Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
41 Chapter 4. Database Development Two potential alternative approaches to crash prediction for roadside trees and utility poles were identified for potential development and application. These were: ⢠Empirical crash prediction models for tree- and utility-pole-related crashes developed with statistical modeling methods ⢠Existing crash prediction methods, adapted and calibrated to predict tree- and utility-pole- related crashes The initial step in developing these approaches was to develop a database for actual roads including roadway geometric and traffic volume data, number and placement of the roadside trees and utility poles, and tree- and utility-pole-related crash data. The development of such databases is described in this chapter. 4.1 Existing Databases and Supplementary Sites A review of existing data available from state DOTs found the most promising existing datasets relevant to the research objectives in Kentucky and Washington. These existing datasets were then obtained and supplemented with additional data to create project databases for rural state highways in Kentucky and Washington. An existing database on the roadside characteristics of selected rural two-lane highways in Kentucky that had been assembled by Souleyrette et al. (2017) using usRAP protocols was obtained and used as a guide for identifying roadways with trees and utility poles present on the roadside. The existing usRAP database focused on Kentucky numbered routes, but was supplemented with sites on both rural two-lane undivided and rural four-lane divided roadways on U.S. numbered routes in Kentucky, which were selected through review of aerial and street- level photographs in Google Earth®. The existing usRAP database identified the presence of trees and poles for 327-ft or 100-m roadway sections, but did not identify the number of trees or poles within a 327-ft roadway segment and did not distinguish between utility poles and other types of poles. This lacking information was obtained in subsequent reviews of aerial and street- level photographs (see Section 4.2). An existing database was obtained from the Washington State Department of Transportation (WSDOT) with an inventory of utility pole locations on selected state highways (primarily rural two-lane highways) in Washington. This database had been used to identify utility poles in need of removal or relocation from among all of the utility poles on these highways. Therefore, the sites identified from this database were studied only for the period before the utility pole inventory was conducted in 2012, since some of the utility poles in the database may have been removed or relocated after that date. No similar database on roadside trees was available for state highways in Washington, so roadways with trees were selected on rural two-lane undivided and rural four-lane divided nonfreeway state highways in Washington through review of aerial and street-level photographs in Google Earth®, much as the supplementary sites in Washington were selected.
42 While rural four-lane divided nonfreeway sites were identified in both Kentucky and Washington, only a limited mileage of such sites was identified, and those sites were found to have very few trees and utility poles and very few tree- and utility-pole-related crashes. For example, 108 centerline-mi of four-lane divided nonfreeways were reviewed in aerial and street- level photographs in Kentucky (about one-third the mileage as reviewed for rural two-lane undivided highways), but only 2.9 mi of road with tree groups, 16 individual trees, and 74 utility poles were found. Furthermore, the average offset from the traveled way to trees and utility poles on four-lane undivided highways was found to be over 30 ft. The mileage of useful four-lane divided nonfreeway sites in Washington was similarly limited. Therefore, the study focused on the rural two-lane undivided roadway sites. For both Kentucky and Washington, roadway, traffic volume, and crash history data were also obtained for the sites of interest from existing databases. For Kentucky sites, these data were obtained from the website of the Kentucky Transportation Cabinet. For Washington sites, these data were obtained from the FHWA HSIS. 4.2 Database Assembly The roadways selected as described above in Section 4.1, were reviewed in Google Earth® to obtain additional roadside information. Information extracted from existing state files for roadway characteristics, traffic volumes, and crash history data were obtained and added to the project databases. For sites in both Kentucky and Washington, a 327-ft roadway segment had to meet the following criteria to be included in the study: ⢠Contain trees or utility poles within 40 ft of the traveled way on the right or left side of the road (or both) that could be struck by a vehicle running off the road (i.e., not behind a guardrail or another object) ⢠Consist of one of the priority road types being studied (rural two-lane undivided highway or rural multilane undivided highway). Undivided roads must have a single or double centerline with no median and no auxiliary lanes. ⢠Does not include an at-grade intersection with a public road The study focused on trees and utility poles within 40 ft of the traveled way so that data would be available for objects both within and outside the typical clear zone widths used in roadside design. We did not want to focus exclusively on objects located within the clear zone, as we did not want to presume that current clear zone criteria were necessarily correct. The first step in assembling the databases for Kentucky and Washington roads was to review the available aerial and street-level photographs to identify specific locations with trees and/or utility poles present and to determine the specific locations and types of roadside trees and utility poles. These data were merged with data on roadway characteristics, traffic volumes, and crash history from existing sources. The specific data elements sought for the project database included: ⢠Location (route, county, milepost) ⢠Type of roadway (e.g., rural two-lane highway, rural multilane undivided highway)
43 ⢠Side of road (left or right in increasing milepost direction) ⢠Type of object (tree, tree group, utility pole) ⢠Lateral offset from edge of traveled way to object ⢠Longitudinal extent of tree groups (milepost range) ⢠Roadway AADT (veh/day) ⢠Posted speed limit ⢠Type of shoulder (paved/unpaved) ⢠Paved shoulder width ⢠Roadway alignment (tangent/curve) ⢠Direction of curve (left or right in increasing milepost direction) Two data elements that might have been useful for empirical modeling could not be obtained. These are the diameter of the roadside tree or utility pole and the roadside slope between the edge of the shoulder and the tree or utility pole. These data elements could not be obtained accurately from review of aerial or street-level photographs. For each selected road segment, data were recorded for individual trees, tree groups, and individual utility poles. These types of roadside objects were defined as follows: ⢠individual trees â individual, isolated trees in roadside areas not near an intersection and not behind a guardrail or other object. Only trees at least 4 inches in diameter have been considered. The minimum 4-inch diameter for trees was assessed visually in review of street-level photographs. ⢠tree groups â continuous lines of trees in roadside areas not near an intersection and not behind a guardrail or other object, in which the trees are closely spaced such that any vehicle running a sufficient distance off the road would, with certainty, strike a tree. Tree groups often represented the edge of a wooded or forested area. ⢠individual utility poles â individual, isolated utility poles (supporting power or telephone cables) in roadside areas not near an intersection and not behind a guardrail or other object The longitudinal location of a tree or pole, the longitudinal extent of a tree group, and the lateral offset distance from the tree, tree group, or utility pole to the edge of the traveled way were documented. In most cases, tree groups were located at a consistent offset from the traveled way but, if variations were present on a given side of the road, the offset to the tree closest to the traveled way was used. These dimensions were determined in aerial photographs using the measuring tool in Google Earth®. All data of interest on the preceding list were obtained from review of aerial and street-level photographs except roadway AADT, speed limit, and shoulder type and width which were obtained from existing databases. The data for trees and tree groups on both Kentucky and Washington roadways and the data for utility poles on Kentucky roadways were structured using 327-ft road segment lengths. The Kentucky data were already structured this way in the existing usRAP database and the Washington tree and tree group data were also structured this way for consistency. The Washington utility pole data were reorganized by roadway segments with a maximum length of
44 327 ft for consistency. Crash history data for each site for a recent five-year period were obtained from existing state databases. The crash study periods were 2015 to 2019, inclusive, for Kentucky sites and 2013 to 2017, inclusive, for Washington tree and tree group sites. The available crash records allowed identification of crashes involving collisions with trees or utility poles. As noted above, the research team obtained and utilized a database of utility pole locations from the Washington State DOT. The data on utility pole locations was gathered in 2012. The research team review of these data, in comparison to Google Earth® images for 2012, indicates that they are accurate. Utility pole locations in this database were given in mileposts, so they are linkable to crash data. Crash data for a five-year period (2008 to 2012) were used with these data. More recent crash data were not used because the Washington State DOT informed the research team that some of these utility poles have been relocated since 2012, but that there is no available list of specific project locations. 4.3 Extent of Available Data The following discussion reviews the extent of the database assembled for Kentucky and Washington sites. Kentucky Data on roadside tree groups, individual trees (not adjacent to other trees), and individual utility poles were obtained for 322 centerline-mi of rural two-lane undivided highways in Kentucky. Data were collected for the right (outside) roadside in each direction of travel for all sites, so a total of 644 mi of roadway edges were included. The Kentucky database consists of 327-ft roadway segments with roadside tree groups, with individual tress, and with utility poles. If a roadway segment was reviewed and found to have none of these three roadside object types present, it was omitted from the database. Roadway segments with both roadside tree groups and individual trees present were omitted from the database, because it would be difficult to separate their effects. Roadway Segments with Tree Groups Table 22 summarizes the data for roadway segments where tree groups within 40 ft of the traveled way were found and coded at the Kentucky sites. Tree groups were present on the roadside for 1,093 roadway segments with a total length of 67.8 centerline-mi. These roadway segments consist of a total length of 135.6 mi of roadway edges; in other words, each of the 67.8 centerline-mi of roadway segments has two roadside edges. Taken together, these sites include 52.6 mi of roadside edges with tree groups. Thus, the roadside edges with tree groups present constitute 39 percent of the total length of roadside edges for their study sites.
45 Table 22. Summary Data for Tree Groups at Kentucky Sites on Rural Two-Lane Undivided Highways Location type Length of roadway edge with tree groups (mi) Exposure (hundred million veh-mi of travel) Number of crashes in 5 years (2015 â 2019) Crash rate per hundred million veh-mi in 5 years (2015 â 2019) Average offset to tree group (ft) Primary Primary & Secondary Primary Primary & Secondary Total FI PDO Total FI PDO Total FI PDO Total FI PDO All 52.59 1.902 21 7 14 44 23 21 11.1 3.7 7.4 23.2 12.1 11.1 18.3 Tangent 32.02 1.441 15 4 11 26 11 15 10.4 2.8 7.6 18.0 7.6 10.4 19.5 Curve 20.57 0.461 6 3 3 18 12 6 13.0 6.5 6.5 39.0 26.0 13.0 16.5 Outside of Curve 6.47 0.162 2 2 0 6 5 3 12.3 12.3 0.0 37.1 30.9 6.2 17.8 Inside of Curve 5.41 0.122 1 0 1 8 5 8.2 0.0 8.2 65.6 41.0 24.6 17.0 NOTE: FI = fatal and injury; PDO = property damage only; primary tree crashes are collisions in which the first harmful event was leaving the roadway and striking a tree; secondary tree crashes are collisions that involve striking a tree after some other harmful event.
46 The roadway segment edges with tree groups present have a mean AADT volume of 1,980 veh/day. The total exposure or vehicle travel passing the tree groups is 1.902 hundred million veh-mi of travel during the five-year study period. Where tree groups are present at the same location on both sides of the road, the travel on that portion of the roadway is included twice, once for each roadway edge with tree groups. During the five-year study period (2015 to 2019), there were a total of 21 crashes involving a primary collision with a tree in a tree group (i.e., collisions in which the first harmful event was leaving the roadway and striking a tree) and an additional 23 crashes involving a secondary collision with a tree in a tree group (i.e., striking a tree after some other harmful event). Table 22 shows the severity distribution for these crash types and the computed values of crash rate per hundred million veh-mi of travel by crash severity level. Table 22 also shows that the average offset from the traveled way to roadside tree groups is 18.3 ft. The range of offsets of tree groups from the traveled way to tree groups 4 to 40 ft. Table 22 also summarizes the data for roadway segments with tree groups on tangents and horizontal curves, for roadway segments with tree groups on the outside of a horizontal curve, and for roadway segments with tree groups on the inside of a horizontal curve. The table shows that, overall, there are more primary crashes with trees in tree groups on horizontal curves than on tangents and more primary crashes with trees in tree groups on the outside of horizontal curves than on the inside. The same relationship holds for secondary crashes on horizontal curves vs. tangents, but there are more secondary crashes on to the inside of curves than the outside. The likelihood of a vehicle running off the road is presumed to be higher on a horizontal curve than on a tangent, but also trees on horizontal curves are, on the average, about 3 ft closer to the traveled way than trees on tangents. Roadway Segments with Individual Trees Table 23 summarizes data for roadway segments with one or more individual trees (but no tree groups) within 40 ft of the traveled way. Individual trees were found on 367 roadway segments with a total length of 22.8 mi and with 570 individual trees present, or an average of 1.55 trees per 327-ft roadway segment. Of these individual trees, 358 were on tangent roadway segments and 212 on horizontal curves; there were 87 trees on the outside of horizontal curves and 65 trees on the inside of horizontal curves. There were a total of 60 trees on curved roadway segments that had trees on both the inside and the outside of a horizontal curve. Table 23 also shows that the average offset from the traveled way to individual trees is 20.7 ft. The range of offsets from the traveled way to tree groups 3 to 40 ft.
47 Table 23. Summary Data for Individual Trees at Kentucky Sites on Rural Two-Lane Undivided Highways Location type Number of individual trees Exposure (hundred million vehicle passes) Number of crashes in 5 years (2015 â 2019) Crash rate per hundred million vehicle passes in 5 years (2015 â 2019) Average offset to trees (ft) Primary Primary & Secondary Primary Primary & Secondary Total FI PDO Total FI PDO Total FI PDO Total FI PDO All 570 25.175 5 4 1 14 11 3 0.20 0.16 0.04 0.56 0.44 0.12 20/7 Tangent 358 17.636 4 3 1 10 7 3 0.23 0.17 0.06 0.57 0.40 0.17 21.6 Curve 212 7.539 1 1 0 4 4 0 0.13 0.13 0.00 0.53 0.53 0.00 19.2 Outside of Curve 87 3.718 0 0 0 1 1 0 0.00 0.00 0.00 0.27 0.27 0.00 19.4 Inside of Curve 65 2.790 0 0 0 2 2 0 0.00 0.00 0.00 0.72 0.72 0.00 20.8 NOTE: FI = fatal and injury; PDO = property damage only; primary tree crashes are collisions in which the first harmful event was leaving the roadway and striking a tree; secondary tree crashes are collisions that involve striking a tree after some other harmful event.
48 Exposure and crash rates are shown in terms of hundred million vehicle passes over a five-year period. A vehicle pass is one vehicle passing one particular roadside object. So, if there are three individual trees on a roadway segment, the number of vehicle passes per day is three times the AADT. The roadway segments with individual trees, overall, experienced five tree-related primary crashes during the five-year study period, or about one-quarter as many crashes as for the roadway segments with tree groups. Roadway Segments with Individual Utility Poles Table 24 summarizes data for roadway segments with one or more individual utility poles within 40 ft of the traveled way. Individual utility poles were found on 1,208 roadway segments with a total length of 74.9 mi and with 1,760 individual utility poles present, or an average of 1.46 utility poles per 327-ft roadway segment. Of these individual utility poles, 1,265 were on tangent roadway segments and 495 on horizontal curves; there were 240 utility poles on the outside of horizontal curves and 153 utility poles on the inside of horizontal curves. There were a total of 102 utility poles on curved roadway segments that had utility poles on both the inside and the outside of a horizontal curve. Table 24 also shows that the average offset from the traveled way to individual utility poles is 19.1 ft. The range of offsets from the traveled way to tree groups is from 2 to 40 ft. The roadway segments with individual utility poles, overall, experienced seven pole-related primary crashes during the five-year study period. Washington Data on roadside tree groups and individual trees (not adjacent to other trees) were obtained for 289 centerline-mi of rural two-lane undivided highways in Washington using the same data collection procedures as were used for roadside objects on Kentucky roads. Data were collected for both right (outside) roadsides for all sites, so data for a total of 578 mi of roadway edges were obtained. Data on roadside utility poles were obtained for 308 centerline-mi of rural two-lane undivided highways in Washington from an existing WSDOT database. Thus, utility pole data were included for 616 miles of roadside edges on rural two-lane highways. WSDOT collected these data on utility poles for a portion of their state highway system to serve as the basis for managing a program of utility pole relocation to improve roadside safety. The utility pole data were accurate as of 2012. Utility pole relocations were made in subsequent years, so the data may no longer represent the roadside conditions as they exist today. Therefore, crash data were obtained from WSDOT for the five-year period from 2008 through 2012. Data on roadway characteristics and traffic volumes were obtained from WSDOT data for 2012.
49 Table 24. Summary Data for Individual Utility Poles at Kentucky Sites on Rural Two-Lane Undivided Highways Location type Number of individual poles Exposure (hundred million vehicle passes) Number of crashes in 5 years (2015 â 2019) Crash rate per hundred million vehicle passes in 5 years (2015 â 2019) Average offset to poles (ft) Primary Primary & Secondary Primary Primary & Secondary Total FI PDO Total FI PDO Total FI PDO Total FI PDO All 1760 84.026 7 0 7 13 3 10 0.08 0.00 0.08 0.16 0.04 0.12 19.1 Tangent 1265 65.642 6 0 6 11 2 9 0.09 0.00 0.09 0.17 0.03 0.14 19.8 Curve 495 18.384 1 0 1 2 1 1 0.05 0.00 0.05 0.10 0.05 0.05 17.0 Outside of Curve 240 8.454 1 0 1 1 1 0 0.12 0.00 0.12 0.12 0.12 0.12 16.8 Inside of Curve 153 5.699 0 0 0 0 0 0 0.00 0.00 0.00 0.00 0.00 0.00 17.0 NOTE: FI = fatal and injury; PDO = property damage only; primary utility-pole crashes are collisions in which the first harmful event was leaving the roadway and striking a utility pole; secondary utility pole crashes are collisions that involve striking a utility pole after some other harmful event.
50 Roadway Segments with Tree Groups and Individual Trees The Washington database for rural two-lane undivided highways consists of 327-ft roadway segments with roadside tree groups and individual trees. If a roadway segment has neither tree groups nor individual trees present, it was omitted from the database. Roadway segments with both roadside tree groups and individual trees present were also omitted from the database, because it would be difficult to separate their effects. Table 25 summarizes the data for roadway segments where tree groups within 40 ft of the traveled way were found and coded at the Washington sites. The roadway segments for this database included the roadside on only one side of the road; thus, if there were tree groups on both sides of the roadway, the two sides of the road were analyzed as separate roadway segments. Tree groups were present on the roadside for 1,570 roadway segments with a total length of 95.6 centerline-mi. These roadway segments consist of a total length of 191.2 mi of roadway edges; in other words, each of the 95.6 centerline-mi of roadway segments has two roadside edges. Taken together, these sites include 57.3 mi of roadside edges with tree groups. Thus, the roadside edges with tree groups present constitute 14 percent of the total length of roadside edges for their study sites. The roadway segment edges with tree groups present have a mean AADT volume of 3,400 veh/day. The total exposure or vehicle travel passing the tree groups is 3.560 hundred million veh-mi of travel during the five-year study period (2013 to 2017). Where tree groups are present at the same location on both sides of the road, the travel on that portion of the roadway is included twice, once for each roadway edge with tree groups. During the five-year study period, there were a total of 30 crashes involving a collision with a tree in a tree group (i.e., a collision in which the first harmful event was leaving the roadway and striking a tree). Table 25 shows the severity distribution for these crash types and the computed values of crash rate per hundred million veh-mi of travel by crash severity level. Table 25 also shows that the average offset from the traveled way to roadside tree groups is 24.5 ft. The range of offsets from the traveled way to tree groups is from 9 to 40 ft. Table 25 also summarizes the data for roadway segments with tree groups on tangents and horizontal curves, for roadway segments with tree groups on the outside of a horizontal curve, and for roadway segments with tree groups on the inside of a horizontal curve. The table shows that, overall, there are more crashes per hundred million veh-mi of travel with trees in tree groups on horizontal curves than on tangents and more crashes with trees in tree groups on the outside of horizontal curves than on the inside. The likelihood of a vehicle running off the road is presumed to be higher on a horizontal curve than on a tangent, but also trees on horizontal curves are, on the average, about 1.5 ft closer to the traveled way than trees on tangents. Table 26 summarizes data for roadway segments with one or more individual trees (but no tree groups) within 40 ft of the traveled way. Individual trees were found on 48 roadway segments with a total length of 2.9 mi and with 55 individual trees present, or an average of 1.15 trees per 327-ft roadway segment. Of these individual trees, 42 were on tangent roadway segments and 13 on horizontal curves; there were seven trees on the outside of horizontal curves and six trees on the inside of horizontal curves. Table 26 also shows that the average offset from the traveled way to individual trees is 26.5 ft. The range of offsets from the traveled way to tree groups is from 14 to 39 ft.
51 Table 25. Summary Data for Tree Groups at Washington Sites on Rural Two-Lane Undivided Highways Location type Length of roadway edge with tree groups (mi) Exposure (hundred million veh- mi of travel) Number of tree-related crashes in 5 years (2013-2017) Crash rate per hundred million veh-mi in 5 years (2013 â 2017) Average offset to tree group (ft) Total FI PDO Total FI PDO All 57.28 3.560 30 13 17 8.43 3.65 4.78 24.5 Tangent 37.71 2.317 16 7 9 6.91 3.02 3.89 25.0 Curve 19.57 1.243 14 6 8 11.26 4.83 6.43 23.5 Outside of Curve 9.39 0.614 11 4 7 17.92 6.52 11.40 23.5 Inside of Curve 10.18 0.629 3 2 1 4.77 3.18 1.59 23.5 NOTE: FI = fatal and injury. PDO = property damage only. Table 26. Summary Data for Individual Trees at Washington Sites on Rural Two-Lane Undivided Highways Location type Number of individual trees Exposure (hundred million vehicle passes) Number of tree-related crashes in 5 years (2013-2017) Crash rate per hundred million vehicle passes (2013 â 2017) Average offset to trees (ft) Total FI PDO Total FI PDO All 55 5.307 0 0 0 0.00 0.00 0.00 26.5 Tangent 42 4.299 0 0 0 0.00 0.00 0.00 26.9 Curve 13 1.008 0 0 0 0.00 0.00 0.00 25.3 Outside of Curve 7 0.416 0 0 0 0.00 0.00 0.00 27.4 Inside of Curve 6 0.592 0 0 0 0.00 0.00 0.00 22.4 NOTE: FI = fatal and injury; PDO = property damage only. Exposure and crash rates are shown in terms of hundred million vehicle passes over a five-year period. A vehicle pass is one vehicle passing one particular roadside object. So, if there are three individual trees on a roadway segment, the number of vehicle passes per day is three times the AADT. The 48 roadway segments with individual trees experienced no tree-related crashes during the five-year study period. Roadway Segments with Individual Utility Poles Table 27 summarizes data for roadway segments with a utility pole on the roadside within 40 ft of the traveled way. Roadway segments for these data were defined as including the roadside on only one side of the road and extending from each pole halfway to the adjacent poles, with a maximum length of 327-ft. A total of 3,455 individual utility poles were found on 3,455 roadway segments with a total length of 133.4 mi. Of these individual utility poles, 2,565 were on tangent roadway segments and 890 on horizontal curves; there were 393 utility poles on the outside of horizontal curves and 497 utility poles on the inside of horizontal curves. Table 27 also shows that the average offset from the traveled way to individual utility poles is 17.4 ft. The range of offsets from the traveled way to utility poles is from 2 to 40 ft.
52 The roadway segments with individual utility poles, overall, experienced 59 pole-related primary crashes during the five-year study period. No data were available for secondary crashes in Washington. Table 27. Summary Data for Individual Utility Poles at Washington Sites on Rural Two-Lane Undivided Highways Location type Number of individual utility poles Exposure (hundred million vehicle passes) Number of crashes pole-related in 5 years (2013-2017) Crash rate per hundred million vehicle passes (2013 â 2017) Average offset to poles (ft) Total FI PDO Total FI PDO All 3,455 395.332 59 23 36 0.15 0.06 0.09 17.4 Tangent 2,565 278.253 37 16 21 0.13 0.06 0.07 17.1 Curve 890 117.079 22 7 15 0.19 0.06 0.13 18.1 Outside of Curve 393 56.384 7 2 5 0.12 0.04 0.08 19.7 Inside of Curve 497 60.695 15 5 10 0.25 0.08 0.17 16.9 NOTE: FI = fatal and injury; PDO = property damage only.
