temporal and spatial patterns of vehicle-pedestrian

This article was downloaded by: [University of Hong Kong Libraries] On: 29 April 2013, At: 01:03 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Asian Geographer Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/rage20

TEMPORAL AND SPATIAL PATTERNS OF VEHICLE-PEDESTRIAN CRASHES IN BUSY COMMERCIAL AND SHOPPING AREAS: A CASE STUDY OF HONG KONG a

Becky P.Y. LOO & M. K. TSUI

a

a

School of Geography The University of Hong Kong Version of record first published: 03 May 2011.

To cite this article: Becky P.Y. LOO & M. K. TSUI (2005): TEMPORAL AND SPATIAL PATTERNS OF VEHICLEPEDESTRIAN CRASHES IN BUSY COMMERCIAL AND SHOPPING AREAS: A CASE STUDY OF HONG KONG, Asian Geographer, 24:1-2, 113-128 To link to this article: http://dx.doi.org/10.1080/10225706.2005.9684124

PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

Loo & Tsui

113

TEMPORAL AND SPATIAL PATTERNS OF VEHICLE-PEDESTRIAN CRASHES IN BUSY COMMERCIAL AND SHOPPING AREAS: A CASE STUDY OF HONG KONG

Downloaded by [University of Hong Kong Libraries] at 01:03 29 April 2013

Becky P.Y. LOO and M.K. TSUI School of Geography The University of Hong Kong Abstract: Vehicle-pedestrian crashes are not purely random events. They occur more frequently at certain locations and during certain time periods. This paper addresses the issue of pedestrian safety by examining the temporal and spatial characteristics of vehicle-pedestrian crashes in two busy commercial and shopping areas of Hong Kong. The local road crash data in 1993 and 2003 were analyzed. Temporal variations were found to be greater at the commercial and business district (CBD). Using geographic information systems (GIs) and the nearest neighbour analysis, the distribution of vehicle-pedestrian crashes was found to be significantly clustered. The findings of this paper give insights about the distinctive patterns of vehicle-pedestrian crashes in busy commercial and shopping areas, where the risk of vehicle-pedestrian crashes are particularly high because of the co-existence of high volume of pedestrians and vehicles. Keywords: vehicle-pedestrian crashes, temporal and spatial analysis, nearest neighbour analysis; commercial and shopping areas, CBD. Introduction Pedestrians constitute the most vulnerable group of road users in crashes because they are not as protected as most other road user groups like drivers and passengers. From 1993 to 2003, there were 164,214 road crashes in Hong Kong. One-third of these crashes involved pedestrians (3 1.5%). Table 1 shows the pedestrian casualties by injury type over the study period. The total casualties had dropped by 25%, from 6,005 in 1993 to 4,517 in 2003. The numbers by different types of injuries had also declined quite substantially. When the total fatalities are analyzed in Figure 1, pedestrians still accounted for the largest share throughout the last decade. Taking 2003 as an example, the shares of pedestrians, drivers and passengers were 50%, 33% and 16%, respectively. In fact, much attention has been paid to the safety of drivers and passengers in aspects like engineering (e.g. road and highway improvements and vehicle safety designs including brakes, wheels and vehicle lights), education and propaganda (e.g. the slogan "smart drivers-well prepared for all situations"), and legislation (e.g. compulsory wearing of seat belts and the ban on hand-held mobile phones while driving). In comparison, targeted policies and measures for safeguarding pedestrians, for example, the installation of pedestrian facilities and other pedestrianization measures, were less numerous. Figure 2 compares the situation of pedestrian safety in Hong Kong with some overseas administrations. It can be seen that Hong Kong's pedestrians were having the highest share in total road fatalities (about 50%). The share was far higher than the United Kingdom (25.1%) and Japan (28.4%). This can rightly be attributable to the fact that Hong Kong is a high-density city, instead of a country. However, this characteristic also underlined the critical importance of pedestrian safety in the highlydense local environment. In fact, the costs of vehicle-pedestrian crashes (e.g. the Asian Geographer 24(1-2):113-128 (2005)

114

Vehicle-pedestrian Crashes in Hong Kong

physical and physiological sufferings of crash victims, additional burden to hospitals, traffic congestion, extra government expenses on social welfare and reduced economic productivity of a society) are very huge and profound. Moreover, walking is increasingly encouraged as a sustainable transport mode worldwide. To make the public accept wallung for short journeys, a safe walking environment is the prerequisite. The safety of pedestrians, especially in busy commercial and shopping areas, is therefore a major concern. Table 1. Pedestrian casualties by injury type in Hong Kong, 1993-2003.


Year

Slight Injury

Serious Injury

Fatal

Total Casualties

1993 4107 1664 234 6005 1994 4241 1554 169 5964 1995 3767 1507 161 5435 1996 3679 1445 142 5266 1997 3760 1430 131 5321 1998 3458 1362 113 4933 1999 3516 1204 110 4830 2000 3503 1186 96 4785 2001 3549 1332 97 4978 2002 3487 1232 86 4805 2003 3349 1069 99 4517 Source: Road Safety and Standards Division, Transport Department, HKSAR (1993-2003).

/

+Wdestrian & Driver

&Passenger

I

Year

Source: Road Safety and Standards Division, Transport Department, HKSAR (1993-2003).

Figure 1. Fatality rates by road user in Hong Kong, 1993-2003.

Many research studies have examined the influence of various risk factors on road crashes. In this paper, two basic questions are asked - they are the "when" and "where" questions. Levine et al. (1995) examined the hourly variations and spatial distributions of motor vehicle crashes in Honolulu. They found that crashes were highly concentrated between 0600 and 0900 and on weekdays. Two spatial patterns of crashes, namely geographically concentrated and dispersed, were also identified. Another study by Schneider et al. (2004) examined the spatial distribution of perceived and actual

Asian Geographer 24(1-2): 113-128 (2005)

Loo & Tsui

115


vehicle-pedestrian crash locations in a university campus. There were several spatial clusters of police-reported and perceived vehicle-pedestrian collisions but this locations were not identical. As in many disease studies, intensity (spatial concentration) and duration of exposure (temporal distribution) are two important safety concerns (Chipman et al., 1982). Therefore, the temporal and spatial patterns of vehiclepedestrian crashes are the focus of this study.

Countries

Key (from left to right): ISL Iceland; B Belgium; NL Netherlands; F France; S Sweden; Germany; N Norway; Austria; Luxembourg; E Spain; FIN Finland; SLO Slovenia; DK Denmark; IRL Ireland; P Portugal; CH Switzerland; CZ Czech Republic; UK United Kingdom; H Hungary; JAP Japan; YU Yugoslavia; BG Bulgaria; MA Morocco; PL Poland; ALB Albania; LT Lithuania; EST Estonia; LV Latvia; AZ Azerbaijan, BY Belarus; RUS Russian Federation; RO Romania; HK Hong Kong. Source: European Conference of Ministers of Transport (2003) & Road Safety and Standards Division, Transport Departmerrf, HKSAR (1993-2003).

Figure 2. Share ofpedestrians in total roadfatalities in selected admirzistrations, 2000'

Recently, geographical information systems (GIs) have become commonly used in crash analysis, particularly in analyzing the spatial distributions of crashes (Kam 2003; Noland and Quddus, 2004; Peled et al., 1996; Schneider et al., 2004). An early study by Braddock et al. (1994) used GIs to map the distribution of child pedestrian injury and identify the high-frequency collision areas. In relation, the nearest neighbour analysis (NNA) is a useful method for spatial analysis. NNA is a classical method originally developed by plant ecologists for the measurement of point patterns (Clark and Evans, 1954). Later, it was employed in transport safety studies. For instance, Schneider et al. (2004) used NNA to analyze the geographical distribution of vehicle-pedestrian crashes. Levine et al. (1995) examined the spatial relationship between motor vehicle crashes and selected variables with the use of NNA and the ellipse method. This paper focuses on the temporal and spatial characteristics of vehicle-pedestrian crashes with special reference to the busy commercial and shopping areas. It aims to identify the hazardous time (crash peaks) and locations of vehicle-pedestrian crashes (crash clusters). In light of the findings, suggestions for improving pedestrian safety within busy commercial and shopping areas will be presented.




Study Area Central, the commercial and business district (CBD) of Hong Kong (Figure 3), and Causeway Bay, the famous shopping district near the CBD (Figure 4), are the study areas. Central and Causeway Bay were chosen because of several reasons. Firstly, the potential of vehicle-pedestrian conflicts is high due to the co-existence of huge volume of both pedestrians and vehicles in these areas. The high-density development within the CBD and the busy shopping area leads to the high concentration of pedestrians and vehicles within limited space. Table 2 shows the crash records in Hong Kong by district board. It is clear that both Central and Causeway Bay suffered from serious pedestrian safety problems. central2 shared 7.3% of the aggregate 10-year (1993-2003) vehiclepedestrian crashes while Causeway ~a~~ accounted for 7.65%. The situation was only slightly better than Kowloon City (9.3%), which is one of the most densely populated areas in the territory. Secondly, Central and Causeway Bay have been the targeted areas of pedestrianization measures, including full-time pedestrianized streets, part-time pedestrianized streets and traffic calming streets (see Figures 3 and 4), since 2000. The effects of pedestrianization on road safety have not yet been fully explored. Thirdly, though commercial land-use dominates in both areas, the activities generated are different in nature. Central is mainly dominated by financial, commercial, banking and insurance services, as well as higher-order retailing activities, whereas Causeway Bay is characterized by a full range of shopping and recreational activities. The former was regarded as an employment center while the latter was a major retail center. The difference in trip nature is likely to result in different rush hours. Office hours and weekdays are the busy traffic time for Central whereas Causeway Bay is likely to be busier during off-work hours and on weekends. The pedestrians attracted to the study areas are also different in terms of their demographic characteristics. Most of the pedestrians found in Central are likely to be worlung adults whereas teenagers and young adults are likely to shop in Causeway Bay.

Data and Methods The major data used in this study are the annual road traffic crash data compiled by the Transport Department of the Hong Kong Special Administrative Region (HKSAR) Government. The period from 1993 to 2003 was chosen as the time frame because it is hoped that the time span is long enough to reflect the influence of external changes (e.g. population growth, population movement, changing lifestyles and government policies) on the characteristics of vehicle-pedestrian crashes. In Hong Kong, the road traffic crash information has been recorded in a computerized and systematic format, which is known as the traffic accident data system (TRADS), since 1990. The crash dataset in TRADS includes many informative attributes such as crash time (year, date and hour), crash locations (x- and y-coordinates), crash details (e.g. collision type, severity, number of vehicles involved and number of casualties), roadway characteristics (e.g. road type and junction type), casualties characteristics (e.g. gender and age), vehicle's and driver's information (e.g. vehicle type and licence type) and natural conditions (e.g. weather). Relevant data such as collision types, crash year, crash hour, day-of-week and x- and y-coordinates can be extracted. In this study, only vehicle-pedestrian crashes within the two study areas are considered. Vehicle-pedestrian crashes refer to road crashes with at least one pedestrian involved.



Loo & Tsui

Figure 3. Central study area.

0 I

-

400

800 Meters I 5

Figure 4. Causeway Bay study area.


118



Table 2. Aggregate 10-year (1993-2003) vehicle-pedestrian crashes by district board. District Board

Count

%

Central and West Eastern Islands Kowloon City Kung Tong Kwai Tsing Mong Kok North Southern Sai Kung Sham Shui Po Sha Tin Tuen Mun Tai Po Tsuen Wan Wan Chai Wong Tai Sin Yuen Long Yau Tsim

3,781 4,025 187 4,820 4,406 2,095 3,987 1,356 1,417 716 4,351 1,934 1,554 1,487 2,806 3,862 2,597 2,278 4,105

7.3% 7.8% 0.4% 9.3% 8.5% 4.0% 7.7% 2.6% 2.7% 1.4% 8.4% 3.7% 3.0% 2.9% 5.4% 7.5% 5.0% 4.4% 7.9%

Total

51,765

100.0%

Source: Based on Road Safety and Standards Division, Transport Department, HKSAR and crash locations Go-validated by Loo (Loo, 2006).

GIs was used to map crash locations and the spatial environmental features such as the network, and pedestrianized streets. All crash locations were geo-validated (Loo, 2006). NNA was used to analyze the spatial patterns of point distributions within the two study areas. Basically, the technique relies on the measurement of the nearest neighbour distance (NND),the shortest distance of one crash location to its nearest neighbour. The spatial point pattern can be analyzed by the average nearest neighbour distance ( N N D ), which is equal to the total nearest neighbour distance divided by the total number of crash points ( n ) (Eqn. 1).

-

NND =

CNND -

A smaller NND means a more clustered distribution. Nevertheless, the size of the study area ( A ) also determines the pattern of spatial distribution. Therefore, R, which takes the density of the points into account, is used to compare the observed distribution ( N N D ) with its theoretical random distribution ( NNDR).

-

NND R== NND,

-

1 n and density = -. The calculated R value can be compared " - 2J;ienrify A to three theoretical R values, namely the random pattern (RR=l),the perfectly clustered pattern (Rc=O) and the perfectly dispersed pattern (&=a. 149). where NND -

Asian Geographer 24(1-2):113-128 (2005)

Loo & Tsui

119

The parametric difference-of-means Z-test can be used to test whether the spatial pattern differs significantly from a random one. In other words, the one-sample difference-ofmeans test mainly compares NND with NNDR for difference (Eqn. 3). - -

z =NND - NND, 0NND


where o ,

=

0.26136

. Four sets of null hypotheses (Ho) and alternative

JW

hypotheses (HI)were developed. Ho(A): NND in Central in 1993 is equal to that of NNDR . H1(A): NND in Central in 1993 is not equal to thatof NNDR. Ho(B): NND in Central in 2003 is equal to that of NNDR . H1(B): NND in Central in 2003 is not equal to that of NNDR . Ho(C): NND in Causeway Bay in 1993 is equal to that of NNDR . H1(C): NND in Causeway Bay in 1993 is not equal to thatof NNDR. Ho(D): NND in Causeway Bay in 2003 is equal to that of NNDR . H1(D): NND in Causeway Bay in 2003 is not equal to that of NNDR. For the above hypotheses, two-tailed tests are used. With a = 0.05, Ho is rejected when Z is greater than + 1.96 or smaller than -1.96. In addition, the two-sample difference-ofmeans test is used (Eqn. 4). - NNDl - NND2

z=

where

(4)

0NNDl -NNDz

-E2

=

. Here, the distribution of vehicle-pedestrian

crashes in Causeway Bay is expected to be more clustered than that of Central because of the high concentration of retail shops along the main roads. In addition, the spatial patterns of vehicle-pedestrian crashes in both Central and Causeway Bay are expected to be less clustered in 2003 due to the efforts of blacksite treatment and other road safety measures. In relation, three additional sets of hypotheses are developed. -

-

Ho(E): NND of Central in 1993 is equal to NNDin 2003. H1(E): NND of Central in 1993 is smaller than NND in 2003. Ho(F): NND of Causeway Bay in 1993 is equal to NNDin 2003. H,(F): NND of Causeway Bay in 1993 is smaller than NND in 2003. Ho(G): NND in Causeway Bay is equal to NNDin Central. HI(G): NND in Causeway Bay is smaller than NND in Central. In these situations, one-tailed tests are appropriate. With a = 0.05, Ho is rejected when Z is smaller than - 1.645. Last but not least, we test whether the spatial pattern of vehicle-pedestrian crashes in 1993 and 2003 is associated with the pedestrianization measures. The purpose is not to conduct a scientific before-and-after analysis with reference to a specific Asian Geographer 24(1-2): 113-128 (2005)


120

implementation year because the pedestrianization measures in Central and Causeway Bay were not introduced in a single year but were implemented in a gradual manner after 2000. It is recognized that a scientific before-and-after analysis would be meaningful for evaluating each location-specific pedestrianization measure and the data can be pooled for an aggregate analysis (Wong et al., 2006). However, the focus of the analysis in this paper is much more limited. Basically, it aims to see whether there is any statistical association between the crash locations and the analysis years (1993 and 2003). Therefore, the less vigorous chi-square (xZ)test is used. The row and the column of the 2*2 contingency table are year (1993: before pedestrianization and 2003: after test basically pedestrianization) and pedestrianized zone (inside and outside). The compares the observed count (0) with the expected count (E). The sum of all the squared difference gives rise to x2. The formula is shown in Eqn.5. The larger is the difference between the observed and expected counts, the greater is the value of x2. The statistical significance of the calculated is determined by comparing the calculated with the critical at a given degree of freedom (df) and probability (p).


xZ

x2

L

With df=l, p=0.05, the critical

xZ

xZ

x2in this case is 3.84.

Results and Findings 1. Temporal variations of vehicle-pedestrian crashes The hourly and weekly distributions of vehicle-pedestrian crashes in Central and Causeway Bay are presented in terms of the percentage of total annual vehiclepedestrian crashes in Figures 5-8. Should pedestrian flow data be available, some useful indicators on the pedestrian crash risk level can be worked out. However, these exposure data, or other meaningful proxies for measuring the exposure, in 1993 and 2003 are unavailable. Hence, the discussion below only reflects the general vehiclepedestrian crash patterns, instead of the relative risk level. In general, vehiclepedestrian crashes distributed unevenly over time. Daytime and weekdays had higher percentages of vehicle-pedestrian crashes. The variations of hourly distributions were more obvious than that of weekly distributions. In 1993, three major vehicle-pedestrian crash peaks were identifiable in both study areas (Figures 5-6). They were the morning peak (0900- 1000), the afternoon peak (1400-1600) and the evening peak (1700- 1900). In 2003, the vehicle-pedestrian crash peaks changed quite noticeably. Firstly, the morning peak in Central (1000 in 1993) was split into two minor peaks at 0900 and 1000 in 2003. The morning peak in Causeway Bay also started earlier at 0900 (1000 in 1993). Secondly, the afternoon peak in Central delayed by two hours to 1600 in 2003. Although the timing of the afternoon peak did not change in Causeway Bay, the situation has gotten worse. The percentage of vehicle-pedestrian crashes at 1400 has increased from 8.04% in 1993 to 10.06% in 2003. Thirdly, the evening crash peak has delayed by about two hours in both study areas (from 1700 in 1993 to 1900 in 2003). In addition, the night (2200-2300) and mid-night (0100-0300) peaks have become more obvious in both Central and Causeway Bay. Figures 7-8 show the weekly variations. For the CBD (Central), the percentages of vehicle-pedestrian crashes stayed high on weekdays and dropped on Saturday. Moreover, the percentage of vehicle-pedestrian crashes on Sunday has risen from 5% in 1993 to about 7% in 2003. For the busy


Loo & Tsui

121


shopping area (Causeway Bay), the difference between weekdays and weekends was not so apparent.

1

3

5

7

11 13 15 17 19 21 23

9

Hour of day

Figure 5. Hourly variations of vehicle-pedestrian crashes in Central, 1993 & 2003.

1

3

5

7

9

11 13 15 17 19 21 23

Hour of day

Figure 6. Hourly variations of vehicle-pedestrian crashes in Causeway Bay, 1993 & 2003.

Mon

Tue

Thur

Wed

Fri

Sat

Sun

Day of week

Figure 7. Weekly variations of vehicle-pedestrian crashes in Central, 1993 & 2003. Asian Geographer 24(1-2): 113-128 (2005)



Mon

Tuc

Thur

Wcd

Fri

Sat

Sun

Day of week

Figure 8. Weekly variations of vehicle-pedestrian crashes in Causeway Bay, 1993 & 2003.

2. Spatial characteristics o f vehicle- pedestrian crashes The spatial patterns of vehicle-pedestrian crashes are analyzed both qualitatively and quantitatively. Qualitative analysis refers to the descriptions of cartographic maps (Figures 9-12) produced by GIs. First of all, the maps help to give some preliminary ideas about the locations of vehicle-pedestrian (PC) and non-pedestrian crashes (NPC). It can be seen that they were clearly segregated. In particular, vehicle-pedestrian crashes were found on all types of roads including major roads, secondary roads and minor streets. Non-pedestrian crashes, however, were largely concentrated along the busy major roads.

-

I

600

1200 Meters

O

Figure 9. Distribution of vehicle-pedestrian and non-pedestrian crashes in Central,1993.



Loo & Tsui

600 -

1200 Meters

-

Figure 10. Distribution of vehicle-pedestrian and non-pedestrian crashes in Central, 2003.

Figure I I . Distribution of vehicle- pedestrian and non-pedestrian crashes in Causeway Bay, 1993. 4sian Geographer 24(1-2):113-128 (2005)



Figure 12. Distribution of vehicle-pedestrian and non-pedestrian crashes in Causeway Bay, 2003.

Then, NNA is used to form a more objective judgment on the spatial patterns of vehiclepedestrian crashes. Three major observations can be made from the difference-of-means tests (Table 3). First, all R values are smaller than 1 and the Z values are smaller than 1.96, meaning that the distributions of vehicle-pedestrian crashes in Central and in Causeway Bay were significantly clustered in both 1993 and 2003. The first four sets of null hypotheses, namely Ho(A), Ho(B), Ho(C) and Ho(D), are rejected with 95% confidence. Second, the R values increased over time in both Central (from 0.64 in 1993 to 0.85 in 2003) and Causeway Bay (from 0.48 in 1993 to 0.65 in 2003), implying a tendency of spatial dispersion. Results of the hypothesis testing also suggest that the differences are statistically significantly. In other words, Ho(E) and Ho(F) are rejected. Third, holding the year constant, the distribution of vehicle-pedestrian crashes in the busy shopping area (Causeway Bay) was significantly more concentrated than that in the CBD (Central). In other words, Ho(G) is rejected. When vehicle-pedestrian crashes are analyzed with respect to the pedestrianization zones, the figure has decreased both inside and outside the pedestrianized zones in Central. However, the improvement was only found outside the pedestrianized zones in Causeway Bay. As described earlier, a chi-square test is used to test whether the difference is statistically significant. The results show that the x2values of both Central (1.68) and Causeway Bay (3.60) are not greater than the critical x2 (3.84). In other words, there is no significant association between the introduction of the pedestrianization measures and vehicle-pedestrian crash locations.


125

Loo & Tsui

Table 3. NNA results. -

A (m2)


Ho(A) Ho(B) Ho(C) Ho(D)

1419377.06 1419377.06 2080880.45 2080880.45

n

NNDR

NND

ONND

209 116 224 158

41.20 55.31 48.19 57.38

26.55 50.03 21.31 37.27

1.49 2.68 1.68 2.39

R

Z

Results

0.64 0.90 0.44 0.65

-9.84 -1.97 -15.97 -8.43

Reject Reject Reject Reject

-87.06

Reject

-72.34

Reject

-40.72

Reject

Discussions and Conclusion 1. Extension o f vehicle-pedestrian crash peaks over the day

From 1993 to 2003, the temporal crash analysis shows that crash peaks occurred earlier in the morning and later in the afternoon and the evening in both Central and Causeway Bay. This extension of vehicle-pedestrian crash peaks may be explained by the longer worlung time, longer commuting time and larger variety of activities at the urban core areas. In recent years, working hours in Hong Kong have been getting longer. In addition, the urban sprawl (residential relocations from the city center to the suburb) has increased the trip time of both daily trips (such as the journeys to work and journeys to school) and occasional trips (such as the social and recreational trips). According to the Travel Characteristics Surveys (Transport Department, Hong Kong, 1993; Transport Department, HKSAR, 2003), the average trip journey time had slightly increased from 34.8 minutes in 1992 to 39 minutes in 2002. In addition, the provision of a larger variety of social and recreational activities in the CBD and the busy shopping area (e.g. the opening of more cafes, bars and gymnasiums) and the night lifestyle also contributed to the extended evening peaks and the more obvious night and mid-night peaks in these areas. 2. Different patterns o f variations within the week Day-of-week is another temporal indicator. In the shopping area, little variations of vehicle-pedestrian crashes were found between weekdays and weekends. It may be related to the more ad hoc nature of recreation and shopping activities. People go shopping and recreation at anytime they want, both on weekdays (especially, after work or school) and weekends. As a result, the number of vehicle-pedestrian crashes in the busy shopping area is consistently high throughout the week. In the CBD, vehiclepedestrian crashes did vary quite noticeably over the week. Daily employment trips attract huge volumes of pedestrians and automobiles to the CBD, leading to higher :rash risk during weekdays. The risk obviously diminishes at weekends because fewer pedestrians and vehicles are on the roads. In 2003, the problem aggravated slightly on Sundays. This may be due to the increased provision of bars and cafes (e.g. in Lan Kwai Fong and Soho areas) and the availability of many tourist spots and heritage sites :e.g. St John's Cathedral, Central Police Station, Hong Kong Zoological and Botanical Sardens and IFC Mall) in Central. Further research on the exposure factor (as discussed isian Geographer 24(1-2):113-128 (2005)

126


above in Section 4.1) and the activity patterns of people at the busy CBD would be beneficial.


3. Some policy implications about the temporal analysis The temporal analysis of vehicle-pedestrian crashes provides valuable insights on road safety planning and policies. In response to the extended hourly crash peaks, part-time pedestrianization measures may need to be lengthened for both the CBD and the shopping area. Moreover, safety measures should be tailor-made for the two districts because of the differences in weekly variations. For instance, access restriction of some vehicle types into the CBD should be considered. For the shopping area, pedestrianization measures should be implemented throughout the week, that is, both on weekdays and weekends, due to the indistinctive weekly variation of vehiclepedestrian crashes.

4. Spatial clusters o f vehicle-pedestrian crashes In this paper, vehicle-pedestrian crashes in the CBD and shopping area were found to be clustered in both 1993 and 2003. The results may be related to the high trip generation rates of the two study areas. There is a large variety of facilities providing different services, e.g. financial and commercial, personal and government services, at the CBD. Moreover, many facilities, such as parks, city halls, clubs, cinemas, retail shops and restaurants, can be found in both the CBD and the busy shopping area. The agglomeration of these activities leads to the high concentration of pedestrians and vehicular traffic at certain locations, which ultimately resulted in the clustering of vehicle-pedestrian crashes. Furthermore, the limited pedestrian space within the CBD fails to accommodate for such a large volume of road users and, sometimes, forces pedestrians to walk on roads and compete with the traffic for the limited road space. Moreover, the clustered pattern implies that some areas are more hazardous than others. These dangerous areas should be the targeted locations of road safety measures with higher priority. In order to find out the relationship between vehicle-pedestrian crashes and spatial environmental factors, further research is needed.

5. Pedestrian safety in the shopping district Although vehicle-pedestrian crashes in both study areas were clustered, their degree of concentration differed. The spatial pattern of vehicle-pedestrian crashes in the shopping district was found to be significantly more concentrated than that of the CBD. It may be attributed to the high concentration of famous landmarks such as Sogo, Times Square and the World Trade Centre along the major roads. Another possible reason is that some pedestrians in the shopping district (especially tourists and occasional shoppers) may be less familiar with the local environment and, hence, less aware of the dangerous road locations. Therefore, vehicle-pedestrian collisions occur repeatedly at certain hazardous locations. Targeted safety measures such as the segregation of pedestrians and traffic, the setting up of pedestrian priority zones, the addition of pedestrian crossing facilities and the launching of pedestrian safety campaigns within the shopping area should be planned to improve pedestrian safety in the shopping district. In conclusion, this study focuses on the characteristics of vehicle-pedestrian crashes in the busy commercial and shopping areas. One limitation is that the "like-like"



Loo & Tsui

127

comparison of Central and Causeway Bay may not completely reflect the characteristics of vehicle-pedestrian crashes in busy commercial and shopping areas. The analysis can be enriched by further comparisons with other lower-density areas, such as the rural districts, new towns, industrial areas and university campus. Nonetheless, this study provides some valuable insights about pedestrian safety in the busy commercial and shopping areas. In both the CBD and the shopping areas, vehiclepedestrian crashes were concentrated in time (crash peaks) and space (crash clusters). Distinctive features were identifiable for both types of commercial and shopping areas. The findings of this paper can facilitate transport engineers, safety researchers and policy-makers to devise effective policies and measures to improve pedestrian safety in these highly-dense areas. A reduction of vehicle-pedestrian crashes means an improvement of the quality of life. It is hoped that the findings can contribute to the creation of a safe and friendly environment for pedestrians, especially in the busy commercial and shopping areas.

Acknowledgements The authors are grateful to the Road Safety and Standards Division of the Transport Department of HKSAR for providing the road crash database. This project is funded by the Research Grant Council (HKU 7296104H).

Endnotes 1. A road crash fatality is defined as death caused by the road crash and occurring within 30 days. 2. Central belongs to the Central and Western District in the road traffic crash dataset. 3. Causeway Bay spans over the Eastern and Wan Chai Districts. Thus, its share of pedestrian crashes in the total is the average of these two districts.

References Braddock, M., Lapidus, G., Cromley, E., Cromley, R., Burke, G. and Banco, L. (1994) Using a geographic information system to understand child pedestrian injury. American Journal of Public Health 84(7): 1158-1161. McGrew, J. Chapman and Monroe, Jr. Charles B. (2000) An Introduction to Statistical Problem Solving in Geography (2ndedition), 305 pp. Boston: McGraw-Hill. Chipman, M.L., MacGregor, C.G., Smiley, A.M. and Lee-Gosselin, M. (1992) Time vs. distance as measures of exposure in driving surveys, Accident Analysis and Prevention 24(6): 679-684. Clark, P.J. and Evans, F.C. (1954) Distance to nearest neighbour as a measure of spatial relationships in populations. Ecology 334): 445-453. European Conference of Ministers of Transport (2003) Statistical Report on Road Crashes in 1999-2000. European Conference of Ministers of Transport. Kam, B.H. (2003) A disaggregate approach to crash rate analysis. Crash Analysis and Prevention 35(5): 693-709. Levine, N., Kim, K. and Nitz, L. (1995) Spatial analysis of Honolulu motor vehicle crashes I: spatial patterns. Accident Analysis and Prevention 27(5): 663-674. Loo, Becky P.Y. (2006) Validating crash locations for quantitative spatial analysis: a GIs-based approach. Accident Analysis and Prevention 38(5): 879-886.



128

Vehicle-pedestrianCrashes in Hong Kong

Noland, R.B. and Quddus, M.A. (2004) A spatially disaggregate analysis of road casualties in England. Accident Analysis and Prevention 36(6): 973-984. Peled, A., Haj, Yehia, B. and Hakkert, A.S. (1996) ArcInfo-based geographical information system for road safety analysis and improvement [online]. Available from: http://ais.esri.com~librarv/userconf/proc96/T05O/PAP005/P5.HTM [Accessed 15 November 20041 Road Safety and Standards Division, Transport Department, HKSAR (1993-2003) Road TrafJic Crash Statistics (1993-2003). Hong Kong: Road Safety and Standards Division, Transport Department. Schneider, R.J., Ryznar, R.M. and Khattak, A.J. (2004) A crash waiting to happen: a spatial approach to proactive pedestrian planning. Accident Analysis and Prevention 36(2): 193-211. Transport Department, Hong Kong (1993) Travel Characteristic Survey: Final Report. Hong Kong: Territory Transport Planning Division, Transport Department. Transport Department, HKSAR (2003) Travel Characteristics Survey 2002 Final Report [online]. Available from: http:llwww.info.gov.hkltdlena/publication/tcs/TCS main.htm [Accessed 20 November 20041 Wong, S.C., Sze, N.N., Yip, H.F., Loo, B.P.Y., Hung, W.T. and Lo, H.K. (2006) Association between setting quantified road safety targets and road fatality reduction. Accident Analysis and Prevention 38(5): 997- 1005.
