This chapter presents the results of the analysis, a discussion of the results, and potential solutions that could mitigate the consequences of pole crashes. The primary focus is passively safe poles, their characteristics, the pole materials, and possible applications.
3.1. Discussion of Analysis Results and Limitations
A total of 38 pole-impact crashes were identified for this analysis through a combination of media reports and police database searches, covering the period from 2016 to 2025. All crashes in this analysis involved conventional roadside poles made of metal, wood, or concrete, none designed as passive-safe structures.
Most crashes (94.7%) involve a single vehicle, with passenger cars being the predominant vehicle type (84.2%). Powered two-wheelers (PTWs) and cargo vehicles were involved in 7.9% of crashes each. Most crashes occurred during nighttime (62.2%), while 37.8% occurred during the day. A substantial proportion (89.5%) occurred within settlement areas. Regarding road geometry, 50.0% of crashes occurred in curves, 30.6% on straight road sections, and 19.4% at intersections.
Weekday crashes accounted for 60.5% of the sample, while 39.5% occurred during weekends. Most crashes (78.9%) were registered in speed zones limited to 50 km/h. Crashes in areas with speed limits of up to 40 km/h accounted for 7.9%, and those in zones exceeding 50 km/h made up 13.2%.
As for crash outcomes, 65.8% of cases resulted in injury or fatal consequences, while 34.2% involved only material damage.
Table 1 shows an overview of potential variables, with the binary crash consequence variable as a dependent one. For analysis, the outcome variable was recoded into a binary format, distinguishing between crashes that resulted in injury or fatal outcomes and those involving material damage only. This binary classification was applied due to the limited sample size and the distribution of cases across the three original outcome categories, which posed challenges for reliable modeling using ordinal regression or decision trees with multiple branches. Similarly, the speed limit variable was grouped into ≤40 km/h, 50 km/h, and >50 km/h. This grouping was based on both the distribution of data and practical road classification standards while ensuring a sufficient number of cases per group to allow for valid statistical comparisons.
Crosstabulation analysis explored potential associations between crash consequences and various contextual factors. This method is suitable for examining relationships between categorical variables. It is particularly appropriate in studies with small sample sizes, where more complex modeling approaches may be limited by statistical power [
26].
With crosstabulation, we investigated whether the distribution of crash outcomes (material damage vs. injury/fatal) differed across categories such as vehicle type, visibility conditions, settlement type, road geometry, weekend occurrence, and grouped speed limits. Chi-square tests were used to assess the statistical significance of observed associations, and measures of effect size (Phi and Cramer’s V) were included to evaluate the strength of relationships.
The results of the crosstabulation analyses are summarized in
Table 2. Among the examined variables, only speed limit showed a statistically significant association with crash consequences in pole-impact crashes (
χ2(2) = 8.98,
p = 0.015). All other variables, including visibility conditions, road section type, location within a settlement, vehicle type, weekend occurrence, and crash circumstances, were not significantly related to the severity of pole crashes.
This result suggests that the speed limit may be an important contextual factor in whether pole crashes result in injury or fatality; however, caution is warranted given the sample size and other limitations. In contrast, the lack of statistically significant associations for the remaining variables may reflect the limited sample size and potentially weaker or indirect influence of these factors on injury outcomes in pole collisions. Additionally, the unbalanced distribution of cases across categories likely reduced the ability to detect more subtle effects.
Given these findings, speed limit was retained as a potential predictor variable in further analyses, particularly in the classification tree modeling (CHAID), which explored whether the severity of pole-impact crashes could be predicted based on posted speed limits.
As shown in
Table 3, the results indicate a differentiation in crash severity based on posted speed limits. Crashes occurring in areas with a speed limit of ≤40 km/h resulted exclusively in material damage, while those in zones above 50 km/h were consistently associated with injuries or fatalities. Crashes at 50 km/h showed a mixed pattern, suggesting a transitional threshold. This pattern highlights the importance of speed as a determinant of injury risk in pole-impact crashes. Also, this differentiation underscores the relevance of posted speed limits in crash severity [
17] and aligns with existing findings that link higher speed zones with increased risk in fixed-object impacts [
27].
To further explore the relationship between speed limits and crash outcomes in pole collisions, a CHAID (Chi-squared Automatic Interaction Detection) decision tree analysis was conducted. The binary variable representing crash consequence (material damage vs. injury or fatal outcome) was used as the dependent variable, while the categorized speed limit variable served as the sole predictor. This method was selected for its suitability in identifying interactions between categorical variables and its ability to produce interpretable, rule-based models that segment the dataset based on statistically significant differences.
The CHAID tree was grown using default parameters with minor adjustments to match the data constraints. The maximum tree depth was set to 3, while the minimum number of cases in parent and child nodes was 5 and 2, respectively. The Pearson chi-square statistic was used for evaluating splits, and Bonferroni-adjusted p-values (α = 0.05) were applied to reduce the likelihood of Type I error due to multiple testing. No validation sample was specified due to the small sample size, and all available cases were used in model building.
The resulting classification tree, shown in
Figure 2, included only one predictor, producing a significant split (adjusted
p = 0.037, χ
2 = 6.264, df = 1). The first node (Node 0) contained all 38 cases, with 34.2% involving only material damage and 65.8% resulting in injury or fatal outcomes. The tree was then split into two terminal nodes based on the speed limit group. Crashes that occurred at or below 40 km/h (Node 2) were exclusively material damage crashes (100%), while crashes at 50 km/h or higher (Node 1) were associated with more severe outcomes—only 28.6% resulted in material damage, and the remaining 71.4% led to injury or death. These results provide evidence for prioritizing the application of passive safe infrastructure on roads where the speed limit exceeds 40 km/h.
The model achieved an overall classification accuracy of 73.7%, correctly classifying all injury/fatal cases but only 23.1% of material damage cases. The risk estimate, which reflects the proportion of misclassified cases, was 0.263 with a standard error of 0.071. These results are consistent with the crosstab analysis and indicate a potential association between speed limit zones and crash severity in pole collisions. However, the findings should be interpreted with caution due to limited data.
The use of CHAID decision tree analysis in this study offered several advantages. First and foremost, the method provides an intuitive and visually interpretable model structure, which is particularly useful when communicating findings to scientific and professional audiences. CHAID is designed to handle categorical predictors and outcomes effectively, making it an appropriate choice for this dataset, which involved grouped speed limits and a binary outcome. It also allows for automatic detection of the most statistically relevant splits without requiring assumptions about linearity or normality, which is especially beneficial when working with relatively small and non-parametric samples. Another strength of the CHAID method lies in its capacity to reveal interaction effects and segment the data into homogeneous subgroups. In this analysis, CHAID successfully identified a speed threshold (≤40 km/h vs. >40 km/h) that separated low-severity from high-severity pole crashes. While the CHAID model identified a potential speed threshold associated with differences in crash severity, this finding requires confirmation in larger, more representative datasets.
However, several limitations must be acknowledged. The most important is the small sample size (N = 38), which restricts the statistical power of the model and increases the risk of overfitting. Although CHAID identified a significant split, the reliability and generalizability of the findings may be limited. Additionally, due to the small number of cases, only a single predictor (speed limit) was included in the final tree, preventing exploration of potentially meaningful interactions with other variables such as vehicle type, visibility, or road section. Also, there is the potential selection bias introduced by relying on media reports to identify pole crash cases. Severe crashes are more likely to be reported in the media, whereas minor incidents involving material damage may be underrepresented. While this may skew the sample toward higher-severity cases, it is important to note that collisions with roadside poles often result in vehicle immobilization and significant public visibility, increasing the likelihood of media coverage, regardless of the driver’s/passenger’s consequences.
Another limitation is the imbalance in outcome distribution. While CHAID correctly classified all injury/fatal cases, the model struggled to accurately classify material damage cases, leading to low sensitivity in that category. On the other hand, enhancing road safety predominantly focuses on eliminating severe injuries and fatalities, which confirms the usefulness of this model.
Several categorical variables in the dataset include fewer than five observations per group, weakening the robustness of statistical associations and contributing to unstable CHAID splits. Given the limited sample size, the CHAID analysis was applied primarily as an exploratory tool to identify preliminary patterns rather than construct a robust predictive model. Consequently, the findings should be interpreted as hypothesis-generating and indicative of potential trends, rather than as definitive conclusions. While exploratory, the results indicate that CHAID may be useful in future research with larger datasets for generating hypotheses about relevant crash risk factors and supporting data-driven safety assessments.
The fact that speed stands out as an influential factor shows that measures to increase safety should first of all be applied on roads with higher speed limits or conditions on the road such that vehicles move fast. It is certainly not always possible to decrease risk by reducing speed limits, especially on roads outside settlements or rural areas where it is more difficult to “control” drivers.
For future research, in addition to a larger sample size, it may be beneficial to incorporate crash impact speed or actual traffic flow speeds rather than relying solely on posted speed limits. By measuring the speed of the traffic flow at specific crash-prone locations, insights into the real needs could be obtained, rather than relying only on the posted speed limit. This can also be used for proactive action, rather than just acting retroactively. This would allow a more accurate assessment of the relationship between speed and crash severity in pole-impact crashes
The limited sample size in this study restricts the generalizability of results, as already stated. Future research should expand data sources by integrating IoT-based traffic monitoring, crowdsourced incident platforms, or detailed police reconstruction data. Additionally, spatial analysis techniques such as kernel density estimation and spatial autocorrelation could offer valuable insights into the clustering of pole crashes and regional crash risk patterns.
Although the sample size limits the generalizability of the results, this study provides a methodological framework for analyzing pole-impact crashes by identifying existing data limitations, proposing improvements in crash data collection practices, and outlining an exploratory approach for detecting influential factors. This framework lays the groundwork for future studies on pole-impact crashes, particularly those based on larger, more systematic datasets.
3.2. Passively Safe Road Infrastructure
Passively safe road infrastructure is designed to minimize injury severity in a collision [
28]. Unlike conventional roadside elements, passively safe structures, such as breakaway or energy-absorbing poles, are designed to deform, yield, or detach upon impact, thereby reducing the forces transferred to vehicle occupants. These features are critical in improving outcomes in single-vehicle crashes, particularly those involving fixed objects like poles or signposts. Increasing the use of such infrastructure is considered an effective strategy for enhancing road safety, especially in areas with a high risk of run-off-road incidents. In this section, a review of passive safe infrastructure based on EN 12767 is made in order to detect application potentials and provide recommendations.
In the European Union, support structures such as light poles must comply with EN 40 and be tested according to EN 12767:2019—Passive Safety of Support Structures for Road Equipment—Requirements and Test Methods [
29]. The EN 12767:2019 standard sets out criteria for passive safety and defines performance requirements for structures like lighting poles, road signs, traffic lights, and other roadside components. The goal is to reduce the severity of injuries in the event of a crash involving these elements. According to the CEN/CENELEC Regulations, the CEN members are bound to implement EN 12767:2019 as a national standard by June 2025 (all EU members). However, that does not imply that it is obligatory to apply the national standard in practice.
Compared to the previous version (EN 12767:2007) [
30], which relied on vehicle impact speed, energy absorption, and occupant safety level, the 2019 revision introduced additional parameters such as the backfill type of the column’s foundation, collapse mode, direction class, and risk of roof indentation. It also replaced the older three-part designation (e.g., 100-NE-3) with a more detailed seven-part classification format (e.g., 100-NE-B-R-SE-MD-0). Structures not tested or not meeting the criteria are classed as 0.
Table 4 presents the meaning of the designation codes used under EN 12767:2019 standard.
The speed class indicates the speed used during full-scale crash tests and can be 50, 70, or 100 km/h. All poles must also be tested at a 35 km/h low-speed impact. The energy absorption category is divided into high energy absorbing (HE), low energy absorbing (LE), and non-energy absorbing (NE) poles, depending on how much energy is dissipated during a crash. The classification is based on the difference between the vehicle’s impact and exit speeds.
The occupant safety level, marked with letters from A to E, reflects the level of safety for vehicle occupants based on the Acceleration Severity Index (ASI) and Theoretical Head Impact Velocity (THIV). ASI measures deceleration forces affecting passengers, while THIV estimates the velocity at which a hypothetical occupant would strike the vehicle interior. Lower ASI and THIV values indicate higher safety.
The backfill type of the foundation influences the pole’s behavior during impact and includes three classes: standard soil (S), rigid (R), and non-standard (X). The collapse mode describes whether the pole separates from the foundation during impact: no separation (NS) or separation (SE). Direction class defines how pole performance changes with vehicle impact angle, categorized as single-directional (SD), bi-directional (BD), or multi-directional (MD).
Finally, the risk of roof indentation distinguishes between Class 0 (indentation < 102 mm) and Class 1 (indentation ≥ 102 mm), indicating whether the pole could pose a threat to occupants by intruding into the passenger compartment.
Non-energy absorbing (NE) poles remain rigid during impact and typically shear near the base, separating from the foundation and falling behind the vehicle (
Figure 3a). Their primary advantage is allowing the vehicle to move post-impact, reducing damage and passenger injury. However, this behavior risks secondary crashes due to the falling pole or the vehicle’s continued motion. NE poles are thus recommended for high-speed rural roads with minimal roadside hazards or vulnerable road users. However, they are unsuitable for urban areas where secondary collisions pose a greater risk [
31]. High energy absorbing (HE) poles plastically deform upon impact, bending around the front and underside of the vehicle while remaining anchored to the foundation (
Figure 3b). This progressive deformation significantly reduces vehicle speed, lowering the risk of secondary collisions with roadside objects or other road users. While HE poles may result in higher forces transmitted to vehicle occupants, especially at higher speeds, these injuries are generally less severe than those resulting from crashes with conventional rigid poles. HE poles best suit low-speed areas with high pedestrian activity or complex roadside environments. Low energy absorbing (LE) poles exhibit characteristics of both NE and HE poles. They deform around the vehicle similarly to HE poles but detach from the foundation like NE poles (
Figure 3c). This allows for moderate deceleration and reduced vehicle damage. LE poles are considered appropriate for a wide range of road environments, including urban roads and areas with moderate speed limits, where a balance between vehicle occupant protection and risk to other road users is needed.
In addition to primary risks for vehicle occupants, crashes involving lighting poles can pose secondary hazards with potentially severe consequences for other road users. This may occur if the pole or its fragments fall onto the roadway after impact, creating unexpected obstacles and increasing the likelihood of further collisions. The danger is especially relevant at night when visibility is reduced, and debris may be less noticeable. However, the existing literature does not report cases where fragments from passively safe poles have caused problems in countries with widespread implementation of such infrastructure [
32]. Another concern arises with non-energy absorbing (NE) poles, where the vehicle may continue moving after impact, increasing the risk of secondary collisions with other road users or roadside objects.
To mitigate these risks, selecting passively safe poles should consider the collapse mechanism of the pole, the safety of both vehicle occupants and other road users, local speed limits, the presence of roadside infrastructure, and potential vehicle damage. Properly implementing passively safe poles can significantly reduce crash severity, particularly in locations with a high likelihood of secondary impacts.
Their use is recommended on all roads outside urban areas where the posted speed exceeds 50 km/h, and road barriers do not protect poles. Additionally, they are suitable for urban roads and all roundabouts where the speed limit exceeds 30 km/h, especially in cases where lateral skidding may lead to pole impact or poles are placed within the working width of roadside barriers [
33].
Where passively safe poles are not feasible, critical locations (especially near pedestrian crossings, cycling paths, and densely trafficked urban areas) should be protected by adequately installed safety barriers.
Passively safe poles are typically made from steel, aluminum, or composite materials. The choice depends on road type, local conditions, costs, aesthetics, and safety requirements. Steel is the most commonly used material due to its affordability, durability, fatigue resistance, and recyclability [
34]. Anti-corrosion protection is crucial, as structural weakening due to corrosion could compromise crash performance. An innovative alternative is Magnelis
® steel, a hot-dip galvanized product with 3.5% aluminum and 3% magnesium, which with zinc offers a superior corrosion resistance and self-healing properties [
35]. It provides a service life of up to 25 years, making it a cost-effective substitute for more expensive materials like stainless steel or aluminum.
Aluminum poles are about one-third lighter than steel, reducing transport and installation costs [
36]. Their service life is approximately 50 years, and they are naturally corrosion-resistant due to a protective aluminum oxide layer. This makes them particularly suitable for coastal regions [
37]. Aluminum is also fully recyclable, adding to its sustainability.
Composite poles, while the most expensive option, offer significant advantages, including high tensile strength, low weight, fire resistance, and excellent electrical insulation. They do not absorb water or conduct salt, making them highly resistant to corrosion and ideal for extreme environments (e.g., hurricanes and wildfires) [
38]. They require no maintenance or hazardous coatings and have an expected service life of up to 80 years [
39].
The material has a direct impact on energy absorption and crash performance. A study involving vehicle crashes at 37 km/h into concrete, steel, and aluminum poles showed that concrete poles produced the highest G-forces, followed by steel [
3]. In contrast, aluminum poles exhibited lower, more uniform G-force values, which favor passenger safety. Vehicles impacting steel poles decelerated rapidly, causing high occupant loads and substantial vehicle damage. In contrast, aluminum poles absorbed energy more gradually, reducing injury severity and vehicle damage. Although not included in the G-force comparison, composite poles are expected to outperform all other materials in energy absorption. Due to their lightweight nature, they often remain upright after a crash, reducing the risk of secondary collisions. Their non-conductive properties also enhance passenger safety in case of electrical failure. While initial costs are higher, composites offer long-term savings through durability and enhanced safety [
38].
3.3. Detection of Risk Zones and Proposal of Passively Safe Columns
Risk zone analysis identifies and assesses potentially hazardous roadside locations where poles are exposed to a high risk of vehicle impacts. This type of analysis supports the proposal of countermeasures to improve road safety, particularly through implementing passively safe lighting poles. Furthermore, it provides a foundation for future research and infrastructure improvements.
Although the current study was limited by sample size and spatial data availability, future research could benefit from incorporating spatial analysis techniques such as Geographically Weighted Regression (GWR) [
40], which can help quantify localized crash risk patterns and improve the identification of high-risk zones. Although full integration with smart infrastructure and detailed spatial risk models remains a long-term goal, such interdisciplinary approaches represent a promising direction for increasing the real-world impact of passive safe infrastructure.
In this study, the identification of risk zones was primarily based on the posted speed limit, which emerged as the only statistically significant factor in the analysis. Other infrastructure characteristics, such as curve radius, pole offset from the roadway, and absence of protective barriers, were considered qualitatively to support contextual interpretation but were not included in the quantitative classification due to data limitations. The mapped locations reflect actual crash sites and serve as preliminary risk indicators. These zones should be considered exploratory and subject to future refinement through analyses based on larger, more comprehensive datasets.
“Pole risk” refers to the risk associated with a particular pole location (based on its environment), while “zone risk” reflects the cumulative risk of a roadway section containing one or more such poles. Based on the available data, risk zones were categorized as low, medium, high, and very high risk. The classification aspired to prioritize intervention and propose locations where passively safe poles or other measures would be most effective.
Figure 4 and
Figure 5 present representative examples of defined pole crash risk zones.
The results obtained from this study’s analysis served as a foundation for proposing appropriate solutions following EN 12767:2019. The selection process considered descriptive crash patterns, classification results, and speed thresholds defined in the standard. The application of passively safe poles is illustrated through two representative locations described below.
For both locations, the recommended solution is the use of energy-absorbing poles. NE poles are not advised at such sites, as they are most appropriately deployed along straight road sections with clear zones exceeding 40 m and where there are no other road users or secondary hazards.
For the location shown in
Figure 4, which is situated outside a settlement and includes sections adjacent to forested areas, the recommended solution is the application of LE poles. Given the limited presence of pedestrians in this area, LE poles are considered more suitable than HE poles, as they offer greater protection for vehicle occupants. Although the posted speed limit is 50 km/h, it is advisable to use poles classified for at least 70 km/h due to the likelihood that pole impacts at such sites result from speeding.
Regarding occupant safety level, the highest feasible classes (B or C) are recommended, since class A is achievable only with deformable bollards. As for the backfill type, no specific requirements are foreseen, and types R, S, or X are acceptable. Regarding collapse mode, LE poles are designed to separate from their foundations (SE). MD poles are recommended for the direction class, as they ensure consistent behavior regardless of the direction of impact. Concerning the risk of roof indentation, class 0 is advised, indicating that any indentation is expected to be less than 102 mm.
In conclusion, recommended pole classifications for this site include: 70-LE-B-NR-SE-MD-0, 70-LE-C-NR-SE-MD-0, or 100-LE-B-NR-SE-MD-0.
At the location shown in
Figure 5, situated within a roundabout on a road with two traffic lanes per direction, adjacent to a shopping center (where traffic volumes are high during certain times of the day and week) the recommended solution is the application of passively safe poles with high energy absorption capacity (HE), for the following reasons. Unlike NE and LE poles, HE poles remain fixed to their foundations during a vehicle impact, reducing the likelihood of secondary crashes caused by pole detachment. Compared to LE poles, HE poles bring the vehicle to a stop more rapidly, which is particularly beneficial in minimizing the risk of secondary collisions with other vehicles in such traffic-congested environments. For this location, the recommended classifications include 70-HE-B-R-NS-MD-0 (for the roundabout zone) and 100-HE-B-R-NS-MD-0 (for the approaching zones). Given that the speed limit on the roads approaching the roundabout is 60 km/h, the appropriate minimum speed class for poles within the roundabout is 70 km/h. However, a higher classification (100 km/h) is more appropriate for poles along the approach roads. Regarding the occupant safety level, class B or C is recommended. Since the existing poles are embedded in concrete, the appropriate backfill classification is rigid (R). As HE poles do not detach from their foundations upon impact, the collapse mode is classified as NS. Given the nature of roundabouts, where collisions may occur from multiple angles, MD poles are strongly recommended to ensure consistent performance regardless of the impact direction. Finally, to minimize the risk of roof intrusion during a collision, the safer roof indentation class 0 is advised, indicating that any indentation is expected to be less than 102 mm.
Further analysis of pole crashes is essential to understand their specific characteristics and contributing factors better, enabling more precise identification of high-risk zones and optimizing safety improvements.
Future studies could benefit from cross-country comparisons between regions that have widely implemented EN 12767-compliant infrastructure and those that have not, to evaluate the standard’s real-world effectiveness in reducing crash severity and roadside fatalities.