In point pattern analysis or spatial hotspot analysis, KDE addresses first-order properties, i.e., the number of events per unit area at a given location (D€uzg€un, 2009). However, validation of the methods is a particular challenge due to the data requirements for the calibration and implementation of these methods. One of the possible causes of RTA is insufficient sight distance on substantially curved roads with limited or obstructed visibility (Gundogdu, 2010).
The risk levels of the six RSA locations are also shown, which are explained in section 2.5. To apply KDEþ, the point shape file of RTAs and a polyline shape file of the road centerline were used. Cluster prioritization was based on the square of the density of RTA points in a given cluster multiplied by the cluster strength; this is a measure of the importance of a cluster as it represents the collective risk of each cluster (Bıl et al., 2013).
Similar to SANET-KDE and KDEþ, STAA is a network-constrained approach, but unlike SANET-KDE, STAA does not require the event points to cross the centerline of the road, thus retaining the initial coordinates of the RTA points. First, the 223 RTA points were transformed into 223 overlapping polygons using a buffer radius of 78 m, i.e. The SSD along Jalan Tutong that was calculated in Section 1.3. Therefore, each hotspot polygon contained attributes related to the sum of deaths, serious injuries, minor injuries and non-injury cases in the RTA.
The STAA method identifies three parameters that contribute significantly to the size of the hotspot, namely frequency (F), severity (S), socio-economic impact (SEI) (Zahran et al., 2017).
RTA SSD L 1
These results indicated that the RTA data are randomly distributed rather than being organized into groups with similar RTA frequency. Therefore, it was concluded that the Getis-Ord Gi hotspot analysis was not appropriate for this particular RTA data set. The STAA method was developed by the Center for Transport Research (CfTR) at Universiti Teknologi Brunei to identify RTA hotspots on Brunei's accident-prone roads.
The approach is a risk-based method that considers RTA frequency, severity and socio-economic impacts to analyze the recorded historical RTA data (Zahran et al., 2017). The next step was to use the "Join" function in ArcGIS to consolidate the RTA point data into the polygons that intersect the data. To evaluate the risk of each hotspot polygon, two additional parameters were established by calculating two 4 x 4 matrices: normalized frequency (NF) versus normalized severity (NS) and normalized frequency (NF) versus normalized socioeconomic impact (NSEI) ( Zahran et al. al., 2017).
X¼Total number of minor injuries within each hotspot zone per SSD Y¼Total number of serious injuries within each hotspot zone per SSD Z¼Total number of fatalities within each hotspot zone per SSD.
S SSD L 1
SEI SSD L 1
The final step was to assign each of the hotspot polygons one of the composite risk levels shown in Tables 1 and 2. The four hotspot identification methods above yielded contrasting results in terms of the spatial distribution of risk levels derived from the analysis of historical RTA data. . Proactive RSA is undertaken to evaluate the safety of the road and its surroundings, without using RTA data (Soames Job, 2012).
A reactive RSA, on the other hand, is made to evaluate the security of the road and its. These sites were selected based on an analysis of the results of the RTA hotspot analysis methods in Sections 2.1, 2.2 and 2.4, which identified six particular stretches of road of high concern. At each of the six locations along Jalan Tutong, safety issues were identified and their expected impact on accident frequency and severity evaluated.
Risk levels in KDE were based on the square of the density of RTA points in a particular cluster multiplied by the strength of the cluster. Therefore, it was decided to use a relative risk level matrix to compare the calculated relative risk levels by each of the RTA hotspot analysis methods at the six RSA locations with the risk levels determined by the RSA. The comparison of the RSA matrix with the matrix of each of the RTA hotspot analysis methods allowed the calculation of a similarity index for each method, which served as a false identification index.
Because RSA location 6 was not assigned a risk level by KDEþ, the similarity index of the KDEþ method was the lowest among the other hotspot analysis methods, as shown in Figure 12. Some parts of the road could be inherently dangerous with one or two fatalities, however, because the RTA frequency is low, risk levels for these road sections of SANET-KDE and KDEþ may not appear. This indicates that STAA based on NF and NSEI is more conservative than STAA based on NF and NS in terms of the probability of underestimating observed risk levels.
Referring to Figure 13, different values for the bandwidth led to different similarity indices for each of the four investigated HSA methods. STAA produced the most consistent results compared to the RSA risk levels in the identification and ranking of RTA hotspots along Jalan Tutong, although this may be due to its inherent consideration of the combined impact of RTA frequency and severity or socio -economic impact. Furthermore, the new validation method showed that STAA resulted in the risk levels most similar to those of the RSA at a bandwidth equal to the SSD along Jalan Tutong, while the risk levels of SANET-KDE are more similar than those of KDEþ to the RSAs. risk levels using the same bandwidth.
A model proposal for determining the traffic accident hotspots on the Turkish highway network: A pilot study. Identification and ranking of black spots: Sensitivity analysis. Transport Research Record: Journal of the Transport Research Council.