Update of Highway Capacity Manual: Merge, Diverge, and Weaving Methodologies (2023)

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.

Introduction This document is the final report of NCHRP Project 07-26, “Update of Highway Capacity Manual: Merge, Diverge, and Weaving Methodologies.” The project focused on updating the methodologies in the Highway Capacity Manual (HCM) Chapter 13: “Freeway Weaving Segments” and Chapter 14: “Freeway Merge and Diverge Segments” for forecasting the opera- tion of merges, diverges, and weaves along general-purpose freeway lanes. In addition to this report, the project produced four new draft chapters for the HCM. • Revised HCM, Volume 2, Chapter 13 – “Freeway Weaving Segments.” • Revised HCM, Volume 2, Chapter 14 – “Freeway Merge and Diverge Segments.” • Revised HCM, Volume 4 (online), Chapter 27 – “Freeway Weaving: Supplemental.” • Revised HCM, Volume 4 (online), Chapter  28 – “Freeway Merges and Diverges: Supplemental.” In addition to the revised HCM chapters, three other products were developed during this project, including the following: • NCHRP Web-Only Document 343: Traffic Modeling Document containing supplemental details on the collected data and modeling results. • Spreadsheet-based, computational engines implementing the proposed methods. • Presentation providing a comprehensive summary of the completed research and the pro- posed enhancements to methods in the HCM. It is intended for a 6-hour training session that could be delivered virtually or in person. NCHRP Web-Only Document 343, the draft revisions to the HCM chapters, and the products associated with this report can be found on the National Academies Press website (nap.nationalacademies.org) by searching for NCHRP Research Report 1038. Background and Motivation The HCM’s procedures for merge and diverge segments were developed by NCHRP Project 03-37, “Capacity and Level of Service at Ramp-Freeway Junctions,” which was com- pleted in 1993. The HCM’s procedures for weaving segments were developed by NCHRP Project 03-75, “Analysis of Freeway Weaving Sections,” with data collection completed in 2008. Both studies, as well as studies conducted by other research projects, were limited by the amount of data that could be reduced within a given budget using the technology avail- able at the time. For example, the HCM weaving methods are based on 13 hours of manually reduced data collected at 14 sites, using a combination of video from fixed-wing aircraft 1   S U M M A R Y Update of Highway Capacity Manual: Merge, Diverge, and Weaving Methodologies 16497-01_Summary-Ch01-3rdPgs.indd 1 5/4/23 3:01 PM

2 Update of Highway Capacity Manual : Merge, Diverge, and Weaving Methodologies (10 sites), ground- or pole-based video (two sites), and detailed data available from FHWA’s Next Generation Simulation (NGSIM) effort (two sites) (Roess 2008). The HCM merge and diverge methods were developed from a larger number of sites (42 to 57 single-lane merge sites and 16 to 18 single-lane diverge sites, depending on the model) but are also based on only approximately 1 hour of data per site (Roess and Ulerio 1994). As a result of these limitations, the current HCM’s methods that existed before this research have several shortcomings. Three big-picture issues include: • Because the HCM basic segment, merge and diverge, and weaving methods were devel- oped through different research efforts, there are discontinuities between the methods that cause illogical results. For example, adding an auxiliary lane between an on-ramp and off-ramp to form a weave has been shown to add capacity, but HCM methods can estimate that a weave has a lower capacity than a merge (Elefteriadou, Kondyli, and St. George 2014; Stanek 2014). Similarly, at new zero ramp-volumes, merge, diverge, and weaving segments should operate similarly to basic segments, but HCM methods may not predict this result. • The HCM’s merge and diverge methodologies, by providing independent and distinct regression equations for predicting flow, speed, and density, do not conform to the funda- mental relationship of traffic flow, namely that flow is the product of speed and density. • The datasets used to develop the HCM merge, diverge, and weaving methods do not include or include very few examples of several common ramp configurations. These configurations include two-lane on- and off-ramps, ramps with lane adds or drops, and complex weaves. These shortcomings are confirmed in the literature, with several sources detailing issues and concerns with the HCM merge, diverge, and weaving methods, including: • The HCM merge method considerably overestimates capacity (Elefteriadou, Kondyli, and St. George 2014; Kondyli, Gubbala, and Elefteriadou 2016; Morris et al. 2011). • The HCM weaving method underestimates speeds in the 50 to 65 mph range (Xu et al. 2020). • The HCM weaving method includes weaving length as an input, but performance mea- sure outputs are insensitive to weaving length for practical purposes (Ahmed et al. 2019). • The HCM weaving method underestimates capacities at high weaving ratios (Skabardonis and Mauch 2014). • The HCM weaving method can indicate that volumes are less than capacity but produce densities greater than the density at capacity. Across the largest U.S. dataset included in the literature review, the method overestimated density by an average of 22% (Skabardonis and Mauch 2014). • The results of the HCM merge, diverge, and weaving methods do not use the same method- ology or model form and therefore can result in inconsistent and counterintuitive results when comparing different segment types. Research Approach NCHRP 07-26 was developed to address the identified shortcomings of the HCM’s merge, diverge, and weaving methods. The project’s objectives were twofold: develop updated methods based on new data collection and pilot the updated methods to demonstrate the full range of their applicability. This project was able to take advantage of the ubiquitous sensor coverage of urban free- ways and probe vehicle coverage of entire roadway networks that have developed over the

Summary 3 last decade. Using sensor-based data sources, this project was able to gather data efficiently from over 120 freeway bottlenecks, generally covering a full year of observation. In total, this project was able to generate over 3 million data points (defined as a 15-minute observa- tion of speed, flow, and density) that were used to calibrate and validate the recommended models. Using probe data, the project was able to estimate weaving traffic flow propor- tions efficiently—a task that has been a considerable data-reduction and budget challenge in previous research efforts. Capacity was determined by following the HCM’s procedure for determining capacity from sensor data, except when the procedure produced unreason- able results due to limited data or a small number of breakdown events. A breakdown event is defined as a notable and sustained reduction in vehicle speed in the segment, typically resulting from the vehicular demand exceeding the available capacity and a queue forming upstream of the bottleneck. The recommended models developed by this project were designed to address the three big-picture issues of the existing HCM methods identified earlier in this summary. • Recommended models are consistent with each other and converge to the operation of a basic freeway segment at low demand flows. • Recommended models conform to the fundamental relationship of speed, flow, and density. • Recommended models have been calibrated and validated to a larger range of ramp con- figurations than were used to develop the current HCM models. The project team drafted updated versions of HCM Chapters 13 and 14 to integrate the draft recommended methods, along with updated versions of the example problems con- tained in HCM Chapters 27 and 28 demonstrating the methods. The methods were also integrated into computational engines to facilitate testing the new procedures. The proposed models and HCM chapters were applied in two pilot implementation test- ing efforts with state departments of transportation (DOTs), as well as presented to the TRB Committee on Highway Capacity and Quality of Service, the steward of the HCM. Based on the pilot testing results, the models and draft HCM chapters were revised into the versions presented with this report. Findings The data collection and analysis effort resulted in the following key findings: • Density at Capacity – Breakdown of merge, diverge, and weaving segments is expected to occur when the average density of all vehicles in the segment exceeds 35 passenger cars per mile per lane (pc/mi/ln). This value represents an average condition based on observed breakdown densities, with some sites showing breakdown at higher or lower density values. This value is substantially lower than the HCM’s current values of 45 pc/mi/ln for merge and diverge segments and 43 pc/mi/ln for weaving segments. • Merges – The HCM considerably overestimates the capacity of simple merges (that is, a single- lane merge without a lane add) and consistently and considerably overestimates the capacity of complex merges (that is, all other merge configurations). – An increase in the acceleration lane length increases a merge segment’s speed and capacity substantially when the acceleration lane is less than 500 ft long. Speed and capacity increase more gradually at lengths between 500 ft and 1,500 ft, while addi- tional length over 1,500 ft provides minimal additional improvement.

4 Update of Highway Capacity Manual : Merge, Diverge, and Weaving Methodologies – The average speed and the capacity of a merge segment decrease as the percentage of segment volume consisting of merging traffic increases. Accordingly, the capacity of a merge segment may change by the time of day as the volume of merging traffic fluctuates. • Diverges – The HCM considerably overestimates the capacity of simple diverges (that is, a single- lane diverge without a lane drop) and consistently and considerably overestimates the capacity of complex diverges (that is, all other diverge configurations). – A diverge segment’s speed and capacity are greater than that of an equivalent merge segment at short deceleration lane lengths, but merge and diverge segments operate similarly when acceleration and deceleration lane lengths exceed 1,000 ft. – Increasing the deceleration lane length above 500  ft provides minimal additional improvement in diverge segment performance. – The average speed and the capacity of a diverge segment decrease as the percentage of segment volume consisting of diverging traffic increases. Accordingly, the capacity of a diverge segment may change by time of day as the volume of diverging traffic fluctuates. • Weaving – The HCM model consistently underestimates the average speed of simple weaves (that is, a single-lane on-ramp connected to a single-lane off-ramp by an auxiliary lane). – Weaving segment speed and capacity are relatively insensitive to the volume ratio (the proportion of segment volume consisting of weaving traffic), showing slight down- ward trends with increasing volume ratio. Complex weaves (that is, weaves where at least one ramp has more than one lane) are somewhat more sensitive to the volume ratio than are simple weaves. – Weaving segment speed and capacity are sensitive to the weaving short length, with the greatest reductions occurring with short lengths less than 1,000 ft; some reductions occurring with short lengths less than 3,000 ft; and little effect above 3,000 ft. The modeling calibration and validation effort resulted in the following key findings: • Merges – The recommended speed model has a similar root mean square error (RMSE) as the HCM for simple merges (3.93 versus 3.81 mph), with the caveat that the HCM estima- tion error would be higher if the volume in the right two lanes was estimated using the HCM method rather than being field measured. – The recommended speed model overestimates speeds for complex merges. No com- parison to the HCM was possible due to a lack of field data for the volume in the right two lanes at these sites. – Compared to the HCM, the recommended model’s capacity estimates for simple merges are much closer to the field-measured values. – The recommended model does a better job than the current HCM at estimating the capacity of complex merges but still tends to slightly overestimate capacity. • Diverges – The recommended speed model has a better RMSE than the HCM for simple diverges (4.04 versus 6.12 mph). The HCM estimation error would be higher if the volume in the right two lanes was estimated using the HCM method rather than being field measured. – The recommended speed model overestimates speeds for complex diverges but to a lesser degree than for complex merges. No comparison to the HCM was possible due to a lack of field data for the volume in the right two lanes at these sites. – The recommended model produces capacity estimates much closer to the field-measured values than does the HCM for simple diverges. – The recommended model does a better job than the current HCM at estimating the capacity of complex diverges but still tends to slightly overestimate capacity.

Summary 5 • Weaving – The recommended speed model has a much lower RMSE for both simple weaves (the most common type of weave) and complex weaves (3.22 and 4.86 mph, respectively), compared to the HCM model (12.92 mph). The recommended model’s estimates are relatively evenly distributed above and below the field-measured speed. – The recommended capacity model produced considerably better estimates of capacity than the HCM for three out of four complex weaves. At the fourth site, the field-measured capacity exceeded that of a basic freeway segment. – The recommended model produced similar estimates of capacity for simple weaves as the current HCM model. The recommended model removed the HCM’s maximum weaving demand flow rate criterion to prevent inconsistencies under boundary condi- tions; for sites with a high weaving volume ratio, the removal of this criterion resulted in the recommended model overestimating capacity. Results The general speed model form for all three segment types shares the same consistent approach in which the average speed of an equivalent basic freeway segment (without ramp turbulence effects) is reduced to account for impedances due to the presence of merging, diverging, or weaving maneuvers. The model formulation is consistent with the fundamental speed–flow–capacity relationship, while providing the necessary sensitivity to factors that influence the average speed of merge, diverge, and weaving segments. The recommended models’ speeds converge to basic freeway segment speeds at low demand volumes, and the models avoid discontinuities at boundary locations. The recommended methods are simpler to apply than the current HCM methods, requiring just four calculation steps: (1) estimating and adjusting volumes within the segment, (2) esti- mating the segment’s average speed, (3) estimating the segment’s capacity, and (4) estimating the segment’s density and level of service. The recommended models require fewer inputs than the HCM models, use inputs that are more likely to be available to practitioners, and are sensitive to the inputs included in the model. With few exceptions, the recommended speed and capacity models developed by NCHRP 07-26 outperformed their HCM counterpart models. Furthermore, the models’ forms generally apply to the more complex segment configurations, including complex weaves, two-lane merges and diverges, and closely spaced merges and diverges. The recom- mended models are based on larger datasets than those used to develop the HCM models or used in any prior research effort. While the methods developed through this effort are considered a notable improvement over the current HCM methods, it is emphasized that the methods remain estimates for average capacity and operations of the respective segment types. The data collection observed and the report documented significant site-to-site variability, suggesting capacity variability between sites. This variability is likely attributable to differences in driver behavior, the effects of pavement conditions, and other site-specific nuances not observed or measured during the project. Practitioners should be aware of this uncertainty when using these methods to conduct an operational or design analysis. This project produced draft HCM chapters to incorporate the recommended models into the HCM and to demonstrate their application. Before the chapters can become an official part of the HCM, they will need to be reviewed and approved by the TRB Committee on Highway Capacity and Quality of Service and subsequently published by TRB in a future version of the HCM (for example, Version 7.1). To assist with this review and sub sequent use, the project developed computational engines to help users perform the required calculations.