Multimodal Level of Service Analysis for Urban Streets (2008)

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.

86 8.1 Model Development Two basic forms were considered for the pedestrian LOS for arterials model. The first was an aggregate model that used the outputs from existing segment and intersection LOS models to determine the arterial LOS. The other was an agglomerate model that considered the independent charac- teristics of the roadway/walkway environment to calculate an arterial LOS for pedestrians directly. Both were preliminarily evaluated during model development. The aggregate model was chosen for refinement for several reasons. The stepwise approach to an aggregate model is use- ful because it allows the practitioner to evaluate the effect of improvements at individual intersections or along specific segments on the overall LOS of the facility. The aggregate model also retains all the terms found both intuitively and mathematically validated to be significant to pedestrians walk- ing within an urban environment. The agglomerate models form was tested during our preliminary models and did not retain all the terms as significant. Consequently, we focused on the aggregate model in our model development efforts. We considered various functional techniques for model de- velopment, including linear regression and ordered probit. We performed linear regression modeling because it is more intu- itive than probit modeling in practice and non-modelers better understand the sensitivity of the regression model. These rea- sons are particularly important in that these models are most frequently used: the development or analysis of specific design options or in the development of pedestrian facility community master plans with presentations to interested citizens and pub- lic officials. To ensure the validity of the results of the linear re- gression modeling results, we evaluated the ordered probit model form as well. The results of both the linear regression and ordered probit modeling efforts are described below. For both modeling efforts the dependent variable, Observed pedestrian LOS, was defined as the score that a par- ticipant assigned to a specific video clip. The scores were on a scale of A (best) through F (worst). For modeling purposes, the letter grades were converted to numerical scores: A=1, B=2, C=3, D=4, E=5, and F=6. Before starting correlations analysis and modeling, we cre- ated two data subsets from the overall dataset. The total dataset was sorted by city and LOS grade responses. A ran- dom sampling of 20% of the data representing each city and LOS grade response was taken from the overall dataset for model validation. The balance of the data, 80% of the total dataset, was used for model development. We used SPSS 14.0 to conduct Pearson correlation analysis on the extensive array of geometric and operational variables. Subsequently, we selected the following relevant variables for additional testing: • Segment LOS—The pedestrian LOS for roadway segments (see below). • Intersection LOS—The pedestrian LOS for signalized intersections (see below). • Midblock Crossing LOS—The LOS associated with mid- block crossings (see below). • Total Pedestrians—The total number of pedestrians encountered in the video clip; a measure of pedestrian space, which is an input to the existing pedestrian LOS methodology in the HCM. • Conflicts per mile—The total conflicts per mile represent the motor vehicle conflicts resulting from motorists turn- ing across the pedestrian facility at unsignalized locations. • Size of the city in which the data collection took place— The MSA population was used to represent the size of each city. The panel asked that MSA be dropped from further con- sideration as a variable. Other variables were dropped from further consideration because of their poor correlation with C H A P T E R 8 Pedestrian LOS Model

87 LOS Minimum Pedestrian Space Per Person Equivalent Maximum Flow Rate per Unit Width of Sidewalk A > 60 SF per person ≤ 300 peds/hr/ft B >40 ≤ 420 C >24 ≤ 600 D >15 ≤ 900 E >8 ≤ 1380 F ≤ 8 SF > 1380 Source: Exhibit 18-3 HCM 2000 [110] Exhibit 93. Pedestrian Walkway LOS (Density). the dependent variable or because of their colinearity with more strongly correlated variables. Also, variables such as traffic volume, sidewalk width, and signal delay, are compo- nents of the segment LOS or the intersection LOS, so we did not model them independently. Several variables were evaluated for inclusion as additional terms in the model. Frequency of unsignalized conflicts (in- tersections and driveways) per mile was tested for its correla- tion and significance to the arterial LOS for pedestrians and was not found to be a significant factor. Additionally, the density of pedestrians on the sidewalk (the current HCM measure of LOS) was not found to be significant for this model, within the low range of density values available in the video clips. 8.2 Recommended Pedestrian LOS Model The proposed pedestrian level of service predicts the mean level of service that would be reported by pedestrians along or across the urban street. The average pedestrian LOS for the urban street facility is a function of the segment level of serv- ice, the intersection level of service, and the mid-block cross- ing difficulty. Overall Pedestrian LOS Model The overall pedestrian level of service for an urban street is based on a combination of pedestrian density and other fac- tors. The level of service according to density is computed. Then the pedestrian LOS according to other factors is com- puted. The final level of service for the facility is the worse of the two computed levels of service. Ped LOS = Worse of (Pedestrian Density LOS, Ped Other LOS) (Eq. 33) Where Ped LOS = The letter grade level of service for the urban street combining density and other factors. Ped Density LOS = The letter grade level of service for side- walks, walkways, and street corners based on density Ped Other LOS = The letter grade level of service for the urban street based on factors other than density Pedestrian Density LOS Model for Sidewalks, Walkways, Street Corners The methods of Chapter 18 of the HCM are used to com- pute the pedestrian density for the sidewalks and the pedes- trian waiting areas at signalized intersection street corners. The LOS thresholds given in that chapter for these facilities are used to determine the level of service. The thresholds for sidewalks and walkways are given in Exhibit 93. Pedestrian Other LOS Model The pedestrian LOS for the facility that is representative of non-density factors is computed according to either of the two models below: Pedestrian Other LOS Model 1 OtherPLOS (#1) = (0.318 PSeg + 0.220 PInt + 1.606) * (RCDF) (Eq. 34) Pedestrian Other LOS Model 2 OtherPLOS (#2) = (0.45 PSeg + 0.30 PInt + 1.30) * (RCDF) (Eq. 35) Where OtherPLOS = Pedestrian non-density (other factors) LOS PSeg = Pedestrian segment LOS value PInt = Pedestrian intersection LOS value RCDF = Roadway crossing difficulty factor The first model provides the better statistical fit with the video lab data. However, this model does not produce LOS F for the streets in the video clip data set. The second model is a manual modification of the parameters of the first model so that the second model will produce a full range of LOS A to F for the streets in the video clip data set. The constant was

88 manually adjusted downward and the other parameters were adjusted upward until the second model produced LOS F for at least one of the streets in the data set. Although none of the video clips actually produced a LOS A or F rating (on average) from the video lab participants, the second model was developed to address potential public agency acceptance issues that might arise with adopting the rst LOS model for pedestrians that might not produce LOS A and LOS F for at least some streets in the jurisdiction. The second model produces a full range of LOS A to F results for a reasonable range of street conditions typical of urban areas of the United States. The output of both of these models is a numerical value, which must be translated to a LOS letter grade. Exhibit 94 pro- vides the numerical ranges that coincide with each LOS letter grade. These thresholds are the same as for the other modes. Pedestrian Segment LOS The segment pedestrian LOS is calculated according to the following widely used equation [111]: Ped Seg LOSS = −1.2276 ln (W ol + W l + fp × %OSP + fb × × W b + fsw × Ws) + 0.0091(Vol15/L) + 0.0004 SPD2 + 6.0468 (Eq. 36) Where Ped Seg LOSS = Pedestrian level of service score for a segment ln = Natural log W ol = Width of outside lane W l = Width of shoulder or bicycle lane fp = On-street parking eect coefcient (= 0.20) %OSP = Percent of segment with on-street parking fb = Buer area coefcient (= 5.37 for trees spaced 20 feet on center) W b = Buer width (distance between edge of pave- ment and sidewalk, in feet) fsw = Sidewalk presence coefcient (= 6 − 0.3Ws) W s = Width of sidewalk Vol 15 = Volume of motorized vehicles in the peak 15 minute period L = Total number of directional through lanes SPD = Average running speed of motorized vehicle trafc (mi/h) Pedestrian Intersection LOS The intersection LOS for pedestrians is computed only for signalized intersections according to the following equation developed by Petritsch et al. [112]: Ped Int LOS (Signal) = 0.00569(RTOR +PermLefts) + 0.00013(PerpTrafVol*PerpTrafSpeed) + 0.681(LanesCrossed 0.514) + 0.0401ln(PedDelay) −RTCI(0.0027PerpTrafVol − 0.1946) + 0.5997 (Eq. 37) Where RTOR +PermLefts= Sum of the number of right- turn-on-red vehicles and the number of motorists making a permitted left turn in a 15-minute period PerpTrafVol*PerpTrafSpeed = Product of the trafc in the outside through lane of the street being crossed and the midblock 85th percentile speed of trafc on the street being crossed in a 15-minute period LanesCrossed= The number of lanes being crossed by the pedestrian PedDelay= Average number of seconds the pedestrian is delayed be- fore being able to cross the intersection RTCI = Number of right turn chan- melization islands on the crossing. Pedestrian Midblock Crossing Factor The pedestrian Roadway Crossing Difculty Factor (RCDF) measures the difculty of crossing the street between signalized intersections. The RCDF worsens the pedestrian LOS if the crossing difculty is worse than the non-crossing LOS for the facility. It improves the pedestrian LOS if the crossing difculty LOS is better than the non-crossing dif- culty LOS. The factor is based on the numerical dierence be- tween the crossing LOS and the non-crossing LOS. The pedestrian RCDF is limited to a maximum of 1.20 and a min- imum of 0.80. RCDF = Max[0.80, Min{[(XLOS#-NXLOS#)/7.5 + )83 .qE(]}02.1,]00.1 Where RCDF = Roadway crossing difculty factor XLOS# = Roadway crossing difculty LOS Number NXLOS# = Non-crossing Pedestrian LOS number = (0.318 PSeg + 0.220 PInt + 1.606) LOS Numerical Score A ≤ 2.00 B >2.00 and ≤ 2.75 C >2.75 and ≤ 3.50 D >3.50 and ≤ 4.25 E >4.25 and ≤ 5.00 F > 5.00 Exhibit 94. Pedestrian “Other” Model LOS Categories. PLOS = -1.2276 ln (fLV x Wt + 0.5Wl + fp x %OSP + fb x Wb + fsw x Ws) + 0.0091 (V/(4*PHF*L)) + 0.0004 SPD2 + 6.0468 (Eq.36) Pedestrian level of service score for a segment Natural log Low volume factor (=1.00 unless average annual daily trac (AADT) is less than or equal to 4,000, in which case fLV =(2 - 0.00025 * AADT) total width of outside lane (and shoulder) pavement Width of shoulder or bicycle lane, or, if there is un-striped parking and %OSP=25 then Wl=10ft to account for lateral isplac ment of trac On-street parking eect coecient (=0.50) Percent of segment with on-street parking Buer area coecient 5.37 for a y continuous barrier at least 3 feet high separating walkway from motor vehicle trac. A discontinuous barrier (e.g. trees, bollards, etc.) can be considered a continuous barrier if they are at least 3 feet high and are spaced 20 feet on center or less. Buer width (distance betwe n edge of pavement and sidewalk, in feet) Sidewalk presence coecient(fsw=6-0.3Ws if Ws=10, otherwise fsw = 3.00) Width of widewalk For widths greater than 10 feet, use 10 feet. Directional volume of motorized vehicles in the direction closest to the pedestrian (vph) Peak hour factor Total number of through lanes for direction of trac closest to pedestrians. Average running speed of motorized vehicle trac (m/h) Ped SegLOS = ln = fLV = Wt = Wl = fp = %OSP = fb = = Wb = fsw = Ws = V = PHF = L = SPD = Where

89 Pseg = Ped. Segment LOS number (computed per equation #20) Pint = Ped. Intersection LOS number (computed per equation #21) The crossing difficulty LOS number is computed based on the minimum of the waiting-for-a-gap LOS number and di- verting-to-a-signal LOS number. XLOS = Min [WaitForGap, DivertToSignal] (Eq. 39) Where XLOS = Crossing LOS score (based on Exhibit 96) WaitForGap = Delay waiting for safe gap to cross. DivertToSignal = Delay diverting to nearest signalized inter- section to cross. The delay is converted into a LOS numerical score based on the minimum of the mean delay waiting for a gap or di- verting to a signal, according to the values given in Exhibit 95. Wait-For-Gap LOS Calculation The Wait-For-Gap LOS is computed based on the ex- pected waiting time required to find an acceptable gap in the traffic to cross the street. The acceptable gap is computed as a function of the number of lanes, their width, and the aver- age pedestrian walking speed, with 2 seconds added. Acceptable Gap = (Number of Lanes * 12 feet/lane) / 3.5 feet/second + 2 seconds (Eq. 40) The expected waiting time until an acceptable gap becomes available is computed as follows: (Eq. 41) Where t = The acceptable gap plus the time it takes for a vehicle to pass by the pedestrian. The average pass-by time = Average Vehicle Length/ Average Speed, converted to seconds. λ = The average vehicle flow rate in vehicles per second. Exp = The exponential function MeanWait t t= ( )−[ ]−1 1λ λexp Using the numerical cutoffs shown in Exhibit 96 the final numerical score is then interpolated between the cutoff val- ues based on the probability of obtaining an adequate gap within the allowed time. For this calculation, the increasing LOS numerical score is assumed to become logarithmic beyond LOS F. Divert To Signal LOS The LOS rating for diverting to the nearest traffic signal to cross the street is computed as a function of the extra delay in- volved in walking to and from the mid-block crossing point to the nearest signal and the delay waiting to cross at the signal. The geometric delay associated with a pedestrian deviation is the amount of time it takes the pedestrian to walk to a con- trolled crossing and back. To calculate this delay, one must first determine the distance to the nearest crossing. For this methodology, this was assumed as one third of the block length. This distance is then divided by the pedestrian’s walk- ing speed (assumed to be 3.5 feet/second) to obtain the geo- metric delay: Ped Geometric Delay = 2/3 * (Block Length)/ Ped Walking Speed (Eq. 42) The control delay at the intersection is calculated as shown in the HCM [113]: Ped Control Delay = (Cycle Length − Green Time)2/ (2*Cycle Length) (Eq. 43) The total delay is the sum of the two: Total Ped Deviation Delay = Ped Geometric Delay + Ped Cycle Delay (Eq. 44) Minimum of Wait or Divert Delay (Seconds) XLOS Score 10 1 20 2 30 3 40 4 60 5 > 60 6 Exhibit 95. Pedestrian Crossing LOS Score. Pedestrian LOS Delay Threshold Seconds Equivalent LOS Numerical Score Range Equivalent LOS Midpoint Score A 10 1.5 1 B 20 >1.5 and 2.5 2 C 30 > 2.5 and 3.5 3 D 40 > 3.5 and 4.5 4 E 60 > 4.5 and 5.5 5 F > 60 > 5.5 6 For this calculation, the increasing LOS numerical score is assumed to become logarithmic beyond LOS F. Exhibit 96. Pedestrian LOS and Delay Thresholds.

90 Clip Location Sidewalk Width (ft) Pedestrian Flow Rate (pph) Outside Lane (ft) Shoulder Width (ft) On-Street Parking (%) Barrier (Y/N) Buffer Width (ft) Dir. Vol. (vph) Traffic Lanes (lanes) Traffic Speed (mph) Video LOS HCM LOS Model #1 LOS Model #2 LOS 215 7th Ave at 15th St, N side 8 60 12 0 50% Yes 7 170 1 25 B E B B 227 Grant Ave at California St, E side 6 200 16 0 0% Yes 4 630 2 30 B B C C 230 3rd St at Mission St, E side 6 220 12 0 0% No 5 220 2 30 B D C D 221 Stockton St at Washington St, E side 4 640 16 0 0% Yes 3 0 1 30 B E B B 224 Grant Ave at Jackson St, E side 4 1320 12 0 100% Yes 2 80 1 30 B E B B 228 Post St at Stockton St, S side 6 180 10 0 40% Yes 1 370 1 30 B D B C 226 Geary Blvd at Divisadero St, S side 9 190 20 0 50% Yes 5 1180 2 40 B D D D 232 Hillsborough, Arm. to Tamp., N side 6 0 16 4 0% No 0 540 1 45 B B D D 229 Post St at Stockton St, S side 6 280 10 0 40% Yes 0 310 1 30 B D B B 205 Alumni Dr at Magnolia Dr, N side 10 0 12 4 0% No 10 200 2 30 C B C C 211 Bearss Ave at North Blvd, N side 4 0 12 0 0% No 5 570 1 45 C A B C 214 Dale Mabry at Tampa Bay, E side 9.5 0 12 5 0% No 35 2030 3 45 C E D E 225 Geary Blvd at Divisadero St, S side 9 280 20 0 50% Yes 5 1050 2 40 C C D D 218 Market St at Kearney St, N side 15 340 12 0 0% No 12 60 1 30 C C B B 222 Stockton St at Broadway St, E side 6 610 16 0 50% Yes 3 220 2 30 C E C C 219 Stockton St at Clay St, E side 7 640 16 0 100% Yes 4 150 1 30 C E B B 220 Stockton St at Clay St, E side 7 820 16 0 100% Yes 4 150 1 30 C D B B 223 Stockton St at Broadway St, E side 6 1600 16 0 50% Yes 3 0 2 30 C D A A 210 Magnolia Dr at Holly Dr, W side 0 0 12 0 0% No 0 160 2 30 C C C C 216 21st St at 7th Ave, W side 6 0 12 0 0% No 0 360 1 30 C A C C 217 21st St at 7th Ave, W side 6 0 12 0 0% No 0 300 1 30 C E B C 203 Collins Blvd at Alumni Dr, E side 10 0 12 4 0% No 15 270 2 30 D D C C 204 Collins Blvd at Alumni Dr, E side 10 0 12 4 0% No 15 160 2 30 D E B C 231 Dale Mabry, State to Carmen, W side 5 0 12 0 0% No 6 570 1 35 D B D D 201 Holly Dr at Magnolia Dr, N side 0 0 10 0 0% No 0 270 2 20 D E D E 209 Fletcher at Bruce B Downs, S side 0 0 12 4 0% No 0 2170 4 45 D A E F 206 Fowler Ave at 56th St, S side 5 0 12 5 0% No 23 1690 4 50 E E D E 208 Fletcher at Bruce B Downs, S side 0 30 12 4 0% No 0 1750 4 45 E C E F % Exact Match to Video Rating 100% 25% 43% 43% % Within 1 LOS of Video Rating 100% 43% 86% 79% Notes: On-Street Parking = Percent of on-street parking lane occupied by parked vehicles. Barrier is presence of trees, or other barrier between pedestrian sidewalk and street. Traffic lanes is number of lanes in direction of travel closest to pedestrian. Video LOS is the mean of the letter grade LOS ratings reported by subjects in video lab. Model LOS is the LOS grade predicted by the proposed pedestrian LOS model. Exhibit 97. Evaluation of Proposed Pedestrian Model and HCM Against Video Lab Results.

The total delay is then converted into a numerical LOS score by linearly interpolating numerical scores on the scale provided in Exhibit 96. 8.3 Performance Evaluation of Pedestrian LOS Model Exhibit 97 compares the performance of the proposed pedestrian LOS model (with the mid-block crossing factor) to the mean LOS rating for each pedestrian video clip. The video clips did not expose lab subjects to any arterial mid- block crossing situations. Although the HCM reproduces the mean video lab ratings for each video clip 25% of the time, the two proposed pedestrian LOS models (1 and 2) both re- produce the mean video clip ratings 43% of the time. The dif- ference is that Model 2 produces LOS A to F results for the streets in the video clip data set. Model 1 produces LOS A to E results for the same streets. 91