Factors Influencing Stopping Locations of Motorcycle Riders on Signalized Urban Intersection Approaches

Abstract

In developing countries, motorcycle riders normally attempt to stop at their desired locations during queue formation on signalized intersection approaches. Under mixed-traffic conditions, motorcycle positioning in a queue affects the operational and safety performance of the intersection. This study aimed to identify factors influencing motorcycle riders’ stopping locations at signalized urban intersections. This study applied Unmanned Aerial Vehicles (UAVs) to observe the stopping behavior of 1413 motorcycle riders on 24 approaches from 10 signalized intersections in Thailand (N = 1413). Multinomial logistic regression analysis was used to determine the relationship between the stopping locations of motorcycle riders and rider- and motorcycle-related variables and traffic- and environmental-related variables. The statistical analyses presented a Cox and Snell R² and Nagelkerke R² of 0.466 and 0.499, respectively, indicating that the model accounted for almost 50% of the variation among the five stopping locations of motorcycle riders. The results showed that, under mixed-traffic conditions in Thailand with left-hand traffic, motorcycle riders intending to turn right, the morning peak period, the presence of shadows, motorcycle riders not wearing helmets, the presence of a larger vehicle in the queue, and the density of desired stopping locations significantly influenced the motorcyclists’ choice of stopping locations on signalized intersection approaches. Practical policy-related recommendations drawn from the findings are provided to improve motorcyclists’ safety on signalized intersection approaches.

1. Introduction

Currently, many developing nations are encountering the problem of overcrowded roads due to overpopulated cities, an overwhelming number of vehicles, and unplanned urban development. Motorcycles have become a popular travel mode on mixed-traffic roads in developing and dense cities. Motorcycles have many advantages compared to four-wheeled vehicles, as they provide higher accessibility, higher reliability, lower prices, and lower operation and maintenance costs. They are daily traveling vehicles for low-income families. Motorcycles in developing countries are not limited to physically static lanes as with other types of four-wheeled vehicles, due to their small size (i.e., an engine size under 200 cc), acceleration rate, and means of control. They generally do not adhere to the “first in first out’’ rule, unlike large motorcycles (i.e., an engine size ≥ 200 cc) [1]. On signalized intersection approaches, most motorcycle riders intend to maneuver ahead of other traffic. The main reasons for the maneuvering of motorcycle riders in a queue are as follows: (1) an attempt to stop at a favorable location (motorcycle riders intend to move forward and position themselves ahead of other traffic so that they can freely depart when the traffic light turns green); (2) an attempt to avoid traveling behind a large vehicle; and (3) preparation for making a turn at an intersection. To stop at the location closest to the stop line [1], motorcycle riders generally filter between two lanes of other traffic on a mixed-traffic signalized intersection approach, which is known as lane filtering [2].

However, most developing countries are now facing a critical challenge—a high motorcycle fatality rate. Specifically, Thailand has the highest rate of deaths due to motorcycle crashes in the world [3]. Motorcycle crashes at intersections are more likely to result in fatal or serious injuries, as compared to those at non-intersections [4]. To account for motorcycle crashes in developing countries, there were a number of research studies focusing on the crashes and risky behavior of motorcycle riders—for example, the injury severity of motorcyclists [4,5,6,7,8,9,10,11], Red Light Running (RLR) behavior [12,13,14], failure to wear a helmet [15], speeding behavior [16], and other risky behaviors [17,18,19,20]. To decrease the number of motorcycle crashes at a signalized intersection, the behavior of motorcycle riders on signalized intersection approaches should be explored. In recent years, many studies [21,22,23] have investigated the behavior of motorcycle riders when approaching or passing unsignalized or signalized intersections. However, there is a lack of study regarding the stopping behavior of motorcycle riders on signalized intersection approaches. This study can fill this knowledge gap. In many developing cities, many motorcycle riders stop in front of a stop line, which is against traffic law, yet they are rarely reprimanded by police officers. Moreover, many motorcycle riders perform RLR by departing before the green light at signalized intersections [12,14]. This risky behavior increases the likelihood of a crash. Factors influencing motorcycle riders’ decisions regarding stopping positions must be explored to determine the most appropriate measures that can encourage motorcycle riders to stop at safer locations according to the safe system approach [24].

The objective of this study was to examine factors influencing the stopping positions of motorcycle riders on signalized intersection approaches in a mixed-traffic context in developing countries. The naturalistic observational study was conducted by using UAVs for data collection. The scope of this study was to explore the stopping behavior of motorcycle riders during daytime at small-sized signalized intersections in an urban area. The study selected Khon Kaen City as a case study. In Thailand, police rarely reprimand motorcycle riders for stopping at illegal locations, particularly in front of a stop line. These conditions are normal in developing countries.

The remainder of the paper is organized as follows. Section 2 discusses the relevant studies. Section 3 describes the research methodology. Section 4 contains the results and discussion. In Section 5, conclusions and recommendations are presented.

2. Literature Review

There has been a number of research studies regarding the riding behavior of motorcycle riders in mixed-traffic conditions, as summarized in

Table 1. Summary of previous studies on mixed-traffic riding behavior of motorcycle riders.

Authors	Riding Behavior	Observation Method	Analytical Technique
Minh et al., 2012 [1]	Maneuvering in queue at signalized intersections	Video cameras installed on road infrastructures and buildings	Dynamic model
Nikias et al., 2012 [25]	Maneuvering in urban arterials	Video cameras installed on road infrastructures and buildings	Linear regression model
Babu et al., 2015 [26]	Maneuvering in urban arterials	Video cameras installed on road infrastructures and buildings	Logit model
Trinh et al., 2018 [29]	Cut tail and giving way behaviors at signalized urban intersections	Video cameras installed on road infrastructures and buildings	Two-player game-theory-based analysis
Das et al., 2019 [27]	Filtering in urban arterials	Video cameras installed on road infrastructures and buildings	Logit model
Dong et al., 2019 [28]	Maneuvering in urban arterials	Video cameras installed on road infrastructures and buildings	Dynamic discrete choice model
Ramlan et al., 2021 [22]	Right-turn movements at unsignalized intersections	Video cameras installed on road infrastructures and buildings	Logit model
Promraksa et al., 2022 [2]	Lane filtering at signalized urban intersections	Lidar onboard device	Multilevel linear regression

. Many of them studied the behavior of motorcycle riders maneuvering on urban arterial corridors [25,26,27,28]. Some of them studied the behavior of motorcycle riders approaching or passing through signalized intersections. A few studies investigated the maneuvering behavior in queues [1,2] and the crossing behavior [29,30] of motorcycle riders at signalized urban intersections. These studies explored various behaviors of motorcycle riders riding along urban road networks. However, study on the stopping behavior of motorcycle riders at signalized urban intersections is lacking. A study regarding this stopping behavior is required to fully understand riding behavior when approaching, stopping, and passing through signalized intersections.

Based on literature reviews and observational studies, seven research hypotheses can be put forward regarding motorcycles stopping on signalized urban intersection approaches in left-hand traffic (LHT) road networks, such as Thailand, which are as follows. Motorcycle riders wish to stop at their favorable location on signalized intersection approaches due to convenience (e.g., earlier start, less conflict, and avoidance of exhaust emissions, sunlight, and hot weather). The majority of motorcycle riders may stop in front of the stop line, while not being reprimanded by police. The research’s null hypothesis (No. 0) is that the number of motorcycle riders stopping in front of a stop line is significantly higher than those in other locations.

Firstly, motorcycle riders intending to make a turn at a signalized intersection would stop at locations where they encounter less conflict with other vehicles. Research hypothesis No. 1 is that the turning movement of motorcycles would significantly influence the stopping locations of motorcycle riders on intersection approaches.

Secondly, in developing cities, traffic police normally reprimand motorcycle riders not wearing a helmet by establishing temporal checking points at intersections, i.e., police inspections [31]. Motorcycle riders who do not wear helmets will stop far away from a stop line if a police inspection is located nearby. Research hypothesis No. 2 is that helmet usage would significantly influence the stopping locations of motorcycle riders.

Thirdly, it is more difficult for motorcycle riders with large motorcycles than small motorcycles to access and stop at their preferred locations on intersection approaches due to their heavy weight, large dimensions, and high power. Research hypothesis No. 3 is that the size of the motorcycle would significantly influence the stopping locations of motorcycle riders.

Fourthly, motorcycle riders who are hurrying to their destinations during peak periods would stop at locations where they can depart before other traffic. Research hypothesis No. 4 is that the time of day would significantly influence the stopping locations of motorcycle riders.

Fifthly, on a sunny day, motorcycle riders would prefer to stop at traffic lights in the shadows of trees or buildings in order to avoid sunlight and hot weather. Research hypothesis No. 5 is that the shadow in queue areas would significantly influence the stopping locations of motorcycle riders.

Sixthly, motorcycle riders maintain longer lateral clearance when they perform lane filtering beside large vehicles [2,25]. Motorcycle riders would not stop near a larger vehicle because they seek to avoid conflict with a larger vehicle, especially in front of and at the lateral side of a large vehicle, truck, or bus, where the drivers sometimes cannot see them (i.e., blind spot). Research hypothesis No. 6 is that a larger vehicle, such as a truck or a bus, in a queue would significantly influence the stopping locations of motorcycle riders.

Lastly, motorcycle riders would not access and stop at their preferred location if it is densely occupied by other motorcycles. Research hypothesis No. 7 is that the density of the desired stopping location would significantly influence the stopping locations of motorcycle riders.

3. Methodology

3.1. Analysis Variables

In this study, the dependent variable was the stopping location of motorcycles on a signalized intersection approach categorized into 5 zones, as displayed in Figure 1. Zone 1 is in front of the stop line, Zone 2 is behind the stop line and in front of a larger vehicle in the queue, Zone 3 is on the left side of a larger vehicle in the queue, Zone 4 is on the right side of a larger vehicle in the queue (close to a line separating traffic from the opposite direction), and Zone 5 is behind a larger vehicle in the queue.

The independent variables were divided into two categories. Traffic- and environment-related variables are the motorcycle density of stopping locations, the presence of a larger vehicle in the queue, the time of day, and the presence of shadows in a queue area. Rider- and motorcycle-related variables are the turning movement of motorcycle riders, the engine size of the motorcycle, and the helmet use of motorcycle riders. The definitions and types of independent variables are presented in

Table 2. List of independent variables.

Variable	Definition	Type
Traffic- and environment-related variables
Time of day	A motorcycle is present and stops at an intersection during morning peak period or evening peak period	Categorical variable 0 = Morning peak 1 = Evening peak
Presence of a larger vehicle in queue	There is a larger vehicle, including a passenger car, a truck, or a bus, in queue	Categorical variable 0 = Yes 1 = No
Presence of shadow	There is a shadow (from trees or structures) in queue areas on an intersection approach	Categorical variable 0 = Yes 1 = No
Motorcycle density in front of a stop line (Zone 1)	The stopping location in front of a stop line (Zone 1) is dense (see Figure 2b)	Categorical variable 0 = Yes 1 = No
Motorcycle density behind a stop line and in front of a larger vehicle in queue (Zone 2)	The stopping location behind a stop line and in front of a larger vehicle in queue (Zone 2) is dense (see Figure 2c)	Categorical variable 0 = Yes 1 = No
Motorcycle density on the left side of a larger vehicle in queue (Zone 3)	The stopping location on the left side of a larger vehicle in queue (Zone 3) is dense (see Figure 2d)	Categorical variable 0 = Yes 1 = No
Motorcycle density on the right ride of a larger vehicle in queue (Zone 4)	The stopping location on the right side of a larger vehicle in queue (Zone 4) is dense (see Figure 2e)	Categorical variable 0 = Yes 1 = No
Motorcycle density behind a larger vehicle in queue (Zone 5)	The stopping location behind a larger vehicle in queue (Zone 5) is dense (see Figure 2f)	Categorical variable 0 = Yes 1 = No
Rider- and motorcycle-related variables
Turning movement of motorcycle	A motorcycle rider is intending to turn right at intersection	Categorical variable 0 = Turning right 1 = Going straight
Helmet use	A motorcycle rider is wearing a helmet	Categorical variable 0 = Not wearing 1 = Wearing
Engine size of motorcycle	A motorcycle rider is riding either a large motorcycle or a small motorcycle	Categorical variable 0 = Large motorcycle (Engine size ≥ 200 cc.) 1 = Small motorcycle (Engine size < 200 cc.)

. The stopping zones are classified as dense if identified zones are densely occupied by other motorcycles; approaching motorcycle can still stop, but it must cross the zone’s boundary line, as displayed in an example of a real-world image recorded by UAV at the studied approach in Figure 2. All zones are not dense, as shown in Figure 2a. Zones 1 to 5 are dense, as shown in Figure 2b–f, respectively.

3.2. Selection of Study Intersections

This study selected 10 signalized urban intersections across Khon Kaen City, Thailand, as the study area. Khon Kaen City is a major city in the northeastern region of Thailand, with a high number of registered motorcycles. The locations of 10 signalized intersections are shown in Figure 3. The traffic flow was controlled by fixed-time signal operation. This study selected 24 approaches from 10 intersections as the study area based on the following criteria: (1) two lanes per approach (one for left turning and one for both right turning and continuing straight); (2) two phases of signal control; (3) division of lanes by pavement marking and no physical islands; and (4) clear pavement markings (stop line and pedestrian crossing). The lane width in the study areas was 3 m, which is a typical lane width in an urban area. The road pavement is either Portland cement concrete or asphaltic concrete.

3.3. Data Collection

The stopping behavior of motorcycle riders in the study area was observed during the morning peak period and evening peak period according to the peak hours reported in a previous study [32].

Since existing observation techniques using video cameras installed on road infrastructures or buildings were unable to provide a high enough perspective to identify clearly the stopping behavior of motorcycle riders, this study applied UAVs equipped with video cameras to monitor it. This is one of the techniques used in naturalistic observational studies, which attempts to observe riding behavior by avoiding any disturbance of the motorcycle riders [33]. Many recent studies applied UAVs to collect traffic data, because UAVs can observe traffic conditions with a bird’s-eye-view perspective and changeable orientation, so that the researcher can analyze details and obtain precise results [34,35,36,37,38,39,40].

This study selected a small-sized UAV model, the DJI Mavic air2, to avoid any disturbance caused by the propeller’s noise to motorcycle riders, as its weight was less than one kilogram. The researcher operated the UAV at a 60 m height. Since its battery can provide power for the UAV for approx. 20 min, the rechargeable battery was replaced twice for one flying hour. Observers then manually determined the stopping behavior of motorcycle riders from video records.

For sample size consideration in logistic regression analysis, a minimum of 10 samples per estimated model coefficient is recommended [41]. This study analyzed 5 groups of stopping locations and included 12 variables to estimate the model coefficients; thus, the minimum required sample size was 600 samples. This study collected the stopping behavior of 1413 motorcycle riders.

3.4. Data Analysis

This study applied multinomial logistic regression for data analysis. Logistic regression is a specialized form of regression that is formulated to predict and explain a categorical variable, rather than a metric-dependent measure. Logistic regression can also accommodate nonmetric independent variables through dummy-variable coding [41]. Multinomial logistic regression analysis has been broadly applied in many fields of transportation study—for example, the determination of factors influencing helmet usage [15,31], the determination of factors increasing motorcycle use [42], the development of a model for the lateral placement and movement of vehicles [43], the determination of factors associated with the Red Light Running (RLR) behavior of motorcycle riders [12], the study of the influence of driver characteristics and the road environment on RLR behavior at signalized intersections [14], and the evaluation of contributing factors to motorcycle crash severity at intersections [44].

This study applied multinomial logistic regression in the following stages. The stopping location of motorcycles was set as a dependent variable that was categorized into five groups. They were assigned (coded) values of 1, 2, 3, 4, and 5. Traffic- and environment-related variables and rider- and motorcycle-related variables were set as independent variables. They were categorical variables and their coding is presented in Table 2. The Pearson chi-square test was applied to assess relations among variables. During the model’s development, independent variables that were significantly related to the dependent variable were included in the model. High correlations among independent variables were removed from the model to account for any multicollinearity problems. The maximum likelihood estimation (MLE) technique was then applied for model estimation, i.e., estimating constant and coefficients. The likelihood ratio test was used to evaluate a reduction in the log likelihood value, i.e., the model estimation fit and statistical significance of the whole model. Cox and Snell R² and Nagelkerke R², i.e., R²-like indices, were applied to assess goodness of fit (they are interpreted as reflecting the amount of variation accounted for by the logistic regression analysis, with higher values indicating greater model fit and 1.0 indicating perfect model fit [41]). Finally, the Wald statistic was used to test the significance of each independent variable’s coefficient on the dependent variable.

4. Results and Discussion

4.1. Results of Data Collection

From a bird’s-eye-view perspective on video recordings, observers accurately classified the stopping location of motorcycle riders. The stopping locations of 1413 motorcycle riders were observed on 24 approaches from 10 signalized urban intersections. The numbers of motorcycles stopping at different locations on the intersection approaches is summarized in

Table 3. The maximum, minimum, and average number of motorcycle riders stopping at different zones per red time.

Descriptive Statistics	Stopping Location
	Zone 1	Zone 2	Zone 3	Zone 4	Zone 5	All Zones
Maximum (veh.)	12	10	7	7	10	46
Minimum (veh.)	0	0	0	0	0	0
Average (veh.)	6	5	3.5	3.5	5	23

. The maximum, minimum, and average number of motorcycles stopping at all zones (Zones 1 to 6) per red time were 46, 0, and 23 vehicles, respectively.

The number and percentage of motorcycle riders stopping at each zone (Zones 1 to 6) classified by independent variables are summarized in

Table 4. Basic statistics on motorcycle riders’ stopping locations with respect to independent variable.

Variable	Stopping Location
	Zone 1	Zone 2	Zone 3	Zone 4	Zone 5	All Zones
Peak period
Morning peak (veh.)	440	183	49	9	28	709
% within stopping location	65.1%	73.2%	17.7%	12.0%	20.7%	50.2%
Evening peak (veh.)	236	67	228	66	107	704
% within stopping location	34.9%	26.8%	82.3%	88.0%	79.3%	49.8%
Engine size of motorcycle
Small motorcycle (veh.)	580	218	233	57	115	1203
% within stopping location	85.8%	87.2%	84.1%	76.0%	85.2%	85.1%
Large motorcycle (veh.)	96	32	44	18	20	210
% within stopping location	14.2%	12.8%	15.9%	24.0%	14.8%	14.9%
Helmet usage
Wearing (veh.)	405	125	127	37	61	755
% within stopping location	59.9%	50.0%	45.8%	49.3%	45.2%	53.4%
Not wearing (veh.)	271	125	150	38	74	658
% within stopping location	40.1%	50.0%	54.2%	50.7%	54.8%	46.6%
Larger vehicle in queue
Have larger vehicle (veh.)	337	60	177	49	82	705
% within stopping location	49.9%	24.0%	63.9%	65.3%	60.7%	49.9%
No larger vehicle (veh.)	339	190	100	26	53	708
% within stopping location	50.1%	76.0%	36.1%	34.7%	39.3%	50.1%
Presence of shadow in queue area
Have shadow (veh.)	263	118	184	44	76	685
% within stopping location	38.9%	47.2%	66.4%	58.7%	56.3%	48.5%
No shadow (veh.)	413	132	93	31	59	728
% within stopping location	61.1%	52.8%	33.6%	41.3%	43.7%	51.5%
Intended direction of motorcycle rider
Continuing straight (veh.)	473	152	233	28	100	986
% within stopping location	70.0%	60.8%	84.1%	37.3%	74.1%	69.8%
Turning right (veh.)	203	98	44	47	35	427
% within stopping location	30.0%	39.2%	15.9%	62.7%	25.9%	30.2%
Density of Zone 1
Dense (veh.)	4	2	46	9	4	64
% within stopping location	0.6%	0.4%	16.6%	12.0%	3.0%	4.5%
Not dense (veh.)	672	249	231	66	131	1349
% within stopping location	99.4%	99.6%	83.4%	88.0%	97.0%	95.5%
Density of Zone 2
Dense (veh.)	294	14	158	52	91	609
% within stopping location	43.5%	5.6%	57.0%	69.3%	67.4%	43.1%
Not dense (veh.)	382	236	119	23	44	804
% within stopping location	56.5%	94.4%	43.0%	30.7%	32.6%	56.9%
Density of Zone 3
Dense (veh.)	4	2	20	5	5	36
% within stopping location	0.6%	0.8%	7.2%	6.7%	3.7%	2.5%
Not dense (veh.)	672	248	257	70	130	1377
% within stopping location	99.4%	99.2%	92.8%	93.3%	96.3%	97.5%
Density of Zone 4
Dense (veh.)	10	2	6	1	6	25
% within stopping location	1.5%	0.8%	2.2%	1.3%	4.4%	1.8%
Not dense (veh.)	666	248	271	74	129	1388
% within stopping location	98.5%	99.2%	97.8%	98.7%	95.6%	98.2%
Density of Zone 5
Dense (veh.)	67	39	87	21	3	217
% within stopping location	9.9%	15.6%	31.4%	28.0%	2.2%	15.4%
Not dense (veh.)	609	211	190	54	132	1196
% within stopping location	90.1%	84.4%	68.6%	72.0%	97.8%	84.6%
Total (veh.)	676	250	277	75	135	1413
% of total	47.8%	17.7%	19.6%	5.3%	9.6%	100.0%

. Among 1413 motorcycle riders observed, 50.2% of them were observed during the morning peak period, while 49.8% of them were observed during the evening peak period. Most motorcycle riders rode small motorcycles (85.1%), while 14.9% of motorcycle riders rode large motorcycles. Unfortunately, 46.6% of motorcycle riders did not wear helmets, while 53.4% of them used helmets.

When motorcycle riders arrived at the intersection, approximately half of them (49.9%) encountered a larger vehicle in the queue. In addition, 43.1% of motorcycle riders encountered the stopping location in Zone 2 (behind a stop line and in front of a larger vehicle in the queue) being occupied by other motorcycles.

Surprisingly, the majority of motorcycle riders, 47.8%, stopped in front of a stop line (Zone 1) and 70.0% of them intended to go straight. This finding is associated to the violation behavior of moped riders in China that the majority of them waited beyond the stop line before going straight. This could be explained by the shorter passing distance when going straight [45]. Further, 19.6% of motorcycle riders stopped on the left side of a larger vehicle in the queue (Zone 3); 17.7% of them stopped behind a stop line and in front of a large vehicle in the queue (Zone 2); 9.6% of them stopped behind a larger vehicle in the queue (Zone 5); and 5.3% of them stopped on the right side of a larger vehicle in the queue (Zone 4). In addition, almost half (48.5%) of the motorcycle riders stopped at locations with shadow.

4.2. Results of Multinomial Logistic Regression Analysis

The results of the multinomial logistic regression analysis are presented in

Table 5. Results of multinomial logistic regression analysis.

	Zone 2 Compared with Zone 1		Zone 3 Compared with Zone 1		Zone 4 Compared with Zone 1
	B	Exp (B)	B	Exp (B)	B	Exp (B)
Intercept	−1.143 **		−1.231 **		−3.194 **
Intending to turn right	0.318	1.375	−1.156 **	0.315	1.381 **	3.981
Large motorcycle	−0.139	0.870	−0.405	0.667	0.210	1.233
Not wearing a helmet	0.338 *	1.403	0.728 **	2.072	0.461	1.586
Presence of shadow	0.524 **	1.690	1.085 **	2.961	0.829 *	2.293
Morning peak	0.278	1.321	−2.128 **	0.119	−2.473 **	0.084
Larger vehicle in queue	−0.720 *	0.487	0.399	1.491	−0.239	0.787
Density of Zone 1	−0.251	0.778	3.247 **	25.728	2.507 *	12.273
Density of Zone 2	−2.314 **	0.099	−0.109	0.897	0.834 *	2.305
Density of Zone 3	0.969	2.635	0.721	2.056	0.498	1.645
Density of Zone 4	−0.887	0.412	−0.410	0.663	−0.825	0.438
Density of Zone 5	1.472 **	4.362	1.216 **	3.377	0.800 *	2.227
	Zone 5 compared with Zone 1		Zone 3 compared with Zone 2		Zone 4 compared with Zone 2
	B	Exp (B)	B	Exp (B)	B	Exp (B)
Intercept	−1.897 **		−0.088		−2.051 **
Intending to turn right	−0.078	0.925	−1.474 **	0.229	1.063 **	2.896
Large motorcycle	−0.212	0.809	−0.265	0.767	0.349	1.418
Not wearing a helmet	0.726 **	2.067	0.389	1.476	0.122	1.130
Presence of shadow	0.597 *	1.817	0.560 *	1.752	0.305	1.356
Morning peak	−1.981 **	0.138	−2.407 **	0.090	−2.752 **	0.064
Larger vehicle in queue	0.674 *	1.964	1.119 **	3.063	0.481	1.617
Density of Zone 1	0.710	2.033	3.498 *	33.071	2.758 *	15.775
Density of Zone 2	0.396	1.485	2.205 **	9.078	3.149 **	23.324
Density of Zone 3	1.105	3.019	−0.248	0.780	−0.471	0.624
Density of Zone 4	0.808	2.244	0.477	1.611	0.062	1.064
Density of Zone 5	−2.002 **	0.135	−0.256	0.774	−0.672	0.511
	Zone 5 compared with Zone 2		Zone 4 compared with Zone 3		Zone 5 compared with Zone 3
	B	Exp(B)	B	Exp (B)	B	Exp (B)
Intercept	−0.754 *		−1.963 **		−0.666 *
Intending to turn right	−0.396	0.673	2.538 **	12.654	1.078 **	2.940
Large motorcycle	−0.072	0.930	0.614	1.848	0.193	1.213
Not wearing a helmet	0.387	1.473	−0.267	0.766	−0.002	0.998
Presence of shadow	0.072	1.075	−0.256	0.774	−0.488 *	0.614
Morning peak	−2.260 **	0.104	−0.345	0.709	0.147	1.159
Larger vehicle in queue	1.394 **	4.035	−0.639	0.528	0.275	1.317
Density of Zone 1	0.961	2.613	−0.740	0.477	−2.538 *	0.079
Density of Zone 2	2.710 **	15.030	0.943 *	2.569	0.504	1.656
Density of Zone 3	0.136	1.146	−0.223	0.800	0.384	1.468
Density of Zone 4	1.696	5.450	−0.415	0.660	1.219	3.383
Density of Zone 5	−3.475 **	0.031	−0.416	0.659	−3.219 **	0.040
	Zone 5 compared with Zone 4		Model Summary
	B	Exp (B)
Intercept	1.296 *		* p-value < 0.05
Intending to turn right	−1.459 **	0.232	** p-value < 0.001
Large motorcycle	−0.421	0.656	−2LL Intercept Only = 1982.129
Not wearing a helmet	0.265	1.303	−2LL Intercept Final = 1095.950
Presence of shadow	−0.232	0.793	Cox and Snell R2 = 0.466
Morning peak	0.492	1.635	Nagelkerke R2 = 0.499
Larger vehicle in queue	0.914 *	2.495
Density of Zone 1	−1.797 *	0.166
Density of Zone 2	−0.439	0.644
Density of Zone 3	0.607	1.835
Density of Zone 4	1.634	5.123
Density of Zone 5	−2.802 **	0.061

. The Cox and Snell R² and Nagelkerke R² are 0.466 and 0.499, which are close to 0.50, indicating that the multinomial logistic regression model accounts for almost 50% of the variation among the five stopping locations of motorcycle riders [41].

The result of the hypothesis No. 0 test showed that the number of motorcycle riders stopping in front of a stop line (Zone 1) is significantly higher than that at other locations when other influencing factors are constant. The result showed that almost half of motorcycle riders (47.8%) stop in front of the stop line (Zone 1) (see Table 4). They intended to stop in front of the stop line (Zone 1), yielding results approximately 3.1-, 3.4-, 6.7-, and 24.4-times higher than the results for Zone 2, Zone 3, Zone 5, and Zone 4, respectively. Motorcycle riders are likely to stop in front of a stop line (Zone 1) because they can freely depart from the intersection during the green light, resulting in less conflict with other larger vehicles in the queue. However, stopping at Zone 1 is illegal. They may conflict with pedestrians and crossing traffic from other directions. In addition, it also increases the likelihood of RLR behavior [12].

The results of the hypothesis No. 1–No. 7 tests showed that factors significantly influencing the stopping locations of motorcycle riders are motorcycle riders’ intention to turn right, motorcycle riders not wearing a helmet, the morning peak period, the presence of shadow, the presence of a larger vehicle in the queue, and the density of the desired stopping location. The size of the motorcycle alone does not significantly influence the stopping locations of motorcycle riders. The significant factors are described in detail as follows.

4.2.1. Intending to Turn Right

Motorcycle riders’ intention to turn right significantly influences the stopping locations of motorcycle riders. The probability that motorcycle riders making a right turn would stop at Zone 4 is 2.9, 4.0, and 12.7times higher than that for Zone 2, Zone 1, and Zone 3, respectively. In addition, motorcycle riders making a right turn have a 4.4 times higher likelihood of stopping at Zone 2 than stopping at Zone 3.

The results showed that 62.7% and 39.2% of motorcycle riders stop at Zone 4 and Zone 2 before making a right turn (see Table 4). When making a right turn at an intersection, stopping at Zone 4 and Zone 2 results in less conflict with other larger vehicles in the queue. However, 15.9% of those stopping at Zone 3 encounter more conflict, particularly with straight-traveling larger vehicles. This reveals that motorcycle riders prefer to stop at a location where they can start to turn right before larger vehicles in the queue when the traffic light turns green, thereby reducing the likelihood of crashes at intersections. Notably, this finding can extend to right-hand-traffic countries. In this case, motorcycle riders may prefer to stop at a location where they can start to turn left before larger vehicles in the queue as well.

4.2.2. Not Wearing a Helmet

Motorcycle riders not wearing helmets significantly influenced the stopping locations of motorcycle riders. The likelihood of non-helmeted motorcycle riders stopping at Zone 5 is 1.5 and 2.1 times higher than that for Zone 2 and Zone 1, respectively. The results showed that 54.8% of non-helmeted motorcycle riders stopped at Zone 5 (see Table 4). They attempt to stop far from a stop line if a police inspection checkpoint is located nearby. This finding is consistent with previous studies, indicating that motorcycle riders riding through intersections with police inspections were more likely to wear a helmet than those riding through intersections without police inspections [15,31].

4.2.3. Time of Day

The time of day significantly influenced the stopping locations of motorcycle riders. During the morning peak period, 73.2% and 65.1% of motorcycle riders stopped at Zone 2 and Zone 1, respectively. Motorcycle riders intend to stop at Zone 2, which is 9.6-, 11.1-, and 15.7-times more likely than stopping at Zone 5, Zone 3, and Zone 4, respectively. In addition, motorcycle riders intend to stop at Zone 1, which is 7.3-, 8.4-, and 11.9-times more likely than stopping at Zone 5, Zone 3, and Zone 4, respectively. During the morning peak period, motorcycle riders are usually hurrying to reach their workplaces and they prefer to stop at locations where they can depart earlier than other vehicles. The morning peak period influences not only motorcyclists’ stopping locations at signalized intersections but also motorcyclists’ choice to wear a helmet [15,31,46,47].

4.2.4. Presence of Shadow

The presence of shadow significantly influenced the stopping locations of motorcycle riders. The majority of motorcycle riders (66.4%) stopped at Zone 3 because this location is more shadowed by trees and buildings than other zones (67% of cases according to the site observation). Motorcycle riders intend to stop at Zone 3, which is 1.6-, 1.8-, and 3.0-times more likely than stopping at Zone 5, Zone 2, and Zone 1, respectively. Motorcycle riders wish to wait for a green light under a shadow because of the hot weather. They are aware of the time remaining until a green light at intersections where a countdown timer is installed. This finding is consistent with the findings of studies on non-motorized modes in a tropical country. They stated that motivators for cycling include planting trees for shade [48], and walking activity is considerably reduced during hot weather [49].

4.2.5. Presence of Larger Vehicle in Queue

The presence of a larger vehicle, including a passenger car, truck, or bus, in the queue significantly influenced the stopping locations of motorcycle riders. When a larger vehicle is stopping in the queue, 63.9%, 65.3%, and 60.7% of motorcycle riders decide to stop at Zone 3, Zone 4, and Zone 5, but only 24.0% of them decide to stop at Zone 2. Motorcycle riders do not wish to stop at Zone 2 because they are at risk of a rear-end crash when a larger vehicle starts to accelerate at the onset of a green light. Specifically, the drivers of heavy vehicles (truck and bus) may not see them when they stop at blind spots. They perceive a higher risk when they stay closer to a heavy vehicle; thus, they wish to maintain longer distances from any large vehicles. This finding is consistent with that in previous studies on vulnerable road users’ riding behavior. A large vehicle led to a higher lateral clearance of lane-filtering motorcycle riders [2]. Motorcycle riders intend to move out of their current lane if they are following a large vehicle when they maneuver in a queue at a signalized intersection [1]. This causes the risk perception of vulnerable road users when they travel in mixed traffic [50,51].

4.2.6. Density of Stopping Location

The density of the desired stopping location significantly influenced the stopping locations of motorcycle riders. Among all motorcycle riders arriving at the intersections, 43.1% found that Zone 2 was densely occupied. In this case, the likelihood that these motorcycle riders chose to stop at Zone 3, Zone 1, Zone 5, and Zone 4 was 9.1-, 10.1-, 15.0-, and 23.3-times higher than stopping at Zone 2 itself, respectively. Moped riders in China intended to wait for the green light beyond the stop line when the waiting area did not have adequate space [45].

4.2.7. Discussion toward Recommendation for Implementation

The findings of this study can be used to improve the safety of motorcyclists at signalized urban intersections in developing countries as follows.

For engineering implementation, the findings revealed that when no larger vehicle is present in the queue, Zone 2 is likely to be the most favorable stopping location among motorcycle riders. In particular, 76% of motorcycle riders stopped at Zone 2, which is legal and safer than other locations. However, 43% of motorcycle riders arriving at intersections found that Zone 2 was already occupied by other motorcycles during peak periods. Therefore, an exclusive stopping space for motorcycles should be installed in Zone 2 in accordance with the concept of exclusive stopping spaces for bicycles (bike box), which could decrease the number of conflicts between bicycles and other larger vehicles [52]. This exclusive stopping space should be designed to provide sufficient space for stopping motorcycles in peak periods and a buffer zone in front of any larger vehicles in the queue. This exclusive stopping space would also increase visibility and the distance between motorcycles and larger vehicles in the queue. This separation could reduce conflicts during start-up time [53]. Moreover, an exclusive stopping space for motorcycles at signalized intersections can reduce the lost start-up time of the traffic flow. The higher the number of stopping motorcycles in the exclusive stopping space during the red light, the greater the reduction in the start-up time of the traffic flow [54]. A filtering lane allows motorcycle riders to filter between a queue of stopping vehicles and access the exclusive stopping space for motorcycles more easily [2]. In addition, the findings indicated that motorcycle riders prefer to stop at locations covered by the shadows of trees or buildings. To encourage motorcycle riders to stop at safer locations, such as an exclusive stopping space for motorcycles, a built environment for shadows, such as tree planting, should be provided to increase the likelihood of stopping [48]. The findings also revealed that some motorcycle riders (15.9%) making a right turn would stop on the left side of a larger vehicle in the queue (Zone 3), where the likelihood of conflicting with other vehicles in the queue traveling straight increases. Therefore, a pavement marking that serves as a guide for motorcycle riders, such as separated arrows for straight traveling and right turns, may guide motorcycle riders to stop at safer locations [55].

For enforcement implementation, the findings revealed that 47.8% of motorcycle riders stop at Zone 1, which is against traffic law, and they have a high likelihood of conflict with vehicles traveling in other directions. In addition, 46.6% of non-helmeted motorcycle riders stop far away from the stop line. Therefore, to enhance law enforcement, automated enforcement by CCTV cameras operating throughout the area should be introduced at these intersections—for example, automated helmet enforcement via CCTV cameras [31].

5. Conclusions and Recommendations

This study aimed to identify factors influencing the stopping locations of motorcycle riders on mixed-traffic signalized intersection approaches. The stopping behaviors of 1413 motorcycle riders from 24 intersection approaches were observed by applying UAVs. The study applied multinomial logistic regression analysis to determine the relationship between the stopping locations of motorcycle riders and relevant variables, including rider and motorcycle characteristics, traffic characteristics, and environmental conditions.

The relevant results of data collection revealed the risky stopping behavior of motorcycle riders. Half of the motorcycle riders stopped in front of the stop line (Zone 1). Some of the motorcycle riders intending to turn right stopped on the left side of a larger vehicle in the queue (Zone 3). These stopping behaviors increase the likelihood of conflicts with other road users.

The statistical analyses showed that the significant factors influencing the stopping locations of motorcycle riders on signalized intersection approaches are the motorcycle rider’s intention to turn right, high traffic during the morning peak period, the presence of shadows, motorcycle riders not wearing a helmet, the presence of a larger vehicle in the queue, and the density of the desired stopping location.

The majority of right-turning motorcycle riders stopped on the right side of a larger vehicle in the queue in Zone 4. The majority of motorcycle riders stopped behind the stop line and in front of a larger vehicle in the queue in Zone 2 during the morning peak period. The majority of motorcycle riders stopped at the left side of a larger vehicle in the queue in Zone 3 because this location is more shadowed by trees and buildings than other locations. More than half of non-helmeted motorcycle riders stopped behind a larger vehicle in the queue in Zone 5. Three-quarters of motorcycle riders did not stop behind the stop line but stopped in front of a larger vehicle in the queue in Zone 2.

The findings of this study can contribute to traffic management, traffic law enforcement, and traffic-simulation applications.

For traffic management, the findings revealed that behind the stop line and in front of a larger vehicle in the queue (Zone 2) are likely to be the most preferable stopping locations among motorcycle riders. An exclusive stopping space with a filtering lane for motorcycles should be introduced to improve the safety of motorcyclists on a signalized intersection approach, with a high mixed-traffic condition. In Southeast Asian countries, such as Indonesia, two types of exclusive stopping spaces for motorcycles have been introduced, i.e., the square type and P type. The square type is used when the number of motorcycles in each lane is almost the same, while the P type is used when the proportion of motorcycles in the left lane, which is the through lane, is considerably higher than that in the right-turn lane. Its width depends on the lane width, while its length is 8–12 m for the square type and 12–16 m for the P type. It depends on the number of stopping motorcycles during the red light [56]. In Thailand, exclusive stopping spaces have been designed for both motorcycles and bicycles. These are only of the square type. Their width depends on the lane width, while their length is 4 m or longer. It also depends on the number of stopping motorcycles during the red light [54]. An exclusive stopping space with colored pavements, as well as information signs and pavement markings, can encourage motorcycle riders to stop in this area. Moreover, warning signs can warn drivers of other larger vehicles not to stop in this area. In addition, an automated enforcement system using CCTV cameras could be introduced to increase the efficiency of law enforcement.

For traffic law enforcement, the findings revealed that half of motorcycle riders stopped illegally in front of the stop line (Zone 1) and half of motorcycle riders who stopped at study intersection approaches did not wear helmets. The automated enforcement system with CCTV cameras should be installed at these approaches.

The limitations of this study are as follows. The study excluded the stopping behavior of motorcycle riders during nighttime, since UAVs cannot clearly observe at night. During nighttime, the number of motorcycle riders stopping at Zone 1 or Zone 2 may be higher than that during daytime due to the lower traffic volume and lower degree of police inspection. Furthermore, this study focused on stopping behavior at small-sized signalized intersections in urban areas. A future study should extend this to other intersection sizes under various circumstances, such as larger signalized intersections in rural areas. It is important to comprehensively investigate how the stopping locations of motorcycles affect traffic flow and crossing conflict at signalized intersections.