Review of road user mobility impacts and criteria for prioritising highway-rail grade crossings for grade separation

ABSTRACT

Road users experience mobility impacts when a train occupies a highway-rail grade (level) crossing. Research has shown that the cost of reduced mobility exceeds safety costs, yet there is little consistency in the integration of mobility-related criteria into approaches for prioritising crossings for grade separation. A synthesis of findings from a review of literature and practice demonstrated the importance of mobility impacts at blocked crossings, identified and compared mobility-related decision criteria and actionable thresholds used within prioritisation approaches to rank crossings for grade separation, and revealed methods to quantify and monetise delay at blocked crossings. The review identified the need for the joint application of traffic microsimulation and intelligent transportation systems to quantify road user delay at blocked crossings. Such work should consider network-level effects, account for the severe consequences of delay for certain road users (e.g. emergency responders), and develop methods for monetising delay impacts associated with different road users. Moreover, a knowledge gap persists in establishing the interrelationship between road user delay at blocked crossings, risky behaviour, and safety impacts. Finally, further work is required to establish and calibrate thresholds for mobility-related criteria within prioritisation approaches used to rank crossings for all types of improvements, including grade separation.

KEYWORDS:

• Grade crossing
• level crossing
• grade separation
• road user mobility
• decision criteria
• prioritisation approaches

Introduction

Managing the multidimensional impacts at highway-rail grade (level) crossings draws considerable attention from transportation agencies, rail companies, regulators, and the public. Despite recognition of numerous impact categories, safety impacts remain the focus of most research and advancements of practice, with comparatively little work addressing broader economic, social, and environmental impacts (Berndt et al., 2019; De Gruyter & Currie, 2016). While available research has shown that costs associated with mobility impacts exceed those associated with safety impacts (AECOM, 2015; Dodgson, 1984; Gitelman, Hakkert, Doveh, & Cohen, 2006; Powell, 1982; USDOT, 2002, 2004), practitioners struggle to meaningfully integrate mobility-related criteria into grade crossing prioritisation decisions. This is a critical shortcoming given the costs and resources required to implement grade separations.

Road user delays occur when a train occupies a grade crossing—an event referred to as a crossing blockage when considered from the road users’ perspective. While crossing blockage delays affect all road users, the impact of those delays varies by road user type, ranging from general annoyance for typical commuters to severe and potentially life-threatening consequences for emergency response (ER) vehicles. Delays imposed on road users, while incremental at the individual level, represent a major portion of road user costs when aggregated for all users and blockage events (Dodgson, 1984; Ghaffari Dolama, Grande, Klassen Townsend, & Kashi Mansouri, 2018). If travel delay exceeds an acceptable level, drivers’ compliance degrades and the safety of both road and rail users may be jeopardised.

The particular motivation for this review stems from a need in Canada to develop guidelines to support the identification and prioritisation of crossings as candidates for grade separation. The Transportation Safety Board of Canada (TSB) articulated this need in its investigation of a fatal accident involving a double-decker bus and a passenger train at a crossing in Ottawa, Ontario in 2013. In its report, the TSB recommended that “[t]he Department of Transport provide specific guidance as to when grade separation should be considered.” It also noted that while government and industry had conventionally accepted a cross product (i.e. the product of the average daily rail movements and the annual average daily traffic at a crossing) of 200,000 as a threshold beyond which grade separation should be considered, no research existed to support that threshold (Transportation Safety Board of Canada, 2015a).

In response to this recommendation, Transport Canada (i.e. the Department of Transport) commissioned the University of Manitoba in 2018 to undertake a literature review to identify criteria and best practices that would assist public transportation agencies and rail companies in determining when grade separation should be considered at a crossing. That review provided the foundation for Canada’s Grade Separation Assessment Guidelines (Transport Canada, 2019) and for some of the content presented in this paper. More specifically, the review documented in this paper seeks answers to the following research questions:

• What decision criteria—particularly mobility-related criteria—are identified by available research and guidance related to grade separation and how are those criteria used to prioritise crossings for grade separation?

• What actionable thresholds have been established for the mobility-related decision criteria?

• What methods exist to quantify mobility impacts at blocked crossings and how have these evolved?

The research underlying this paper comprised two phases. Phase 1 involved a comprehensive literature review of relevant research concerning mobility impacts at grade crossings, current grade separation prioritisation practices (guidelines, standards, or criteria), and methods to quantify and monetise delay at grade crossings. The literature search identified documents published internationally over approximately the past decade (since around 2010), although some older documents deemed particularly relevant were also included. Given the initial aim to inform new Canadian guidance for Transport Canada, the review of grade separation practices focused on documented guidance used by jurisdictions in Canada and the United States, but also encompassed practices in other countries considered relevant to the Canadian context (e.g. Australia, the United Kingdom, Israel).

The review included scholarly articles identified on publicly-available research repositories (Google Scholar, Transportation Research Information Documentation, ScienceDirect, ASCE Research Library, Taylor & Francis Online), published government reports (e.g. produced by Transport Canada, the United States Federal Highway Administration (FHWA), and the United States Federal Railroad Administration (FRA)), and internal agency manuals of practice. Principal keywords used for on-line searches included: highway rail grade separation, grade crossing prioritisation, user cost grade crossing, collision grade crossing, emergency responders crossing, and cross product grade crossing. When necessary the term grade was replaced with level, which is more commonly used in the United Kingdom, Australia, and New Zealand.

The search identified more than 100 relevant documents. An examination of the abstracts or executive summaries of those documents produced a refined subset of documents, which were subject to a full text review. The full text review was followed by backward snowballing to identify additional relevant documents; in limited cases, forward snowballing was also practiced (Wee & Banister, 2016). The curation process resulted in approximately 70 documents advancing to Phase 2.

Phase 2 involved synthesising the information from Phase 1 to identify: the relative magnitude of mobility impacts at blocked crossings, approaches for prioritising grade crossings for separation, the use of mobility-related criteria within those approaches, and methods to quantify and monetise delay at grade crossings. To facilitate comparison of practices documented in the literature, this paper presents tabular summaries of findings on these topics. Each table is sorted chronologically and identifies the jurisdiction to which the research applies.

In the last five years, two review documents summarised grade crossing impacts and the implementation of grade separation to address those impacts. De Gruyter and Currie (2016) described various impacts of highway-rail crossings (i.e. grade separated and at-grade) and identified two principal knowledge gaps: the need to support crossing impact assessments with empirical evidence and the need for additional research on certain impacts of crossings (i.e. impact on land use, crime, disability access, travel time variability, traffic flow, rail vehicle delay, and safety). Their review provided a useful foundation for the more detailed assessment of mobility impacts addressed in this paper.

Berndt et al. (2019) conducted a literature review on methods, data, and tools used in the United States and abroad to evaluate and/or rank crossings for grade separation, with the goal of developing a tool for prioritising crossings along a rail corridor. Their review, which occurred simultaneously to the review commissioned by Transport Canada, identified various decision criteria and prioritisation approaches, but did not specify thresholds associated with those criteria or provide details on mobility-related impact assessments. In fact, the tool produced by Berndt et al. (2019) and subsequently published by Mathew, Benekohal, Berndt, Beckett, and McKerrow (2021) ultimately assumed an average delay per vehicle of 2.5 min/hour, despite acknowledging the need for more comprehensive methods for quantifying delay. Consequently, the report by Berndt et al. (2019) constitutes an important contribution to the literature (largely from the perspective in the United States) and emphasises the need for the review documented herein.

The remainder of this paper is structured as follows. The next section provides evidence of the magnitude of the mobility problem at blocked grade crossings. This is followed by a review of approaches and criteria for prioritising crossings for grade separation. The subsequent section describes methods to quantify and monetise delay at crossings. Finally, the closing sections discuss the contributions of this paper, identify remaining knowledge gaps, and offer concluding remarks.

Magnitude of the mobility problem at grade crossings

Safety concerns at grade crossings—particularly high-profile fatal accidents like the one noted at the beginning of this paper—often catalyse crossing upgrades and regulatory advancements (Transport Canada, 2019; Transportation Safety Board of Canada, 2015b). Recent statistics reveal the ongoing need to address crossing safety. In Canada, for example, the annual number of crossing accidents increased from 133 in 2016 to 178 in 2019, and the number of fatalities increased from 19 to 28 over that same period. Those parameters decreased to 129 and 18, respectively, in 2020 (Transportation Safety Board of Canada, 2021); however, that decline could be due to the reduced levels of road and rail activity during the COVID-19 pandemic rather than an underlying safety improvement. Likewise, the FRA reported increases in the number of crossing collisions (from 2050 in 2016 to 2229 in 2019) and fatalities (from 255 in 2016 to 298 in 2019) (Operation Lifesaver, 2021).

There is a relationship between crossing safety and the delays experienced by road users who encounter a blocked crossing. A report by the United States Government Accountability Office (2018) indicates that risky driver behaviour or poor judgment contributed to 94% of crossing crashes in the United States between 1994 and 2003. According to this report, delay at crossings instigates drivers’ impatience, which in turn motivates them to illegally bypass the crossing gates. Leibowitz’s early research on the psychological motivations for risk taking behaviour at crossings finds that uncertainty in the duration of crossing delays motivates drivers to ignore warnings and sometimes to undertake illegal manoeuvres. Every success in such an illegal crossing rewards and reinforces the risky behaviour (Leibowitz, 1985).

Other research on road users’ behaviour at grade crossings identifies that delayed road users take risky and operationally non-compliant actions such as running a red light, zigzagging around gate arms, speeding, and driving on the sidewalk or wrong side of the road in order to jump the queue (Delmonte, Tong, & Limited, 2008; EU Agency for Railways, 2016; FHWA, 2020; Tey, Ferreira, & Wallace, 2011). The FHWA also notes problems associated with trucks rerouting to streets that are not designed to accommodate them (FHWA, 2020).

Even when road user delay does not lead to a safety problem, when monetised, the literature nearly always finds that the costs associated with road user delay at grade crossings outweigh safety costs. Table 1 summarises findings from the literature. While comparisons between studies are not straightforward and studies employ various analytical methods, the evidence in the table indicates that road user delay costs can exceed safety costs by at least a factor of three, and can be higher by an order of magnitude or more. Notably, one recent study by Evans and Hughes (2019), which utilised newly-available crossing traverse data at railway-controlled and automatic crossings, found that the type of crossing control may influence the safety-delay trade-off.

Table 1. Comparison of road user delay and safety cost estimates at grade crossings.

Study	Jurisdiction	Road user delay cost	Safety cost	Context
Powell (1982)	Indiana, United States	32,000 per crossing (1980 USD)	9,400 per crossing (1980 USD)	Calculated delay using traffic microsimulation models and statistical analysis techniques. Monetised delay costs based on then-standard recommendations.
Dodgson (1984)	Ontario, Canada	72% to 94% of total cost	4% to 25% of total cost	Used five case studies to demonstrate the application of benefit-cost analysis for justifying grade separation. Integrated input from road and rail authorities to monetise benefits and costs.
USDOT (2002)	United States	Estimated increase in annual delay cost in 2020 compared to 2002 between USD 5.5 billion (uniform) and USD 7.8 billion (simultaneous peak) (2002 USD)	Estimated a USD 0.5 billion increase in annual safety cost in 2020 compared to 2002 for both scenarios (2002 USD)	Used GradeDec 2000 software to predict the increase in vehicle delay assuming two types of daily train and road traffic distributions: (1) the simultaneous occurrence of peak train traffic and peak highway traffic, and (2) a uniform hourly distribution of both train and highway traffic.
USDOT (2004)	United States	Estimated increase in annual delay cost in 2024 compared to 2004 between USD 7.8 billion (uniform) and USD 8.8 billion (simultaneous peak) (2004 USD)	Estimated a USD 0.7 billion increase in annual safety cost in 2024 compare to 2004 for both scenarios (2004 USD)	Same as USDOT (2004).
Gitelman et al. (2006)	Israel	USD 0.05–1.6 million per annum (2000 USD)	USD 8,000–37,000 per annum (2000 USD)	Predicted reduced accident cost after implementing crossing improvements along with the user delay cost, which included stopping delay and speed reduction delay.
AECOM (2015)	Minnesota, United States	USD 6.2 million (2015 USD)	USD 0.83 million (2015 USD)	Used methods underlying the GradeDec.Net software to support a benefit-cost analysis of a rail improvement project. Benefits comprised savings in rail inventory and rail operation costs, the salvage value of infrastructure after its useful life, and the incremental travel time, safety, and emissions reduction benefits associated with the improvement. Costs comprised capital costs and operation and maintenance costs.
Evans and Hughes (2019)	Great Britain	£73,919 per year at railway-controlled crossings (2010 GBP) £2,365 per year at automatic crossings (2010 GBP)	£1,509 per year at railway-controlled crossings (2010 GBP) £6,698 per year at automatic crossings (2010 GBP)	Compared safety and delay costs at railway-controlled (majority of crossings) and automatic crossings through a hypothetical analysis. Demonstrated that the trade-off between safety and delay costs hinges on the type of crossing.

Approaches and criteria for prioritising grade crossings for improvement

Literature and practice provide evidence of numerous approaches and criteria for prioritising grade crossings for improvement. Grade separation is one of several engineering treatments available for an existing crossing requiring improvement. Generally, because of the cost and complexity of such a project, the literature regards grade separation as the last resort after considering less aggressive options. These options include (Ogden & Cooper, 2019; Taylor & Crawford, 2009; Transport Canada, 2019):

• upgrading the crossing to a more active system (i.e. upgrading passive crossings by adding active components such as flashing lights, bells, and/or gates);

• utilising traffic control features (e.g. pavement markings, signs, signals, pre-emption technologies, variable message signs) to control and/or reroute traffic;

• performance monitoring of the operation of the crossing and/or various traffic control components installed at the crossing;

• redirecting road traffic away from the crossing through road network changes; and

• closing or eliminating the crossing.

This section reviews approaches and decision criteria evident in literature and practice for prioritising crossings for improvements, focusing on the consideration of mobility-related decision criteria and accompanying thresholds.

Prioritisation approaches and decision support tools

The literature and current practice propose or utilise approaches to systematically prioritise crossings for grade separation. Table 2 presents a chronological, document-by-document review of prioritisation approaches. The table reveals a common recognition that crossing improvement decisions need to be based on multiple criteria. The integration of those criteria within a prioritisation approach may rely on binary eligibility assessments, the assignment of criteria weights to facilitate ranking, or the application of monetisation techniques to support benefit–cost analysis. Most approaches consider some mobility-related criteria, such as traffic and rail volume or measures of delay, though establishing thresholds, weights, or monetary values for those criteria is neither straightforward nor consistent. Given the level of effort in gathering data for crossings under consideration, several prioritisation approaches include initial screening using easily-obtained criteria (e.g. highway or track class or cross product). After prioritisation, some of the literature acknowledges the need for site-specific technical feasibility analysis based on initial design considerations, costs, constructability, a variety of socio-economic factors, and input from stakeholder consultation. Finally, most of the prioritisation approaches documented in the literature appear to stem from cross-sectional studies of crossings, rather than longitudinal studies which could better characterise the impacts of various treatments on safety and mobility.

Table 2. Summary of approaches for prioritising grade crossings for grade separation.

Study	Jurisdiction	Description of prioritisation approach and study details	Consideration of mobility-related criteria?
Bates and Lee (1989)	Los Angeles, United States	Multi-criteria approach used for identifying the crossing improvement type that would alleviate light rail transit (LRT) impacts at grade crossings. Cross-sectional study of five crossings in 1986.	Yes
Nichelson and Reed (1999)	United States	Four-phase, multi-criteria prioritisation approach composed of initial screening, preliminary technical analysis, detailed technical analysis, and estimating expected benefits of grade separation. Guidance document.	Yes
ICC (2002)	Illinois, United States	Simplified benefit-cost methodology to estimate delay imposed on motorists at grade crossings and prioritise crossings based on the user delay cost. Cross-sectional study of 1732 crossings (analysis period unspecified).	Yes
Gitelman et al. (2006)	Israel	Benefit-cost screening tool based on predicted safety benefits, vehicle delay reduction benefits, and construction cost. Cross-sectional study of 216 crossings in 2000.	Yes
Bien-Aime (2009)	North Carolina, United States	Two methodologies to estimate the safety benefits of grade crossing closures and improvements along a high-speed rail corridor, which could be used to support prioritisation. Longitudinal study of 189 crossings during the period of 1995–2004.	No
Taylor and Crawford (2009); Crawford (2010)	Melbourne, Australia	Four-stage, multi-criteria weighting-and-ranking analysis tool to prioritise crossings for grade separation. Cross-sectional study of 177 crossings circa 2007.	Yes
Ben Aoun et al. (2010)	France and Germany	Benefit-cost analysis to identify optimum safety improvement measures for grade crossings. Cross-sectional study of crossings in the national railway systems (analysis period unspecified).	No
Wilbur Smith Associates (2011)	Kern County, California	Multi-criteria weighting-and-ranking methodology to identify potential grade crossing projects that create the most significant traffic and safety improvements. Cross-sectional study of 176 crossings (analysis period unspecified).	Yes
InfraConsult LLC (2012)	Riverside County, California	Multi-criteria weighting-and-ranking methodology for prioritising crossings for grade separation. Cross-sectional study of 46 crossings circa 2006.	Yes
FRA (2014)	United States	Benefit-cost methodology embedded into the GradeDec.Net software designed to support corridor and network level analysis of grade crossing improvements, grade separation, and crossing closures. Guidance document.	Yes
Minnesota Department of Transportation (2014)	Minnesota, United States	Multi-criteria weighting-and-ranking methodology to prioritise crossings for upgrade based on three criteria: population density adjacent to crossings, FRA accident prediction index, and existing physical conditions at the crossing. Cross-sectional study of 683 crossings in 2014.	No
Peel Goods Movement Task Force (2014)	Region of Peel, Canada	Multi-criteria approach for prioritising grade crossings for upgrade to improve the efficiency of goods movement. Cross-sectional study of 12 crossings identified in 2012.	Yes
Hans et al. (2015)	Iowa, United States	Weighted index method and an associated spreadsheet tool for identifying and ranking grade crossings as candidates for closure or elimination. Guidance document.	Yes
Rezvani et al. (2015)	North Carolina, United States	Safety-focused benefit-cost methodology, which considers delay as a secondary impact of collisions. Guidance document.	Yes
Jensen and Purkis (2016)	Toronto and Hamilton, Canada	Multi-criteria weighting-and-ranking methodology. Cross-sectional study of 185 crossings in 2016.	Yes
City of Edmonton (2017)	Edmonton, Canada	Initial screening process and multi-criteria weighting-and-ranking methodology used for LRT crossings. Longitudinal study of crossings along three LRT lines in 2015.	Yes
HDR (2016)	Montana, United States	Two-tiered, multi-criteria weighting-and-ranking methodology used to prioritise grade crossing improvements. Cross-sectional study of more than 5600 crossings in 2016.	Yes
Wu, Rilett, and Chen (2018)	Lincoln, Nebraska	Methodology to determine travel time reliability at the route and network levels, which could ultimately enrich prioritisation approaches. Cross-sectional study of three crossings.	Yes
Nebraska Department of Transportation (2019)	Nebraska, United States	Multi-criteria approach for identifying crossings eligible for grade separation. Guidance document.	Yes
Ogden and Cooper (2019)	United States	Multi-criteria approach recommended by the FRA for selecting appropriate types of crossing upgrades. Guidance document.	Yes
Transport Canada (2019)	Canada	Multi-criteria assessment approach for identifying crossings eligible for grade separation. Guidance document.	Yes
Prodan et al. (2021)	Ohio, United States	Multi-criteria weighting-and-ranking methodology. Cross-sectional study of 5,700 crossings (analysis period unspecified).	Yes

Given the complexity associated with prioritisation, software tools have been developed to support practitioners with the decision-making process. Examples include GradeX in Canada (Fu, 2016; Saccomanno, Park, & Fu, 2007), GradeDec.Net (FRA, 2014) and RCAT (i.e. the railroad crossing assessment tool) (Berndt et al., 2019) in the United States, and ALCAM (i.e. the Australian Level Crossing Assessment Model) in Australia and New Zealand (National ALCAM Committee, 2016).

Decision criteria

The literature increasingly recognises the need to consider a wide range of criteria when making grade separation decisions. De Gruyter and Currie (2016) and Berndt et al. (2019) provide recent tabular summaries of the various impacts and factors related to grade separation decisions. De Gruyter and Currie (2016) categorise crossing impacts as: (1) transport and economic, (2) social, and (3) environmental. The transport and economic impacts include safety, road and rail vehicle delay, vehicle operating costs, traffic volume, accessibility/connectivity, crossing operation costs, and grade separation costs. Social impacts include land use, community cohesion, geographic distribution, crime, visual amenity, noise, and sites of social significance. Environmental impacts include air quality (emissions), water quality, and sites of environmental significance. Building on this review, Berndt et al. (2019) identify factors within six categories, namely: (1) safety, (2) traffic and delay, (3) location and crossing geometry, (4) environmental, (5) community liveability, and (6) economic.

Given these recent reviews and the magnitude of mobility impacts associated with crossing blockages, this sub-section examines how literature and practice have integrated mobility-related criteria in grade crossing prioritisation decisions and, where available, identifies thresholds related to those criteria that trigger grade separation. In this way, the review presented herein extends the previous work and contributes to a more detailed understanding of road user mobility impacts at blocked crossings, particularly in the North American context.

Table 3 provides a summary of the quantitative thresholds specified in the literature for six categories of mobility-related decision criteria.

• Road traffic volume criteria: This category includes annual average daily traffic (AADT), average daily traffic (ADT), the temporal distribution of traffic volume, truck traffic volume, vehicle classification distribution, and traffic growth. The literature reports a range of road traffic volume thresholds used to justify grade separation, from an average daily volume of 15,000 up to 100,000. This range varies depending on the stage of crossing assessment, whether the crossing is in an urban or rural setting, and the type of rail traffic at the crossing (e.g. LRT lines have different thresholds than non-LRT lines). Notably, the most recent federal guidance in the United States (Ogden & Cooper, 2019) specifies lower traffic volume thresholds than those in earlier versions of that guidance (Ogden & Korve Engineering, a Division of DMJM+Harris, 2007), which were more consistent with current Canadian federal guidance (Transport Canada, 2019).

• Road traffic speed criteria: This category includes posted speed limit and vehicular speed approaching a crossing. Federal guidance in both the United States and Canada indicates that a crossing should be considered as a candidate for grade separation if the speed limit exceeds approximately 90 km/h (55 mph) (Ogden & Cooper, 2019; Transport Canada, 2019).

• Road traffic delay and related criteria: This category includes measures of current or future vehicular delay, travel time savings due to improvements, crossing blockage duration, train presence (binary variable), train length, gate down time, added travel time due to rerouting, congestion relief, queue length and queue release time, road capacity, level of service (LOS), and service reliability. The federal guidance in the United States and Canada specifies grade separation thresholds of 30 or 40 vehicle-hours per day (FHWA, 2002; Ogden & Korve Engineering, a Division of DMJM+Harris, 2007; Ogden & Cooper, 2019; Transport Canada, 2019). Otherwise, there are relatively few quantitative thresholds reported for delay related criteria in the literature, possibly owing to the recognition that establishing such measures requires an in-depth engineering study.

• Rail traffic volume criteria: This category includes measures of train volume (rail movements per day), train traffic distributions, the type of rail traffic (i.e. passenger or freight), and tonnage (commonly expressed as million gross tons or MGT). Rail traffic volume thresholds for grade separation range widely from as low as 12 trains per day to as high as 150 trains per day, depending on the urban/rural context and the type of train considered (FHWA, 2002; Ogden & Korve Engineering, a Division of DMJM+Harris, 2007; Ogden & Cooper, 2019; Transport Canada, 2019).

• Rail traffic speed criteria: This category includes rail design speed and maximum train speed. There is consistency between previous federal guidance in the United States and current Canadian guidance, which both specify that consideration should be given to grade separation when the maximum train speed exceeds 177 km/h (FHWA, 2002; Ogden & Korve Engineering, a Division of DMJM+Harris, 2007; Transport Canada, 2019). More recent guidance in the United States has lowered this threshold to 127 km/h (Ogden & Cooper, 2019).

• Cross product (Canada) or crossing exposure (United States): These parameters, which are definitionally consistent, combine measures of road and rail traffic volume by multiplying the AADT and the average daily rail movements at a crossing. Despite certain shortcomings (Grande, Rempel, & Regehr, 2020), this parameter has historically been used as an input for grade crossing improvement decisions—including grade separation—in both Canada and the United States. Table 3 provides evidence of this, but reveals inconsistent thresholds ranging 15,000 in rural areas (Nichelson & Reed, 1999) to 4,500,000 in urban areas when considering transit trains (Ogden & Cooper, 2019). In Canada, the most commonly referenced cross product value is 200,000. However, recent federal guidance indicates that grade separation studies may be warranted when the cross product exceeds 1,000,000. This threshold aligns with previous federal guidance in the United States (Ogden & Korve Engineering, a Division of DMJM+Harris, 2007), but not with the most recent guidance in the United States (Ogden & Cooper, 2019), which considers separate thresholds for freight, passenger, and transit trains. Those factors are identified as criteria without thresholds in the Canadian guidance.

Table 3. Summary of mobility-related grade separation decision criteria and thresholds.

Study	Jurisdiction	Criteria value or threshold that triggers grade separation
Road traffic volume
Bates and Lee (1989)	Los Angeles, United States	ADT exceeds 20,000 (LRT grade separation might be needed) or ADT exceeds 30,000 (LRT grade separation should be seriously considered).
Institute of Transportation Engineers (1993)	North American jurisdictions with LRT systems	Crossings with an ADT above 50,000 and 55,000 are likely to require grade separation for LRT headways of 3 and 6 min, respectively.
FHWA (2002); Ogden & Korve Engineering, a Division of DMJM+Harris (2007)	United States	In urban areas, AADT exceeds 100,000 or exceeds 50,000 if grade separation economically justified. In rural areas, AADT exceeds 50,000 or exceeds 25,000 if grade separation economically justified.
Wilbur Smith Associates (2011)	Kern County, United States	In initial screening, grade crossings with ADT greater than 15,000 (in 2011) or 20,000 (projected for 2035) are considered candidates for grade separation; these include crossings in both rural and urban areas. For subsequent prioritisation, crossings with ADTs greater than 40,000 are given highest priority.
Ogden and Cooper (2019)	United States (current federal guidance)	In urban areas, AADT exceeds 30,000. In rural areas, AADT exceeds 20,000.
Transport Canada (2019)	Canada (current federal guidance)	AADT exceeds 100,000.
Road traffic speed
Nichelson and Reed (1999)	Unspecified state	Exceeds 80 km/h (50 mph) if cross product exceeds 15,000 in rural areas.
FHWA (2002); Ogden & Korve Engineering, a Division of DMJM+Harris (2007)	United States	Posted highway speed equals or exceeds 113 km/h (70 mph) or 88 km/h (55 mph) if grade separation economically justified.
Ogden and Cooper (2019)	United States (current federal guidance)	Posted highway speed equals or exceeds 88 km/h (55 mph).
Transport Canada (2019)	Canada (current federal guidance)	Posted/unposted highway speed equals or exceeds 90 km/h.
Road traffic delay and related measures
FHWA (2002); Ogden & Korve Engineering, a Division of DMJM+Harris (2007)	United States	Exceeds 40 vehicle hours per day or exceeds 30 vehicle hours per day if grade separation economically justified. “An engineering study indicates that the absence of a grade separation structure would result in the highway facility performing at a level of service below its intended minimum design level 10 percent or more of the time”, if that separation is economically justified.
Wilbur Smith Associates (2011)	Kern County, United States	For prioritisation after initial screening, crossings with average delay greater than 300 s per vehicle are given a full score (i.e. highest priority). For prioritisation after initial screening, crossings with average queue length per lane greater than 150 feet are given a full score (i.e. highest priority).
Ogden and Cooper (2019)	United States (current federal guidance)	Exceeds 30 vehicle hours per day with consideration of cost effectiveness.
Transport Canada (2019)	Canada (current federal guidance)	Exceeds 40 vehicle hours per day. The highway/roadway facility is performing at a level of service below its intended minimum design level 10 percent or more of the time. Existing crossings where there are known queuing issues and an entranceway or intersection is within 30 m of the nearest rail of the crossing.
Rail traffic volume
FHWA (2002); Ogden & Korve Engineering, a Division of DMJM+Harris (2007)	United States	Average of 150 or more trains per day or 300 million MGT per year or, if grade separation economically justified, average of 75 or more trains per day or 150 MGT per year. In urban areas, average of 75 or more passenger trains per day or, if grade separation economically justified, average of 50 or more passenger trains per day. In rural areas, average of 30 or more passenger trains per day or, if grade separation economically justified, average of 12 or more passenger trains per day.
Wilbur Smith Associates (2011)	Kern County, United States	In initial screening, grade crossings with greater than 36 trains per day are considered candidates for grade separation. For subsequent prioritisation, crossings with more than 34 trains per day are given a full score (i.e. highest priority).
Ogden and Cooper (2019)	United States (current federal guidance)	Average of 30 or more trains per day. In urban areas, average of 75 or more passenger trains per day or 150 or more transit trains per day. In rural areas, average of 30 or more passenger trains per day or 60 or more transit trains per day.
Transport Canada (2019)	Canada (current federal guidance)	Average of 150 or more trains per day.
Rail traffic speed
FHWA (2002); Ogden & Korve Engineering, a Division of DMJM+Harris (2007)	United States	Maximum authorised train speed exceeds 177 km/h (110 mph) or exceeds 161 km/h (100 mph) if grade separation economically justified.
Minnesota Department of Transportation (2014)	Minnesota, United States	Train speeds are 64 km/h (40 mph) or more, the road has more than four lanes, and (a) the road speed limit is 48 km/h (30 mph) and the road has an ADT of 5,000 or more, or (b) the road speed limit is 88 km/h (55 mph) or more and the road has an ADT of 3,000 or more.
Ogden and Cooper (2019)	United States (current federal guidance)	Maximum authorised train speed exceeds 127 km/h (79 mph).
Transport Canada (2019)	Canada (current federal guidance)	Maximum train speed exceeds 177 km/h (110 mph).
Cross product (or crossing exposure)
Nichelson and Reed (1999)	Unspecified states; suggested thresholds may apply generally in the United States	In one state, crossing exposure exceeds 15,000 in rural areas and 30,000 in urban areas. In another state, crossing exposure exceeds 50,000 for state highways and 100,000 for all other highways. In another state, crossing exposure exceeds 15,000 in rural areas where speeds exceed 80 km/h (50 mph). Suggested crossing exposure thresholds for initial screening are 25,000 for rural areas and 48,000 for urban areas.
FHWA (2002); Ogden & Korve Engineering, a Division of DMJM+Harris (2007)	United States	In urban areas, crossing exposure exceeds 1,000,000 or 500,000 if grade separation economically justified. In urban areas, passenger train crossing exposure exceeds 800,000 or 400,000 if grade separation economically justified. In rural areas, crossing exposure exceeds 250,000 or 125,000 if grade separation economically justified. In rural areas, passenger train crossing exposure exceeds 200,000 or 100,000 if grade separation economically justified.
McCormick Rankin Corporation (2009)	Oakville, Canada	Cross product exceeds 200,000.
British Columbia Ministry of Transportation (2014)	British Columbia, Canada	Cross product exceeds 200,000.
PGMTF (2014)	Peel Region, Canada	Cross product exceeds 200,000.
The City of Ottawa (2017)	Ottawa, Canada	Cross product exceeds 200,000.
Jensen, Percy, and Pfeifer (2017)	Toronto/Hamilton, Canada	Unspecified, but reference is made to a cross product threshold of 200,000.
Nebraska Department of Transportation (2019)	Nebraska, United States	Crossing exposure must exceed 50,000 to be identified as a candidate location for grade separation, but this could be waived if justified based on the combined consideration of crossing exposure, crash costs, and the expected elimination of vehicular delay at multiple crossings in a corridor.
Ogden and Cooper (2019)	United States (current federal guidance)	In urban areas, freight train crossing exposure exceeds 900,000, passenger train crossing exposure exceeds 2,250,000, or transit train crossing exposure exceeds 4,500,000. In rural areas, freight train crossing exposure exceeds 600,000, passenger train crossing exposure exceeds 600,000, or transit train crossing exposure exceeds 1,200,000.
Transport Canada (2019)	Canada (current federal guidance)	Cross product exceeds 1,000,000.

The review identified two additional categories of mobility-related decision criteria, though neither had accompanying quantitative thresholds. First, the literature acknowledges that rail traffic delay (including rail vehicle delay and rail passenger delay) may need to be incorporated in economic analyses of grade separations (De Gruyter & Currie, 2016), even though available estimates suggest that the reduction in rail vehicle delay may contribute little to total benefits (Dodgson, 1984). Second, the literature increasingly recognises the importance of mobility impacts on special road users, such as non-motorised traffic, trucks carrying hazardous materials, school buses, and ER vehicles (De Gruyter & Currie, 2016; Ogden & Cooper, 2019; Transport Canada, 2019).

Methods to quantify and monetise delay at grade crossings

Quantification of delay at blocked crossings (or, conversely, travel time savings benefits associated with crossing improvements) necessitates the measurement and aggregation of incremental delays experienced by numerous road users. Table 4 summarises methods for quantifying delay evident in the literature. The methods include the use of site-specific field measurements, estimating delay using average gate down time, application of queuing theory to calculate delay experienced by stopped vehicles and vehicles required to slow down, and limited application of traffic microsimulation models.

Table 4. Summary of methods to quantify delay at blocked grade crossings.

Study	Jurisdiction	Method to quantify delay
Powell (1982)	Indiana, United States	Quantified delay using a simulation model (developed in Fortran) and statistical techniques.
Dodgson (1984)	Ontario, Canada	Applied queuing theory to quantify two types of delay: vehicle stopping delay at blocked crossings and vehicle slowing delay approaching a crossing.
Rymer et al. (1988)	United States	Used a traffic microsimulation model (NETSIM) to quantify vehicular delay at LRT crossings.
Roper and Keltner (1999)	United States	Applied queuing theory to quantify two delay components: vehicle slowing delay while approaching a crossing and crossing blockage delay.
ICC (2002)	Illinois, United States	Considered delay equal to gate down time. Calculated time train occupies a crossing as the product of train length and train speed, based on field observations.
United States Department of Transportation (2002, 2004)	United States	Appears to apply queuing theory to predict the increase in vehicle delay (e.g. passenger cars, trucks, and buses) assuming two types of daily train and road traffic distributions: (1) the simultaneous occurrence of peak train traffic and peak highway traffic; and (2) uniform daily distribution of both train and highway traffic.
Chandler and Hoel (2004)	Virginia, United States	Used traffic microsimulation model (Vissim) to quantify vehicular delay caused by LRT crossings.
Gitelman et al. (2006)	Israel	Used field data to quantify two delay components: vehicle stopping delay at a blocked crossings and vehicle slowing delay. Input data were obtained at 20 sample crossings and from train timetables.
Appiah and Rilett (2008)	Nebraska, United States	Used traffic microsimulation model (Vissim) to quantify crossing delay.
Taylor and Crawford (2009)	Australia	During primary screening, assumed delay equal to crossing blockage duration. In final assessment, applied a microsimulation tool to quantify delay at a single crossing and to generate “network effect multipliers” that could be applied to other similar crossings.
Wilbur Smith Associates (2011)	California, United States	Applied estimation methods based on queuing theory to estimate vehicle delay, as described in the Highway Capacity Manual (2000), Taggart, Lauria, Groat, Rees, and Brick-Turin (1987), and Schrader and Hoffpauer (2001). The last two references used train length, speed, and gate lowering time to estimate crossing blockage duration.
InfraConsult LLC (2012)	California, United States	Estimated delay imposed on impacted vehicles using total gate down time.
FRA (2014)	United States	Applied model developed by Lawson, Lovell, and Daganzo (1997) to estimate delay, based on queuing theory.
AECOM (2015)	Minnesota, United States	Assumed average vehicle delay equal to the average crossing blockage duration per day. Used the methodology embedded in GradeDec.Net software to quantify user delay cost.
Rezvani et al. (2015)	North Carolina, United States	Used expert judgment to estimate increased travel time of vehicles which reroute due to a grade crossing collision.
Nguyen-Phuoc et al. (2017)	Melbourne, Australia	Used traffic microsimulation and four step travel demand models to quantify the percentage change in network travel time.
Wu et al. (2018)	Nebraska, United States	Used traffic microsimulation to quantify both route and network level travel time reliability.
Chen and Rilett (2018)	Nebraska, United States	Used traffic microsimulation to quantify delay as an input to optimise signal timing plans at corridors with multiple grade crossings.
Berndt et al. (2019)	United States	Used a fixed value for average delay of 2.5 min per hour.
Evans and Hughes (2019)	Great Britain	Applied adaptation of queuing theory to quantify delay at typical railway-controlled and automatic crossings.

Literature published as early as 1982 envisages the possibility of using traffic simulation tools for measuring potential delay reduction benefits of grade separation. Two relatively early examples model network level delay caused by blocked LRT crossings (Chandler & Hoel, 2004; Rymer, Urbanik, & Cline, 1988). At these locations, crossing blockage times and durations may be estimated from known train schedules.

In contrast, crossing blockage patterns at freight rail crossings are more difficult to characterise. Consequently, as shown in the Table 4, quantifying delay at these crossings has mainly relied on queuing theory or fixed assumptions for average delay or gate down time. Neither of these methods considers the stochastic nature of blockages or network level delay impacts. One of the early applications of advanced microsimulation software (e.g. Vissim®) at freight rail crossings is a case study in Lincoln, Nebraska that investigates the usefulness of supplementing railroad pre-emption operations with variable message signs (VMSs) (Appiah & Rilett, 2008). More recent advancements using this same test bed have further demonstrated the capability of microsimulation tools: (1) to quantify the impact of crossing blockages on travel time reliability along routes and throughout a local sub-network (Wu et al., 2018); and (2) to quantify delay and optimise signal timing plans along a corridor with multiple grade crossings (Chen & Rilett, 2018). Similar work by Nguyen-Phuoc, Currie, De Gruyter, and Young (2017) used traffic microsimulation and four-step travel demand models to quantify the percentage change in network travel time based on analyses in Melbourne, Australia. More recently, Grande et al. (2020) analyse data obtained from pilot deployments of crossing blockage sensors developed by TRAINFO® in Winnipeg, Canada to characterise the distributions of blockage times and durations.

While progress has been made in quantifying crossing blockage delay, recent work indicates the need for further research about the application of microsimulation tools for estimating the delay reduction benefits of grade crossing improvements (especially at freight rail crossings), under various physical and operational scenarios (e.g. varying blockage timing and duration) (Appiah & Rilett, 2008; Berndt et al., 2019; De Gruyter & Currie, 2016; Ghaffari Dolama et al., 2018). Moreover, existing guidance in Canada and the United States recommends the assessment of secondary network impacts when considering grade separation decisions (Ogden & Cooper, 2019; Transport Canada, 2019). In the United States, GradeDec.Net provides the basis for such analysis by applying queuing theory; however, no such analytical tool is prescribed for use in the Canadian context.

Though not evident in Table 4, a small body of literature has recently identified the need to quantify delay imposed by blocked crossings on ER vehicles and to include this as a consideration in grade crossing prioritisation (Berndt et al., 2019; FRA, 2006; Hans, Albrecht, & Johnson, 2015; Ogden & Korve Engineering, a Division of DMJM+Harris, 2007; Ogden & Cooper, 2019; Transport Canada, 2019; Wilbur Smith Associates, 2011). Those documents recognise the potentially catastrophic consequences that may result from a delayed ER vehicle trip, including incremental fire damage to property, unnecessary complications or more severe outcomes for a delayed ambulance patient, or the delayed apprehension of a suspected criminal (FRA, 2006). To date, analytical contributions focus on the impact of grade crossing upgrades (including the implementation of train detection systems to better quantify blockage duration) (FRA, 2006; Goolsby, Vickich, & Voigt, 2003; Lee, Gay, Carroll, Hellman, & Sposato, 2004; Park et al., 2016), but not on grade separation. To further highlight the importance of travel time on ER trips, consider that in the case of a structural fire, the National Fire Protection Association (NFPA) in the United States limits the maximum initial full alarm for fire engines to eight minutes for low and medium hazard fires, and ten minutes for high hazard or high rise structure fires (NFPA, 2016). Given this directive, there appears to be a need to better quantify the negative impacts of crossing blockages on ER operations and the concomitant benefits of grade separation.

Finally, regardless of the approach to quantifying delay, the literature recognises the need to monetise vehicle delay to support grade separation prioritisation decisions (AECOM, 2015; Ben Aoun, El Koursi, & Lemaire, 2010; Berndt et al., 2019; Crawford, 2010; De Gruyter & Currie, 2016; Dodgson, 1984; FRA, 2014; Fu, 2016; Gitelman et al., 2006; Illinois Commerce Commission, 2002; Powell, 1982; Rezvani, Peach, Thomas, Cruz, & Kemmsies, 2015; Roper & Keltner, 1999; Rymer et al., 1988; Taylor & Crawford, 2009). As the quantification of delay increasingly accounts for network level effects, temporal variations, and the disparate values of time associated with different road users (e.g. ER vehicles), there is a need to similarly advance tools that incorporate probabilistic methods to monetise road user delay and other road user costs (e.g. vehicle operating costs, safety costs, emissions costs).

Road user cost is a decision support metric used, inter alia, to justify accelerated bridge construction methods, to optimise work zone configurations, and to identify contractual incentives and disincentives (Ghaffari Dolama, Cowe Falls, & Regehr, 2020; Sadasivam & Mallela, 2015; Salem, Salman, & Ghorai, 2018). More advanced probabilistic tools for monetising road user costs associated with crossing blockages have not yet emerged within mainstream practice. However, such methodologies, even if developed for other applications, are considered adaptable to quantify user cost reductions associated with grade crossing improvements. In particular, probabilistic approaches could assist in quantifying the network level impacts of both extreme delays (rare but lengthy delays that impact public perception and sometimes lead to non-compliant behaviour) and incremental delays (frequent but short delays that may be negligible at an individual level but considerable when aggregated for all road users at the network level).

Discussion of contributions and knowledge gaps

This section synthesises the findings from the literature by presenting six principal contributions and associated remaining knowledge gaps. First, road user delay costs instigated by crossing blockages can outweigh the cost of collisions at grade crossings by at least a factor of three, according to the reviewed literature. This highlights a key and long-standing challenge in supporting grade separation decisions: the need to measure, quantify, and monetise benefits accruing from network level travel time savings (i.e. reduction of vehicle delay). This necessitates measurement and aggregation of a wide range of delays (from extreme to incremental delays) experienced by various road users on the impacted road network, and the post hoc evaluation of delay reductions following grade separation. While the literature recognises this need, the quantification of delay at grade crossings has relied on relatively imprecise measures and has been spatially constrained to the immediate vicinity of the crossing due to the lack of granular data about the times and durations of grade crossing blockages and the inability of analytical tools to simulate traffic responses to blockages in real-time. Recent advancements in the capabilities of traffic simulation software (e.g. multi-resolution simulation, dynamic route assignment, and integration of real-time and offline data input from field sensors) offer improved tools for quantifying network wide delay and the travel time savings realised through crossing upgrades, crossing closure, or grade separation. Likewise, emerging ITS technologies (e.g. train detection systems) offer potential to better measure the times and durations of crossing blockages—two critical inputs for network level traffic microsimulation models. Further research and development are needed to better establish this analytical capability.

Second, the literature identifies the interrelationship between road user delays and safety problems at grade crossings. As safety science advances to enable evidence-based prediction of the potential safety benefits of grade separation, it appears that those predictions will need to incorporate a robust understanding of how delays motivate risky behaviours at grade crossings, and ultimately the extent to which delay reduction might mitigate grade crossing collisions.

Third, and related to the previous points, the literature identifies a need for a more detailed understanding of the types of road users impacted by a crossing blockage. Existing work has focused on motorised vehicles as a broad category, but the literature notes that neglecting certain road users may lead to underestimation of the consequences of delay. While several categories of road users are mentioned (e.g. buses, trucks carrying hazardous materials, and non-motorised road users), some recent literature focuses on quantifying the risk imposed by crossing blockages on ER vehicles, the delay experienced by those vehicles, and the potentially severe consequences of those delays. The integration of these considerations within a prioritisation method emphasises the importance of monetising delays in ways that account for the relative value of time for different road users. Here, there appears to be value in leveraging and adapting research developed in other transportation fields to probabilistically quantify road user costs. This capability would facilitate a more robust understanding of the relative contribution of delay costs and other road user costs.

Fourth, while specific calculations, weighting values, and criteria differ, the literature identifies that principal steps within grade crossing upgrade prioritisation tools include:

• an assessment of whether a crossing is a candidate for closure;

• an initial ranking or screening using easily-obtained criteria (e.g. highway or track class, road or rail traffic, or cross product) and/or more comprehensive multi-criteria weighting-and-ranking or cost–benefit analysis methods; and

• site-specific technical feasibility analysis based on initial design considerations, costs, constructability, additional socio-economic factors, and input from public consultation processes.

Moreover, while grade separations may minimise or eliminate certain negative impacts of grade crossings, it is evident that they are costly and time-consuming to construct and have relatively under-researched social impacts. Recent advances in traffic simulation models and ITS—such as the implementation of predictive crossing blockage warning systems integrated with a VMS to reroute traffic away from a blocked crossing or within optimised real-time traffic signal timing plans—represent a new class of crossing upgrades that could be explored as an alternative for grade separation. Prioritisation methods need to be tailored to determine the conditions for which this class of solutions is preferred over grade separation.

Fifth, the consideration of mobility-related criteria in grade separation decisions involves both simplistic comparisons with established thresholds that trigger grade separation as well as more detailed network-level analysis of mobility impacts. The review identified six categories for which quantitative thresholds have been defined in the literature: road traffic volume, speed, and delay; rail traffic volume and speed; and cross product. When thresholds are defined, there is evidence that they evolve over time and that inconsistencies between jurisdictions can arise—including current federal level guidance in Canada and the United States. To advance the state-of-the-practice, there is a need to better demonstrate the rationale for using disparate criteria and thresholds, to identify the pertinent contextual factors (e.g. rural or urban, passenger or freight rail) that should be considered in the application of a particular criterion and threshold, and to formulate practical mechanisms for harmonising criteria and thresholds amongst similar jurisdictions where appropriate.

Finally, the literature offers insights on the benefits of grade separation and the methodologies implemented to support grade separation decisions. Countries around the world face a common challenge in whether and how to implement grade separations, particularly in constrained urban environments. However, researchers and practitioners should carefully consider the transferability of findings in the literature to their subject region. Decisions to implement grade separations are complex and often depend on numerous site-specific factors. Broadly speaking, consideration should be given to the following questions:

• What is the regulatory context within which the grade separation decision is being considered? How does this context compare with the subject situation?

• How does infrastructure ownership affect grade separation decisions? How do the methods used to calculate benefits change if the rail line is owned privately or publicly?

• How is the grade separation decision-making process impacted by the presence of different types of road and rail traffic at a crossing? In other words, how can the presence of different types of road users and rail traffic (freight, passenger, transit) be reliably and systematically measured and incorporated into decision processes?

Conclusion

This paper reviewed and synthesised international research and practice on the consideration of mobility-related impacts within grade separation decisions. Focusing on the North American context, the review extended the current body of knowledge by demonstrating the relative importance of mobility impacts at blocked crossings, comparing the mobility-related decision criteria and actionable thresholds used within documented prioritisation approaches, and describing the evolution of methods to quantify delay at blocked crossings. Beyond the limited geographic scope of the review, the literature search and curation process were subject to limitations stemming from the multidisciplinary and dynamic nature of the topics addressed, and the necessary integration of peer-reviewed research with non-academic documentations of practice.

Despite these limitations, the review revealed knowledge gaps in need of further research and technological development. These gaps concern:

• the integration and application of traffic microsimulation and ITS to better quantify road user delays at blocked crossings;

• the interrelationship between road user delay, risky behaviour and operational non-compliance of road users at blocked crossings, and safety impacts;

• the monetisation of road user delays to account for varying types of road users, especially ER vehicles; and

• the calibration of thresholds for mobility-related criteria used in prioritisation approaches to accommodate local conditions and which consider opportunities to implement ITS-based upgrades as alternatives to grade separation.

Collaborative pursuit of solutions for these gaps by the research community, public sector agencies, and private technology firms will support more robust and defensible approaches for prioritising crossings for grade separation.

Acknowledgement

The authors gratefully acknowledge Transport Canada for financially supporting the research that led to the development of this review paper.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by Transport Canada.