Accessibility in a Post-Apartheid City: Comparison of Two Approaches for Accessibility Computations

Abstract

Many authors argue that issues related to interpretability, lack of data availability, and limited applicability in terms of policy analysis have hindered a more widespread use of accessibility indicators. Aiming to address these aspects, this paper presents two accessibility computation approaches applied to Nelson Mandela Bay in South Africa. The first approach, a household-based accessibility indicator, is designed to account for the high diversity both among the South African society and in terms of settlement patterns. Besides OpenStreetMap (OSM) as its main data source, this indicator uses a census and a travel survey to create a synthetic population of the study area. Accessibilities are computed based on people’s daily activity chains. The second approach, an econometric accessibility indicator, relies exclusively on OSM and computes the accessibility of a given location as the weighted sum over the utilities of all opportunities reachable from that location including the costs of overcoming the distance. Neither a synthetic population nor travel information is used. It is found that the econometric indicator, although associated with much lower input data requirements, yields the same quality of insights regarding the identification of areas with low levels of accessibility. It also possesses advantages in terms of interpretability and policy sensitivity. In particular, its exclusive reliance on standardized and freely available input data and its easy portability are a novelty that can support the more widespread application of accessibility measures.

Appendices

Appendix: A: Sensitivity Test for the Scale Parameter

As pointed out in Section 3.1, the MATSim default scoring function (Nagel et al. 2016) has supplied the scale parameter μ and the utility of traveling $V^{trav}_{ij}$, which are required to compute the econometric accessibility scores according to Eq. 1. These values are based on model-theoretic considerations based on the well-known Vickrey bottleneck scenario (Vickrey 1969). It was pointed out that using MATSim default values has proven to be a valid approach in a number of scenarios when region-specific utility values are absent. Also recall that μ and $V^{trav}_{ij}$ are dependent on each other and can, therefore, not be estimated separately (Train 2003).

As the scale parameter is highly influential, a sensitivity analysis has been conducted. Figure 5 shows accessibilities to education facilities computed based on the MATSim default scoring function that was used to create the results shown in Section 3.4, for which the scale parameter was set to one (μ = 1.0).

In Fig. 6, the scale parameter was increased to μ = 2.0. While the overall accessibility pattern does not change, one can detect some changes in Kwa Nobuhle (the township in the northwest, just south of Uitenhage). Some measure points now fall into the higher-scored deciles (i.e. are drawn in green colors). By increasing the scale parameter, the presence of facilities on a very local scale is given a higher weight. A measure point that is in the direct vicinity of very few facilities (education facilities in this case), will rank in the upper deciles of the evaluation. This is confirmed by the observation that the heterogeneity of accessibility values (i.e. different colors) in a given areas is somewhat higher than in Fig. 5.

Fig. 6

Scale parameter μ = 2.0; decile color ramp

In line with this, the opposite effect can be observed when the scale parameter is decreased to μ = 0.5 (cf. Fig. 7). Here, Kwa Nobuhle is quite homogeneously red, i.e. people residing in any part of Kwa Nobuhle are affected by low accessibilities. The reduction of the scale parameter has effected that a higher number of facilities is required to reach a good score, but that longer trips are accepted to reach these facilities. Accordingly, the suburbs to the Southwest of the center of Port Elizabeth receive a better evaluation with a scale parameter of μ = 0.5. The availability of a larger number of facilities in central Port Elizabeth that are reachable within an acceptable travel time brings them into the scope of people who reside in these suburbs and, thus, results in a quite good accessibility score.

Fig. 7

Scale parameter μ = 0.5; decile color ramp

The results confirm that the taken approach of using utilities out of model-theoretical reason is viable for a study with a comparative focus. When the capability of the econometric measure is applied to monetize results, a region-specific estimation of the model parameters is required.

Appendix: B: Considering Time-Dependent Accessibilities

Some authors (e.g., Moya-Gómez et al. 2017) highlight that accessibilities may change during the course of a day due to time-dependent variance in the transport and land-use systems. As pointed out in Sections 2 and 3, all computations carried out for this study are implemented within the MATSim transport simulation framework. As described in more detail in Section 3.3, this offers the opportunity to run accessibility computations that respect time-dependent (i.e. potentially congested) traffic conditions without significant alterations of the procedure. The reason why the results in Section 3.4 have been created under un congested traffic conditions (i.e. using free-flow network speeds) is the goal to exclusively use freely available and standardizes input data (OSM).

For the computation of congestion-based accessibilities, the synthetic population that was described in Section 2.2, which relies on a (not publicly available) travel survey, is required. The results of the congestion-based accessibility computation for the morning peak (8 o’clock) of a regular day are shown in Fig. 9. It can be observed that there are some detectable, but no significant changes in accessibilities when compared to un congested conditions as depicted in Fig. 8.

Fig. 8

Accessibility to education facilities under un congested traffic conditions; decile color ramp

Fig. 9

Accessibility to education facilities under congested traffic conditions at 8 o’clock; decile color ramp

Under congested conditions, the accessibility values in the area southwest of Port Elizabeth (Mount Pleasant, Broadwood) range in worse quantiles than under free-flow traffic conditions. This seems plausible because from these locations only smaller roads (M9 – Buffelsfontein Road/Heugh Road and M7 – Main Road/RiverRoad), which are more prone to be affected by congestion, lead to the center of Port Elizabeth, while from the other directions bigger thoroughfares (R102 and N2 from the West; R75 from the Northwest, and R102 and N2 from the North) lead to the center of Port Elizabeth.

In line with the municipality’s Comprehensive Integrated Transport Plan (CITP Nelson Mandela Bay Municipality 2011), it can, therefore, be concluded that congestion effects do not seem to play an overly important role in Nelson Mandela Bay and that accessibility computations under free-flow conditions yield viable results.