What is a pattern?

Patterns represent a way to understand driver behavior, roadway attributes, and crash outcomes based on crash data. There are generally three types of patterns; temporal, spatial, and attribute. Each of these pattern types help us understand different crash elements in order to determine if a hazardous situation exists and what could be done to mitigate it.

Pattern Cohesion

When exploring potential safety problems, ideally a site will show all three patterns; a temporal pattern of crashes happening consistently, a spatial pattern of where they are happening, and an attribute pattern in the events of the crashes. When all three patterns are present there is a higher level of confidence that a countermeasure will be effective. The presence of only one pattern may be sufficient depending on the situation, but ideally multiple patterns will be observed before recommending a countermeasure.

Temporal Patterns

Temporal patterns show that a specific event is happening at specific hours of the day, days of the week, or season/month of the year. A specific hour may have more crashes or higher severity crashes. A specific time of year may have more crashes due to schools, public events, or weather. Temporal patterns also show consistency. A few similar crashes happening one month may not represent a pattern if that crash type is not also happening the rest of the year. Or it may suggest a unique condition that existed for a short period.

Spatial Patterns

Spatial patterns relate to specific locations or areas where crashes are happening. These are most obvious at intersections where vehicle movements and crashes are concentrated. These may also happen on congested urban corridors where a specific travel direction has a concentration of crashes approaching an intersection. Roadway departure crashes observed consistently at one rural curve is a spatial pattern. Several roadway departure crashes within a few miles may also reveal a spatial pattern. Spatial patterns help us identify exactly where a countermeasure may need to be applied.

Attribute Patterns

Another pattern type is consistent or related sets of attributes. A rural highway may experience roadway departure crashes, but the other events of the crash will help identify the pattern that can be mitigated. For example, if these crashes all include a fixed object, clear zone improvements are probably needed. If they all involve rollover/overturning, shoulder improvements or barriers could be considered. If they vary in type and event, rumble strips may be the best solution. At intersections, travel direction and vehicle movements are especially important for attribute pattern identification. Crashes that involve similar movements and directions tell us what design elements or signal characteristics should be changed to prevent future crashes.

How Many Crashes Make a Pattern?

There is no specific number of crashes that constitutes a pattern. When exploring crash data, focus on determining if the patterns lead to a clear countermeasure opportunity. There may only be a few crashes but if they all have the same patterns there is likely an opportunity for mitigation. In other situations dozens of crashes may be happening at one location but with no consistent patterns.

Comparing Patterns vs Average Distributions

The presence of a pattern does not always mean there is a safety problem. The safe system approach tells us that crashes are inevitable but serious injury or death is not. Patterns that result in serious injury or death present a likely opportunity for mitigation, but patterns that only include low-severity crashes may indicate that the system is performing well by avoiding serious harm. Comparing patterns at one site against statewide average distributions helps reveal what patterns might be “normal” and what patterns are unique to that site or situation.

UDOT has developed severity and crash type distributions to better understand what “normal” trends are. By comparing your site to these averages you can identify unique characteristics of the site to be addressed. If your site has a higher proportion of left-turn crashes you can focus on a left-turn mitigation. If there is a high proportion of injury crashes, you may need to look at reducing speeds or separating movements to avoid injury.

If the distributions match statewide distributions, this does not mean a site is “safe”, only that it matches expected patterns. In this case you will probably need a site visit or a deeper look into the attribute data to determine what countermeasures might be appropriate.