Cover Image

Not for Sale

View/Hide Left Panel
Click for next page ( 30

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement

Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 29
29 2. Add the numbers of before-treatment crashes, injuries identify correctable crash patterns, conduct an economic and fatalities from each lane- departure site and divide by analysis to ensure a minimum B/C ratio, and develop a com- the number of years of before data to produce a total num- bined treatment program which maximizes the benefits that ber of "potentially treatable lane-departure crashes and can be gained from a given total treatment budget. The crash injuries per year." network-screening tools within SafetyAnalyst provide a good 3. Multiply these totals by 20 percent to get the number of approach for applying Procedure 1. lane-departure crashes, injuries and fatalities that are The primary emphasis of this guide is on planning site- expected to be reduced per year by your current program. specific projects at high-crash locations. If preliminary analy- (This assumes an average Crash Reduction Factor of sis indicates that even an enhanced and expanded high-crash 20 percent for all lane-departure strategies. This is probably location program will not meet the goal, then the users too high, but in the ballpark of reality, and good enough will need to add systems-based or systemwide treatment for this exercise.) programs to the effort. This guide is not specifically intended 4. Compare the numbers of crashes and injuries reduced and for planning such systemwide treatments, but the four lives saved to your statewide lane-departure goal and procedures described earlier and detailed below can be ap- calculate the proportion of your total goal that this plied by the user to identify roadway systems or corridors (or represents (e.g., 20 percent or 0.20). even large numbers of individual segments) to treat and to 5. To calculate approximately how much you will have to help define the treatments that should be implemented for expand the lane-departure part of your HCL program to those systems or corridors. Again, the choice between which meet your goal, divide 1.0 by the proportion from the pre- procedure is appropriate is defined by three factors whether vious step. For example, if the fatality and injury savings or not treatment effectiveness is known, whether the jurisdic- from your current program is 20 percent of your goal, tion has inventory data that can be linked to their crash data, then you will have to identify and treat five times as many and whether the crashes are "mileposted" or not. Exhibit IV- lane-departure sites in the future (i.e., 1.0 / 0.20 = 5). 1 will guide the user to the appropriate procedure. The user will then need to make the determination of Procedure 1 Choosing Roadway-Based whether enough sites with high numbers of lane-departure Treatments and Target Populations crashes can be identified. Usually the HCL program identifies When Treatment Effectiveness Is more sites than can be treated. This full "census" of potential Known, and Both Crash and Non-Crash HCL sites can be examined to determine whether enough Data Are Available sites with high numbers of lane-departure crashes are avail- able. In most cases, if a stretch goal has been set, the answer The following text identifies the data needed for conducting will be "no." In that case, the user should consider adding sys- Procedure 1, followed by the individual steps in the procedure. tem improvements to the plan. While all states and some local jurisdictions have proce- Data Needs dures in place to identify and treat high-crash locations, it is noted that an improved methodology is currently being devel- The following are the specific data needed to use Proce- oped by FHWA in the SafetyAnalyst program described in the dure 1 when choosing and targeting roadway-based treatments. preceding Section (also see This set of the software tools for safety management of specific A specified effectiveness level (CRF or AMF) for each highway sites includes a series of procedures that will allow the treatment to be examined. user to identify high-crash locations or sites with potential for The "Treatment Effectiveness" section under each treat- safety improvement, diagnose potential treatment sites to ment in each NCHRP Report 500 series guide provides a Inventory Data Available and Linkable to Crashes? Treatment Yes No Effectiveness Mileposted Crashes Mileposted Crashes Unmileposted Crashes Known? Yes Procedure 1 Procedure 2A Procedure 2B No Procedure 3 Procedure 3 Procedure 3 Some known, Procedure 4 Procedure 4 Procedure 4 some unknown Exhibit IV-1. Guide to choice of procedures based on knowledge of treatment effectiveness and crash data quality.

OCR for page 29
30 description of what is known about CRFs. It is important Most state DOTs have computerized roadway inventory that the user review the material in the guides for a given files for the full state highway system that can be linked to treatment. Valuable information about the stability of the crashes, since both the homogeneous segments in the CRF, cautions about the use of the treatment and other inventory file and the crashes are identified by "ad- valuable information is included there, but will not be dresses" usually route and milepost or GIS coordinates. repeated here. Most local jurisdictions (i.e., counties, towns, townships, For a significant proportion of the treatment strategies cities) do not have such an inventory system. For jurisdic- defined in the six guides covered in this section, a specific tions that do not have an inventory file, Procedures 2A and AMF is not presented. Since the preparation of some of the 2B below can be used. earlier guides, additional information on treatment CRFs A network screening computer program which will ex- has been developed in both NCHRP 17-25, and in prelim- amine a fixed-length portion of each route (e.g., 1 mile) inary work for the Highway Safety Manual. The AMFs and calculate the number of target crashes (e.g., run-off- from NCHRP Project 17-25 have been published in road crashes, curve-related crashes) that have occurred NCHRP Research Results Digest 299 (27). The AMFs devel- in each "window" in the past 3 to 5 years. oped in NCHRP Project 17-27 will be incorporated in the This program exists in some jurisdictions, but may not forthcoming Highway Safety Manual. exist in others. If not, a knowledgeable computer analyst A computerized crash data file which includes sufficient can build one. The process/program will require that crash details to isolate target crash types (run-off-road, counts of target crashes can be made and "attached to" head-on crashes, and run-off-road on curves), and potential each homogeneous section on a route, and that each seg- target populations that will be affected by each treatment. ment includes a "segment length" variable. The program Here, the user will need to examine the data formats for must then be able to examine each segment, starting from variables in their crash file to identify variables and codes the first segment on a route, and accumulate both target within variables that can be used in determining whether or crash counts (by adding up the numbers in the segment- not each crash in the file is a "target crash." Crash databases specific counters) and segment length (by adding up the often categorize crash data for a given crash into up to three individual segment lengths). When the total accumulated subfiles general accident/crash variables, variables for segment length reaches the window length (e.g., 1 mile), each vehicle in the crash, and variables for each occupant in the total number of crashes is recorded in an output file, the crash. The variables needed to determine whether a along with the route number, the beginning milepost of crash is a "target crash" or not for roadway-segment-based the first homogeneous segment analyzed, and the ending crashes can usually be found in one of the first two milepost of the final segment analyzed in the current win- subfiles crash or vehicle data. Crash files differ from juris- dow. The program would then begin again the accumu- diction to jurisdiction. While certainly not always the lation of target crash counts and segment lengths on the case, the following variables (or similar variables) listed next homogeneous section of the route, and would repeat in Exhibit IV-2 will be used in this determination. the same process until the full route has been completed. Computerized roadway inventory data and/or intersec- The process would be repeated for all routes in the system. tion inventory data that can be linked to the crash data by Most current highway agency network screening programs location of the crash. operate in a database environment. Efforts are underway Crash Type Variable Crash Database Subfile Run-Off-Road (ROR) Crashes Accident/Crash Type Crash Manner of Collision Crash Sequence of Events Crash or Vehicle First Harmful Event Crash Most Harmful Event Vehicle Crash Location (Off-road) Crash Number of Vehicles or Units Crash Run-Off-Road Crashes into Same as ROR plus Trees and Utility Poles Object Struck Vehicle (Sometimes Crash) Most Harmful Event Vehicle Lane-departure Crashes on Same as ROR Crashes plus Curves Location Type Crash (Sometimes Vehicle) Head-On Crashes Accident/Crash Type Crash Sequence of Events Crash or Vehicle First Harmful Event Crash Most Harmful Event Vehicle Exhibit IV-2. Crash variables and subfile location by crash type.

OCR for page 29
31 to develop network screening programs that operate in a types/classes of interest, such as two-lane rural road seg- GIS environment as well. ments, multi-lane urban segments (perhaps by number of Computerized traffic count data that can be linked to the lanes), or rural interstate segments. If desired, these roadway inventory data (unless they are already con- potential treatment segments could be further screened tained in the same database). by AADT level (e.g., only "high-traffic" segments). While the procedure can be operated without comput- 2. Develop critical crash frequencies for each candidate erized count data, these data are almost always available in treatment type (e.g., shoulder rumble strips) for each state DOT files that have a roadway inventory system. This roadway class of interest. The "critical frequency" is the is not always the case in urban systems. If available, the frequency of target crashes per mile that, if treated, will traffic count information can be used to further target the result in crash-injury reductions whose economic ben- potential treatment sites in two ways. First, if the user only efit will exceed implementation costs by some factor. In wishes to treat "high-traffic" sites, these data can be used to the example presented in the FHWA Sample Plan (24), screen out "low-traffic" roadway segments prior to run- the target B/C ratio used was 2.0 or greater. ning the network screening program. Second, after the These "critical frequencies" must be developed for program has been run, the identified sites can be further each candidate treatment being examined. If the same screened by a given AADT level, or the sites can be sorted treatment is to be used on different roadway classes, it by AADT to assist the user in final site choice. will be necessary to develop different critical frequencies Unit cost for each treatment both original implemen- for each treatment by roadway class if the treatment cost tation costs and annual maintenance costs per unit length or treatment effectiveness varies by road- The guides do not provide treatment costs due to differ- way class. The following formula is used: ences between states and expected changes over time. The CF = (Ann. Cost)(Target B/C)/(Eff)(Avg. Crash Cost) user will need to obtain information on such costs, either from vendors or from other jurisdictions that have used Where: the treatment. The guides do provide "Information on CF = Critical annual frequency of target crashes per Agencies or Organizations Currently Implementing This unit length to consider the strategy to be cost effective. Strategy" that could be contacted for help under each of the Ann. Cost = The annual cost of the improvement per treatment strategies. (The guides provide only early users, unit length (e.g., per mile). If it is a construction im- and surrounding jurisdictions may have implemented the provement, it is the construction costs annualized over treatment after the guide was completed.) Finally, the user the expected life of the improvement. Note that if the will need an estimate of annual maintenance cost per mile treatment strategy is related to horizontal curves, then for each treatment to be analyzed. Since maintenance in- the cost required here is annual cost per unit length of cludes "replacement after a crash" in some cases, the user curve (e.g., cost per mile of curve). If the treatment may have to make estimates of the number of expected strategy is an education or enforcement treatment, the crashes per year and the amount of expected damage. annual cost can be expressed on a per-unit-length basis Again, past users of the treatment can be of assistance here. if the treatment is to apply to a specific road segment or corridor, but may also cover an entire geographic area or road system to which the treatment is applied Procedure (e.g., if the treatment for run-off-road [ROR] crashes The general procedure for choosing and targeting treat- is jurisdiction-wide in nature). ments with known effectiveness levels was provided in Sec- Target B/C = The B/C ratio defined by the user. It is tion III above. The following text will expand that description usually between 1.0 and 2.0. In the FHWA Sample Plan while focusing on roadway-segment treatments designed to (24), a value of 2.0 is used. reduce lane-departure crashes. Because the user needs to Eff = The estimated effectiveness of the treatment strat- understand the computerized procedure in order to input the egy in reducing targeted crashes, expressed as a propor- correct values and tailor it for their own jurisdiction, the tion (i.e., the CRF/100). This can be extracted from the following provides the details of each step and sub-step. FHWA Sample Plan (24) or from other sources. Avg. Crash Cost = The average economic cost per crash 1. Specify the types/classes of roadway segments that are for the target crash type that will be affected by this treat- potential targets for the treatments. ment strategy. The following were extracted from Table Because the choice of treatments, the treatment effec- 10 of Crash Cost Estimates by Maximum Police-Reported tiveness, and the treatment cost per unit length may dif- Injury Severity Within Selected Crash Geometries (22) fer by roadway class/type, the user will need to specify the and represent "comprehensive costs" in terms of 2001

OCR for page 29
32 Crash Type Speed Limit Comprehensive Here the user will need to have the network screening Category Cost/Crash* program output the locations of the originally chosen Run-off-road crashes < 45 mph $67,000 treatment sites, enter these sites into some type of spread- involving trees or other > 50 mph $107,000 sheet one treatment segment per row, and then sort the roadside objects Run-off-road crashes < 45 mph $148,000 rows by route and beginning milepost. By scanning involving rollover as the primary impact type > 50 mph $256,000 down this listing for each route under consideration, the < 45 mph $60,000 user can determine where the treatment gaps are located Head-on crashes > 50 mph $613,000 along each route the missing segments in the listing. * Costs in 2001 dollars (Note that some of these missing segments would result Exhibit IV-3. Crash cost by crash type and posted from the fact that the roadway class changed within the speed limit (22). route e.g., from two-lane to four-lane. This determina- tion will have to be made by comparison with informa- dollars. Comprehensive cost estimates include not only tion from the inventory file on the route in question.) the monetary losses associated with medical care, other This correction for missing segments within the same resources used, and lost work, but also non-monetary roadway class will be needed because the network screen- costs related to the reduction in the quality of life. The ing will only detect segments along a given route that cost for each crash type is shown in Exhibit IV-3 for two exceed the critical crash-frequency threshold. In almost ranges of speed limits 45 mph and 50 mph. The all cases, this will leave treatment gaps along a given route former should be useful for urban crashes, and the lat- within the same roadway class segments that do not ter for rural crashes. meet the threshold. The user will need to determine 3. Using the inventory file, stratify potentially treatable whether or not these below-threshold segments should roadway segments by roadway class. be treated. The logical first answer is "no," since the seg- This stratification will result in a file of roadway seg- ments did not meet the critical threshold. However, there ments sorted by route number for each of the roadway may be times when all or some of these gaps should be classes under consideration for treatment. included in the treatment program. 4. Link target crashes with roadway segments from the First, it may be illogical to leave isolated gaps untreated appropriate inventory data file, and then perform a on a given route if the gaps have basically the same road- computer screening of all segments on all routes that way and roadside characteristics and AADT as the adjacent are potential treatment locations to determine which treatment sites. The network screening program is exam- segments have crash frequencies that exceed the criti- ining a certain past period of crashes (e.g., the past 5 years). cal crash frequencies calculated above. Crashes, particularly run-off-road and head-on crashes, This will be done using the network screening pro- are usually low-frequency events per mile, and the loca- gram described above, and will be done independently tion of a crash in a given time period may be somewhat for each of the roadway types under consideration. The random. Thus, if a different 5 years of data were chosen, network screening program will need to output the route the chosen treatment locations might be slightly different. number and beginning and ending milepost for each 1- These factors will produce such treatment gaps. The user mile segment that exceeds the critical crash frequency. will have to make a judgment concerning whether to treat Note that if the treatment being considered is for hor- the gaps, and there are no precise guidelines for making izontal curves (i.e., the user is searching for a "system" of this judgment. As general guidance, if a long section of a horizontal curves to correct with, say, improved curve route has very few treatment gaps (i.e., most of the warnings), this step will require that the user's roadway segments on the route are identified by the network inventory system can identify the locations (routes and screening program), and if the user knows that the gaps begin/end mileposts) for horizontal curves. If no curve in- are very similar to the surrounding treatment sections in ventory data are available (as will unfortunately be the terms of AADT and roadway and roadside characteristics, case in most jurisdictions), then the user will have to use then it would appear logical to treat those gaps also. If either Procedure 2A or 2B instead of this Procedure 1. there are more "gaps" than "treatments" on a given route 5. Correct the output for "treatment gaps" along the same but the gaps and treatment segments are similar in terms route resulting from the network screening computer of AADT and characteristics, then the user might either program. decide to just treat the originally chosen sections, or not This correction will require that the user manually to treat this route at all. examine each of the routes under consideration within Second, there may also be situations where the logical each roadway class to detect possible "treatment gaps." length of the treatment may be greater than 1 mile. For

OCR for page 29
33 example, an agency may decide that shoulder rumble number of such target crashes for all unit-length segments strips will be installed only with repaving, and that chosen for treatment. Users with full crash and inventory repaving is not done in roadway sections of less than X systems who have developed the network screening pro- miles (with X greater than 1 mile). The first option here gram will have the ability to link such goal-oriented tar- would be to change the length of the window in the net- get crashes to each segment chosen and to sum the total work screening program to the minimum project length. over all segments. The best annual estimate will be one However, since the window jumps from point to point based on more than 1 year of past data (usually 3 to 5 along the route (e.g., from the beginning of mile 1 to the years) and then dividing by the number of years used. beginning of mile 2), the longer the window, the more 7. Repeat the above steps for each potential treatment type. likely that concentrations of target crashes that fall at the The above steps are then repeated for the second and ends and beginnings of adjacent sections will be split into subsequent potential treatment types. In each case, criti- two parts, and thus each part will fall below the critical cal crash frequencies are calculated for each roadway frequency. An alternative strategy is to use the 1-mile class, the network screening program is used to identify window as originally suggested, and to conduct the same treatment segments, and corrections are made for treat- examination as above if there are limited gaps among ment gaps. However, a final correction is needed for many treatment segments. The user could also examine segments that have been identified for more than one the identified treatment sites to see where groups of 1-mile treatment type, as detailed in the following step. section equal or exceed the minimum project length, and 8. Correct for multiple treatments on the same segment. then add additional adjacent segments if their AADT and Since many segment-based treatments affect the same characteristics are similar to the treatment group sites. type of target crash (e.g., shoulder rumble strips and In summary, there are no hard-and-fast guidelines shoulder widening both can affect run-off-road crashes), concerning how to correct for treatment gaps within the the above procedure will identify the same segment as a same roadway class. The decision will have to be made potential for treatment in many cases. In these cases, the by the user, and the rules for making the decision may user has two options: (1) choose only one treatment for change from project to project and treatment to treat- each of these segments, or (2) choose to implement two ment. In general, it would appear that the best decision or more treatments on the same segment. will be at least partially based on similar AADT and sim- Under Option 1, the user would compare the lists of ilar characteristics. potential treatment segments (after correction for gaps) 6. Estimate the expected crash/injury reductions on all from Step 5 above, and would decide which treatment to the identified target locations. place on each segment where two or more treatments The results of this step will be used in Step 9 below to could be implemented. That segment (and its related determine whether or not the goal is reached. Here, for goal-oriented crashes or injuries) is then removed from each treatment segment within a given roadway class the list of segments for all other treatments. identified at the end of Step 5 (i.e., after correction for Under Option 2, the user must develop some measure treatment gaps), the user will need to determine the of combined effectiveness for the two or more treatments to number of crashes and injuries that will be reduced by be applied to a given segment. Since the combined effec- this treatment. This will be done by summing up all per- tiveness of two treatment strategies on the same location tinent crashes or crash injuries for all segments to be will not be the simple sum of the two effectiveness levels, treated, and then multiplying this total by the estimated some correction must be applied for the second and all effectiveness level for the treatment under consideration. subsequent treatments that are applied to the same seg- CI reduction = (CI on segments) Eff ment. Unfortunately, there is little knowledge available about the combined effects of multiple treatments. Until Where: that knowledge is developed, it is suggested that the effec- CI = "Goal-related" crashes or crash injuries tiveness level (Eff) of the second treatment applied to a Eff = treatment effectiveness given section be reduced by 50 percent, and the effective- The definition of "goal-related" crashes or injuries is, as ness of the third treatment and subsequent treatments ap- implied, based on the nature of the overall goal that has plied to the same segment be reduced to 25 percent of their been established. If the goal is oriented to fatal and injury expected effectiveness when used alone. For example, target crashes, then these will be accumulated. If the goal assume that the first treatment for a given segment has an is total target crashes, then these will be accumulated. effectiveness level of 0.2, the second has an effectiveness The summing of goal-related crashes or injuries will be level of 0.15, and the third has an effectiveness level of 0.10, done by using a computer program to estimate the annual and the fourth and subsequent treatments add no addi-