Home Skip to main content

UGPTI Research Will Help Transportation Agencies Improve Safety at Railroad Crossings

Posted: May 16, 2023

Photo Courtesy of NDDOTWhen trains and traffic meet, the result is often catastrophic and fatal. There were 203 railroad crossing crashes in North Dakota between 2011 and 2021. Twenty-five of those crashes resulted in fatalities and 67 resulted in injuries.

Consequently, safety at highway-rail grade crossings is a priority for transportation agencies, but with more than 3,223 miles of railroad operated in North Dakota and 3,725 public at-grade crossings, improving safety is huge undertaking.

Researchers at the Upper Great Plains Transportation Institute at North Dakota State University are developing tools to help those agencies prioritize where and how to spend limited safety improvement dollars. UGPTI Researcher Pan Lu is leading efforts to develop an innovative computer modeling technique that can serve as an efficient decision-making system to predict both crash occurrence and severity. The technique is also able to identify contributing factors to crashes, quantify their effects, and evaluate proposed countermeasures.

Of the 203 railroad crossing crashes in North Dakota from 2010 to 2021, 28 crossings had multiple crashes in the 12 years. “The likelihood of a crash at any single location is low, but if crashes do occur, the severity is likely to be very high” Lu noted. In fact, North Dakota has one of the lowest highway-rail grade crossing crash rates in the nation, but when crashes do happen in North Dakota, about 15 percent of them are fatal compared with just over 10 percent nationally. “That high fatality rate was one of the main drivers for us to combine these models in a way to help agencies compare scenarios to reduce risk at crossings,” Lu said.

Previously, Lu noted, most research efforts focused either on crash likelihood or crash severity, but not both together. “We wanted to fill that gap,” she said. In the research, Lu used a statistical technique called competing risk modeling, a form of machine learning, to evaluate both crash likelihood and crash severity together over time.

“The low incidence of crashes at any one crossing makes it difficult to direct improvements based only on crash history,” Lu said. “However, if we can look at what factors contribute to crashes and how much they contribute, the picture becomes much clearer. We want to know the marginal contribution of individual factors to crash risk and severity. That knowledge allows us to make improvements with the largest safety impact.”

The model, based on research into highway rail grade crossing crashes, considers the volume of train traffic and highway traffic, but also considers the type and size of the highway (Paved or unpaved? Two lanes or four lanes?), number of tracks, percent of truck traffic on the highway, the angle that the roadway crosses the tracks, the location of the nearest intersection, and many other factors. In addition, the model can evaluate the effectiveness of adding safety countermeasures such as stop signs, audible warning systems, flashing lights, and crossing gates.

Some factors, like nighttime train operations and multiple-track crossings, contribute to higher crash rates, just as one would expect. However, other factors such as the type and placement of warning signs and devices may have an opposite effect from what is expected or may reduce overall crash risk while increasing the severity of the crashes that do happen. For example, crossing gates often reduce crash risk significantly, but for those who choose to go around the gates, crashes are often severe.

Recently, the system has been incorporated into a web app that transportation agencies can use to evaluate crossings. Crossing characteristics such as road type, crossing angle, control devices, signage, and more can be entered into an “input panel” with resulting changes in crash likelihood and severity displayed on a “prediction panel.” The app illustrates the impact of various characteristics on expected crashes and their severity.

“Transportation agencies only have a limited amount of funding to direct toward safety,” Lu said. “The app allows them to evaluate where those dollars would have the highest impact.”

Safety at highway rail grade crossings is not just a North Dakota issue. There are more than 205,000 at-grade crossings in the United States. The Federal Highway Administration has recently given the research conducted by Lu’s team a relatively high reliability rating and added the work to its clearinghouse of crash modification factors. Federal, state, and local agencies use the factors to develop benefit-cost ratios for assessing safety investments.

The research has been underway for nearly a decade, and Lu and her colleagues at the Upper Great Plains Transportation Institute are now recognized as leaders in the field. Denver Tolliver, director of the institute, collaborates in the research. He has decades of experience in railroad operations and economics as well as transportation planning and highway systems modeling. Lu now serves as a research coordinator on a national committee focused on improving highway-rail grade crossing safety.

NDSU Dept 2880P.O. Box 6050Fargo, ND 58108-6050