial and multivariate analyses. Four GIS-based spatial analysis methods were used to identify clusters of pedestrian crashes within the study area. These methods include the planar kernel density estimation (KDE), the Anselin local Moran’s I, the Getis-Ord Gi* and the Optimized Hot Spot Analysis (OHA). Two modelling techniques, the Generalised Linear Modelling (GLM) and Geographically Weighted Regression (GWR) modelling were used to relate the built environment and population variables to total; intersection; and killed and seriously injured (KSI) pedestrian crashes. For this analysis, the data was aggregated and analysed at the census suburb level. Among other results, it was found that population; land use mix; traffic signals; roundabouts/mini-circles; industrial use; four- and multi-legged intersections; and high mobility roads are associated with greater numbers of pedestrian crashes. The study also revealed that pedestrian crashes are positively related to socio-economic deprivation. In addition, spatial variations of the associations in the models were investigated and discussed. Hotspots of pedestrian crashes were identified mostly in the South Eastern regions of Cape Town which are also areas where economically-disadvantaged residents are concentrated. The presented models can be used to predict future pedestrian crashes using information that is easily available at the city level. The models are also crucial for the planning of safe walking environments which are particularly needed in South Africa and other developing countries.

AFRIKAANSE OPSOMMING: Voetgangers is die kwesbaarste padgebruikers in die padomgewing, veral in die ontwikkelende wêreld. Om ʼn beter begrip van voetgangerbotsing oorsake te verkry, is baie aandag aan die geboude omgewing gegee in internasionale verkeerveiligheid navorsing. Navorsing van hierdie aard is egter steeds skaars in die ontwikkelende wêreld, insluitend Suid-Afrika. Hierdie studie ondersoek die skakel tussen die geboude omgewing en die voorkoms van voetgangerbotsings. Die studie het voetgangerbotsing data, ingesamel in Kaapstad oor ʼn drie-jaar periode tussen 2012 en 2014, gebruik. Die navorsingsmetode het sifting, geo-kodering en die aanvulling van lae-kwaliteit sekondêre data oor voetgangerbotsings ingesluit. Verder pas hierdie studie ʼn verskeidenheid van analitiese metodes toe, insluitend univariante, bivariante, geo-ruimtelike en multivariante analises. Vier GIS-gebaseerde ruimte-analise metodes is gebruik om voetgangerbotsing-bundels in die studiearea te identifiseer. Hierdie metodes sluit die planêre kerndigtheidsberaming (KDB), die Anselin plaaslike Moran's I, die Getis-Ord Gi * en die Geoptimaliseerde Warmkol Analise (GWA) in. Twee modelleringstegnieke, die Veralgemeende Lineêre Modellering (VLM) en Geografies Geweegde Regressie (GGR) modellering is gebruik om die geboude omgewing en populasie veranderlikes te verbind aan totale, interseksie, en noodlottige en ernstig beseerde (NEB) voetgangerbotsings. Vir hierdie analise is die data gesommeer en geanaliseer op die sensus voorstad vlak. Dit is, onder andere, bevind dat populasie, grondgebruik samestelling, verkeerseine, verkeersirkels/mini-sirkels, industriële gebruik, vier- en multi-been interseksies, en hoë mobiliteitspaaie geassosieer word met groter hoeveelhede voetgangerbotsings. Die studie het ook onthul dat voetgangerbotsings positief verbind is aan sosio-ekonomiese ontneming. Daarbenewens is ruimtelike variasies van die assosiasies in die modelle ondersoek en bespreek. Warmkolle van voetgangerbotsings is meestal in die Suid-Oostelike streke van Kaapstad, wat ook areas is waar ekonomies benadeelde inwoners gekonsentreerd is, geïdentifiseer. Die voorgestelde modelle kan gebruik word om toekomstige voetgangerbotsings te voorspel deur inligting te gebruik wat maklik op die stadsvlak beskikbaar is. Die modelle is ook noodsaaklik vir die beplanning van veilige loop omgewings wat veral in Suid-Afrika en ander ontwikkelende lande benodig word.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/105105
This item appears in the following collections: