🌧️ Wet Weather and Nighttime Cornering on Rural Roads: Lessons from the Calvados Department

‍The Context

Nestled in the heart of Normandy, Calvados is a tapestry of rolling hills and ancient hedgerows, where every winding road tells a story centuries in the making. These rural routes snake through landscapes that witnessed the Norman invasion of England and the D-Day landings, carrying whispers of history in their worn edges and weathered surfaces.

‍

Along the coast, the routes départementales offer breathtaking views of the English Channel, curving dramatically as they follow the contours of the chalk cliffs. Inland, the roads undulate through a countryside dotted with stone-built farmhouses and Calvados distilleries, where the famous apple brandy that bears the department's name is crafted.

‍

‍

The challenge of maintaining these poetic but precarious routes is uniquely Calvadian. Modern trucks and agricultural machinery must navigate paths originally carved for horse-drawn carts. Sharp turns that once slowed mounted messengers now test the limits of contemporary vehicles, particularly in the rain-slicked conditions common to this maritime department.

‍

The Value of Cornering Data

Rural highway safety presents unique challenges that differ significantly from urban road management. Over a wide area, it is not possible to maintain sensors and traffic cameras at scale, and therefore large information gaps exist. Connected vehicle data offers a unique window into the behavior of drivers in rural areas such as Calvados. Because the data is being generated and collected before the department even knows that it is needed, there is significant cost savings when compared to the installation and maintenance of expensive hardware. And because vehicles travel all over the road network, behavior and usage information is captured for virtually every road, even with a relatively small sample of the total number of vehicles.

‍

Vianova works with a number of data providers who have data in both urban and rural contexts. Some inputs are similar and useful in both- for example the volume of vehicles on a road and the average or 85th percentile speed on the segment. But some inputs are particularly insightful for one location vs another. For example, on rural roads, the number of rapid deceleration events (or “harsh braking”) is relatively rare, as there are fewer vehicles spread out over a wider area than there are in urban areas. Additionally, on rural roads, there are relatively few conflicts between vehicles and other modes, such as cyclists and pedestrians.

‍

However, connected vehicles can also provide unique insights about swerving and harsh cornering behavior. Cornering events occur when a vehicle experiences significant lateral force while navigating a curve. Our dataset captures these events by measuring both the vehicle's speed and the g-forces experienced during the turn. When a vehicle enters a curve at 50 km/h or higher and experiences lateral forces exceeding predetermined thresholds, the system logs it as a cornering event.

‍

Dangerous cornering events are a relatively common occurance on rural roads, given few design cues to slow down. In a data sample over six months, we analyzed more than 1.3 million hard cornering events  in Calvados, nearly 60% occurring on department-managed roads. The analysis reveals patterns that highway departments can use to enhance road safety and reduce accident risk.

‍

‍

Cornering in Calvados

Our study analyzed approximately 1.37 million vehicle cornering events in Calvados, France. Each cornering event in our dataset contained essential information including precise location coordinates, timestamps, speed measurements, and force measurements that indicated the severity of the cornering maneuver.

‍

We were most interested in understanding the impact on departmental roads. We pulled a map from The Institut National de l’information Géographique et Forestière (IGN) of departmentally managed roads in the department. These 9,422 road segments from the Calvados departmental road network connect towns and villages with each other, but largely avoid the center of large cities such as Caen and Bayeux. More than 60% of the total number of cornering events fell onto this set of roads, with the rest primarily lying on larger roads (limited access highways or routes nationales).

‍

‍

Weather conditions for each event were obtained through the Meteostat API, which provides historical weather observations. Our system queried the nearest weather station for each event, collecting data on precipitation levels in millimeters and visibility conditions. To ensure accuracy, we matched weather data within a one-hour window around each event's timestamp. Additionally, we implemented a simplified daylight hours model specific to northern France, with monthly sunrise and sunset times allowing us to classify events as occurring during either daylight or nighttime hours.

‍

Events were characterized through multiple dimensions: severity based on force measurements, speed categories ranging from low (under 30 km/h) to very high (above 70 km/h), environmental conditions including daylight/night and rain/dry distinctions, and detailed turn characteristics describing both direction and sharpness. To maintain data quality, we implemented several control measures, including excluding events with speeds exceeding 130 km/h, requiring valid speed readings above 0 km/h, employing GPS accuracy checks for location data, and applying temporal validation for weather data matching.

‍

‍

Learnings and Actions

The statistical analysis component of our study focused on comparing various metrics between departmental and non-departmental roads, examining differences in average cornering forces, speed distributions, severe event rates, weather condition impacts, and time-of-day patterns.

‍

The severity classification of these events proves particularly illuminating. Of the recorded events, 39.3% were classified as severe, while 14.2% reached extreme levels. These aren't just numbers – they represent potential near-misses and dangerous situations that could result in accidents.

‍

Speed ranges tell an important story. At lower speeds (0-30 km/h), only 24% of events were classified as moderate. However, in the 91-130 km/h range, a staggering 46.6% of cornering events were extreme, with the remaining 53.4% classified as severe. This progression clearly demonstrates how higher speeds amplify the risks associated with sharp turns.

‍

Our findings revealed distinct temporal risk patterns, with high-risk periods occurring during early morning (2-4 AM) and late evening (8-11 PM), despite these being low-volume periods. Event volume peaked during normal rush hours, but severity rates were actually lower during these high-traffic periods. This suggests that time of day may be a more significant risk factor than traffic volume alone. Force magnitudes showed clear daily patterns, with higher magnitudes recorded during low-traffic periods. This pattern suggests a need for enhanced safety measures during pre-dawn hours, possibly through improved lighting or signage.

‍

‍

Road-specific analysis identified several notable patterns. The D30 emerged as the highest-risk road segment, with a 30.5% severe event rate (±2.1% confidence interval). Event density varied dramatically across roads, ranging from 191.7 to 5,923.8 events per kilometer, indicating significant variations in risk exposure. Interestingly, we found no strong correlation between event volume and severity rate, suggesting that high-usage roads aren't necessarily more dangerous for abnormal cornering events.

‍

‍

Left turns consistently present higher risks than right turns, with severity rates of 10.1% compared to 8.1% for right turns. This difference is not only statistically significant but is accompanied by systematically higher force magnitudes (446.3 vs 438.7) and higher average speeds (40.4 km/h vs 36.3 km/h) during left turns. The consistency of these findings across different metrics suggests a fundamental difference in risk profiles between left and right turning maneuvers.

‍

Individual road segments show notable variations in risk patterns. The D30 road stands out with an unusually high rate of severe right turns (36.8%), while the D263 shows the highest risk for left turns (35.2%). The fact that seven of the ten highest-risk segments involve left turns reinforces the broader pattern of left turn risk, but the presence of some high-risk right turn locations (like D30) suggests that local road geometry and conditions can override general patterns. This indicates that while systemic approaches to left turn safety are important, location-specific factors must also be considered in safety improvements.

‍

The speed differential between left and right turns (4.1 km/h on average) is particularly interesting. Higher speeds during left turns may partly explain their increased risk, but the relationship isn't straightforward - the correlation between speed and force is relatively weak (r=-0.024 for left turns, r=-0.007 for right turns), suggesting that other factors such as turn geometry, visibility, or driver behavior patterns play important roles. The boxplots of force distribution show similar patterns of outliers for both turn directions, indicating that extreme events occur regardless of turn direction, even though they're more frequent in left turns.

‍

‍

Environmental factors played a less significant role than might be expected. Rain conditions showed minimal impact on event severity, and night conditions demonstrated similar risk patterns to daytime. This suggests that road characteristics and time of day may be more influential factors than weather or lighting conditions in determining the likelihood and severity of abnormal cornering events. However, the fact that risky cornering events did not decrease during wet weather is itself a warning sign, as these unsafe cornering behaviors could result in a collision when combined by the surface road conditions and visibility limitations during wet weather events.

‍

These findings have significant implications for road safety and infrastructure planning. The identification of specific high-risk road segments, combined with clear temporal patterns, suggests the need for targeted, time-specific safety measures. The complex relationship between speed and event characteristics challenges simple speed-based risk assessments and suggests the need for more nuanced approaches to road safety. High event density locations warrant particular attention for infrastructure review, while the observed turn direction preferences at different speeds should inform road design considerations.

‍

Roads with high numbers of cornering events were largely consistent regardless of the conditions, but the order of priority varied slightly.

‍

‍

Outcomes

Moving forward, these insights can help inform targeted interventions in road safety and infrastructure planning. The clear patterns in temporal risk and road-specific characteristics suggest opportunities for dynamic safety measures and focused infrastructure improvements. Further analysis might explore the specific characteristics of high-risk road segments and the factors contributing to their elevated risk profiles.

‍

For rural highway departments, this data presents several actionable insights. First, curves that consistently generate high-severity cornering events may need redesign or enhanced warning systems. Second, the correlation between speed and severity suggests that strategic placement of speed reduction measures before challenging curves could significantly improve safety. Finally, the temporal patterns indicate when increased patrol or monitoring might be most effective.

‍

By understanding these patterns, highway departments can move from reactive maintenance to proactive safety management. Rather than waiting for accidents to identify dangerous curves, cornering event data allows for preventive interventions before accidents occur.

‍

‍

About Vianova

Vianova is the data analytics solution to operate the mobility world. Our platform harnesses the power of connected vehicles and IoT data, to provide actionable insights to plan for safer, greener, and more efficient transportation infrastructures.

‍

From enabling regulation of shared mobility to transforming last-mile deliveries, and mapping road risk hotspots, Vianova serves 150+ cities, fleet operators, and enterprises across the globe to change the way people and goods move.

‍

Ready to learn more?  Visit our page to get in touch.