Roadside design assessment in an urban, low-density environment in the Gulf Cooperation Council region

Abstract

Objectives: Proper roadside design is crucial in order to mitigate the consequences associated with single-vehicle run-off-road (SVROR) crashes. However, the Gulf Cooperation Council (GCC) region lacks in-depth, detailed information on its roadside design status. Hence, there is a need for an in-service evaluation of roadside design in the GCC region. The objective of this study is to assess the existing roadside design in a medium-sized, low-density city in the United Arab Emirates (UAE).

Methods: A multiyear crash database was used to identify 116 locations where SVROR injury crashes occurred between 2013 and 2016 in the city of Al Ain in the UAE. Visits to these locations were made in order to assess their roadside design. Subsequently, the collected data were analyzed. Roadside design was classified based on whether or not it deviated from roadside design guidelines. The guidelines adopted as a benchmark were those contained in the 2012 Abu Dhabi Department of Transport Roadside Design Guide (RDG) and/or those in the 2011 American Association of State Highway and Transportation Officials (AASHTO) RDG. It is worth stressing that local guidelines are heavily based on the 2011 AASHTO RDG.

Results: The study found that almost one quarter of all SVROR crashes resulted in injuries. The study also found that a staggering 80.17% of the SVROR injury crashes investigated occurred at locations where roadside design deviated from the benchmark. Lack of an adequate clear zone was the main cause of noncompliant locations. Most SVROR injury crash locations containing roadside design with deviations from the benchmark were located on roads with posted speed limits of 100 kph or higher. Light poles, trees, curbs, and barriers were the most harmful objects most often struck, and tree collisions accounted for the highest number of severe crashes. Ninety-four and 86% of all studied locations containing light poles and trees, respectively, were found to be noncompliant with the benchmark. Twenty-eight percent of all SVROR injury crashes involved a rollover. All rollovers were preceded by a collision with a tree, pole, guardrail, or curb. Forty-four percent of all rollover crashes resulted in severe injuries.

Conclusions: Significant revision of the existing roadside design not only in the area studied but throughout the UAE is recommended. The authors propose measures that may be useful in making roadside design in the area studied better align with the benchmark requirements.

Keywords:

• Run-off-road crashes
• roadside design
• GCC region

Introduction

Background

Single-vehicle run-off-road (SVROR) crashes have been found to account for a significant number of all road crash fatalities (American Association of State of Highway and Transportation Officials [AASHTO] 2011; Federal Highway Administration 2017, n.d.; Kloeden et al. 1999; Liu and Subramanian 2009; Neuman et al. 2003; NHTSA n.d.; Palamara et al. 2013; Roque et al. 2015; Van Petegem and Wegman 2014; Wu et al. 2016).

In the United States, the number of fatalities caused by SVROR crashes have historically been elevated (FHWA 2017; Liu and Subramanian 2009; NHTSA n.d.). A study conducted in The Netherlands found that approximately one quarter of all traffic fatalities resulted from SVROR crashes, which were found to occur more frequently in rural areas (Van Petegem and Wegman 2014). Indeed, more recent research has shown that SVROR crashes occurring in rural areas tend to be more severe than those occurring in urban environments, as well as that different crash contributing factors affect SVROR crashes on urban and rural roadways at varying levels (Wu et al. 2016). In Portugal, SVROR crashes have been associated with half of all freeway fatalities. Portuguese crash data from 2007 to 2010 showed that slopes, embankments, and ditches contributed to over half of all fatal and severe SVROR events (Roque et al. 2015). In Thailand, SVROR crashes account for approximately 45% of all fatal crashes (Somchainuck et al. 2013). A study conducted in Australia found that though SVROR crashes in nonurban areas tended to be more severe, over 20% of all SVROR crashes that occurred in metropolitan Perth resulted in injuries (Palamara et al. 2013).

Thus, the literature suggests that in order to improve road safety, greater focus needs to be placed on preventing SVROR crash-related injuries. As such, roadside areas must be properly treated based on adequate guidelines. However, though widely used roadside design guidelines are available (Abu Dhabi Department of Transport 2012; AASHTO 2011), it is unknown whether, as well as to what extent, these guidelines have been implemented not only in the Emirate of Abu Dhabi, part of the United Arab Emirates (UAE), but also in the Gulf Cooperation Council (GCC) region. Furthermore, roadside safety and design-related literature pertaining to the GCC region is extremely scarce. Previous research has indicated that SVROR crashes accounted for approximately 20% of all crash types that occurred in Dubai between 1995 and 2006. Half of these crashes involved fixed-object collisions, and the other half resulted in rollovers (Al Dah 2010). Another study investigated SVROR crashes that occurred between 2007 and 2013 in the Emirate of Abu Dhabi and found that SVROR crashes accounted for 22 and 25% of all fatal and serious crashes, respectively. It also found that speeding, passenger cars, Emirati drivers, rural roads, and nonintersection locations were all factors that were found to significantly contribute to increased SVROR crash severity (Shawky et al. 2016). Though these are all valuable research findings, previous research studies have focused on investigating either SVROR crash characteristics (e.g., location, object hit, and road type) or the relationship between SVROR crash frequency/severity and its contributing factors. To date, there has been no study conducted in the GCC region that has attempted to investigate actual roadside design compliance with selected benchmark.

Research objectives

The objectives of this research are 2-fold: to (1) assess the roadside design in an urban, low-density environment located in the GCC region in terms of its compliance to the adopted benchmark and (2) provide data-driven recommendations aimed at bringing an existing, citywide roadside design into compliance with the roadside design guideline benchmark, which in this study was chosen to be both the 2011 AASHTO Roadside Design Guide (RDG) and the 2012 Abu Dhabi Department of Transport RDG.

Research significance

The findings of this research may be relevant not only in the UAE but also in other GCC countries (i.e., Bahrain, Kuwait, Oman, Qatar, and Saudi Arabia) whose roads and roadside designs, as well as traffic characteristics and driving cultures, may be similar. In addition, the findings from this study may be relevant to government authorities, decision makers, practitioners, researchers, and road safety equipment suppliers.

Methods

The present study conducted an assessment of the roadside design at a number of locations where SVROR crashes had occurred between 2013 and 2016 within the boundaries of the city of Al Ain in the UAE. In order to carry out such an assessment, site visits were made to the crash locations. Site locations were identified using the Global Positioning System coordinates contained in a multiyear crash injury database received from the Abu Dhabi traffic police. Finally, design assessment data collected during visits to crash locations and crash data collected from the Abu Dhabi traffic police were compared with roadside design guidelines contained both in the AASHTO RDG (AASHTO 2011) and in the Abu Dhabi Department of Transport RDG (Abu Dhabi Department of Transport 2012). In this way, compliance of the actual roadside design was evaluated. It is worth noting that the Abu Dhabi Department of Transport RDG appears to be heavily based on the AASHTO RDG, including its clear zone distance policy.

Study area

The study investigated all SVROR injury crashes that occurred within the boundaries of the city of Al Ain, part of the Emirate of Abu Dhabi. Located approximately 120 km south of Dubai, Al Ain has many tree-lined, high-speed, 6-lane, median-divided avenues. These are also typical road characteristics of other urban areas in the Emirate of Abu Dhabi such as Al Shamkha, Khalifa City, and Mohammed Bin Zayed City.

Until mid-2018, drivers in the Emirate of Abu Dhabi were legally allowed to drive 20 kph above the posted speed limit before they were fined for speeding. However, in mid-2018, this 20 kph speed gap between posted speed limits and legally allowed speeds was removed by raising posted speed limits by 20 kph (Duncan 2018; Zacharias 2018). With generous road infrastructure supply and relatively low traffic volumes, roads in Al Ain are uncongested most of the time. Congestion tends to occur in a few spots (i.e., usually close to schools) and lasts for short periods of time. As such, traffic usually flows under free-flow conditions and, before the 20 kph speed buffer removal, motorists usually traveled close to or at legally allowed speeds, rather than at posted speed limits. Therefore, data collected in this study, in terms of roadside design compliance, may not have been affected by the removal of this 20 kph speed buffer.

Data collection

Crash data were provided by the Abu Dhabi traffic police in 2 databases: no-injury and injury crash databases. Only the injury crash database contained information regarding the Global Positioning System coordinates of crash locations. As a result, no-injury crash locations could not be identified and, therefore, could not be visited and assessed. As such, the roadside design assessment conducted in this study relied on injury crash locations only.

Three types of data were used: field, crash, and traffic data. Field data were collected during the crash site visits. These data included descriptions and measurements on fixed object type, clear zone distance, hazard lateral offset, curb height, barrier length, barrier end terminal type, hazard-to-barrier offset, roadside terrain topography, as well as breakaway device and crash cushion use/type. The crash data included not only crash descriptions and diagrams but also data on the first object struck, crash severity, and posted speed limit. All descriptions and diagrams pertaining to injury crashes were reviewed in order to better understand crash dynamics, as well as to extract data on the sequence of crash events and rollover occurrence outcome. Traffic data pertaining to the average daily traffic of road locations of interest was provided by the Al Ain municipality. Finally, field, crash, and traffic data were merged into a single crash injury data set.

Lastly, data were retrieved based on the number of vehicles involved and the crash type. That is, all crashes classified as having involved only one vehicle and falling into the off-road collision with fixed/moveable object category were selected for inclusion in the study, resulting in 64 cases. It was later realized that the crash type coding scheme was not very reliable and that other SVROR crashes could have been classified in a variety of different ways. An effort was made to estimate the real number of SVROR crashes based on a keyword (e.g., barrier, tree, pole, sign, fence, wall, and curb) search within the crash description field, resulting in an estimated 210 other cases. Based on resource constraints in terms of project budget and duration, a sample of 52 locations was randomly selected out of the 210 locations. Visits were also made to these 52 extra locations. In order to investigate whether this sample would be a fair representation of the entire 210 cases, chi-square tests were conducted to determine whether the legally allowable traveling speed distributions (see Table 1 for a breakdown of legally allowed traveling speed categories) were statistically different or not. It was found that distributions were not statistically different (i.e., P = .16).

Table 1. Crash severity distribution by legally allowable traveling speed.

			Legally allowable traveling speed (kph)
			≤80	≥100	Subtotals
Crash severity					n	%
	Not severe	n	28/11^a	43/9^a	71/20^a	76.34/86.96^a
		%	82.35/84.62^a	72.88/90.00^a
	Severe	n	6/2^a	16/1^a	22/3^a	23.66/13.04^a
		%	17.65/15.38^a	27.12/10.00^a
	Subtotals	n	34/13^a	59/10^a
		%	36.56/56.52^a	63.44/43.48^a

a Compliant locations.

Roadside design guidelines

The benchmark adopted in this study provides a set of design priorities regarding the treatment of roadside areas. The first priority is to, whenever possible, remove the hazard if it is located within the recommended minimum clear zone width. In situations where completely removing the hazard is not possible, the second design priority may be pursued, which consists of redesigning the obstacle so that it can be safely traversed. The third design priority consists of relocating the hazard farther from the roadway edge such that the hazard may be outside the recommended minimum clear zone. The fourth design priority consists of reducing the impact severity by making use of breakaway devices in the case of utility poles and traffic signs. If, and only if, none of the first 4 design priorities is a viable option, shielding the hazard with a roadside barrier (i.e., the fifth design priority) should be considered. If none of these alternatives are technically and/or economically feasible, delineation (i.e., the least desirable design option) should be considered, as a minimum.

In addition to these design priorities, a few other design considerations that are also contained in the benchmark were taken into account to assess whether a roadside design deviated from design guidelines or not: (a) Curb height, (b) barrier installation in curb vicinities, (c) roadside barrier length of need, and (d) barrier working width requirements.

Roadside design assessment

Table A.1 (see online supplement) provides a few cases of roadside design assessments performed in this study. All 3 cases pertained to locations with flat roadside terrain and daily traffic volumes of over 6,000 vehicles. The left-most column shows a case where none of the design priorities described previously were properly implemented. Therefore, this design deviated from the benchmark adopted. On the other hand, the middle column shows a case where the hazard (i.e., traffic sign) was neither removed (i.e., the first design priority) nor relocated (i.e., the second design priority) but the hazard was made traversable (i.e., the third design priority) by equipping it with a breakaway device installed at the base of the traffic sign. Thus, this case was considered to be in compliance with the benchmark. The right-most column illustrates a scenario where the design also deviated from the benchmark due to curb barrier installation design-related criteria. That is, whereas the distance from the front face of the curb to the front face of the guardrail was just 0.9 m, the selected benchmark recommends that guardrails not be placed within a distance shorter than 2.5 m behind a curb on roads with operating speeds above 60 kph. On the other hand, on roads with operating speeds above 85 kph, guardrails should be installed with the curb flush with the face of the guardrail, considering that curbs are sloping faced and no higher than 100 mm.

Data analysis

Data were cross-tabulated to show the distribution of crash severity in relation to a number of relevant road-, roadside-, and crash-related variables. Crash severity was classified into 2 levels: “not severe” and “severe.” Injuries reported as minor or moderate were included in the not severe category, and injuries reported as disabling or fatal were included in the severe category. Thus, 4 crash severity categories were combined into 2. It is worth stressing that breaking a sample size of 116 into too many categories would cause some cells to have too few observations. In addition, there is always a chance that police officers may have classified minor or moderate in an indiscriminate manner in some cases. Indeed, there is a chance for error even between the severe and fatal categories, because those severely injured in a crash may later die as a result of these injuries.

Bivariate analyses using cross-tabulation were carried out (Agresti 2007). Within each cross-tabulation category, percentage values refer to percentages in relation to the column totals. For example, in Table 2, the underlined value (86.96) is a percentage, the product of the ratio between the number of not severe crashes that occurred at locations that did not contain deviations from the benchmark (i.e., 20) and the total number of crashes that occurred at locations that did not contain deviations from the benchmark (i.e., 23).

Table 2. Crash severity distribution by roadside design compliance.

			Roadside design compliance
			Yes	No	Subtotals
Crash severity					n	%
	Not severe	n	20	71	91	78.45
		%	86.96	76.34
	Severe	n	3	22	25	21.55
		%	13.04	23.66
	Subtotals	n	23	93
		%	19.83	80.17

Results

The study found that 1,184 injury crashes and 274 SVROR injury crashes occurred in Al Ain between 2013 and 2016. Hence, almost one quarter of all SVROR crashes resulted in injuries, which is in line with previous research conducted in urban environments (Al Dah 2010; Palamara et al. 2013).

Table 2 shows that 80.17% of all SVROR injury crashes occurred at locations containing design deviations from the benchmark. A significantly higher percentage (i.e., 23.66 versus 13.04%) of severe crashes occurred at locations containing design deviations from the benchmark. It is worth stressing, however, that other potential injury-contributing factors (e.g., impact speed and hazard type) were not controlled for. Nevertheless, the fact that 93 (see number in bold) out of 116 roadside locations contained design deviations from the benchmark suggests that there may be a need to better match actual roadside design with recommended technical guidelines. Of these 93 sites found to be noncompliant, 40 contained light poles or traffic signs installed within the recommended clear zone area. These poles or signs were neither equipped with breakaway devices nor shielded by a barrier. Thirty-four out of the 93 sites found to be noncompliant contained trees or fences installed within the recommended clear zone area. These trees or fences were also not shielded by a barrier. The remaining noncompliant sites can be broken down as follows: 16 sites that contained curbs found to be excessively high compared to recommendations from the adopted benchmark, 1 site that did not provide adequate offset between the curb and front face of the guardrail, 1 site that contained an improper barrier installation due to insufficient lateral offset between the hazard and barrier, and 1 site that contained an improper barrier installation due to insufficient barrier upstream length. Table 1 shows that over half of the studied sites were located on roads with a legally allowed traveling speed of at least 100 kph. This is a worrisome finding because most of the roadside locations containing design deviations are located on higher-speed roads (i.e., 59 out of 93) where roadside design compliance may be even more important, because errant vehicles leave the roadway at higher speeds, potentially leading to more severe injuries. Table 1 also indicates that, within the noncompliant locations, there was a significant difference in the percentage of severe crashes between the 2 traveling speed categories. That is, 17.65% of the injury crashes that occurred on roads with legal traveling speeds of 80 kph or less resulted in severe injuries, whereas 27.12% of the injury crashes that occurred on roads with legal traveling speeds of 100 kph or higher resulted in severe injuries. This may reinforce the importance of revising the roadside design, especially on higher-speed roads, potentially yielding measures such as the adoption of adequate clear zones or the installation of breakaway devices.

Table 3 indicates that light poles, trees, barriers, and curbs were the most harmful struck objects in 83% of all SVROR injury crashes. Light poles were the most harmful object most often struck, accounting for 31.03% of all SVROR injury crashes, and collisions with trees proved to be the most severe because 27.59% (see number in bold) of all collisions with trees resulted in severe injuries. It is relevant to point out that most collisions with tree involved palm trees, which may be considered unforgiving, rigid roadside hazards. The “other” category involved collisions with objects such as traffic signs, signals, or fences. Curbs were found to be the most harmful object struck in 18% of the cases, which is not surprising because this study covered urban locations only. Twelve out of the 21 curb crashes involved curbs no shorter than 20 cm. Fourteen out of the 21 curb crashes resulted in rollovers. All severe curb crashes resulted in a rollover event after the curb was struck.

Table 3. Crash severity distribution by most harmful object struck.

			Most harmful object struck
			Tree	Light pole	Barrier	Curb	Other
Crash severity
	Not severe	n	21	29	9	16	16
		%	72.41	80.56	81.82	76.19	84.21
	Severe	n	8	7	2	5	3
		%	27.59	19.44	18.18	23.81	15.79
	Subtotals	n	29	36	11	21	19
		%	25.00	31.03	9.48	18.10	16.38

Tables 4 and 5 show the distribution of crash severity by design compliance (i.e., compliant vs. noncompliant) for light pole and tree crash locations, respectively. As can be seen in Table 4, though 19.44% of all injuries due to collisions with light poles were severe, 94.44% of all injuries due to collisions with light poles occurred at locations containing designs not in line with the benchmark. This can be attributed to the fact that the majority of light poles located within the minimum clear zone width recommended by the benchmark were not equipped with breakaway devices or any other sort of energy-absorbing mechanism and/or were not shielded. It is worth noting that there was one pole that was shielded by a barrier, but the roadside design was still found to be noncompliant with the benchmark because lateral offset from the barrier to the pole was not sufficient to accommodate barrier deflection and working width requirements. Table 5 shows that though over one quarter of all injuries due to collisions with trees were severe, 86.21% (see number underlined) of all injuries due to collisions with trees occurred at noncompliant locations.

Table 4. Crash severity distribution by light pole injury crash location design compliance.

			Light pole injury crash locations
			Compliant locations		Noncompliant locations		Subtotals
Crash severity							n	%
	Not severe	n	2		27		29	80.56
		%	100.00	6.90	79.41	93.10
	Severe	n	0		7		7	19.44
		%	0.00	0	20.59	100
	Subtotals	n	2		34		36
		%	5.56		94.44

Table 5. Crash severity distribution by tree injury crash location design compliance.

			Tree injury crash locations
			Compliant locations	Noncompliant locations	Subtotals
Crash severity					n	%
	Crash severity				n	%
	Not severe	n	2	19	21	72.41
		%	50	76.00
	Severe	n	2	6	8	27.59
		%	50	24.00
	Subtotals	n	4	25	29
		%	13.79	86.21

Crash site locations were also segregated by crash severity, barrier usage, and barrier requirements. It was found that approximately 64% of all SVROR injury crashes occurred at locations where a barrier was not installed where it should have been. These locations mostly included those where fixed object hazards, such as poles (i.e., those not equipped with a breakaway device), as well as trees, were within the recommended minimum clear zone width. In addition, 10% of the locations had a roadside barrier installed when one was warranted, and 26% of the locations did not have a roadside barrier installed when one was not required. Thus, 64% of all crash sites investigated did not meet roadside design guidelines for barrier placement. Just over one quarter of the collisions that occurred at these sites resulted in severe crashes, which was greater than the proportion (i.e., 15%) of collisions that occurred at sites where recommended barrier placement guidelines were properly followed based on the adopted benchmark.

Finally, almost 28% of all SVROR injury crashes involved a rollover. Out of the 32 rollovers, 14 were preceded by a curb impact, 7 by a pole impact, 5 by a tree impact, 3 by a barrier impact, 2 by a sign impact, and 1 by a fence impact. Though almost 44% of all rollover crashes resulted in severe injuries, only 13% of all nonrollover events resulted in severe injuries. Hence, this may be an indication that SVROR crashes involving rollovers may be prone to result in severe injuries, which is in line with previous research (Albuquerque and Sicking 2013).

Discussion

The findings of this research show that the existing roadside design in the area studied has been found to be noncompliant with the benchmark in the majority of the locations studied and, therefore, the authors recommend it to be revised, especially in terms of (1) clear zone provision, (2) curb height, and (3) light pole design. In addition, in order to increase roadside compliance and safety in the area studied, as a minimum, the authors recommend the following:

• Lowering posted speed limits in conjunction with an effective enforcement scheme to ensure compliance with posted speed limits. This measure may significantly decrease clear zone requirements in terms of lateral width, while also reducing injury risk. This is important in a cost-conscious attempt to increase roadside design compliance because narrower recommended clear zone lateral widths may demand less expensive on-site design modifications due to lower roadside fixed-object relocation costs.

• Removing thick-trunked trees located within the clear zone or replacing them with fragile trees or bushes.

• Equipping light poles with breakaway devices or energy-absorbing features. This measure may be coupled with increases in pole lateral offset, increases in pole spacing, the installation of utility poles on the inside of curves, and a combination of pole usage with multiple utilities.

• Aligning curb heights with guidelines contained in the benchmark, especially on roads with higher posted speed limits.

• Installing median barriers in situations where median crossover crashes are a concern and/or recommendations listed previously do not suffice. In this case, the first recommendation could still be followed but the second and third recommendations would be waived.

Acknowledgments

The authors thank the staff at the Abu Dhabi traffic police for sharing a large, good-quality crash database, as well as the staff of the Al Ain municipality for providing average daily traffic data.

Additional information

Funding

The authors thank the United Arab Emirates University for funding this research effort under grant number 31N262.