A Correlation Between Road Surface Conditions and Road Traffic Incidents with Personal Injuries in Cold-Climate Areas

Low road surface temperature (RST) and precipitation can make driving challenging, especially in cold climate areas. Slippery roads increase delays in road traffic and reduce road traffic safety, which can expose motorists to the danger of accidents with injuries and fatalities. To prevent road incidents during winter, it is important to find a correlation between severe road accidents, weather conditions, and road surface conditions (RSCs).

As there is much research, including different types of analysis, to show the correlation between road incidents and road and weather conditions. Hagiwara et al. [1] investigated the rate of road traffic incidents due to road conditions during winter. Their conclusion indicated that the number of accidents reduced when there was no snowfall. Norman et al. [2] utilized a quantitative method to find a relationship between winter road maintenance (WRM), traffic incident risk, and slippery roads. They used data from road weather information in southern Sweden. The results showed that despite high-quality WRM, there is a high risk of road accidents when the road is slippery because of rain and sleet. Andrey et al. [3] analyzed data from six different cities in Canada (Waterloo, Ottawa, Halifax, Hamilton, Regina, and Québec). They conducted a sensitivity analysis to determine the risk of road incidents based on 4 scenarios. They concluded that precipitation is 75% associated with road traffic collisions, and 45 % with injuries, compared to warm seasons; however, the risk level differs according to the weather conditions in different cities. Usman et al. [4] presented a model to find the incident frequency in snowstorms due to slippery road surfaces or friction, and poor visibility. The results showed that RSCs played an important role in road accidents during winter. Yannis and Karlafis [5] used an integer model to estimate the impact of weather on four road incident categories: vehicle incidents, vehicle incidents with injuries, pedestrian incidents, and pedestrian incidents leading to fatalities. They used 21 years of data from Athens, Greece, and the results showed that daily precipitation is a significant variable in road accident occurrence. Andersson and Chapman [6] identified the direct relationship between road incidents and temperature in the West Midlands County in England. Theofilatos and Yannis [7] reviewed the impact of weather conditions on road safety. They found precipitation led to increasing road incidents; however, it does not have a consistent impact on the severity of the incidents. Perrels et al. [8] evaluated the relationship between weather variation and road traffic incidents in Finland. The assessment showed that severe weather conditions can significantly increase road accidents. Black and Mote [9] defined a pair analysis to explore the links between road accident risk and precipitation in winter, including sleet, snowfall, and freezing rain in 13 cities in the United States. The results demonstrated that intense winter precipitation can increase the risk of collisions and injuries. Heqimi et al. [10] proposed a spatial interpolation approach to analyze the impact of snowfall on road traffic accidents in Michigan. They used 10 years of data from road weather stations during winter (December, January, and February). The conclusion indicated that there is a direct link between snowfall and all types of road accidents in winter.

Multi-sensor integration is a powerful approach to enhance the reliability of road condition monitoring. Combining and analyzing data from various sensors provides a more accurate analysis of RSC under different environmental and weather conditions. Casselgren et al. [11] classified road conditions using data from cameras and a laser system to enhance the accuracy of RSC assessments under different winter weather conditions. Diaby et al. [12] used a combination of data from two sensors (camera and microphone) to identify dangerous RSCs under challenging winter conditions to improve decision-making for WRM in cold climate areas. Hong et al. [13] utilized multi-sensor data (round force sensor, water pressure sensor, ultrasonic sensor, and temperature/humidity sensor) to contribute to preventing road accidents due to black ice and improve road safety management. Dilorenzo and Yu [14] provided a review study of the existing sensors to explore icy RSCs to minimize winter road accidents and improve winter road safety. Jiang et al. [15] proposed a methodology to detect RSC using multi-sensor (temperature sensor, road condition sensor, and microwave measurement sensor) data analysis and a neural network algorithm to increase winter road safety and improve WRM strategies. Zhang [16] used data from OBDII sensors and cameras for a precise assessment of RSC and road safety in winter to enhance winter road management strategies.

While several studies have focused on finding the relationship between RSCs, weather conditions, and road safety during winter using multi-sensor data analysis, there is no study to use data points from four different sources (road weather station, sensor mounted in the wheel track, sensor mounted in the middle of the road, and optical sensor) to explore the possible effects of sleet or freezing rain, using the dataset with minimum data loss to find the link between winter road conditions and road accidents with personal injuries. Hence, this study uses datasets from various sources recorded at Testsite E18 in Sweden. Then, data from icy RSCs were extracted from the different datasets to select the dataset with the minimum data loss. In addition, the chosen dataset was used to extract information about possible sleet or freezing rain during autumn and winter. Finally, road accidents with personal injuries were analyzed to determine the correlation between weather conditions, RSC, and severe road accidents.

The remainder of this paper is arranged as follows. Section 2 defines the research question. The road weather data is explained in Section 3. The sensors are described in Section 4. In Section 5, sleet and freezing rain conditions are defined. Road accidents with personal injuries are analyzed in Section 6. The conclusion is drawn in Section 7.

In this study, data points recorded every 10 minutes from the year 2019 to 2023 during autumn and winter at the testsite E18 (Trafikverket [17]) are studied. Reviewing the dataset showed that the data points for the years 2019, 2021, and 2022 are available for different months in autumn and winter (January, February, March, October, November, and December). However, the year 2020 missed the data points in November, and the year 2023 missed the data points in October, November, and December.

3.1 Testsite E18 in Sweden

Level 1 Testsite E18 is a permanent and unique research station for meteorology and traffic data measurement. Testsite E18 is located midway between Enköping and Västerås in Sweden [17]. The goal of constructing this station is to improve winter road maintenance, maximize road safety, develop decision-making, and minimize cost and environmental impacts via detailed information. Various weather and road variables can be measured and recorded using modern digital equipment (sensors, instruments, etc.), which utilize artificial intelligence and machine learning algorithms [18]. Testsite E18 is led by Trafikverket (“Swedish Transport Administration”), involving KTH (“Royal Institute of Technology”), VIT (“Swedish National Road and Transport Research Institute”), and SMHI (“Swedish Meteorological and Hydrological Institute”). Approximately 20,000 vehicles pass this research station every day [19].

3.2 Traffic data

Traffic data is recorded at a 10-minute resolution for light (passenger car and passenger car with trailer) and heavy vehicles (truck, bus, and truck with trailer) in both lanes. The peak of traffic for light and heavy vehicles is at 7:00 am and 4:00 pm, where the afternoon peak is higher than the morning peak for light vehicles, and the morning peak is higher than the afternoon peak for heavy vehicles. In total, average hourly traffic for light vehicles varies over the year from approximately 270 vehicles in January to 365 in May. In addition, the average hourly road traffic for heavy vehicles is at the maximum value in May and the minimum value in December. Therefore, there is more road traffic during spring and summer than in autumn and winter. It is important to mention that in May and September; the traffic data loss is large [18].

3.3 Average temperature in Enköping and Västerås

Average air temperature in Enköping and Västerås, respectively. These two figures depict that the average air is approximately the same, and there is no huge weather condition difference between these two locations [20, 21].

Freezing rain and sleet can lead to very dangerous driving conditions. Freezing rain happens when rain freezes due to a shallow freezing layer close to the road surface. Sleet is snow, which turns to rain when it reaches a warmer air layer, and it refreezes due to the thin layer of cold air above the road surface [24].

Due to the highest number of RSC data points recorded by the optical sensor, it was decided to use icy RSC data points reported by this sensor. In addition, the variable analysis was extended based on precipitation type (rain and snow) reported by the road weather station (RWS). It is important to mention that light snow/rain has not been considered in this extraction. Table 1 shows the number of icy RSC data points due to rain and snow for different months in different years. As shown, December 2021 had the highest icy RSCs.

Table 1. Number of icy RSCs due to rain and snow for different months in different years

Filtering data points according to the mentioned variables (RSC and precipitation type) helped us to extract freezing rain and/or sleet conditions. Extracting information in January for the different years showed that in January 2019, 41 icy driving conditions were related to snowy weather, however, the weather was rainy at 1 datapoint. Table 2 shows the timestamp, air temperature, and RST recorded by two mounted sensors on the road. As is shown in Table 2, the air temperature was -0.3℃, while RST was -1℃ on the wheel track and -1.5℃ in the middle of the road.

Table 2. Possible freezing rain/sleet in January 2019

No rainy weather was reported in icy RSCs in January 2020 and 2021.

Table 3 shows the road conditions in January 2022. 59 datapoints were reported icy RSCs; however, in 8 of them, the weather was rainy (51 datapoints demonstrated snowy weather). As is shown in Table 3, the air temperature was between -1.1℃ and 0.6℃, but both sensors missed recording RST.

Table 3. Possible freezing rain/sleet in January 2022

Table 4. Possible freezing rain/sleet in January 2023

Table 4 shows RSCs in January 2023. Between 19 data points related to icy RSC, rainy weather was reported in 4 cases, and the rest (15 cases) reported snowy weather. In the 4 rainy cases, the minimum and maximum air temperatures were -0.5℃ and 0℃, respectively. In addition, RST measured by the sensors showed slightly higher temperatures on the wheel track. This can be due to friction between the tires and the road surface.

Table 5. Possible freezing rain/sleet in February 2019

Table 5 shows the recorded RSCs in February 2019. 4 icy RSCs reported rainy weather amongst 55 data points, which means that the weather was snowy in 51 icy RSCs. In these 4 data points, the air temperature was around -0.3℃ and the RST was approximately between -2.1℃ and -0.6℃.

No rain was reported in icy RSCs in February 2020, 2021, and 2023.

Table 6. Possible freezing rain/sleet in February 2022

In February 2022, 42 cases were reported in icy RSCs, where 2 cases were reported in rainy weather (Table 6). Unfortunately, no RST data was recorded by two sensors, but the air temperature was around 0.6℃.

No rainy condition was reported in March 2019 and 2023, and no icy RSC was reported in March 2020 and 2022.

Table 7. Possible freezing rain/sleet in March 2021

Table 8. Possible freezing rain/sleet in November 2022

Table 9. Possible freezing rain/sleet in December 2019

Tables 7-9 show two possible freezing rains during December 2019. In 29 icy cases, 2 cases reported rainy weather. In these cases, the air temperature was almost -0.3℃, and RST was slightly higher than -1℃.

No rain was reported in November 2019, and no icy RSC was reported in October 2019, 2020, 2021, and 2022.

In November 2022, 20 icy RSCs were due to snowy weather, however, 2 icy RSCs were related to rainy weather (Table 8). RWS reported air temperature was between -0.2℃ and 0℃, and RST was around -0.8℃.

In March 2021, the weather was rainy in 5 icy RSCs (Table 7), and snowy in 11 cases. RWS showed that the air temperature was between 0.2℃ and 0.4℃, and the sensors showed that RST was -0.6℃ and -0.4℃.

No icy RSC was reported in December 2020, and no rainy condition was reported in December 2022.

In December 2021, RSC was icy in 96 cases, where 14 cases reported rainy weather (Table 10). In these cases, the air temperature was between -1℃ and -0.2℃. Unfortunately, no RST was recorded by the sensors.

Table 10. Possible freezing rain/sleet in December 2021

This study has identified a correlation between road weather conditions and road accidents with personal injuries during autumn and winter at Testsite E18 in Sweden. Road weather condition data were measured by three sensors: one optical sensor, and two sensors mounted on the road (in the wheel track and one in the middle of the road). Analyzing data obtained by these three sensors showed that the optical sensor had the minimum data loss. Although there was not much difference between measurements from the three sensors, the dataset from the optical sensor was chosen due to minimum data loss. Icy road surface conditions extracted from the optical sensor dataset and precipitation type from road weather station data were used to find the average road surface temperature and the possibility of sleet or freezing rain at this location. In addition, road accidents with personal injuries at this location were analyzed based on the date and time of the year to extract weather and road surface conditions during the road accidents.

The results indicate that icy road surface conditions, together with light snow, had the second-highest number of road accidents leading to personal injuries. Moreover, despite a higher average daily traffic in spring and summer, higher accident rates were registered in autumn and winter. Among the reported 83 road accidents during 2019-2023, 51 accidents occurred during the cold months (January, February, March, October, November, and December), and 32 accidents occurred during warm months (April, May, June, July, August, and September).

Information for 40 road accidents during the cold months was available for more analysis, while information about 11 accidents was missed due to a lack of weather and road datasets in October 2023, November 2023, and December 2023 for 8 accidents, and information from 3 accidents in October 2021, December 2022, and March 2023. Among the 40 accidents with personal injury identified, 24 accidents occurred when the road temperature was between -3℃ and 3℃. Furthermore, the number of road accidents when the road surface temperature during this cold period was less than -4℃ was equal to 8, with the same number when the temperature was above 4℃. Thus, the number of road accidents that occurred when the road surface temperature was between -3℃ and 3℃ was three times the number of road accidents when the road surface temperature was less than -4℃ or above 4℃. Note that the average temperature during December, January, February, and March was -2℃ or below. Thus, there is a strong correlation between road accidents and weather and road surface conditions in winter at Testsite E18, with a high number of accidents when the road temperature is around the freezing point.

Winter road maintenance is needed when the temperature is close to zero, particularly when near-zero temperatures and freezing rain or sleet are forecasted. During very cold periods, the needs are lower. In addition, snow removal and salting can be prioritized in high-risk areas. Moreover, the installation of road signs helps drivers reduce speed in risky conditions, particularly on curves and slopes.

For future research, it should be possible to identify a relationship between winter road maintenance conditions and road traffic accidents.