Roadside Encroachment Database and Analysis (2025)
Chapter: 6. Findings and Recommendations
6. Findings and Recommendations
6.1 Findings
6.1.1 Review of Existing Encroachment Studies and Methods
The primary methods used to characterize roadside encroachments have been field collection (i.e., regular observation of roadside soil for physical encroachment evidence) and in-depth crash cases. The Hutchinson and Kennedy (1966) and Cooper (1980, 1981) field collection encroachment studies conducted in the 1960s and 1970s continue to serve as the basis for the guidelines in the RDG (AASHTO 2011) and the basis for the base encroachment frequencies and many of the encroachment frequency adjustments in the latest version of RSAP (Ray et al. 2012). Also, the ROR crash vehicle trajectories present in the NCHRP 17-22 in-depth dataset (Mak et al. 2010) serve as the basis for modeling encroaching vehicles in RSAP. The third primary encroachment data collection method, the ROR state crash data approach, has been less widely used, but a similar approach has been used to generate some of the base encroachment modifiers present in RSAP.
Based on a review of the available existing encroachment data, the following is a list of observations regarding existing encroachment data limitations:
- The most recent data used to determine encroachment rate are now 20 years old, and the vast majority of the available encroachment data are now in excess of 30 years old.
- Although traffic volume varied from approximately 1,000 to 161,000 vehicles per day, the vast majority of the collected encroachment data to date, approximately two thirds, were on roadways with traffic volumes of 40,000 vehicles per day or less.
- Very little is known about encroachment characteristics of other vehicle types (i.e., large trucks, buses, and motorcycles). Although the Hutchinson and Kennedy (1966) and Cooper (1980, 1981) studies note differentiating between passenger vehicle and heavy vehicle encroachments, little raw data were available for examination. No motorcycle- or bus-specific encroachment data were found.
- The vast majority of the existing encroachment data were collected on higher speed roadways with posted speed limits of 50 mph or higher.
- Almost all of the existing data collection methods to date exclude encroachments that do not extend beyond a paved shoulder area.
While field collection of encroachment data has continually been viewed as the standard method of encroachment data collection, this method has serious shortcomings, primarily due to the time/cost required to collect a sufficient number of encroachments across a variety of roadway geometric and operational characteristics and the exclusion of encroachments contained within a paved shoulder. The Hutchinson and Kennedy and Cooper field collection encroachment studies have less than 2,500 encroachment events combined.
6.1.2 Potential Encroachment Data Collection Methods and Selected Collection Method
The research team explored potential new methods to collect encroachment data based on current and near-future technologies, which revealed at least seven potential methods, including use of stationary cameras or other imagery, infrastructure-based instrumentation methods, or use of
vehicle-based methods. Of these seven methods, only three were found to be feasible at present for the current project: NDS data, EDR data, and Google Street View methods. NDS data were chosen for this project due to the ability to capture encroachments without a collision on many roadway types.
Given the project funds available, the research team’s approach was to collect encroachment events from various existing data sources either previously used for encroachment studies (e.g., in-depth data and state crash/maintenance data sources) or not yet used for encroachment studies (i.e., NDS data) to provide updated information on all three types of encroachment events:
- Encroachments without a collision event
- Encroachments with an unreported collision event
- Encroachments with a police-reported collision event
Data from nine different sources, as indicated in Table 64, were used to collect information on the three different encroachment types.
Table 64. Data Sources for the NCHRP 17-88 Encroachment Database
| Database | Data Collection Period | Primary Encroachment Type(s) |
|---|---|---|
| SHRP 2 NDS | 2010-2013 | A) Encroachment without collision |
| SHRP 2 RID | 2010-2013 | |
| NCHRP 17-43 Road Departure Database | 2011-2015 | C) Encroachment with police-reported collision |
| NCHRP 22-26 Motorcycle Departure Database | 2010-2016 | |
| MCCS | 2011-2016 | |
| LTCCS | 2001-2003 | |
| Iowa State Datasets | 2012 – 2017 | B) Encroachment with unreported collision and C) Encroachment with police-reported collision |
| Washington State Datasets | 2012-2018 | |
| Tennessee Datasets | 2015-2019 |
6.1.3 NCHRP 17-88 Encroachment Database and Comparison to Existing Encroachment Data
The research team assembled a dataset of more than 100,000 encroachment events spanning each of the three types of encroachment events. Further, these encroachments involved various vehicle types, including passenger vehicles, large trucks, and motorcycles. The in-depth data sources provide a means of capturing detailed trajectory information for each vehicle type. For the state data, specific routes were selected prior to filtering for encroachment-related crashes to provide a range of roadway types and roadway characteristics, more closely mimicking the traditional field collection approach for encroachment data. A high-level summary of the NCRHP 17-88 encroachment database compared to previously available encroachment datasets is provided in Table 65.
Table 65. Summary of NCHRP 17-88 Database Relative to Previously Available Encroachment Datasets
| Limitation Category | Existing Encroachment Data | 17-88 Encroachment Database |
|---|---|---|
| Age of data | 20 to 30+ years old | Majority of data available within 15 years of study completion |
| Traffic volume coverage | Majority up to 40,000 vehicles per day | Segments up to 250,000 vehicles per day |
| Vehicle type coverage | Primarily passenger vehicles | Includes large truck and motorcycle trajectories |
| Encroachment lateral extent | Omits encroachments contained within paved shoulders | SHRP 2 data include encroachments contained within paved shoulders |
| Posted speed limit | 50+ mph | Includes some lower speed roadway segments |
6.1.4 Encroachment Frequency and Influence of Roadway Characteristics
Based on an examination of the available Washington encroachment-related crash data (i.e., police-reported and unreported crashes with select roadside hardware), the following observations were made with respect to the influence of roadway characteristics on encroachment-related crashes:
- Larger variations in encroachment-related crash rates were found at lower segment traffic volumes, and this variation appears to be larger for rural roadway segments compared to urban roadway segments.
- VMT, horizontal curve presence, and roadway type (e.g., divided) appear to have the largest influence on the frequency of encroachment-related crashes on a specific roadway segment.
- Higher posted speed limits, narrower shoulder widths, narrower median widths, narrower lane widths, higher levels of service values (i.e., closer to LOS A), and rural area type were found to increase encroachment-related crash frequency.
- While controlling for other factors, a one-unit increase in the natural log of VMT of a segment increases encroachment-related crash frequency by 93%.
- Compared to tangent roadway segments, presence of a medium radius horizontal curve (500' < R ≤ 2,800') or small radius horizontal curve (R ≤ 500') increased encroachment-related crash frequency by 34% and 264%, respectively.
6.1.5 Encroachment Characteristics
The encroachments extracted from SHRP 2 were compared to normal, non-encroachment driving on the same roadways. EMFs were computed based on the road and roadside features on these roadways. The following observations were made based on a comparison of SHRP 2 EMFs to those used in RSAPv3:
- The SHRP 2 encroachments occurred at a much higher rate than those captured in the Cooper study. This is primarily due to small encroachments that crossed the road edge line but did not leave the paved surface. These encroachments occur often and were unable to be captured in previous encroachment data collection efforts.
- For roadways with characteristics similar to those included in the Cooper study, the EMFs determined in SHRP 2 were similar to those used in RSAPv3.
- For roadways with lower speed limits, the RSAPv3 does not have applicable EMFs due to the sampling criteria in the Cooper study. The EMFs based on SHRP 2 encroachments on lower speed roads were much higher than those on higher speed roads.
6.1.6 Vehicle Types and Encroachments
Using the in-depth datasets available in NCHRP 17-88, the impact conditions of motorcycles, passenger vehicles, and large trucks were compared. Observations from the comparison are as follows:
- Passenger cars and pickup trucks had 85th percentile impact angles and impact speeds similar to those prescribed in MASH.
- The large truck impacts had an 85th percentile impact angle that was higher than those prescribed in MASH. This is likely much larger than the practical worst-case scenario for all large truck impacts because the source LTCCS dataset only contained crashes that resulted in serious injury.
- Motorcycles had similar impact conditions to passenger vehicles. Motorcycles were typically upright during the first impact event and had an 85th percentile impact angle of 24 degrees.
6.1.7 Unreported Crashes
Matched police-report and roadside hardware maintenance records for w-beam guardrail, guardrail end terminals, impact attenuators, and cable barrier were available from Washington State. Combining these encroachment records with the associated police-reported crashes along the same Washington routes provided a lower bound estimate on unreported crash rates. Generally, the unreported crash rates were between 4% and 15% of the police-reported crash rates, depending on the specific roadway type. In terms of unreported crash rates as a function of roadside hardware device, the unreported crash frequency was between 30% and 70% of the total police-reported impacts with the same device. In general, end terminals and cable barriers had a higher frequency of unreported impacts compared to w-beam guardrail and crash cushions.
6.2 Recommendations
6.2.1 Future Data Collection Efforts
Data Element Availability Challenges
Many of the roadside data elements of interest to study encroachments were not readily available. While there was some effort to collect these data elements on an ad hoc basis for the current study, future encroachment studies would greatly benefit from enhanced datasets available and maintained by the states. These data elements of interest include the following:
- Roadside Slope: None of the state agency partners or in-depth data sources had roadside slope data readily available. Roadside slope data are difficult to collect on an ad hoc basis. While there has been significant recent research examining roadside slope and ditch configuration using computer simulation, roadside slope data for in-service roadside segments would be useful to future encroachment studies. State agencies with LiDAR data may be able to leverage that data to develop a database of roadside slope. For the in-depth
- cases, newer data sources such as the CISS have begun to incorporate scene measurements using a total station, which could be mined to determine roadside slope. Sufficient data points off the roadway would, however, need to be captured.
- Device Lateral Offset: Many states now have GIS-based inventory data of roadside hardware devices such as longitudinal barriers and crash cushions, but these typically lack information on lateral offset from the roadway. Some lateral offset data for Washington State were collected as part of this study, but this ad hoc collection is time intensive. Consideration of inclusion of device offset in these state agency datasets would greatly enhance future encroachment studies.
- Access Density: While two of the participating states had data on intersecting roadways in separate datasets, neither had access density information that included driveways. None of the previous encroachment studies considered the influence of access density on encroachments.
- LOS: This study used available traffic and roadway data to estimate LOS for the roadway segments and examined encroachments relative to LOS level. Although based on the HCM procedures for LOS, several assumptions were made to compute LOS. Future studies would benefit from better LOS data for the included roadway segments.
One additional consideration relates to the coordination between maintenance and asset management within an agency. Using the same identifier in the maintenance records that is used in the inventory would greatly improve the ability to merge maintenance records with any roadside hardware inventory and allow better estimates of unreported crash rates.
Encroachment Data for Other Vehicle Types
The NCHRP 17-88 database currently represents the most comprehensive encroachment dataset, including both passenger vehicles, large trucks, and motorcycles. For motorcycles in particular, there are still relatively few cases with detailed trajectory data. The current project extracted motorcycle trajectory data for every existing roadside crash event with the required information and only had 68 cases. Similarly, there were only 124 large truck cases available with detailed trajectory information, and none specific to buses. Therefore, the research team recommends future studies focused on increasing data collection efforts for large trucks, buses, and motorcycles.
One promising aspect for a future retrospective data collection effort is the increased sampling criteria in the new CISS database collected by NHTSA. Due to the passing of the bipartisan infrastructure bill in 2021, NHTSA had additional funding to expand the CISS dataset to include pedestrians, motorcyclists, and large trucks over the coming years.
6.2.2 MASH Impact Conditions
Future updates to MASH will consider whether the impact conditions in the fleet have changed. Based on the impact conditions observed in the NCHRP 17-88 database, the MASH test impact conditions for passenger vehicles are near the 85th percentile. This represents the practical worst-case scenario desired in the MASH tests. For large trucks, the impact conditions specified in MASH are below the 85th percentile impact conditions observed in the NCHRP 17-88 database. However, this was likely due to the higher crash severity of the large trucks sampled in the source LTCCS dataset. Therefore, actual practical worst-case scenario for large trucks likely has a much lower impact angle. Based on the available data, the research team does not recommend any changes to MASH test conditions for passenger vehicles or large trucks.
6.2.3 Motorcycles
The NCHRP 17-88 database contains trajectory information for all of the publicly available U.S. in-depth motorcycle roadside crash data that exist. However, there are relatively few motorcycle roadside crashes in the dataset. The research team recommends that any future data collection efforts (recommendation 6.2.1) also include motorcycles in the data collection effort.
Additionally, the research team recommends future impact test conditions where the motorcyclist is in the upright position to better match the impact conditions observed in real-world crashes. With increased data and additional impact tests, a standard for motorcycle-barrier impact testing may be able to be addressed.
6.2.4 EMFs
Previous encroachment data collection efforts were unable to capture encroachments in which the vehicle did not leave the paved surface. The incorporation of SHRP 2 encroachments in the NCHRP 17-88 database allows EMFs to account for these small encroachments. In addition, these SHRP 2 encroachments occurred on many road types including low-speed, undivided roadways. The research team recommends that future research uses the NCHRP 17-88 database to compute EMFs for additional roadway types and that the EMFs include small encroachments. These studies could use a negative binomial regression technique, similar to Section 4.1.7, to account for many road and roadside features simultaneously.
