4.0 Accident Data Analysis
This analysis intended to assess the potential impacts of RWIDS on the occurrence of traffic accidents in the winter. It was hypothesized that the ITD use of RWIDS information could result in better and safer roads during the winter season. In addition, with RWIDS, the general public could be better informed of the road conditions thus reducing the likelihood of accident occurrence.
4.1 Methodology
This analysis compared the before-and-after changes in winter weather-related accident rates using Idaho statewide accident data between 1996 and 2004. The accident data was obtained by querying the accident database by a number of relevant criteria:
- Occurred during winter (December-February)
- Winter weather/pavement-related accidents
- Occurred on interstate and state routes (excluding local streets)
While it is desirable to assess safety impacts by analyzing the accident statistics, several threats to validity in this and other accident data analyses need to be acknowledged. The identification of winter weather-related accident records can be obscured by the fact that the presence of snow or ice does not always indicate that the weather or pavement condition is the cause of the accident. For example, driver behavior was often cited as the main contributing factor regardless of the environmental conditions and therefore the reported presence of snow may or may not be a contributing factor. Since accidents on rural freeways are relatively rare events, arbitrations in accident data coding could have potentially significant impacts on the analysis outcome.
When assessing safety impacts, the accident frequency (counts) must be normalized by the exposure. The exposure data in this analysis include Vehicle Miles Traveled (VMT) and winter precipitation level. Ideally, vehicular volume data is needed for all segments of roadway of interest (i.e., state routes). However, the ITD database (like most other states) only contains Annual Average Daily Traffic (AADT) which is derived and adjusted for seasonal variation using a few days worth of traffic counts on various freeway segments.
The winter weather severity data is obtained from the National Climatic Data Center (NCDC) operated by the National Oceanic and Atmospheric Administration (NOAA). A winter weather severity index is computed by normalizing the accumulative precipitation(1) over the winter season (i.e., December to February).
One limiting factor in this analysis is the ITD outdated route number and mile post definitions. ITD has failed to update the route numbering and the correspondent mile post definition as new freeways were built and some freeways merged/connected. The inconsistent route number and mile post definitions have impacted the location reference in accident reporting on rural freeways. This limits the ability to perform analysis of the accident statistics based on specific geographic areas (e.g., maintenance districts and foreman areas).
Given the above limitations, the accident analysis was performed at statewide level using winter weather-related accident data and Vehicle Miles Traveled (VMT), as exposure data, aggregated over all Transportation Analysis Zones (TAZ) in the state of Idaho. VMT is derived based on the AADT counts and adjusted for the total roadway mileage within an area and thus an adequate exposure measure for state level analysis. The analysis compared the before accident rates (normalized for winter weather severity) since 1996 with one year (winter 2003-2004) of post-deployment accident rate data. The analysis results are presented in the following section.
4.2 Results
Approximately fifty-six (56) percent of accidents occurred during the winter months (i.e., October-March). Among those, about thirty-eight (38) percent are winter weather-related. The accident rates were computed by dividing the accident frequency by the exposure data, total vehicle miles traveled (VMT). The unit of the accident rate is the number of accidents per 100 million vehicle miles traveled. The post-deployment winter 2003/2004 shows an increase in the accident rate compared to the previous years. However, winter 2003/2004 experienced more precipitation than the previous three years. When winter severity is considered, the post-deployment accident rate is slightly decreased from the previous year.
| Winter |
Before RWIDS |
After |
||||||
|---|---|---|---|---|---|---|---|---|
| 1996/ 1997 |
1997/ 1998 |
1998/ 1999 |
1999/ 2000 |
2000/ 2001 |
2001/ 2002 |
2002/ 2003 |
2003/ 2004 |
|
No. of Winter-Related Accidents | 1,759 |
1,736 |
1,620 |
1,795 |
1,884 |
1,308 |
1,623 |
2,153 |
VMT (100 millions) | 12,924 |
13,112 |
13,644 |
14,328 |
14,299 |
14,303 |
14,400 |
14,825 |
Accidents per 100 million VMT | 0.14 |
0.13 |
0.12 |
0.13 |
0.13 |
0.09 |
0.11 |
0.15 |
Total Winter Precipitation (inches) | 9.00 |
5.83 |
6.99 |
7.91 |
6.19 |
4.41 |
5.20 |
6.83 |
Winter Severity Index | 1.38 |
0.89 |
1.07 |
1.21 |
0.95 |
0.67 |
0.79 |
1.04 |
Accident Rate Normalized by Winter Severity) | 0.099 |
0.149 |
0.111 |
0.104 |
0.139 |
0.136 |
0.142 |
0.139 |
Figure 4-1 shows the trend of the winter weather-related accident rates between years 1996 and 2004. The vertical bars in the figure show the range of plus and minus one standard deviation from the mean.
Figure 4–1. Statewide Winter Weather-Related Accident Rates
Figure 4-1 shows that the accident rate of winter 2003-2004 (post-deployment of RWIDS) decreased slightly from the previous winter. However, 2003-2004 accident rate falls within one standard deviation from the statistical mean and thus is not considered statistically different from the previous years.
As discussed in the methodology section, it is difficult to assess the impacts with very small sample size (7 years before and 1 year after). In addition, arbitration in accident data coding and the lack of accurate exposure data may prevent the detection of relatively subtle changes in the rare events like accidents.
4.3 Conclusions
- Inconclusive evidence of RWIDS’ impact on accident reduction. There was no statistically significant change in the statewide winter weather-related accident rate during the one year period (2003-2004) after the completion of RWIDS. The post-deployment winter weather-related accident rate decreased slightly from the previous year. However, the decrease was found to be within the statistical variance using accident data since 1996. Due to the coding problem, the accident data did not support the analysis by smaller geographical areas such as the ITD maintenance district or foreman area.
- Exogenous factors. In this evaluation, the hypothesized safety benefits are expected from two sources. First, the “better” winter maintenance practice of ITD resulting from the use of RWIDS is expected to provide safer roadways. Second, upon consulting with the RWIDS information, the general public would be more informed of the road-weather conditions, thus reducing the likelihood of accident occurrence. However, there are exogenous factors that could have direct and more significant impacts on the safety outcome. For example, there has been much change in winter maintenance practice during the past years as a result of ITD’s aggressive promotion of freeze-point depressant chemical (magnesium chloride). Arguably, the change of road maintenance doctrine might have more direct impacts on the road safety than the improved road-weather information brought about by the RWIDS (though good information is essential in freeze-point depressant applications). Because the adoption of freeze-point depressant took place incrementally around the state and over the study period, it created additional complications in measuring the safety improvement. Similarly, it is difficult if not impossible to isolate the safety benefits resulting from better provision of road-weather information to the general public.
- Recommendations for future safety analyses. A number of fundamental problems must be addressed in the Idaho statewide accident reporting system to allow geographic-based accident analysis. The current accident report uses route number and mile post for location reference in rural areas. However, ITD’s mile post and route numbering systems are grossly outdated, inconsistent, and confusing due to the changes of the rural freeway system in the past decades and the failure of ITD to revise the definition system accordingly. A revamp of the route and mile post definitions is needed to support any geographic-based accident data analysis using Geographic Information System (GIS) tools. The same problem was encountered by the ITD in their recent effort of assessing the safety benefits of the use of freeze-point depressant chemical.
1 Monthly precipitation data for state of Idaho from National Climatic Data Center (NCDC): http://lwf.ncdc.noaa.gov/oa/climate/research/cag3/id.html