Freeway Weaving Sections: Comparison and Refinement of Design and Operations Analysis Procedures

Transcription

1 TRANSPOKIATION RESEARCH RECORD 1091 distributional assumptions on the model errors. The standard assumptions are that the dependent variable is normally distributed and that the errors are independent and have homogeneous variances. In this study, the possible consequences of overlooking the distributional assumptions about the error variances have been examined. The results of this limited study suggest that nonconstancy of the error variances (heteroscedasticity) may result in regression models that substantially underestimate left-tum departure headays. It is hoped that the discussion relating to model estimation and validation ill encourage others to address these basic issues in the literature. ACKNOWLEDGMENTS The authors ish to thank the folloing individuals for their contributions to the study: D. L. Pugh, J. dejong, D. A. Max- ell, L. J. Ringer, and A. M. Elmquist, all ith Texas A&M University; D. W. Hall, City of Austin; J. R. Black, City of College Station; and W. E. Rensch, City of Houston. The authors remain solely responsible for the contents of this paper. REFERENCES 1. P. V. Webster and B. M. Cobbe. Traffic Signals. Her Majesty's Stationery Office, London, England, R. W. Stokes. Saturation Flos of Exclusive Double Left-Turn Lanes. Ph.D. dissertation. Texas A&M University, College Station, SAS User's Guide: Statistics. SAS Institute, Inc., Cary, N.C., J. Neter, W. Wasserman, and M. H. Kutner. Applied Linear Reg res.don Models. Rio.hard D. Irin, Inc., Homeood, m., Publication of this paper sponsored by Committee on Highay Capacity and Quality of Service. 101 Freeay Weaving Sections: Comparison and Refinement of Design and Operations Analysis Procedures JOSEPH FAZIO AND NAGUI M. ROUPHAlL Weaving sections represent the common right-of-ay that occurs hen to or more crossing freeay traffic streams are traveling In the same general direction. In conjunction ith the development of the 1985 Highay Capacity Manual (HCM), several procedures have evolved for the purpose of updating, revising, and replacing the 1965 HCM procedure for design and operations analysis of freeay eaving sections. The objectives of this paper are tofold: to present and revie the latest three eaving procedures available to highay and traffic engineers, and to propose specific refinements to a simple eaving section procedure to account for the lane distribution of traffic upstream of the eaving section. These adjustments primarily Involve the development of a lane-shift variable, hich represents the average amount of peak-period passenger car lane shifts occurring under a given geometric configuration and prevailing traffic volumes. Statistical testing of the refined procedure against the three procedures at more Urban Transportation Center and CEMM Department, University of Illinois, Chicago, Ill than 50 sites nationide Indicated that the proposed procedure tends to predict observed average running eaving and noneaving speeds more closely than do the other procedures in most cases. A eaving section represents the physical space along a freeay here to (simple eaving) or more (multiple eaving) traffic streams traveling in the same general direction cross each other. Four basic movements are serviced in a simple eaving section, to eaving and to noneaving (outer flos), as indicated in Figure la. Weaving traffic originating from the freeay mainline is denoted V 2 and noneaving traffic is denoted V 1. Weaving traffic originating from the minor approach or entrance ramp is denoted V 3 and noneaving traffic is denoted V 4. The length of a eaving section (L) and the number of lanes (N) are the to design parameters that dictate the mode of traffic operation to be expected, as illustrated in Figure lb. (Note in this figure that Nb is the basic

2 102 TRANSPORTATION RESEARCH RECORD 1091 MAJOR 1 MAJOR/MINOR APPROACH 2~ EGRESS MINOR APPROACH 3 /ENTRANCE RAMP 4 a) MOVEMENT IDENTIFICATION b) CONFIGURATION MINOR/MAJOR EGRESS FIGURE 1 Simple freeay eaving section: movement identification and configuration. nnber of lanes on a major approach to a eaving section. The nnber of lanes entering the eaving section from the minor approach or entrance ramp is denoted Nr.) As Lor N decreases, drivers must execute their lane changes in a relatively short space, thus resulting in a general decrease in speed and level of service (LOS) for all traffic. In addition, lane changes originating from the outer lanes (i.e., median lane on the main approach and shoulder lane on the minor approach) ill tend to create increased disruption to traffic operations in the eaving section compared ith the situation resulting from a presegregated traffic stream, hen eaving traffic is essentially confined to the boundary lanes beteen the to traffic streams. Design procedures for eaving sections are aimed at determining the minimum length and nnber of lanes in the eaving section needed to meet a prespecified LOS in the analysis period. Operations analysis involves the determination of LOS and average running speeds for an existing eaving section. BACKGROUND One of the earliest eaving procedures for the design and operations analysis of eaving sections for the nation's first freeays appeared in the 1950 Highay Capacity Manual (1). The development of this procedure as based on field data collected at six eaving sites. In 1957, the 1950 HCM procedure as updated ith additional field data (2). A major data collection effort as undertaken by the Bureau of Public Roads in 1963, hich resulted in a ne eaving analysis procedure in the 1965 HCM (3). This methul has been idely used during the past to decades and constitutes the current state of practice for design and analysis of eaving sections. In an effort to keep abreast of changes in traffic composition and characteristics that took place since the Bureau of Public Roads data ere collected, Project 3-15, Weaving Area Operations Study, as initiated in 1969 through the National Cooperative Highay Research Program (4). This study, conducted by the Polytechnic Institute of Ne York (PINY), included field data collection at 17 northeastern sites. The results of Project 3-15 ultimately led to an interim eaving procedure published in 1980 (5). In an independent effort, a nomographic eaving procedure initially published in 1979 (6), as also included in the Interim Materials for Highay Capacity; this procedure as further modified in 1984 [see Leisch (7)). Recently, the FIIWA, U.S. Department of Transportation, initiated a study to evaluate the to procedures; the study resulted in yet another procedure for analying eaving sections [see JHK (8)). Based on the conclusions of the JHK report, the PINY procedure as revised and eventually adopted as the eaving procedure for the 1985 HCM (9). Since that time, both the Leisch and JHK procedures have undergone further revisions. The procedures revieed in this paper are the 1985 HCM (PINY) procedure, the revised JHK procedure (based on eaving study memoranda by W. Reilly and P Johnson, JHK and Associates, November 1984), and revised Leisch procedure (based on information letter from J. Leisch of J. Leisch and Associates, February 1985). COMPARISON OF WEAVING ANALYSIS PROCEDURES Tables 1 and 2 give summaries of the input requirements and output obtained for each of the three procedures. Of the three eaving procedures, the JHK procedure is the simplest to use. TABLE 1 COMPARISON OF INPUT REQUIREMENTS FOR THREE WEAVING PROCEDURES Method Configuration Nb N L V Y Y 2 V 4 JHK Leisch X 1985 HCM X x x x x x xx xx x x x x Note: N = number of lanes ithin eaving section, L = length of the eaving section measured from the point at hich the entrance gore is 2 ft ide to the point at hich the exit gore is 12 ft ide, V = total volume of traffic in the eaving section= V 1 + V 2 + V 3 + V 4, V 1 =volume of noneaving traffic stream originating from the major approach to the eaving section, V 2 = volume of eaving traffic stream originating from the major approach to the eaving section, V 3 = volume of eaving traffic stream originating from the entrance ramp or minor approach to the eaving section,v 4 =volume of noneaving traffic stream originating from the entrance ramp or minor approoch to the eaving section, V =volume of eaving traffic in the eaving section= V 2 + V 3, and VW2 =volume of smaller of the to eaving traffic streams [min (V 2, V 3 )]. In essence, this procedure utilies to equations for average running speeds, one for eaving, and the other for noneaving traffic, in Equations 1 and 2: x S = 15 + [50/(l + (2,000[1 + (Y4/V)]2.7 [1 + (V/V)]0.9 X nr11nj1.7\10.6/'ll.81 '\ l L 'llot"jj J/J SNW = 15 + (50/(1 + (100 [1 + (V 4 1V)f 4 x [1 + V!V)]l.8 [V/(QN)]0.9/Ll.SJ)] x 11\ VJ here Q is the heavy vehicle factor, and the other variables are as defined in Tables 1 and 2. To use the JHK equations, hourly volumes must be adjusted to passenger car equivalents via the heavy vehicle factor (Q). (2)

3 FAZ.IO AND ROUPHAIL 103 TABLE 2 COMPARISON OF OUTPUT GENERATED BY THREE WEAVING PROCEDURES Operation Method S SNW s LOS LOS NW LOSr N SF Mode JHK x x x x Leisch x x x x x 1985 HCM x x x x x x Note: S c average rurming speed of eaving traffic in the eaving section (mph), SNW =average running speed of noneaving traffic in the eaving section (mph), S = average running speed of all traffic in the eaving section (mph), LOS = level of service for eaving traffic, LOS NW= level of service for noneaving traffic, LOST = level of s~ice for all traffic ithin the eaving section, N =theoretical number of lanes used by eaving traffic in the eaving section, and SF = service Ho (pcphpl). After the average running eaving and noneaving speeds are calculated from Equations 1 and 2, eaving and noneaving levels of service are read out from appropriate tables. The Leisch procedure is nomograph-oriented, as shon in Figure 2. [Note in Figure 2 that R is the eaving ratio (V!V ). All other variables are defined elsehere in the paper.] To nomographs are used for one-sided eaving sections and to for to-sided sections. Configuration is accounted for in the procedure by (a) categoriing the eaving section as one sided or to sided, (b) specifying the presence or absence of lane balance (lane balance occurs hen the combined number of exit lanes on the freeay and ramp is one more than the number of lanes on the freeay ithin the eaving section), and (c) providing for an approximate reduction in traffic speeds hen the section configuration is concomitant ith an excessive amount of lane shifts. Peak-hour factor values are built into the procedure, thus requiring no adjustments for peak-hour flo, except for vehicle composition. The procedure derives the average running speed for eaving traffic and overall average running speed ithin the eaving section. Also determined by the procedure are service flo [service volume in passenger cars per hour per lane (pcphpl)], eaving intensity factor (k), LOS, and LOST as defined in Table 2. The 1985 HCM eaving procedure uses the folloing equation to estimate average running eaving and noneaving speeds. Sor SNW = 15 + [50/(1 + [a [1 + V!V)b V x (V!Nf/LdJ)] 2: ~ I I L v -+ FIGURE 2 Leisch procedure: nomograph outline. N (3) here a, b, c and d are calibration constants based on section configuration and type of operation. The method categories eaving sections into Types A, B, and C as a function of the minimum number of lane shifts performed by a driver in each of the to eaving traffic streams, as indicated in the HCM Weaving Chapter (9, Table 4-1). A key element in the HCM procedure is hether traffic operation is constrained or unconstrained. This is determined by comparing the number of lanes required for unconstrained operation, N. ith the theoretical value N (Max) [see HCM Table 4-4 (9)]. Weaving and noneaving speeds are then determined from Equation 3 based on configuration type and operation mode for the section under consideration. Finally, to levels of service are determined separately for eaving and noneaving traffic [see HCM Table 4-6 (9)]. Another important aspect of all procedures is the range of operating conditions in hich a solution can be found. Inherent limitations in the procedures include section geometry [i.e., maximum values for L, N, or Nb, section capacity V, SF, eaving frequencies VR (here VR is volume ratio= V!V), R (here R is eaving ratio= V 2 N). and running speeds S, SNW]. A comparison of the three procedures in that respect is given in Table 3. LANE SHIFT CONCEPT A noticeable difference beteen the JHK and HCM procedures is that the latter introduces the configuration of the eaving section into the speed equation (9). The param.eters a, b, c, and din the HCM method are determined in part from the minimum number of lane shifts required by a driver in each of the eaving; streams (as indicated in HCM Table 4-1), thus implying that eaving traffic is completely segregated on entering the eaving section. Field measurements collected recently indicate that eaving traffic is not fully segregated on entering the eaving section (based on information letter from Eric Ruehr, JHK and Associates, October 1984). For N 0 = 2, it as found that an average of 93.4 percenl of Movement 2 traffic entered by ay of Lane B (Figures la and lb), hile 6.6 percent entered via Lane C. For NB ;;?: 3, only 90.5 percent of Movement 2 traffic entered the eaving section by ay of Lane B, ith almost 10 percent of all traffic arriving in Lanes C and D. A negligible percentage of vehicles arrive in the outer lanes E, F, and so forth. A summary

4 104 TRANSPORTATION RESEARCH RECORD 1091 TABLE 3 COMPARISON OF PROCEDURE LIMITATIONS Procedure Parameter Limitalion Comments Leisch JHK 1985 HCM S1 SF S SNW L V VIN VR - s 55 mph - s 55 mph pcphpl 15 mph S 65 mph 15 mph SNW 65 mph s 4,000 ft Type A, 1,800 pcph Type B, 3,000 pcph Type C, 3,000 pcph 1,900 pcph (A, B, C) Type A: N = 2, 1.00 N= 3, 0.45 N= 4, 0.35 N= 5, 0.22 Type B: 0.80 Type C: 0.50 Inilial average running speed of eaving traffic out of realm of eaving Same as above, for final eaving speed Service flo beyond nomograph boundary Outside the realm of eaving Weaving section capacity Lane capacity Volume ratio limits R L Type A: 0.50 Type B: 0.50 Type C: 0.40 Type A: 2,000 ft Type:; B and C: 2,500 ft 15 S SNW 65 Weaving ratio limits Length out of realm of eaving of the observed distribution of Movement 2 vehicles is given in Table 4. As a logical extension of the results just given, an index as developed that talces into account the interaction of the folloing Weaving volumes V 2 and V 3, Distribution of V 2 and V 3 across lanes, and The minimum number of lane shifts by lane of entry. A lane shift multiplier has been developed that represents the minimum number of lane shifts that must be executed by the driver of a eaving vehicle from his lane of origin to the closest destination lane. This parameter can be determined directly from a sketch of the existing or proposed eaving section. To examples, ith balanced and imbalanced sections, hich demonstrate the computation of the lane shift multipliers A, B, C, and D are shon in Figures 3a and 3b, respectively. Note the folloing in Figure 3: A = lane shift multiplier for entering lane A (LS/veh); B = lane shift multiplier for entering lane B (LS/veh); C = lane shift multiplier for entering lane C (LS/veh); D = lane shift multiplier for entering lane D (LS/veh); From the previous analysis, the total number of pealc-hour lane shifts performed in the eaving section can be calculated When adjusted for variations in vehicle and driver population, peak-hour factor (PHF), and lateral clearances, the resulting TABLE 4 OBSERVED LANE DISTRIBUTION OF TRAFFIC UPSTREAM OF WEAVING SECTIONS JHK Percent Movement 2 Traffic in Indicated Lanea Site N11.=2 N =3 No. B c B c D b b Avg Source: Information letter from Eric Ruehr, IBK and Associates, October 8 See Figure l for lane designation. bmulljp!e observations per ilc. index, termed passenger car lane shifts per hour (pclsph) provides a means for integrating several operating parameters of the eaving section into a single variable. The index also avoids the artificial designation of eaving sections into Type A, B, and so forth; rather, it provides the traffic engineer ith a single numeric value that is indicative of the level of maneuvering difficulty encountered by all drivers in the eaving section.

5 FAZIO AND ROUPHAIL 105 MAJOR APPROACH ENTRANCE RAMP I L~ ---- C1\ ~ -D2----J ; ~~ 02 \ o;- - - ~ _.c-\-\--\ ~ (~ LANE BALANCE LANE SHIFT MUL TIPLIERI (Ll/veh.): A::=J, 1:1, C:2, D::3 S= 15 + {50/[(1 + {[l + cv 3 + v 4 )tv]3.o4s (V!N) 60s x (LSILJ "90 2 } )/ [1 + (LSJIV)]3 94] } (4) and SNW = 15 + { 50/(1 + ([[1 + (V JV}]S.08 [1 + (V!V)J2 019 x (V!Nf 523 }/( [1 (5) MAJOR APPROACH It should be noted that the calibration data set for the speed models consisted of 56 cases, including 35 sites from the Bureau of Public Roads study (3) and 6 from the JHK study [Reilly and Johnson (8), and eaving study memoranda by Reilly and Johnson, JHK and Associates, November 1984]. ENTRANCE RAMP PRELIMINARY MODEL EVALUATION FIGURE 3 multipliers. US) LANE IMBALANCE LA lltwt MUL Tl,.LIERI (LS/veh.l: A:1, 1:1, C::2, D::8 Examples of determining lane shift Equations for detennining the lane shift index are given in Table 5 for different configurations of eaving sections. Note the folloing in Table 5: LS= average number of lane shifts performed by the drivers of eaving vehicles = LS 2 + LS 3 [passenger car lane shifts per hour (pclsph)]; LS 2 = average number of lane shifts performed by the drivers of the Movement 2 vehicles (pclsph); and LS 3 =average amount of lane shifts performed by the drivers of the Movement 3 vehicles (pclsph). Initial testing of the lane shift index consisted of a correlation analysis beteen the index and average running eaving and noneaving speeds observed at six sites comprising a total of 12 cases. Examination of the data indicated an inverse relationship beteen the to parameters, as suspected. Further testing pertaining to the form of the variable indicated that LS!L and LS 3 /V correlated ell ith average running eaving speeds, hereas LS 3 /LS correlated ell ith average running noneaving speeds. The resulting speed models, hich represent an extension of the JHK and 1985 HCM models, are expressed by the folloing equations: To determine that the proposed speed models are an improvement over other procedures available, a comparison ith the JHK models as performed. This task required the recalibration of the JHK models using the set of 56 data points mentioned in the previous section. After both models ere calibrated, they ere utilied to predict average eaving and noneaving speeds in 11 validation cases [Reilly and Johnson (8), and eaving study memoranda by Reilly and Johnson, JHK and Associates, November 1984]. A simple regression model beteen field and predicted average running speeds as developed to test the predictive poer of each model. The results are presented in Table 6. As can be observed, the proposed model exhibited higher correlations ith observed eaving speeds compared ith the recalibrated JHK model. Both models exhibited modest correlations ith average noneaving speeds; the JHK procedure had a slight edge over the proposed model. COMPARISON OF THE FOUR PROCEDURES The proposed model has been expanded to a step-by-step procedure for the design and operations analysis of simple eaving sections. Details of the procedure may be found elsehere (10). In addition, an interactive, rnicrocomputerbased program has been developed that performs all of the calculations necessary to carry out the four procedures described in this paper (11). A set of 67 cases representing the full data base available to the research staff as processed TABLE S LANE SHIFf INDEX EQUATIONS 1 V 2 BJ(PHF "'fhv * f * fp) 2 (0.934V2'J VC)l(PHF "'fhv "'f "'fp) ~ 3 (0.905V 2 B V 2 C V 2 D)/(PHF * fhv "'f * fp) fhv = heavy vehicle adjustment factor; f = lateral clearance adjustment factor; fp = driver propulation Note: adjustment factor; and all other variables are as defined in the text or previous tables.

6 106 TABLE 6 FIELD VERSUS PREDICTED SPEEDS: RESULTS OF TWO MODELS Parameter Weaving speeds,a Slope Intercept Noneaving speeds r Slope Intercept 8 56 cases. bn cases. "Recalibrated model. dconelation coefficient. Data Set Type Calibration a Proposed JHKC Model Validationb Proposed JHK Model through the microcomputer program [Reilly and Johnson (8), and eaving study memoranda by Reilly and Johnson, JHK and Associates, November 1984]. A detailed breakdon of the study sites is given in Table 7. A comparative snmary of the results obtained is given in Table 8. As can be observed, the proposed procedure produced the highest correlation ith field eaving (r = 0.72) and noneaving speeds (r = 0.53). In contrast, the HCM procedure ranked last in correlation ith eaving (r = 0.56) as ell as noneaving speeds (r = 0.31). The JHK and Leisch procedures yielded almost identical correlations for bot.li speeds. Further- TRANSPORTATION RESEARCH RECORD 1091 more, the proposed procedure produced the loest intercept (for perfect correlation, intercept approaches ero) and second highest slope (for perfect correlation, slope approaches unity) compared ith the other three procedures. An assessment of the applicability of each procedure, is presented in Table 9. In this table, the number of sites for hich a solution could not be found as a result of inherent operational limitations in each procedure (given in Table 3) is listed for each procedure. Of the 67 cases making up the data base, the 1985 HCM procedure could only be applied to 39 cases. It appears that the limitation on eaving section capacity of 1,800 pcph for Type A configuration and 3,000 pcph for Types B and C resulted in the rejection of many sites in the data base. It is interesting to note that a solution that disregards these limitations produces an estimate of speeds that is in close agreement ith some of the field observations. To confirm that the majority of invalid cases are indeed reflective of the HCM eaving capacity limitations and not due to site anomalies such as excessive length and or number of lanes, the frequencies of all Type A sections ere compared ith the frequencies of those invalid Type A configuration cases in hich V exceeded 1,800 pcph. The compa_tisons ere made ith respect to length (Figure 4) and number of lanes ithin the eaving sections (Figure 5). In both instances, the invalid case frequencies closely parallel all Type A frequencies; in other ords, the frequencies of invalid cases did not progressively increase as length or number of lanes increased. Similar patterns ere observed ere observed hen the frequencies of Types Band C configurations ere compared. No such limitation problems ere encountere.d ith the other pro- TABLE 7 GEOGRAPHICAL DISTRIBUTION OF STUDY SITES Method(s) That No. of Used Case~) for No. of Location Cases a Calibration Sites Arlington, Virginia 2A L,H,P 2 la L Atlanta, Georgia lb J, p 2 lb L, I Boston, Massachusetts la 1 Chicago, Illinois 13A L,H,P 14 lb L, J la L Goanus Expressay, Ne York la 1 Long Island, Ne York 4A L,H,P 3 Los Angeles, California 6A L,H,P 6 Ne York, Ne York SA L,H,P 5 la San Diego, California 2A 2 San Francisco, California SA L,H,P 9 9B J, p Washington, D.C. 3A L,H,P 5 2B J, p White Plains, Ne York la 1 Yonkers, Ne York la _l Total Source: Reilly and Johnson (8) and eaving study memoranda by Reilly and Johnson, JHK and Associates, November A= pre-1970 data and B = post-1970 data. hr. = Leisch WCl!Ving proooc:lurc, H c 1985 HCM eaving procedure, J = JHK eaving procedure, and P = proposed eaving procedure.

7 FAZIO AND ROUPHAIL 107 TABLES SUMMARY EVALUATION OF FIELD VERSUS METHOD AVERAGE RUNNING SPEEDS Weavin~ Speed (meh) Noneaving Speed (mph) Standard Standard Absolute Standard Standard Absolute Inter- Deviation Deviation Mean Inter- Deviation Deviation Mean cept Field Method Difference cept Field Method Difference Method r Slope (mph) (mph) (mph) (mph) r Slope (mph) (mph) (mph) (mph) Proposed JHK Leischa HCM Noneaving speeds in this procedure cannot be directly estimated. Observed noneaving speeds are correlated ith overall speeds, as recommended by the author. TABLE 9 SUMMARY EVALUATION OF METHOD APPLICABILITY No. of No. of Invalid Cases Due Toa No of Valid V> SF> V> Out of L> Method Cases Cases 1,800 pcph 2,000 pcphpl VR > ,000 pcph Realmb R > 0.4 2,500 ft 4,000 ft VR > 0.5 Leisch JHK HCMC Proposedc Jn hich al lcul one constraint as violaled. bconsldered to be beyond the realm of eaving. cbased on S min peak Oo rales. cedures: the JHK procedure as applicable in 63 cases, the Leisch procedure in 65 cases, and the proposed procedure in all 67 cases making up the data base. FINAL NOTE ON THE ANALYSIS PERIOD Although the decision to calibrate speed models based on hourly or peak flo rates is highly controversial, a simple rule exists hen the final models are to be tested: follo the appropriate input requirements stipulated by the method. In this study, hourly volumes ere not adjusted for peak periods in either the Leisch or JHK procedure; the former procedure automatically performs PHF adjustments in the ~ > i (.) 20 I I / ~ 5 15 ff: 10 ~ 'l [.. " '1. J ~ TYPE A(V > 1800) l!ll TYPE ACALL) y ~ ~ 'I -, 1; "t, ~. ~ I '.. r 6 t~ 0 ~ ; ; f7l > LENGTH ( 102 FT) FIGURE 4 Distribution of section length: rejected HCM cases (V > 1,800 pcph) versus au Type A > 30 (.) :::> 0 a: u ,.:.TYPE A(V>1800) l'./atype ACALL) /.,._,_ NUMBER OF LANES (N) FIGURE 5 Distribution of No. of lanes: rejected HCM cases CV > 1,800 pcph) versus all Type A. nomographs, hereas the latter does not consider any automatic peak-period adjustments. The proposed procedure and HCM procedure require peak-period adjustments for 5- and 15- min peak flo rates, respectively. Hoever, due to the lack of 15-min data, both procedures ere tested based on 5-min peak flo rates. Although it is anticipated that some cases may no longer be invalid under the 15-min assumption, results indicated that the majority of cases that ere rejected under the original test V > 1,800 pcph for Type A, V > 3,000 pcph for Types Band C) remain invalid even hen a PHF of 1.0 is assumed (12 out of 18 cases). The true number of rejected cases ill probably range beteen 12 and 18. Even in the absence of such information, it is evident from Equation 3 that PHF (in the term VIN) does not signifi-

8 108 TRANSPOKfATION RESEARCH RECORD 1091 x a.. 70! CASE >- CASE 2 a.. If) so CJ CASE 3 > 40 3: CJ 30 ::::> 20 a: CJ 10 a: > PHF FIGURE 6 Sensitivity of HCM average eaving speed to peak-hour factor for N = 2. cantly affect the operation of the eaving section. Results for three cases from the data base plotted in Figures 6 and 7 for N = 2 (the most critical value for PHF) indicate little variation in predicted eaving and noneaving speeds in response to variations in the peak-hour factor. -::c a..! CASE1 a.. 60 If) CASE 2 CJ 50 > --- CASE3 40 3: 0 30 CJ 20 ::::> a: 10 CJ a: > PHF FIGURE 7 Sensltlvlty of HCM average noneavlng speed to peakhour factor for N = 2. CONCLUSIONS AND RECOMMENDATIONS This study as designed to investigate several freeay eaving analysis procedures that ere contemplated for the 1985 HCM. It has resulted in the development of a eaving procedure that is superior in predicting eaving and noneaving speeds compared ith existing procedures. The folloing conclusions are offered. The total number of lane shifts required by drivers in eaving sections affects both eaving and noneaving speeds. Negative correlations beteen speeds and lane shifts ere observed in the field. The inclusion of lane shift as an independent variable in average running eaving and noneaving speed models enhanced the predictive ability of the models considerably. When compared ith the latest procedures developed by JHK, Leisch, and the 1985 HCM (PINY), the proposed models yielded the highest correlations ith field eaving and noneaving speeds. The 1985 HCM procedure appears to be severely limited in its application; more than 41 percent of all cases analyed in this study did not meet the constraints stipulated by the method. The majority of the cases failed to satisfy the constraints on eaving section capacity; a majority of these cases ould still have been rejected even if hourly rates had been used instead of peak 5-min flo rates. A loer bound on the proportion of rejected cases is estimated at 33 percent in this study. Considerable research remains to be done in the area of freeay eaving sections design and analysis. Four recommendations follo. Fundamental ork on vehicle dynamics in freeay eaving sections is needed. The procedures described in this paper are primarily empirical (data based) and do not capture the essence of vehicle i...91.teraction and its L"llpact on average eaving and noneaving speeds. Microscopic simulation modeling, using Th!RAS (i2j or a similar package is recommended as a cost-effective tool for conducting such analyses. A persistent problem throughout this study as the inadequate sample sie of ne (post-1970) field data. The reliability of empirical procedures can be greatly enhanced ith additional data points for both calibration and validation purposes. Although t.li.e eaving procedure proposed in this paper has yielded superior results compared ith the other three procedures, it is recommended that all four procedures be tested to solve the same problem. The final design decision must still rest ith the engineer, ho may select the procedure yielding the most conservative design, average out all the results, and so forth. The interactive, microcomputer program developed in this study greatly simplifies this task (14 ). There is great need to tie in the safety characteristics of eaving sections (i.e., accident frequencies, type, location, and so forth) to the design and operations analysis procedures. This may result in defining loer bounds on section length and number of lanes based on accident experience. ACKNOWLEDGMENTS This study as sponsored by a grant from the National Highay Institute. The authors ish to thank Guido Radelat, contract manager, for his guidance and assistance throughout the study. The conclusions and recommendations presented here reflect solely the opinions of the authors a...-id not necessarily the opinions of the National Highay Institute or the Federal Highay Administration. REFERENCES 1. Highay Capacity Manual. U.S. Government Printing Office, Washington, D.C., 1950, pp Norman. Operation of Weaving Areas. HRB Bulletin167. HRB,

9 TRANSPORTATION RESEARCH RECORD National Research Council, Washington, D.C., 1957, pp f/rb Special Report 87: Highay Capacity Manual. HRB, National Research Council, Washington, D.C., 1965, pp Weaving Area Operations Study. NCHRP Project 3-15, Final Report. Department of Transportation Planning and Engineering, Polytechnic Institute of Brooklyn, Brooklyn, Ne York, 1971, unpublished. 5. Transportation Research Circular 212: Interim Materials on Highay Capacity. TRB, National Research Council, Washington, D.C., Jan. 1980, pp J. E. Leisch. A Ne Technique for Design and Analysis of Weaving Sections on Freeays. /TE Journal, Vol. 49, No. 3, March J. E. Leisch and J. P. Leisch. Procedure for Analysis and Design of Weaving Sections. Report FHWA-RD-82/54, FHWA, U.S. Department of Transportation, W. Reilly, H. Kell, and P. Johnson. Weaving Analysis Procedures for the Ne Highay Capacity Manual. JHK and Associates, Aug Special Report 209: Highay Capacity Manual. TRB, National Research Council, Washington, D.C., 1985, Chapter 4, pp. 4-1 to J. Faio. Development and Testing of A Weaving Operational Analysis and Design Procedure. M.S. thesis. University of Illinois at Chicago, Chicago, J. Faio et al. Users Guide 10 the Microcomputer Program Version of Four Weaving Operational Analysis and Design Procedures. 2nd ed. University of filinois at Chicago, Chicago, Aug A. G. Bullen and P. Athol. Development and Testing of JNTRAS, A Microscopic Freeay Simulatiori Model. University of Pittsburgh, Piusburgh, Feb. 1976, Vol. 2. Publication of this paper sponsored by Committee on Highay Capacity and Quality of Service. A Comparison of the 1985 Highay Capacity Manual and the Signal Operations Analysis Package 84 DANE!SMART The primary objective of this paper is to determine if the signalied intersection procedure as described in Chapter 9 of the 1985 Highay Capacity Manual (HCM) 1U give results consistent ith the microcomputer version of the SlgnaUed Operations Analysis Package 84 (SOAP 84). Each procedure as used to analye the lntersectjon In Chapter 9, Calculation 3, of the 1985 RCM. Average stopped delay as calculated for the intersection by each method and as used as the basis for comparing the 1985 HCM and SOAP 84. For through movements and protected- restricted left t urns, tbe to procedures produced slm.uar results for calculating stop delay, X ratios, and effectlve green ratios. Hoever, for the results to be consistent, the saturation fto as calculated by the HCM method must be used In SOAP 84 as the capacity (saturation fto) for through movements and the protected-restricted left turns. For protected- permissive and unprotected lell turns, the to methods produce significantly dlfferent results. Federal Highay Administration, U.S. Department of Transportation, HHP-21, 400 Seventh Street, S.W., Washington, D.C Described is an effort to compare the microcomputer version of SOAP 84 ith the methodology in Chapter 9, Signalied Intersections, of the 1985 Highay Capacity Manual (HCM). The Signal Operations Analysis Package (SOAP 84) is a computeried method for developing control plans and evaluating the operations of individual signalied intersections. As the basis for the comparison beteen SOAP 84 and the 1985 HCM, delay ill be calculated by each method. SOAP 84 determines average delay, hich includes delay incurred during deceleration and acceleration as ell as stop delay. The 1985 HCM calculates average stop delay as the basis for determining level of service. To make a comparison beteen the to methods, average delay ill be converted to average stop delay by using the folloing formula (1): Average delay/1.3 = average stop delay (1) As the first step in the analysis, Calculation 3: Operational Analysis of a Multiphase Actuated Signal from Chapter 9 of