Global and Local Spatial Autocorrelation of Motorcycle Crashes in

Chile

Carola Blazquez

and María José Fuentes

Department of Engineering Sciences, Universidad Andres Bello, Quillota 980, Viña del Mar, Chile

Universidad Andres Bello, Quillota 980, Viña del Mar, Chile

Keywords: Motorcycle Crashes, Spatial Autocorrelation, Clusters.

Abstract: In Chile, the usage of motorcycles as a mode of transport is growing in unison with the number of crashes

that have arisen in recent years. Spatial statistical methods were used in this study to determine whether

motorcycle crashes showed spatial clustering over time from a global and local perspective. The global spatial

autocorrelation results indicate that high intensity clusters of collisions at intersections with traffic signals and

curved road sections resulting in fatalities persisted during the five-year study period. Locally, recurrent high

spatial patterns of motorcycle collisions arose along straight road sections and on sunny days due to the loss

of control of the vehicle, or the imprudence of the driver or pedestrian. Communes located in the centre zone

of Chile, particularly in the city of Santiago and the surrounding areas, presented a large number of highly

clustered crash attributes. The findings of this study may help authorities to target efforts towards policy

measures to improve motorcycle safety in Chile.

1 INTRODUCTION

According to the World Health Organization, traffic

crashes cause 1.2 million fatalities every year and are

the main cause of death of young adults between 15

and 29 years of age worldwide. Approximately 23%

of these deaths are motorcyclists, 22% are

pedestrians, and 4% are cyclists (WHO, 2015). In

Chile, 2,178 people were killed as a result of traffic

crashes in 2016, presenting an increase of 4.9% with

respect to 2010. This high mortality rate is partly due

to the exponential increase of vehicles in the last few

years. Additionally, Chile is the OECD member

country with the worst fatality rate with 11.9 per

100,000 inhabitants (IRTAD, 2017).

In Chile, almost 19,000 crashes occurred between

2011 and 2015 that involved motorcycles. The

national statistics indicate that deaths caused by such

crashes are ranked third and that the total number of

injuries are placed fourth with respect to other types

of crashes (CONASET, 2016). Being vulnerable road

users, motorcyclists are 27 times more frequently

killed in crashes per travelled vehicle mile than motor

vehicle passengers (NHTSA, 2012).

The motorcycle market increases every year in

many countries worldwide, and it is expected to

continue increasing in Chile as well (ANIM, 2015;

MT Motores, 2016). On average, the total number of

motorcycles has increased in 65% between 2011 and

2015. In 2016, 175,019 motorcycles were registered

throughout the country with approximately 9.6

motorcycles per 1,000 inhabitants (INE, 2016).

Motorcycles are deemed as an economical and

convenient transport mode with respect to congestion,

fuel consumption, etc. Therefore, it may be

anticipated that the number of motorcycle crashes

will grow in time. Thus, there is a need for a spatial

and temporal analysis of these crashes in Chile.

Recent studies have analysed motorcycle crashes

employing different approaches. A multiple

correspondence analysis was performed by Jalayer

and Zhou (2016) to conclude that light conditions,

time of day, driver condition, and weather conditions

are the key factors contributing to the frequency and

severity of at-fault motorcycle-involved crashes in

the state of Alabama. Flask et al (2014) employed

Bayesian multi-level mixed effects models to analyse

motorcycle crashes at the road segment level. The

authors concluded that among different

characteristics of the road segments, smaller lanes

and shoulder widths, larger horizontal degree of

curvature and larger maximum vertical grades will

increase the prediction of crashes. In another study, a

Blazquez, C. and Fuentes, M.

Global and Local Spatial Autocorrelation of Motorcycle Crashes in Chile.

DOI: 10.5220/0007716701590170

In Proceedings of the 5th International Conference on Vehicle Technology and Intelligent Transport Systems (VEHITS 2019), pages 159-170

ISBN: 978-989-758-374-2

159

deep learning framework was developed to predict

motorcycle crash severities, which were related to

rider ejection, two-way roads, curved roads, and

weekends (Das et al, 2018). Lee et al (2018)

employed a flexible mixed multinomial logit

fractional split model to analyse the proportions of

crashes by vehicle type (including motorcycles). This

study concluded that the total employment density

has the most significant and negative influence on the

motorcycle crash proportion, and that the proportions

of households with no vehicle negatively impacts the

proportion of motorcycle crashes. Lastly, Chung and

Song (2018) employed multivariable statistical

methods to identify the critical factors associated to

age, motorcycle speed, curved sections, among others

that impact motorcycle crash severity in Korea.

Other researchers have studied the spatial problem

using statistical methods such as spatial

autocorrelation to identify spatial clusters of crashes.

For example, Dezman et al (2016) analysed hotspots

of traffic crashes at the census tract level in Baltimore

using spatial autocorrelation techniques. Spatial

autocorrelation was used to examine hotspots of time

of occurrence, severity, and location of traffic crashes

aggregated to the traffic analysis zonal level in Shiraz,

Iran (Soltani and Askari, 2017). In another study,

Pour et al (2018) applied spatial autocorrelation to

detect any dependency between time and location of

vehicle-pedestrian crashes in Melbourne. Blazquez et

al (2018) performed a spatial autocorrelation of cargo

trucks on Chilean highways at the global and local

level. Yet another study was performed by Aghajani

et al (2017) to identify spatial and temporal patterns

of traffic crashes, and to determine hotspots of fatal

and injury outcomes in Iran.

The authors are not aware of any study performed

in Chile that employs spatial autocorrelation methods

to analyse recurrent spatial clustering of motorcycle

crashes through time. The objective of this study is to

employ spatial statistical indicators to distinguish

significant patterns of motorcycle crashes at the

commune level in Chile, and to assess whether a

spatial dependence of such patterns exists with

respect to the main crash attributes (e.g., type of

crash, relative location, contributing factors, and

weather) that persisted during the 2011-2015 period.

The results of this macroscopic crash study provides

a decision-making tool for helping authorities and

safety professionals allocate resources and apply

policy based countermeasures.

2 METHODOLOGY

The spatial statistical methods were applied to

determine the spatial association of the value of a

certain variable at a given location with values of that

variable at neighbouring locations at the global and

local level (Mitra, 2009). First, the global Moran´s I

index was employed to test the general spatial

autocorrelation of the main crash attributes for each

year of the studied period. Second, a local Moran´s I

statistic was employed to detect statistically

significant clusters with respect to each of these crash

attributes. The following subsections describe each

statistic.

2.1 Global Spatial Autocorrelation

The global Moran´s I indicator is used to identify

statistically significant spatial patterns of crashes by

quantifying the magnitude of clustering or dispersion

of these crashes with Equation 1.

 



 











 







 













 





 











(1)

where x

is the variable value at a particular location

i,  is the mean of the variable, w

are the elements of

a spatial matrix with weights representing proximity

relationships between location i and neighbouring

location j, S

is the summation of all elements w

, and

n is the total number of locations.

The values of Moran´s I may range between -1

(representing perfect dispersion with a strong

negative autocorrelation) and 1 (indicating perfect

clusterisation with a strong positive autocorrelation).

A random spatial pattern exists when the value of

Moran´s I is near zero. The results of the spatial

autocorrelation are interpreted within the context of

its null hypothesis, which denotes that an attribute is

randomly distributed among features in the study

area. The Z score method is employed to compute the

statistical significance of the Moran´s I index. A

positive Z score for a feature indicates that the

neighbouring features have similar values, whereas a

negative Z score denotes that the feature is

surrounded by dissimilar values.

2.2 Local Spatial Autocorrelation

While the global Moran´s I provides a single value to

measure the overall spatial pattern of a certain

attribute throughout a complete study area, Anselin´s

local Moran´s I examines the existence of local

spatial clusters of similar high, low, or atypical values

VEHITS 2019 - 5th International Conference on Vehicle Technology and Intelligent Transport Systems

160

(e.g., high value surrounded by low attribute value

location, and low values with high attribute value

neighbouring features) at certain locations, as

described in Anselin (1995). Thus, the results of this

statistic shows the value similarity of a location to its

neighbours, and in addition, tests the significance of

this similarity (Meng, 2016). The local Moran´s I

index is expressed by Equation 2.





















  







 









(2)

where x

is the variable value at location i,  is the

mean of the variable, w

is the spatial weight between

locations i and j, S

is the sum of the weights, and n is

the total number of locations.

Similarly to the global indicator, the spatial

patterns are associated to Z score values to determine

the statistical significance of the results. Positive Z

score values imply that neighbouring values are

similar and negative values indicate that near values

are dissimilar (Manepalli et al, 2011). This study will

focus on identifying locations of clusters of

motorcycle crashes with particularly high crash

attribute values.

3 DATA DESCRIPTION

The 2011-2015 crash database employed in this study

was provided by the National Commission of Traffic

Safety (CONASET, acronym in Spanish). A total of

18,826 motorcycle crashes were successfully

aggregated into 343 communes, as shown in Figure 1.

This figure shows that five communes (Arica,

Antofagasta, Copiapó, La Serena, and Coquimbo)

present a high number of such crashes in the north

zone of Chile, many crashes prevailed in several

communes of the centre zone of the country, and the

communes of Coyhaique and Punta Arenas in the

south zone have the largest number of motorcycle

crashes. Note that over 40% of the motorcycle crashes

occurred in the Metropolitan Region during the

studied period, where most of the Chilean population

resides in the capital city, Santiago. Figure 2 shows

an increase in the total number of motorcycle crashes

over time and that this number almost doubled in

2015 with respect to previous years.

a) North Zone

b) Centre Zone

c) South Zone

Figure 1: Motorcycle crashes for the 2011-2015 period

aggregated at the commune level for the a) North Zone, b)

Centre Zone, and c) South Zone.

Santiago

Arica

Antofagasta

Copiapó

La Serena

Coquimbo

Punta

Arenas

Coyhaique

Global and Local Spatial Autocorrelation of Motorcycle Crashes in Chile

161

Figure 2: Number of motorcycle crashes for the 2011-2015

period.

The main attributes of the motorcycle crashes

were classified into six groups (type of injury, type of

crash, relative location, contributing factors, type of

zone, and weather conditions), as shown in Table 1.

This table indicates that most crashes occurred in

urban areas (85%) and on sunny days (86%). The

imprudence of the driver was the main contributing

cause of these crashes, representing 40% of the total

number of crashes. On average, collision between two

or more moving vehicles (56.8%) was the most

frequent type of crash, followed by impacts with

static vehicles or objects (19.9%), pedestrian crashes

(10.3%), and rollovers (8.9%). With respect to the

relative location of motorcycle crashes, 38.0% of

these crashes occurred on straight road segments and

6.7% on curved road sections, whereas 22.7% and

4.4% of motorcycle crashes arose at intersections

with traffic signals and without signage, respectively.

Regarding the type of injury, 405 victims were

killed, and 2,450 people suffered serious injuries as a

result of motorcycle crashes during the studied

period. Male victims were more involved in

motorcycle crashes (80%), and young adults between

19 and 33 years of age (42.6%). Approximately 22%

(4,089) motorcycle crashes occurred between 6 pm

and 9 pm. Friday is the day of the week with the

largest occurrence of motorcycle crashes, which

accounted for 16.3% of all studied crashes. On

average, almost 60% of the crashes occurred between

January and June with the highest number of

motorcycle crashes arising in March (2,069).

4 RESULTS

An incremental spatial autocorrelation analysis was

first performed to obtain a distance threshold or

bandwidth value for each analysed crash attribute and

year. This parameter value maximizes the spatial

autocorrelation (Z score), meaning that a cluster

exists up to this calculated distance with a statistical

significance of 0.01. Both global and local spatial

autocorrelation analyses employed these distance

thresholds, as shown in the following subsections.

Additionally, notice that in both spatial

autocorrelation, Z score values greater than 1.96 with

a 95% confidence were utilised to determine the

statistical significance for each value of the

motorcycle crash attributes.

Table 1: Number of motorcycle crashes for each analysed

variable per year.

Variable

2011

2012

2013

2014

2015

Type of injury

Fatalities

151

Seriously

Injured

146

285

349

334

1336

Slightly

Injured

2127

2647

2588

6569

5339

Type of crash

Collision

1150

1393

1811

1826

4508

Impact

579

649

854

899

779

Pedestrian

crash

318

410

431

407

370

Rollover

156

170

217

175

950

Relative location

Straight

section

1250

1513

1850

1783

3660

Curved

section

138

162

242

265

461

Intersection

with

signage

654

822

967

1019

1877

Intersection

without

signage

111

123

147

148

324

Contributing factors

Imprudence

of driver

866

1079

1756

1433

2428

Imprudence

pedestrian

242

296

312

357

467

Loss of

control

225

282

272

309

1900

Driving

under

influence

alcohol

197

213

237

244

748

Other

causes

420

533

639

732

944

Type of zone

Urban

1962

2341

2903

2928

5868

Rural

353

445

560

517

949

Weather conditions

Sunny

2154

2323

2907

2799

5939

Drizzly

Foggy

Rainy

133

150

180

332

Cloudy

272

351

408

495

VEHITS 2019 - 5th International Conference on Vehicle Technology and Intelligent Transport Systems

162

4.1 Global Moran´s I

Table 2 shows the global spatial autocorrelation

results. The average and standard deviation values of

the global Moran’s I index were computed only for

those crash attributes that presented a recurrent

clustering of at least three years during the studied

period.

Different strength measures of persistent global

spatial patterns are observed among the different

crash attributes. For example, motorcycle crashes that

occurred at intersections with traffic signals presents

a stronger positive spatial pattern with an average

Moran´s I value of 0.088 during the five years of the

study than any other analysed attribute in the table.

With respect to the type of injury, victims that

were seriously injured have the lowest clustering

intensity with an average Moran´s I value of 0.021

compared to the remaining attributes. Crash-caused

fatalities clustered during all five years of the 2011-

2015 period with a high average clustering intensity

of 0.073, followed by slight injury outcomes with an

average global positive autocorrelation of 0.050.

Regarding the type of crash, collisions between

moving vehicles present a global positive

autocorrelation of 0.053 and are clustered during the

five years of the study period, followed by impacts

with static vehicles (0.036) and pedestrian crashes

(0.034). Whereas, rollovers of motorcycle crashes

were insignificant during the study period.

The relative locations of motorcycle crashes show

positive spatial dependences of arising along straight

(0.053) and curved (0.066) road sections, and at

intersections with and without traffic signals for all

studied years. Insignificant results were obtained for

crashes that occurred on rural zones, and thus, these

were not listed in Table 2. Whilst motorcycle crashes

that arose in urban zones tend to cluster during four

years of the study period with an average clustering

intensity of 0.051.

The contribution cause related to driving under

the influence of alcohol shows the largest clustering

intensity (0.064) among the other contributing

factors, as in the results of Blazquez et al (2015).

However, the loss of control of the vehicle, the

imprudence of pedestrians and the imprudence of

drivers are other causes with high average Moran´s I

values of 0.059, 0.055, and 0.051, respectively, which

persisted for three or more years. Finally, the weather

conditions of motorcycle crashes tend to cluster

during all five years of the study period for sunny

days with an average statistic value of 0.064, and

during four years for drizzly and foggy days with an

average global Moran´s I value of 0.085 and 0.044,

respectively. No significant results were observed for

motorcycle crashes that occurred on rainy or cloudy

days.

Table 2: Results of recurrent global spatial autocorrelation

of motorcycle crashes.

Variable

Average

Moran´s I

Standard

Deviation

Moran´s I

Number of

Clustering

Years

Type of injury

Fatalities

0.073

0.029

Seriously

injured

0.021

0.019

Slightly

injured

0.050

0.041

Type of crash

Collision

0.053

0.027

Impact

0.036

0.015

Pedestrian

crash

0.034

0.036

Relative location

Straight

section

0.053

0.027

Curved

section

0.066

0.031

Intersection

with signage

0.088

0.017

Intersection

without

signage

0.053

0.028

Contributing factors

Imprudence

of driver

0.051

0.041

Imprudence

of pedestrian

0.055

0.041

Loss of

control

0.059

0.036

Driving

under

influence

alcohol

0.064

0.039

Other causes

0.037

0.023

Type of zone

Urban

0.051

0.035

Weather conditions

Sunny

0.064

0.023

Drizzly

0.085

0.023

Foggy

0.044

0.023

4.2 Local Moran´s I

As aforementioned, the local Moran´s I statistic

identifies high and low value clusters, and spatial

outliers. This subsection presents the local spatial

autocorrelation results regarding the location of

motorcycle crash clusters of high attribute values

surrounded by high attribute values (High-High local

spatial pattern, HH).

The number of HH spatial clusters for each

analysed motorcycle crash attribute per year are

shown in Table 3. This table indicates that the largest

Global and Local Spatial Autocorrelation of Motorcycle Crashes in Chile

163

total number of HH crashes (287) during the studied

period are related to motorcycle crashes that occurred

on road segments with straight sections, followed by

roads with curved sections and fatality outcomes.

Overall, HH clusters appeared in four or five years of

the studied period. However, motorcycle crashes that

resulted in rollovers on rainy or cloudy days along

rural areas present a low existence or lack of

clustering of HH values over time, concurring with

the global spatial autocorrelation results.

A considerable increase in the number of HH

spatial crash clusters that arose in rural areas are

detected in the last couple of years. This increase

should be further investigated using crash data from

more recent years to identify any additional trend.

Also notice that the total number of HH clusters for

all contributing causes of motorcycle crashes is

greater than 200, which highlights the importance of

these factors among the generation of these crashes.

On average, there are several large number of HH

clusters of motorcycle collisions that occurred on

sunny days along straight or curved road segments

caused by the loss of control of the vehicle or the

imprudence of the driver or the pedestrian generating

fatality outcomes.

Note that although approximately 7% of all

reported crashes occurred on curved road segments,

these tend to locally cluster with high values over

time. Similarly, very few motorcycle crashes

occurred on drizzly days compared to the other

weather conditions. However, 226 HH clusters of

such crashes arise on drizzly days during the 2011-

2015 period.

Figures 3-8 present spatial clusters at the

commune level for each analysed motorcycle crash

attribute that persisted for three, four, or five years of

the studied period using the local Moran´s I statistic.

These figures depict that the communes belonging to

four regions of the country (Region of Valparaiso,

Metropolitan Region, and regions of O´Higgins and

Maule) represent statistically significant HH spatial

patterns. This result may be explained by the high

population and the substantial increase in the usage of

motorcycles as a transport mode in these four regions

between 2011 and 2015.

Figure 3 shows the HH clusters of fatality

outcomes that persisted for the whole studied period

in the communes situated in the Metropolitan Region,

the Region of Valparaiso, and the Region of

O´Higgins. Whereas HH clusters of seriously or

slightly injured victims are recurrent for less number

of years and only in some communes of these regions.

Similarly to the results presented in Table 3,

Figure 4 shows that collisions represent the largest

Table 3: Number of HH spatial clusters of motorcycle crash

attributes that arose during the 2011-2015 period.

Variable

Year

∑HH

2011

2012

2013

2014

2015

Type of injury

Fatalities

(46.4)

(44.9)

(48.4)

(58.4)

(61.5)

253

Seriously

Injured

(31.5)

(33.0)

(38.4)

(44.7)

149

Slightly

Injured

(44.7)

(20.8)

(12.8)

(45.7)

(56.4)

223

Type of crash

Collision

(41.6)

(31.3)

(33.6)

(59.8)

(42.5)

224

Impact

(32.6)

(70.3)

(80.7)

(52.2)

4 (72.3)

164

Pedestrian

crash

(28.4)

12 (4.4)

7 (11.6)

(45.4)

(45.3)

184

Rollover

(23.4)

7 (16.3)

14 (5.9)

(39.6)

Relative location

Straight

section

(34.7)

(38.2)

(37.1)

(35.9)

(53.7)

287

Curved

section

(41.2)

(39.4)

(37.4)

(53.2)

(55.0)

253

Intersection

with signage

(83.3)

(63.8)

(52.2)

(75.1)

(84.1)

176

Intersection

without

signage

(42.5)

(37.5)

(54.8)

(52.1)

(77.6)

183

Contributing factors

Imprudence

of driver

(44.5)

(17.1)

(37.5)

(58.9)

(20.2)

240

Imprudence

of pedestrian

(24.3)

(39.7)

(43.1)

(53.5)

(58.3)

240

Loss of

control

(13.3)

(61.2)

(36.9)

(49.1)

(53.7)

245

Driving

under

influence

alcohol

(12.1)

(31.3)

(15.1)

(56.2)

(54.6)

232

Other causes

(62.9)

(38.4)

(21.3)

(63.9)

(44.3)

206

Type of zone

Urban

(46.9)

4 (13.3)

(11.3)

51 (5.9)

(20.8)

195

Rural

8 (9.8)

5 (5.1)

(39.1)

(20.4)

123

Weather conditions

Sunny

(49.6)

(42.5)

(36.2)

(52.4)

(53.1)

240

Drizzly

(48.1)

(50.7)

(59.5)

(57.2)

226

Foggy

(26.4)

(41.9)

(44.2)

(35.2)

150

Rainy

(43.2)

(52.5)

(61.4)

Cloudy

(72.8)

(59.9)

Note: Average local Moran´s I values are shown in parenthesis.

number of HH spatial clusters among other types of

crashes. These clusters are mostly located in

communes of the Region of Valparaiso and

Metropolitan Region. Notice that rollover-related

crash HH clusters are not shown since such clusters

in all communes were positive and significant for less

than three years.

Figure 5 presents the recurrent HH clusters of

motorcycle crashes with respect to their relative

VEHITS 2019 - 5th International Conference on Vehicle Technology and Intelligent Transport Systems

164

location at the commune level. This figure shows that

HH clusters in some communes in the Region of

Valparaiso and in the Metropolitan Region, and in the

commune of Rancagua in the Region of O´Higgins

persisted for the five years of the studied period, and

for all types of relative locations. Conversely,

recurrent HH clusters appeared in many communes in

the centre of the Metropolitan Region (coinciding

with the city of Santiago) along straight and curved

road sections.

Communes in the four regions shown in Figure 6

present persistent HH clusters due to the imprudence

of the pedestrian, whereas clustering of crashes due to

imprudence of the driver that persisted for all five

years of the 2011-2015 are concentrated in the city of

Santiago, a couple of communes in the Region of

Valparaiso, and Rancagua in the Region of

O´Higgins. This figure also suggests that motorcycle

crashes due to the loss of control and driving under

the influence of alcohol are highly clustered during

the five years of the studied period in a few

communes in the city of Santiago and Region of

Valparaiso. HH clusters appear in a lesser degree a a

result of other causes.

Recurrent spatial clustering of crashes that

occurred in urban areas are recurrent for 3 or 4 years

during the studied period in communes of the regions

of O´Higgins and Maule, as shown in Figure 7. No

HH clustering of crashes in rural zones was perceived

in any commune for three or more years.

Regarding the weather conditions, Figure 8

depicts HH spatial clusters of crashes that arose on

sunny and drizzly days that persisted for three or more

years. Concurring with the results in Table 3, this

figure shows that communes with clustering of

crashes on sunny days persisted for three to five

years, whilst more communes are displayed with

crash clusters during drizzly days, however for three

and four years of the studied period.

a) Fatalities b) Seriously injured c) Slightly injured

Figure 3: HH spatial clusters for each type of injury.

a) Collision b) Impact c) Pedestrian crash

Figure 4: HH spatial clusters for each type of crash.

REGION OF VALPARAISO

METROPOLITAN REGION

REGION OF O´HIGGINS

REGION OF MAULE

REGION OF VALPARAISO

METROPOLITAN REGION

REGION OF O´HIGGINS

REGION OF MAULE

REGION OF VALPARAISO

METROPOLITAN REGION

REGION OF O´HIGGINS

REGION OF MAULE

REGION OF VALPARAISO

METROPOLITAN REGION

REGION OF O´HIGGINS

REGION OF MAULE

Global and Local Spatial Autocorrelation of Motorcycle Crashes in Chile

165

a) Straight section b) Curved section c) With signage d) Without signage

Figure 5: HH spatial clusters for the relative location of motorcycle crashes.

a) Driver´s imprudence b) Pedestrian´s imprudence c) Loss of control d) Alcohol of driver

Figure 6: HH spatial clusters for the contributing factor attribute.

e) Other causes

Figure 6 (continued): HH spatial clusters for the

contributing factor attribute.

a) Urban zone

Figure 7: HH spatial clusters of crashes in urban zones.

REGION OF VALPARAISO

METROPOLITAN REGION

REGION OF O´HIGGINS

REGION OF MAULE

REGION OF VALPARAISO

METROPOLITAN REGION

REGION OF O´HIGGINS

REGION OF MAULE

REGION OF VALPARAISO

METROPOLITAN REGION

REGION OF O´HIGGINS

REGION OF MAULE

REGION OF VALPARAISO

METROPOLITAN REGION

REGION OF O´HIGGINS

REGION OF MAULE

REGION OF VALPARAISO

METROPOLITAN REGION

REGION OF O´HIGGINS

REGION OF MAULE

REGION OF VALPARAISO

METROPOLITAN REGION

REGION OF O´HIGGINS

REGION OF MAULE

VEHITS 2019 - 5th International Conference on Vehicle Technology and Intelligent Transport Systems

166

a) Sunny days

b) Drizzly days

Figure 8: HH spatial clusters for weather conditions.

Figure 9: Communes with the largest number of HH

attribute clusters during the 2011-2015 period.

The top ten Chilean communes with the largest

number of HH for the 23 analysed crash attributes in

the five-year period are shown in Figure 9. This figure

shows that 8 out of these 10 communes with the most

number of HH crash attributes are located in the

Metropolitan Region. The communes in the regions

of O´Higgins and Valparaiso (Rancagua and Quilpue,

respectively) are located in the vicinity of the

Metropolitan Region. The large number of HH

clusters of crashes in this area is due to the fact that

over 50% of Chile´s population resides in the

Metropolitan Region and the surrounding areas, and

approximately 65.5% of the total number of

motorcycles nationwide are registered in these three

regions, which are more prone to be exposed to traffic

crashes.

Table 4 presents the average value of the local

Moran´s I index and average Z score in parenthesis

for recurrent crash attributes for the ten communes

depicted in Figure 9. Those communes with no values

indicate that that particular variable was significant

for less than three years. Therefore, HH crash clusters

are not present for those vehicles that rolled over in

rural areas on cloudy or foggy days for any of the ten

analysed communes. HH clusters only exist for the

attributes associated to urban areas and rainy days in

Rancagua and Santiago, respectively.

Overall, Santiago has the highest intensity of HH

clusters in 13 of the crash attributes when compared

to the rest of the communes, similarly to the findings

in Blazquez and Celis (2013) and Blazquez et al

(2016). This result may be attributed to that this

commune has the largest number of registered

motorcycles in Chile with a total of 5.571

motorcycles in 2015, and an increase of 23.6% in the

number of registered motorcycles between 2011 and

2015. Additionally, Santiago is a commune that has a

daily floating population of approximately 2 million

people due to its strong political, economic, and

commercial activities in an area of only 23.2 km

, and

a residential population of 404,495 inhabitants (INE,

2017). Authorities and CONASET should prioritize

the promotion and education of the community about

road safety in this commune.

5 CONCLUSIONS

In this study, the global Moran´s I index was

employed to detect overall spatial autocorrelation of

motorcycle crashes in Chile at the commune level. In

addition, a local spatial autocorrelation was

performed with Anselin´s local Moran´s I to identify

statistically significant crashes that are recurrent

during the 2011-2015 period. Six groups of

motorcycle crash attributes were examined in the

autocorrelation analysis. Certain crash attributes tend

to be located closer together than randomly over time.

REGION OF VALPARAISO

METROPOLITAN REGION

REGION OF O´HIGGINS

REGION OF MAULE

METROPOLITAN REGION

Quilpue

Ñuñoa

Las Condes

La Florida

Puente Alto

Santiago

Quinta Normal

Pudahuel

Providencia

Rancagua

REGION OF O´HIGGINS

REGION OF VALPARAISO

METROPOLITAN REGION

REGION OF O´HIGGINS

REGION OF MAULE

Global and Local Spatial Autocorrelation of Motorcycle Crashes in Chile

167

Table 4: Number of HH spatial clusters of motorcycle crash attributes per commune that persisted for three or more years

during the 2011-2015 period.

Variable

Commune

Florida

Las

Condes

Ñuñoa

Providencia

Pudahuel

Puente

Alto

Quilpue

Quinta

Normal

Rancagua

Santiago

Type of injury

Fatalities

96.9

(4.3)

107.2

(4.2)

118.1

(4.2)

117.2 (4.4)

62.4

(4.2)

162.5

(4.2)

84.3

(4.1)

56.8

(4.3)

120.0

(4.0)

207.1

(4.2)

Seriously

Injured

83.2

(2.8)

42.3

(2.7)

41.3

(2.8)

102.8

(2.6)

80.4

(2.5)

50.2 (2.7)

48.7

(2.7)

Slightly

Injured

62.9

(3.7)

71.5

(3.4)

138.3

(3.8)

85.8 (3.8)

41.4

(3.4)

147.5

(3.8)

42.9

(2.9)

36.0

(3.9)

85.8 (3.7)

113.7

(3.3)

Type of crash

Collision

66.6

(3.5)

58.6

(3.6)

96.2

(3.7)

81.7 (3.6)

30.9

(3.5)

124.7

(3.6)

79.3

(3.7)

38.4

(3.8)

84.7 (3.0)

141.1

(3.5)

Impact

35.5

(3.4)

82.1

(3.4)

72.9

(2.7)

86.7 (3.6)

26.5

(3.3)

71.1

(3.7)

105.7

(3.6)

73.0 (3.2)

180.4

(3.6)

Pedestrian

crash

51.9

(3.6)

105.8

(3.5)

40.3

(3.6)

108.7 (3.5)

58.8

(3.6)

117.4

(3.5)

41.7

(3.4)

27.2

(3.7)

62.8 (3.5)

149.6

(3.5)

Relative location

Straight

section

88.0

(3.7)

74.4

(3.7)

108.6

(3.8)

91.9 (3.7)

49.4

(3.6)

163.2

(3.7)

73.7

(3.7)

48.1

(3.6)

95.5 (3.6)

134.4

(3.6)

Curved

section

88.0

(3.8)

86.1

(4.0)

87.9

(3.9)

114.2 (4.1)

50.2

(3.8)

129.1

(3.8)

91.1

(3.9)

52.7

(4.0)

84.1 (3.5)

176.0

(3.8)

Intersection

with signage

106.2

(12.4)

132.5

(15.6)

122.3

(14.3)

128.5

(15.0)

80.8

(9.4)

169.9

(19.8)

108.3

(12.9)

41.0

(4.8)

136.7

(16.0)

216.5

(25.2)

Intersection

without

signage

88.5

(10.3)

53.3

(6.3)

151.2

(17.6)

83.2 (9.7)

84.2

(9.8)

88.8

(10.4)

46.7

(5.6)

84.6

(9.9)

84.6 (9.9)

117.3

(13.6)

Contributing factors

Imprudence

of driver

61.4

(3.4)

64.2

(3.6)

58.2

(3.3)

82.1 (3.4)

32.5

(3.4)

68.5

(3.4)

71.8

(3.7)

38.5

(3.4)

97.5 (3.2)

153.7

(3.1)

Imprudence

of pedestrian

96.0

(4.2)

98.0

(3.8)

104.1

(4.0)

103.1 (4.1)

50.7

(3.7)

142.1

(4.1)

75.0

(3.7)

43.3

(3.9)

86.3 (3.5)

187.1

(4.1)

Loss of

control

58.9

(3.9)

124.0

(4.2)

111.7

(3.8)

137.5 (3.7)

69.1

(4.2)

174.8

(4.2)

38.1

(3.8)

42.4

(4.4)

70.8 (4.0)

194.6

(4.2)

Driving

under

influence

alcohol

75.9

(3.7)

105.3

(3.7)

103.9

(3.6)

98.5 (3.6)

44.8

(3.5)

111.3

(3.7)

69.4

(3.7)

58.3

(3.6)

95.7 (3.5)

149.3

(3.5)

Other causes

74.3

(3.2)

59.9

(3.4)

122.3

(3.3)

76.5 (3.4)

92.4

(3.4)

93.4

(3.2)

109.4

(3.4)

32.0

(3.4)

67.2 (3.3)

62.8

(2.9)

Type of zone

Urban

57.0 (2.6)

Weather conditions

Sunny

96.2

(4.0)

98.1

(4.1)

71.9

(4.0)

84.6 (4.4)

78.7

(3.9)

161.8

(3.9)

72.9

(4.5)

52.5

(4.0)

94.7 (3.9)

133.5

(3.9)

Drizzly

113.3

(4.7)

122.5

(4.5)

136.0

(4.7)

133.4 (4.6)

74.6

(4.6)

171.5

(4.5)

104.0

(4.5)

65.5

(4.7)

129.0

(4.2)

235.8

(4.5)

Rainy

62.3

(2.3)

From a global perspective, the results indicate that

crash attributes associated with intersections with

traffic signals as a relative location, collisions, and

fatality outcomes are spatially autocorrelated for the

whole study period with the largest intensities among

the remaining analysed attributes. In addition, driving

under the influence of alcohol on drizzly days

strongly clustered during four years of the study

period.

The findings from the local spatial autocorrelation

technique revealed similarities and differences

among the communes. The communes with the

largest number of HH clusters are portrayed,

indicating the persistence and intensity of this

clustering for each group of crash attributes. In

particular, communes located in city of Santiago are

smaller in size and closer together, and present high

number of spatial clustering of motorcycle crashes. In

addition, the presence and persistence of HH spatial

VEHITS 2019 - 5th International Conference on Vehicle Technology and Intelligent Transport Systems

168

clusters of crash attributes in the communes located

in the Region of Valparaiso, Metropolitan Region,

Region of O´Higgins, and Region of Maule are

particularly distinguished.

Overall, a large number of HH clusters of

collisions due to the loss of control are present on

sunny days in the aforementioned regions. Most

motorcycle crashes also tended to spatially cluster

along straight and curved road segments. Drivers tend

to increase their travel speeds as straight road sections

are encountered, which may increase the likelihood of

causing crashes with serious outcomes. Motorcyclists

are more prone to crashes at curves, which may

generate a significant impact on crash severity or

fatality, as shown in Chung and Song (2018).

There is a considerable increase in the use of

motorcycles nationwide in the last few years,

particularly in the centre part of Chile. However, the

number of motorcycle crashes has also presented a

dramatic increase. Future research should prioritize

those communes with high clustering of motorcycle

crashes, in order to implement specific interventions

that help improve traffic safety.

The Road Coexistence Law became effective in

November of 2018 with the aim of equating road

spaces and imposing an equality among all transport

modes (motorised vehicles, bicycles, pedestrians,

etc.). This law enforces road users to become aware

of their rights and obligations when travelling, and

thus, increasing road safety. Further investigation is

required to analyse the impact of this law in the traffic

crashes that involve motorcycles.

REFERENCES

Aghajani, M., Dezfoulian, R., Arjroody, A., Rezaei, M.,

2017. Applying GIS to identify the spatial and temporal

patterns of road accidents using spatial statistics (case

study: Ilam Province, Iran). Transportation Research

Procedia, 25:2126-2138.

Asociación Nacional de Importadores de Motocicletas,

ANIM, 2015. Actualidad del mercado en Mundo Motos

2016 http://www.anim.cl/httpwww-anim-clp1820/

(Accessed on September 22, 2018)

Anselin, L., 1995. Local indicators of spatial association-

LISA. Geographical Analysis, 27: 93–115.

Blazquez, C. and Celis, M., 2013. A Spatial and Temporal

Analysis of Child Pedestrian Crashes in Santiago,

Chile. Accident Analysis & Prevention, 50:304-311.

Blazquez, C., Puelma, I. and Khan, G., 2015. Spatial

Analysis of Bicycle Crashes in Chile. In Proceedings

18th IEEE International Conference on Intelligent

Transportation Systems, Las Palmas de Gran Canaria,

Spain, 2745-2750.

Blazquez, C., Lee, J. S., and Zegras, C., 2016. Children at

Risk: An Initial Comparison of Child Pedestrian Traffic

Collisions in Santiago, Chile and Seoul, South Korea.

Traffic Injury Prevention, 17(3):304-312.

Blazquez, C., Picarte, B., Calderon, J.F., and Losada, F.,

2018. Spatial Autocorrelation Analysis of Cargo Trucks

on Highway Crashes in Chile. Accident Analysis &

Prevention, 120:195-210.

Chung, Y., and Song, T-J, 2018. Safety analysis of

motorcycle crashes in Seoul metropolitan area, South

Korea: An application of nonlinear optimal scaling

methods. International Journal of Environmental

Research and Public Health, 15(12):2702.

CONASET, 2016. Siniestros de Tránsito de ocupantes de

motocicletas y consecuencias. https://www.conaset.cl/

programa/observatorio-datos-estadistica/biblioteca-

observatorio/informes-tematicos/#motocicletas

(Accessed on Abril 13, 2018)

Das, S., Dutta, A., Dixon, K., Minjares-Kyle, L., and

Gillette, G., 2018. Using deep learning in severity

analysis of at-fault motorcycle rider crashes. Journal of

Transportation Research Board, In press.

Dezman, Z., de Andrade, L., Vissoci, J., El-Gabri, D.,

Johnson, A., Hirshon, J., and Staton, C., 2016. Hotspots

and causes of motor vehicle crashes in Baltimore,

Maryland: A geospatial analysis of five years of police

crash and census data. Injury, 47(11):2450-2458.

Flask, T., Schneider, W., and Lord, D., 2014. A segment

level analysis of multi-vehicle motorcycle crashes in

Ohio using Bayesian multi-level mixed effects models.

Safety Science, 66:47-53.

Instituto Nacional de Estadística, INE, 2016. Anuario

Estadísticas vitales. http://www.ine.cl/estadisticas/

demograficas-y-vitales (Accessed on July 31st, 2018)

Instituto Nacional de Estadística, INE, 2017. Censo

https://www.censo2017.cl/ (Accessed on April 15,

2018)

International Traffic Safety Data and Analysis Group,

IRTAD, 2017. Road Safety Annual Report. https://

www.oecd-ilibrary.org/transport/road-safety-annual-

report-2017_irtad-2017-en (Accessed on July 31, 2018)

Jalayer, M., and Zhou, H., 2016. A multiple correspondence

analysis of at-fault motorcycle-involved crashes in

Alabama. Journal of Advanced Transportation, 50:

2089-2099.

Lee, J., Yasmin, S., Eluru, N., Abdel-Aty, M., and Cai, Q.,

2018. Analysis of crash proportion by vehicle type at

trac analysis zone level: A mixed fractional split

multinomial logit modeling approach with spatial

eect. Accident Analysis and Prevention, 111: 12-22.

Manepalli, U., Bham, G., Kandada, S., 2011. Evaluation of

hot spots identification using Kernel Density

Estimation (K) and Getis-Ord (Gi*) on I-630. In

Proceedings 3

International Conference on Road

Safety and Simulation 17.

Mitra, S., 2009. Spatial autocorrelation and bayesian spatial

statistical method for analysing intersections prone to

injury crashes. Journal of Transportation Research

Board, 2136: 92–100

Global and Local Spatial Autocorrelation of Motorcycle Crashes in Chile

169

Meng, Q., 2016. The spatiotemporal characteristics of

environmental hazards caused by offshore oil and gas

operations in the Gulf of Mexico. Science of the Total

Environment, 565: 663-671.

MT Motores, 2016. Espacial Mercado de motos en Chile.

Diario La Tercera, https://www.pressreader.com/chile/

la-tercera-mt-motores/20160930/281505045711190

(Accessed October 2, 2018)

Pour, A., Moridpour, S., Tay, R., Rajabifard, A., 2018.

Influence of pedestrian age and gender on spatial and

temporal distribution of pedestrian crashes. Traffic

Injury Prevention, 19(1): 81-87.

National Highway Traffic Safety Administratio, NHTSA,

2012. Traffic Safety Facts. https://crashstats.nhtsa.

dot.gov/Api/Public/ViewPublication/812016

(Accessed on November 10, 2018)

Soltani, A., and Askari, S., 2017. Exploring spatial

autocorrelation of traffic crashes based on severity.

Injury, 48(3): 637-647.

World Health Organization, WHO, 2015. Global status

report on road safety. https://www.who.int/

violence_injury_prevention/road_safety_status/2015/

GSRRS2015_Summary_EN_final2.pdf (Accessed

September 30, 2018).

VEHITS 2019 - 5th International Conference on Vehicle Technology and Intelligent Transport Systems

170