Cover Image

Not for Sale



View/Hide Left Panel
Click for next page ( 137


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 136
METHODS Method 1. Analysis of national data A very basic method for estimating the safety impacts of transportation system changes involves using national data on crashes to determine the effect of a transportation improvement. This method can be used to assess how road upgrades in the common activity spaces of protected populations would affect both the safety of road users and pedestrians. It also can be applied to make generalized assessments of the comparative safety of roads serving the activity spaces of protected populations and those serving other members of the community. The method is simple to apply and requires only nominal data collection once the activity spaces of minority populations have been identified (see Chapter 2). When to use. These very general crash rates can be used to assess changes to facilities in areas of the community with a concentration of protected populations. Would safety improvements to users of the facility make the prevailing level of safety in areas with such concentrations comparable to other areas? Analysis. To compare an existing road with a proposed upgrade using national data, it is necessary to obtain data on crash rates by roadway functional class. These data are available in Highway Statistics from the FHWA at http://www.fhwa.dot.gov/ohim/hs98/roads.htm. Table 6-2 contains crash rates by functional class of road for 1997. The crash data in Table 6-2 are presented in rates of crashes per 100 million VMT. Using these rates allows you to estimate safety impacts of improving the current roadway by multiplying the current annual VMT by the appropriate crash factor and then subtracting the result from the forecast VMT on the upgraded road times the correct crash factor. For example, if a 10-mile urban principal arterial has 15.2 million VMT per year and is to be upgraded to an urban interstate with a forecast 29.1 million VMT per year, the change in fatal crashes would be 29.1 million VMT/100 million times 0.56, minus 15.2 million VMT/100 million times 1.30. The difference between the value of the upgraded road and the existing road represents the safety benefits and costs. In this example, there would be 3.5 fewer fatal crashes per year, even taking into account the increase in traffic volume. In Chapter 2, we discussed spatial data that can be used to identify areas of the community with relative concentrations of protected populations. Once areas with such concentrations are defined, roadway upgrades within them can be compared to those elsewhere to assess the extent to which safety improvements are being distributed equitably. Data needs, assumptions, and limitations. This method relies on the use of aggregated data that represent an average for the nation. Consequently, it assumes that any roadway conversion to a different functional class will follow the same path as the national average. It should be stressed that a VMT rate that is higher or lower than average (i.e., different from the national traffic density) may substantially affect crash rates. The results of this analysis should be considered to be a general approximation. 140

OCR for page 136
Table 6-2. Motor vehicle traffic fatalities and injuries by functional class, 1997 (per 100 million VMT) Injury crashes Persons injured Most serious Pedestrians injured Highway category Fatal Nonfatal Fatal Nonfatal injuries Fatal Nonfatal RURAL Rural Interstate 1.05 25.08 1.26 41.11 6.38 0.09 0.60 Other principal arterial 1.96 50.87 2.35 87.85 12.69 0.14 1.04 Minor arterial 2.33 70.52 2.73 118.25 16.00 0.18 1.24 Major collector 2.51 86.79 2.85 135.33 18.94 0.15 1.59 Minor collector 3.16 106.02 3.52 159.57 18.83 0.16 2.04 Local 3.52 147.49 3.89 222.82 20.14 0.32 4.31 VMT-weighted 2.15 69.10 2.49 110.35 14.15 0.16 1.51 average--rural URBAN Urban Interstate 0.56 46.56 0.63 72.48 5.24 0.10 1.18 Other freeways & 0.75 68.60 0.82 107.20 7.49 0.14 2.68 expressways Other principal arterial 1.30 124.69 1.40 199.06 14.57 0.35 5.42 Minor arterial 1.08 126.89 1.17 197.95 16.26 0.25 6.72 Collector 1.00 104.95 1.07 159.18 14.31 0.18 7.42 Local 1.33 194.40 1.42 295.74 15.86 0.36 16.78 VMT-weighted 1.01 109.50 1.09 170.48 12.17 0.24 6.19 average--urban Source: FHWA 1998, Table FI-1. The data presented here do not include rates for estimating increases or decreases in PDO crashes because they are not available in the annual FHWA publication, Highway Statistics. If you choose to include estimates for PDO crashes, these estimates should be in the form of a rate of per 100 million VMT rather than in raw numbers. Results and their presentation. A simple table can easily be constructed to depict the appropriate national crash rates for the current road and for the upgraded road. The table also can present the estimated number of crashes of each type per unit of time (e.g., a 1- or 5-year time frame), taking into account VMT before and following the upgrade. This summary table can give a general idea of the changes in road user and pedestrian safety that may result if the upgrade is completed. Assessment. Despite the concerns associated with using aggregate national data, this computation of safety benefits from roadway conversions is an easily implemented method that does not require significant technical skills. It presents clear, easily understandable results in the form of the differences in crash rates for each functional class of roadway. Overlaid on a GIS 141

OCR for page 136
representation of income and race, this approach provides a general sense of whether the safety benefits of transportation investments are equitably distributed. Method 2. Comparison approach A comparison approach can partially overcome the limitations associated with using national data. This method entails comparing crash rates on a roadway where potential changes are being considered--and other roadways comparable to it--with existing roads in the region that are representative of the improved road. It has the advantage of allowing you to focus on facilities serving low-income populations and minority populations to assess the extent to which they are less safe than those serving other members of the community. When to use. This approach is particularly appropriate for determining whether roadways serving areas of the community that fall within the common activity space of members of protected populations are less safe than roadways elsewhere in the community. It also enables you to evaluate the extent to which a particular improvement project would be likely to enhance the safety of those traveling on it. Analysis. The first step of the comparison approach involves collecting information on the road where the improvement is being considered. This includes data on traffic volume, capacity, and road geometry. Crash data are then obtained on roads that share similar characteristics and surrounding land uses. The idea is to obtain a large enough pool of comparable roads to enable a meaningful sample of crash data to be assembled. Finally, a series of roads with characteristics comparable to those the road will have when improved are identified. Crash data for these roads are assembled to facilitate comparison. Once a database is assembled, you can use it to assess many different projects. The first step of the analysis involves setting up a base case for the road that is the focus of the proposed improvements. The base case includes information on the number and types of crashes currently seen on the road, as well as its physical and geographical characteristics. Because crashes occur infrequently, it is a good idea to assemble data for a 3- to 5-year period. The base case is then compared with the example roads to determine whether the alternative improvements are likely to produce safety benefits. This comparison involves considering whether the rate of crashes will increase or decrease and what types of crashes can be expected to occur. Data needs, assumptions, and limitations. The analysis presents estimates of expected crashes from a road improvement project by comparing the roadway with improved roads that have similar characteristics. The resulting estimates can be expressed as reductions in crashes per 100 million VMT or as reductions in crashes on a given roadway per year. This approach requires data on other regional roads for comparison purposes. If such data are available, the method provides a simple means of evaluating safety impacts. It overcomes the limitations associated with using aggregate national data by concentrating on regional data. You need data on a sufficient number of road segments of both functional classifications to enable reliable crash rates to be estimated. Crashes are a rare event on any type of roadway, so stable 142

OCR for page 136
rates require data on numerous segments. Multiyear data files greatly improve the accuracy of crash rates because many more cases are included. Results and their presentation. The most direct way to present the results of this analysis is a table displaying crash rates (fatal, personal injury, and PDO) for one or more road standards within the region alongside the rates for relevant roads within the activity space of protected populations. The table would thereby facilitate direct comparisons to address the issue of whether roads commonly traveled by protected populations tend to be less safe than other roads. If so, specific improvements can be identified to bring the safety performance of these roads more in line with others across the region. Assessment. As discussed earlier in this chapter, roadways of a given standard are unlikely to be less safe in areas where members of protected populations live, work, or travel. The more likely environmental justice issue is whether particular facilities in such areas need to be upgraded because they currently have unacceptable crash rates. This approach is a reasonably effective means for making such an assessment. Method 3. Regression analysis Regression analyses are a more advanced technique for estimating changes in crash rates if a transportation project were undertaken. Data on road segment characteristics (e.g., grade, curvature, lane width, pavement quality, shoulder composition and width, and traffic volume) are merged with data on crashes occurring on each segment. When to use. Using crash rates as the dependent variable, it is possible to predict these rates on the basis of road segment characteristics. The strength of the approach is that one can change the various characteristics and see how these changes influence crash rates. If, for example, a road serving an area of the community frequently traveled by protected populations (see Chapter 2) has an unfavorable safety record, this method can be used to estimate the probable effects of making specific improvements to that road. Analysis. We present an equation derived using the approach just described, as well as the procedure for estimating such an equation in a particular state. This equation was estimated using data on the 17,767 two-lane and four-lane (non-Interstate) rural primary road segments (average length of about 0.4 mile) in Iowa. Data on a total of 21,224 crashes over a 3-year period were included. The relationship between roadway attributes and crash rates is probably quite similar in other states, so the existing equations can provide a preliminary estimate of safety effects. The crash-rate predictive model was estimated as a semilog regression equation. It was necessary to transform the dependent variable to a natural logarithm because almost one-third of the road segments had no crashes over the 3-year period analyzed and a standard linear regression model would have been inappropriate. Full documentation of the analysis methods is contained in Forkenbrock and Foster (1997). 143

OCR for page 136
Table 6-3 contains the dependent and independent variables included in the regression model. The seven independent variables pertain to physical characteristics of the 17,767 road segments. Each of these characteristics can be changed by a project to upgrade a road. Table 6-3. Variables used in regression model of crash costs Dependent variable Natural log of number of crashes (fatal, injury, and PDO) per million VMT. Independent variables PSR: present serviceability rating, ranging from 0 (poor) to 5 (excellent), is a measure of the general surface quality of a road segment. TOPCURV: the number of degrees of arc subtended by a 100-foot length for the sharpest curve on the segment (see AASHTO 2001). Scaling of the variable is as follows: 0 = no curve, 1 = 0.11.4 degrees, 2 = 1.52.4 degrees, 3 = 2.53.4 degrees, 4 = 3.54.4 degrees, 5 = 4.55.4 degrees, 6 = 5.56.9 degrees, 7 = 7.08.4 degrees, 8 = 8.510.9 degrees, 9 = 11.013.9 degrees, 10 = 14.019.4 degrees, 11 = 19.527.9 degrees, and 12 = 28.0 degrees or more. PASSRES: a dummy variable coded 1 if a passing restriction exists anywhere on the road segment and 0 if no passing restriction exists. ADTLANE: average daily traffic in thousands per lane. RIGHTSH: width of the right shoulder in feet. LANES: a dummy variable coded 1 if the road segment has 4 traffic lanes and coded 0 if it has 2 lanes. TOPGRAD: the change in elevation, as a percentage of the horizontal distance traversed for the greatest slope in the segment. Scaling of the variable is as follows: 0 = no grade, 1 = 1.01.9 percent, 2 = 2.02.9 percent, 3 = 3.03.9 percent, 4 = 4.04.9 percent, 5 = 5.05.9 percent, 6 = 6.06.9 percent, 7 = 7.07.9 percent, 8 = 8.08.9 percent, 9 = 9.09.9 percent, 10 = 10.011.9 percent, 11 = 12.014.9 percent, and 12 = 15.0 percent or more. Source: Forkenbrock and Foster 1997, Table 1. After fitting a semilog regression equation (dependent variables transformed to a natural log) to the data just described, we took antilogs of the result. The latter step restored the dependent variable to its original form, thus allowing crash rates to be predicted. The crash-rate equation is as follows: crashes million VMT ( )( )( = 0.517 0.972 PSR 1.068TOPCURV 1.179 PASSRES 1.214ADTLANE )( ) (0.974RIGHTSH)(0.933LANES)(1.051TOPGRAD). All coefficients are statistically significant at the 0.001 level except PSR and LANES, which are significant at the 0.100 level. The r2 is 0.66. Example. The crash rate model allows you to compare the expected crash rate per million VMT of the current standard roadway with the expected crash rate if the roadway were upgraded. To illustrate, we apply a case in which a two-lane highway is a candidate for upgrading to four lanes. Table 6-4 presents the attributes of the base case and improved roadway. 144

OCR for page 136
Table 6-4. Application of the cost model to a typical upgrade Variable Base two-lane Improved four-lane PSR 3.0 4.0 TOPCURV 5 3 PASSRES 1 0 ADTLANE 2.5 1.25 RIGHTSH 7.0 10.0 LANES 0 1 TOPGRAD 4 2 Crash rate (per million VMT) 1.28 0.56 Source: Forkenbrock and Foster 1997, Table 4. Plugging the values of each case into the equation allows expected crash rates to be derived: ( )( )( )( )( )( 1.28 = 0.517 0.9723.0 1.0685.0 1.1791.0 1.2142.5 0.9747.0 0.9330.0 1.0514.0 .)( ) In this case, the crash rate would fall from 1.28 to 0.56 crashes per million VMT. Multiplying these values by the annual VMT of the roadway allows you to predict the change in crashes per year. Suppose that a 30-mile stretch of the two-lane highway with the characteristics of the base case in Table 6-4 is being upgraded to a four-lane highway as in the improved case in the table. The highway has an average annual daily traffic (AADT) of 8,000; after the upgrade, it is forecast to have an AADT of 10,000. Using the same crash data upon which the regression model is based, Table 6-5 shows the breakdown of crashes by type. We can use the crash cost data from Table 6- 6 to construct a weighted estimate of the annual crash costs of the base and improved cases. The cost difference reflects the annual crash cost savings that the improvement would bring about. Table 6-5. Types of crashes by number of lanes* (Values in parentheses are row percentages) Crash type Number of lanes Fatal Personal injury PDO Total 2 369 5,491 13,552 19,412 (1.9) (28.3) (69.8) (100.0) 4 18 476 1,318 1,812 (1.0) (26.3) (72.7) (100.0) *The figures in this table are 3-year totals for 1989, 1990, and 1991 on two- and four-lane rural primary non- Interstate segments. Two-lane roads account for 96.0 percent of the system mileage and 89.2 percent of the VMT on the road segments represented in this table. 145

OCR for page 136
Table 6-6. Estimates of crash costs by police-reported severity Per person Per crash Severity (2003$) (2003$) Fatal 3,359,212 3,820,635 Incapacitating injury 237,872 320,756 Evident injury 46,628 67,826 Possible injury 23,897 35,403 Property damage 2,433 6,299 Crash unreported to police 2,247 5,815 Source: Miller et al. 1991, p. 39, updated by the guidebook authors. For the two-lane base case, the weighted crash cost is 0.019 ($3,820,635) + 0.283 ($67,826) + 0.698 ($6,299) = $96,184. For the four-lane improved case, the weighted crash cost is 0.010 ($3,820,635) + 0.263 ($67,826) + 0.727 ($6,299) = $61,980. With an AADT of 8,000, the annual VMT on the base case highway is 8,000 vehicles/day x 365 days/year x 30 miles = 87.6 million vehicle miles/year. With a crash incidence of 1.28 per million VMT, there are 112.1 crashes per year. With a weighted average crash cost of $96,184, the annual crash cost for the base case is $10.78 million. The improved case would have an annual VMT of 109.5 million. A crash incidence of 0.56 for the improved case yields 61.3 crashes per year. Applying the weighted average crash cost for the four-lane improved case of $61,980, the annual crash cost is $3.80 million. The annual savings in crash costs resulting from the improvement would be $6.98 million. Data needs, assumptions, and limitations. To use the regression equation presented in this guidebook, you will need data on the current characteristics of each road segment to be improved. The analysis also requires information on the changes in characteristics that would result from the project. The data should be segment-specific. Fortunately, most state DOTs maintain data files on the primary roads within their states. Likewise, most DOTs maintain crash data files that link crashes to specific road segments. Results and their presentation. The approach just discussed can be applied at two levels. You can use actual road system and crash data for a particular state to estimate a regression equation or you can use the Iowa equation as an approximation. Because the equation provided above was estimated using many observations, it is quite stable. We should emphasize that it is suitable for rural primary roads, not for interstate highways or urban streets. Although a four-lane urban street may share certain specifications (e.g., lane width) with its rural counterpart, the nature of traffic flows and the general operating environments are sufficiently different that it would be 146

OCR for page 136
inappropriate to use the equation to predict crashes in an urban setting. Note that the predictive regression equation does not address intersections, per se; intersections were treated as part of the nearest road segment in the data file on rural primary roads. Assessment. The primary advantage of this method is that one can estimate the effect of changing one or more road attributes while holding all other attributes constant. Few other methods to estimate safety effects have this capability. Because urban streets vary considerably in such important characteristics as curbs cuts (i.e., driveways, alleys, and parking facilities), a model of this sort may not be a useful tool within urban areas. Method 4. Bicycle safety index A few methods are available for estimating the likely effects of roadway projects on the safety of nonmotorized transportation. The bicycle safety index (BSI) is perhaps the best approach for estimating how bicycle safety might be affected by changes in a series of road attributes. The original BSI was developed by Davis (1987) and modified by Epperson (1994). We present the modified version. When to use. If a roadway through an area inhabited by protected populations is a candidate for upgrading, it is appropriate to evaluate the extent to which the safety of these populations would be affected as they move about. It is not unreasonable to expect that in low-income neighborhoods, especially those with relatively high densities, bicycles are a relatively common means of conveyance. There are occasions when upgrading a roadway to make it easier for traffic to flow through an area of the community will have a deleterious effect on local residents' ability to move about safely in their own activity space. In such instances, it often is users of nonmotorized transportation, particularly bicycles, whose safety is affected. This method enables you to assess the extent to which bicyclists' safety would be reduced if a project were to go forward. The method also is useful in assessing how much bicyclists' safety would be enhanced if specific improvements were to be made to a roadway serving them. Analysis. The BSI is estimated using the following function: BSI = [AADT/(L 3100)] +(S/48) + {(S/48) [(4.25 - W) 1.635]} + PF + LF where BSI = Bicycle safety index for a specific road segment AADT = Average annual daily traffic L = Number of traffic lanes S = Speed limit (kilometers per hour) W = Width of the outside lane (meters) PF = Sum of pavement factors (derived from Table 6.7) LF = Sum of location factors (derived from Table 6.8) 147

OCR for page 136
Pavement factors include elements such as pavement surfaces and conditions that may constitute a hazard to cyclists. Epperson has assigned a value to each of these factors, as shown in Table 6-7. The table indicates that cracks in the pavement, rough railroad crossings, and the presence of drainage grates are the most serious pavement-related hazards to cyclists. Table 6-7. Pavement factor values Factor Value Cracking 0.50 Patching 0.25 Weathering 0.25 Potholes 0.25 Rough road edge 0.25 Railroad crossing 0.25 Rough railroad crossing 0.50 Drainage grates 0.50 Source: Epperson 1994, Table 2. Location factors pertain to conditions that affect the generation of cross traffic, limit sight distance, or restrict the safe operation of bicycles (see Table 6-8). A lower total score indicates that the road segment is comparatively safe for bicycle travel. Negative location factors imply that a feature would improve bicycle safety. For example, a raised median restricts left-turning cross traffic. The most serious location factor is angled parking, and the best safety feature is a paved shoulder. The appropriate factor values are plugged into the BSI function, and an index value is obtained. Table 6-9 provides a basis for interpreting the resulting index value. Example. A roadway is upgraded in the following ways: (1) a center turn lane is added (reduction of 0.20), (2) a solid raised median is added (reduction of 0.50) and angled parking is converted to parallel parking (0.75 down to 0.25), (3) the speed limit is reduced from 50 km/hr to 40 km/hr, and (4) the outside lane is increased from 3 to 4 meters. Other parameters remain unchanged (AADT = 5,000 and L = 4). Let us assume that the sum of pavement factor values before the project is 0.00 (i.e., there would be no pavement-related problems if the road upgrade were to be completed), and some of the location factors would remain unchanged (moderate grades, frequent curves, restricted site distance, numerous drives, and industrial land use). The improvement would reduce the sum of location factor values by 1.20 (i.e., -0.20 [center lane] -0.50 [raised median] -0.50 [angled parking changed to parallel parking]). 148

OCR for page 136
Table 6-8. Location factor values Factor Value Angled parking 0.75 Parallel parking 0.25 Right-turn lane (full length) 0.25 Raised median (solid) 0.50 Raised median (left turn bays) 0.35 Center turn lane (scramble lane) 0.20 Paved shoulder 0.75 Grades, severe 0.50 Grades, moderate 0.20 Curves, frequent 0.35 Restricted sight distance 0.50 Numerous drives 0.25 Industrial land use 0.25 Commercial land use 0.25 Source: Epperson 1994, Table 2. Table 6-9. Interpretation of BSI values Index range Classification Description 0 to 3 Excellent Denotes a roadway extremely favorable for safe bicycle operation. 3 to 4 Good Refers to roadway conditions still conducive to safe bicycle operation but not quite as unrestricted as in the excellent case. 4 to 5 Fair Pertains to roadway conditions of marginal desirability for safe bicycle operation. 5 or above Poor Indicates roadway conditions of questionable desirability for bicycle operation. Source: Epperson 1994, Tables 1 and 2. Original case: BSI = [5000/(4 3100)] + (50/48) + {(50/48) [(4.25 - 3) 1.635]} + 0.00 + 2.30 149

OCR for page 136
BSI = 5.9, in the "poor" category Improved case: BSI = [5000/(4 3100)] + (40/48) + {(40/48) [(4.25 - 4) 1.635]} + 0.00 + 1.25 BSI = 2.8, in the "excellent" category Data needs, assumptions, and limitations. As the list of variables indicates, to apply the BSI you must assemble data on the physical attributes and general condition of the roadway in question and estimate the same measures for the roadway following the proposed improvements. Data also are needed on AADT and flow speeds. None of these data are difficult to acquire. The BSI enables a composite index to be estimated with and without a transportation project that would entail several improvements to a roadway. It is possible to estimate the effects of individual improvements that might be included in the project to see whether these improvements would materially improve bicycle safety. Results and their presentation. The simple equation at the heart of this method enables sensitivity tests to be carried out that assess the efficacy of various features that may be added to make a road safer for cyclists. An initial use of this method can be diagnostic--it can enable you to assess the extent to which a safety problem currently exists on one or more roadways in an area commonly traveled by protected populations. Simple tables can show the safety improvements that specific features are likely to provide. Assessment. The BSI is a simple indicator that helps you gain a sense of how specific changes to a roadway may affect the safety of cyclists within a particular area. Epperson (1994, p. 12) cautions, however, that his index explained only 18 percent of the variation in severe bicycle crashes on various roadways in his test area. He attributes this limited predictive ability to differences in bicycle use patterns and the diverse nature of cyclists. Regarding the latter point, Epperson suggests that the BSI is likely to more accurately predict crash rates of experienced cyclists than those of young children riding bicycles. It does have the advantage of pointing to specific features that can be included in a road upgrade to make a facility safer for cyclists. Method 5. Bicycle compatibility index As discussed in the previous method, nonmotorized transportation safety may be a significant environmental justice issue in some communities. If a community is safe for motorists but relatively unsafe for pedestrians and bicyclists, further road investments may worsen the safety differential and thus bring about an environmental justice problem. Such a problem could be at least partially ameliorated by making that roadway or others more compatible with bicycle traffic. When to use. Various standards are available for evaluating cycling facilities, including those of American Association of State Highway Transportation Officials (AASHTO 1999). Consistent with these standards, the Bicycle Compatibility Index (BCI) (Harkey et al. 1998), developed for the FHWA, can be used to evaluate cycling conditions on road links. It also can be used to 150

OCR for page 136
estimate the effects of transportation projects on bicycle travel within a particular geographic area of interest. Analysis. The BCI consists of an equation into which the relevant values are inserted: BCI = 3.67 - 0.966 BL - 0.41BLW - 0.498CLW + 0.002CLV + 0.0004OLV + 0.022 SPD + 0.506 PKG - 0.264 AREA + AF where BL = presence of bicycle lane or paved shoulder ( 0.9 m. No = 0; Yes = 1) BLW = bicycle lane or paved shoulder width (to nearest tenth meter) CLW = curb lane width (to nearest tenth meter) CLV = curb lane volume (vehicles per hour [VPH] in one direction) OLV = other lane volume(s) (same direction VPH) SPD = 85th percentile speed of traffic (kilometers per hour) PKG = presence of parking lane with more than 30 percent occupancy (no = 0; yes = 1) AREA = type of roadside development (residential = 1; other = 0) AF = adjustment factors, ft + frt + fp NOTE: ft is the truck volume adjustment factor found in Table 6-10, frt is the right turns adjustment factor shown in Table 6-11, and fp is the parking turnover adjustment factor from Table 6-12. Table 6-10. Truck volume factor (ft) Truck* volume (per lane hourly) ft 120 0.5 60-119 0.4 30-59 0.3 20-29 0.2 10-19 0.1 < 10 0.0 *Trucks are defined as all vehicles with six or more tires. Source: Harkey et al. 1998, Table 1. 151

OCR for page 136
Table 6-11. Right turns factor (frt) Right turn volume (hourly)* frt 270 0.1 <270 0.0 *Includes total number of right turns into driveways or minor intersections along a roadway segment. Source: Harkey et al. 1998, Table 1. Table 6-12. Parking turnover factor (fp) Parking time limit (minutes) fp 15 0.6 16-30 0.5 31-60 0.4 61-120 0.3 121-240 0.2 241-480 0.1 >480 0.0 Source: Harkey et al. 1998, Table 1. Once the BCI has been calculated, it is possible to determine the compatibility level and the level of service (LOS) using Table 6-13. The standard BCI values are intended to represent the abilities and preferences of average adult cyclists. The authors of this method therefore suggest that only LOS C or better be considered suitable for casual cyclists. Table 6-13. Average adult cyclist compatibility level and LOS of roadways by BCI BCI range Compatibility level LOS 1.50 Extremely high A 1.51-2.30 Very high B 2.31-3.40 Moderately high C 3.41-4.40 Moderately low D 4.41-5.30 Very low E >5.30 Extremely low F Source: Harkey et al. 1998, Table 2. 152

OCR for page 136
Example. Suppose that a current roadway has the following: No dedicated bicycle lane, A curb lane width of 3.2 m, A traffic volume of 600 VPH in both lanes in the same direction, An 85th percentile speed of 40 km/hr, A parking lane with more than 30 percent occupancy, Residential development along the roadway, A truck volume per lane of 35 VPH, A parking turnover rate of 30 min, and 200 right turns per hour. The improved roadway would have the same attributes except that a 1.2-m dedicated bicycle lane would be added; the curb lane width increased to 3.5 m; the parking turnover rate increased to 50 min; and the parking lane alongside the road decreased to less than a 30-percent occupancy. The change in bicycle LOS can be easily calculated. The original condition is: BCI = 3.67 - 0.966(0) - 0.41(0) - 0.498( 3.2) + 0.002(600) + 0.0004 (600) + 0.022( 40) + 0.506(1) - 0.264 (1) + (0.3 + 0.5 + 0.1) = 5.538. The improved condition is: BCI = 3.67 - 0.966(1) - 0.41(1.2) - 0.498( 3.5) + 0.002(600) + 0.0004 (600) + 0.022( 40) +0.506(0) - 0.264 (1) + (0.3 + 0.4 + 0.1) = 3.325. Referring to Table 6-8, the BCI for this facility was originally "extremely low" (LOS F), but with the improvements it would become "moderately high" (LOS C). Data needs, assumptions, and limitations. Harkey's BCI requires data that are routinely available in planning or public works agencies. As the variable definitions indicate, many of the data are geometric: they define roadway and curb features. Other data pertain to traffic flow and roadside land patterns. All of these data are likely to be easily acquired. Results and their presentation. GIS can be used to produce maps that show existing cycling conditions; identify problems and barriers; assess the effects of a proposed project or policy; and suggest how these correlate with indicators of cycle demand. These maps can be overlaid on maps indicating the common activity space of protected populations. Roadway suitability ratings can also be used to identify preferred cycling routes; these routes can be compared to the same activity space. Collectively, this information can be used to prioritize cycling facility improvements by identifying problems in the road and path network on corridors with relatively high cycling demand that serve protected populations. 153

OCR for page 136
Assessment. The BCI can be a useful technique for measuring and evaluating roadway conditions for cyclists in activity spaces frequented by protected populations. This rating system can be used to assess existing bicycle travel conditions and how a particular project or policy would affect these conditions. A low LOS implies poor safety and convenience, both of which are bound to discourage travel by this mode. The technique is simple and easy to apply, and it gives approximations that should be adequate for most applications. Method 6. Pedestrian street crossings An environmental justice issue is likely to arise when a road project increases the volume and speed of traffic through an area where pedestrians must cross the affected road. This method is a good means for assessing the ability of various types of pedestrians to safely cross a road. It uses a form of LOS rating to evaluate roadway crossing conditions for pedestrians that is similar to LOS ratings used by transportation engineers to evaluate roadway performance for motorized traffic. The most important consideration in terms of pedestrian service and safety is intersection performance. It is there that pedestrianmotor vehicle conflicts are the most likely to occur. When changes are being considered for roadways that are frequently crossed by protected populations, environmental justice considerations dictate that every effort be made to make crossings as safe as possible. This method is a good way to assess how safe a crossing would be with and without pedestrian safety features. Analysis. A logical method of assessing pedestrian LOS for street crossings is pedestrian delay. Wellar (1998) has suggested a rather basic rating system, shown in Table 6-14. The table implies that when delays become relatively long, the likelihood increases that pedestrians will not always comply with signals or yield to traffic. In short, they will occasionally place themselves in harm's way. The implication is that by reducing average pedestrian delays at intersections, two positive effects are possible: encouragement for more short trips to be taken on foot and greater safety for those walking across intersections. Table 6-14. Pedestrian road crossing LOS (Values are average delays in seconds per pedestrian crossing) Signalized Unsignalized Pedestrian LOS intersection intersection noncompliance likelihood A <10 <5 Low B 10-20 5-10 Low to moderate C 20-30 10-20 Moderate D 30-40 20-30 Moderate to high E 40-60 30-45 High F 60 45 Very high Source: Derived from Milazzo et al. 1999, Tables 5 and 7. Data needs, assumptions, and limitations. The current performance of an intersection and its expected performance after a transportation project in terms of pedestrian crossings are the key 154

OCR for page 136
to LOS for foot traffic. It is not difficult to compare the actual time available for crossings with the generally accepted time requirements: crosswalk walking speeds are 1.2 meters per second (m/s) for most areas and 1.0 m/s for crosswalks serving large numbers of older pedestrians. Time available is affected by signal cycles and, in the case of nonsignalized intersections, traffic speed and volume. Results and their presentation. GIS can be used to produce maps that show existing pedestrian conditions, the effects of a proposed project or policy, and how these effects correlate with indicators of pedestrian demand. For example, the city of Portland, Oregon, used GIS mapping to prioritize pedestrian improvements. Planners performed a survey of existing pedestrian facilities, such as sidewalks, and identified barriers and missing links in the network. They also identified areas with a relatively high demand for walking, taking into account factors such as population density; attractions such as schools and commercial districts; and current nonmotorized travel. With this information incorporated into a GIS system, it was relatively easy to identify barriers and links in areas with high pedestrian demand, which were assigned the highest priority for improvement. Assessment. Pedestrian conditions can be evaluated based on sidewalk, path, and roadway crossing conditions. It is possible to estimate the likelihood that pedestrians will venture into dangerous conditions by examining the probable delays at a point of crossing. In hazardous situations, measures such as pedestrian signals can be installed to improve convenience and safety. Other indicators of pedestrian convenience, such as circuitous routes between common origin-destination pairs, can best be examined in the field. Method 7. Pedestrian danger index Conflicts with motor vehicles constitute one of the greatest safety hazards for pedestrians. Approximately 5,900 pedestrians are killed by automobiles annually (NSC 2003). Of all pedestrian fatalities, about 22 percent occur at intersections, as do approximately 44 percent of all pedestrian injuries (FHWA and ITE 2004). If a proposed transportation project would increase the factors that contribute to such incidents and if this would occur in an area of the community that is within the common activity space of protected populations, an environmental justice issue would need to be addressed. The pedestrian danger index is a basic method for assessing the danger that a roadway may pose to pedestrians. When to use. This approach can be used to identify areas adjacent to a proposed project that are most sensitive to environmental justice issues with respect to pedestrian safety. Comparing pedestrian danger index values that pertain to the preproject situation with pedestrian index values based on estimates of the postproject situation will reveal areas where the project would compromise pedestrian safety the most. Postproject estimates are based on index values computed for comparable areas. Estimating the pedestrian danger index values requires pedestrian crash data, population data, and pedestrian exposure data. If reliable data for each of the three variables are available at an appropriate level (i.e., community, neighborhood), an index can be developed from data at that level and used to make relevant comparisons. The method can be especially helpful when a transportation project passes through multiple neighborhoods and there is a question as to where to focus available funds to improve pedestrian safety. 155

OCR for page 136
Analysis. The Surface Transportation Policy Project (STPP) that developed this method used it to measure pedestrian safety at the county level in California. The relative ranking of counties with respect to per capita pedestrian injuries was used as a method for identifying counties that are most hazardous to pedestrians. There is no conceptual reason why this method could not be applied at a much smaller scale of analysis. The pedestrian danger index is determined by dividing the number of pedestrian injuries and fatalities in each area of analysis by the area's population and then dividing that number by a number representing the overall level of pedestrian activity in the area. The quotient is adjusted (normalized) to a scale ranging from 1 to 100, with 100 being the most dangerous (Ohland et al. 2000, p. 7). The steps in deriving the index are as follows: 1. Number of injuries/1,000 people = (pedestrian death and injury rate / population) x 1,000. 2. Pedestrian exposure rate = number of employed residents walking to work / Total number of workers. 3. Unadjusted index value = number of injuries per 1,000 people / pedestrian exposure rate. 4. Pedestrian danger index value adjusted to relative 1 to 100 scale = (unadjusted index value / maximum unadjusted index value within sample) x 100. These values should be acquired for as many separate and comparable observational areas as possible. A proposed transportation project will usually pass through multiple areas for which corresponding pedestrian danger index values have been calculated. Index values for various areas affected by the project can be compared to determine the portions of the route that contribute most to the hazards faced by pedestrians in the vicinity. If data can be obtained for comparable areas in several different communities, a more meaningful assessment can be made of the relative pedestrian danger in the area under study. The index values account for changes in population and pedestrian exposure; therefore, the index is a good measure of relative danger and can be used to compare areas with diverse land use and population density attributes. To determine how a project will impact pedestrian safety, it is necessary to compare current index values with index values that are calculated using projected pedestrian exposure data and projected pedestrian death and injury data. Projections are based on actual numbers collected from one area or (preferably) several areas where a similar project has been completed in the past. Note that pedestrian trip distance, area demographics, road geometry, and traffic volume should be reasonably similar; specifically, hazards created or relieved by the proposed project should be as similar as possible to those in other areas that are included in the comparative analysis. Data needs, assumptions, and limitations. The first requirement for constructing a pedestrian danger index is collecting data from several areas for each of three variables. The first variable, as mentioned above, pertains to pedestrian injuries and fatalities. The STPP researchers, for example, procured these data from the California Highway Patrol. The second variable, area population, can be obtained from census files (e.g., block group data). The third variable, pedestrian exposure rate, is estimated using census data pertaining to the variable "journey to work." Specifically, the category of this variable that reflects the number of people walking to 156

OCR for page 136
work divided by the total number of workers. This pedestrian exposure rate may be regarded as a reasonable surrogate for overall levels of pedestrian activity (Ohland et al. 2000, p. 7). Of course, the more observational areas that are included in the analysis, the more valid the findings will be. Results and their presentation. The pedestrian danger index is a method that can be used to approximate changes in the level of pedestrian safety that would result from a proposed transportation project and can be couched as a way to determine which portion of the proposed project should receive the most attention with respect to pedestrian safety. The benefit of the measurement is that, once calculated, it is an intuitive and concrete measurement that people can easily understand. It is important to present demographic information identifying protected populations along with the pedestrian danger index. This can be done by overlaying census block group scores for the pedestrian danger index on appropriate demographic data (relative presence of minority populations and low-income populations). When planning the proposed transportation project, particular attention needs to be devoted to locations with relative concentrations of protected populations and especially great dangers to pedestrians. Assessment. This method is a practical way of creating a relative measurement of hazard to pedestrians for various geographical areas. The method offers flexibility in terms of application--the index can be applied as a comparison between counties, communities, or neighborhoods, depending on the nature of the data being used. Collecting data that are consistent in scale and scope is crucial to the successful application of this method. For example, data collected on pedestrian deaths and injuries in Community A must be collected according to criteria similar to those used for Community B. The main limitation of the method is the availability of data on pedestrian crashes and pedestrian exposure. The acquisition of quality comparison data will, in effect, determine the quality of the overall method. Also, the comparison areas used to develop the postproject pedestrian danger index projections should match the study area in two important ways. First, and most importantly, the demographics of the residents, and hence the degree to which the area provides residence to protected populations, should be similar. Second, the nature of the completed transportation project that serves as a comparison should closely resemble the proposed project being analyzed. These two conditions can be difficult to meet, which constitutes the most serious limitation of the method. Method 8. Barrier effect analysis The negative effect that highways and vehicle traffic can have on nonmotorized mobility is sometimes called the "barrier effect." Swedish and Danish highway agencies have developed methods for quantifying the barrier effect in terms of additional travel delay experienced by pedestrians and cyclists, similar to the way traffic congestion delays to motor vehicles are quantified. Rintoul (1995) has suggested a reasonably direct method for estimating the barrier effect. When to use. The goal of a road upgrade is generally to move increased volumes of traffic at higher speeds. A consequence of faster and heavier traffic, however, may be that pedestrians and cyclists have increased difficulty crossing the roadway. In an area that constitutes the activity 157

OCR for page 136
space of protected populations, this reduction in mobility can constitute an environmental justice problem. In locations where substantial pedestrian and nonmotorized transportation crossings occur, this method can be applied to assess the change in crossing difficulty that would result from a proposed upgrade. Analysis. Rintoul (1995) suggests three steps to quantify the barrier effect described in turn below: Step 1 Calculate barrier size. Calculate the barrier size based on traffic volumes, average speed, share of trucks, number of pedestrian crossings, and length of roadway under study. B = q kl kh where B = barrier size q = average annual daily traffic kl = correction factor for trucks, 0.667 + 3.33 x percentage of trucks kh = (v/50)4 where v = average traffic flow speed (km/h) For example, let q = 13,600 AADT, the percentage of trucks = 8.1%, and the average traffic flow speed = 60km/hr: Barrier size = (13,600) x (0.667 + [3.33 x .081]) x ([60/50]4) = 26,417. Step 2 Calculate crossing potential. Calculate the demand (i.e., crossing potential) for road and street crossing based on the number of residential, commercial, recreational, and municipal destinations within walking and bicycling distance of the road. The resulting estimate represents the maximum possible number of nonmotorized trips, assuming that there is no traffic barrier to walking and cycling. For a small study area, this can be done using maps to mark major origins (e.g., housing) and pedestrian destinations (e.g., schools, parks, transit stops, and commercial areas). ( CP = d p cpf ) where CP = crossing potential d = population density (persons per km2) p = portion of total population for each age range cpf = crossing potential factor for each age range, indicated in Table 6-15 Continuing our example, let the population density be 741 persons per square kilometer and the population age distribution be as shown in Table 6-16. Then the crossing potential can be obtained from Table 6-15. CP = 741 (.07 .042) + (.12 5.0) + (.07 7.0) + (.82 2.6) + (.12 .74) = 2, 089 158

OCR for page 136
Table 6-15. Crossing potential factor (cpf) Age range cpf Infant/Toddler (0-4 yrs) 0.42 Elementary (5-12 yrs) 5.0 Secondary (13-17 yrs) 7.0 Adult (18-65 yrs) 2.6 Senior (more than 65 yrs) 0.74 Source: Rintoul 1995, p. 9. Values are based on experimental data. Table 6-16 shows in tabular form this example calculation of total crossing potential for an area with a population density of 741 persons per square kilometer. The values for "crossing potential" represent the expected number of crossings per day, in this case, 2,089. Table 6-16. Crossing potential factor example Portion of Crossing Age total potential Population Crossing range population* factor density potential Infant/Toddler (0-4 yrs) 0.07 0.42 741 22 Elementary (5-12 yrs) 0.12 5.0 741 444 Secondary (13-17 yrs) 0.07 7.0 741 363 Adult (18-65 yrs) 0.62 2.6 741 1,194 Senior (more than 65 yrs) 0.12 0.74 741 66 Total 1.00 2,089 *These are example values. They may not be representative of a given community. Step 3. Calculate disruption site. The barrier size and the potential daily crossings are combined to yield a measure of total disruption per kilometer of barrier. The total disruption represents the amount of exposure of pedestrians and cyclists to vehicular traffic. TD = A CP R B where TD = total disruption per kilometer of barrier A = adjustment for controlled crossing (A= 1 percent utilization of the crossing) CP = crossing potential, as previously discussed R = relative disruption factor, an approximate weighting by age (infant = 24, elementary age child = 16, secondary education child = 4, adult = 1, and senior citizen = 4) 159

OCR for page 136
B = barrier size, as previously discussed The relative disruption factor takes into account the fact that street crossing causes different levels of disruption for various age groups. This difference is due to such factors as ability to correctly assess risk, mobility, and ability to use other transportation modes. Although somewhat arbitrary, it provides a greater degree of realism. Suppose that observation leads to an estimate that the use of controlled crossings is 30 percent, so the adjustment factor is 1 0.30 = 0.70. Using this estimate, the total disruption is displayed in the far right column of Table 6-17. A total of 32,602,280 units of disruption results in our example. This value can be compared with the total for the base case or various alternatives. Table 6-17. Total disruption per kilometer of barrier Portion of Total Age total Crossing Crossing Disturbance Barrier disruption range population utilization potential factor size (1,000s) Infant/Toddler (0-4 yrs) 0.07 0.70 22 24 26,417 683.41 Elementary (5-12 yrs) 0.12 0.70 444 16 26,417 15,764.08 Secondary (13-17 yrs) 0.07 0.70 363 4 26,417 1,879.57 Adult (18-65 yrs) 0.62 0.70 1,194 1 26,417 13,689.29 Senior (more than 65 yrs) 0.12 0.70 66 4 26,417 585.93 Total 32,602.28 Data needs, assumptions, and limitations. A barrier effect analysis requires routinely available data. These data pertain to road systems (e.g., number of pedestrian crossings, AADT, average traffic flow speed, and vehicle mix), demographic characteristics of the served population, and land use patterns. Results and their presentation. The results of a barrier effect analysis are presented in terms of total units of disruption. The best use of this numerical result is to compare it with a parallel analysis of an upgraded roadway (or pedestrian facility) to see in fractional terms how much the amount of disruption per kilometer would change. Assessment. Barrier effect analysis was developed in Europe as a means of gauging the impediment to pedestrian and bicycle travel posed by an intervening roadway. It is especially useful in estimating how great a change in barrier effects would result from a proposed transportation system project. Two key assumptions contained in the analysis influence the outcome: the crossing potential factor (i.e., the relative likelihood of risk-taking by age group) and the utilization rate of signalized crossings. The latter factor, of course, can be varied by age group to reflect actual behavior. Best estimates of the two key assumed values by age group can be arrived at through observation, preferably at the actual site where a change in the transportation environment is being contemplated. 160

OCR for page 136
It would not be difficult to construct a spreadsheet to perform sensitivity analyses regarding the importance of assumed values on the actual estimates. This technique, coupled with user surveys, generally will allow good insight into the effects of a project on pedestrian and bicycle crossing behavior. The implications are considerable, in terms of both modal choice and safety. Method 9. User demand and evaluation surveys User demand and evaluation surveys are helpful in gathering information from consumers who may be inclined to use a particular transportation alternative. These surveys can also be used to obtain feedback on the specific barriers and problems facing people who currently walk or cycle on a particular facility or in a specific area. When to use. This method is appropriate when you need to identify specific attributes of roadways and their environs that make them especially conducive to travel by means other than the automobile. The National Highway Institute (1996, Chapter XVI.B) provides information on user surveys for evaluating bicycle and pedestrian conditions. Analysis. A crucial part of this analysis involves identifying specific problems that travelers encounter when walking and cycling, such as streets with inadequate sidewalks, roads with inadequate curb lane widths or shoulders, and dangerous railroad crossings. These problems can then be addressed during the design phase of transportation projects in the area. The following questions might be included in nonmotorized travel surveys: How much do you rely on walking and cycling for transportation and recreation? How do you rate walking and cycling conditions in the study area? What barriers, problems, and concerns do you have related to walking and cycling in the study area? What improvements or programs might improve walking and cycling conditions? For purposes of environmental justice assessment, it is necessary to collect information on the demographic characteristics of the survey respondents. Suggested questions are provided in Chapter 2. Data needs, assumptions, and limitations. User surveys can be distributed to walkers and cyclists at a study site (e.g., survey forms can be passed out along a sidewalk or trail), distributed through organizations (e.g., hiking and cycling clubs) and businesses (e.g., bicycle shops), or mailed to area residents. Note that in some circumstances results may be skewed by the fact that club members, people who frequent bicycle shops, and people most inclined to return surveys may not be representative of the entire user population. Pedestrian and bicycle travel surveys should attempt to gather the following information: Origin and destination of trips, including links by other modes (such as transit); 161