acceleration models

Published on February 2017 | Categories: Documents | Downloads: 29 | Comments: 0 | Views: 266
of 96
Download PDF   Embed   Report

Comments

Content

 

Technical Report Documentation Documentation Page

1. Report No.

 2  2.. G Go overnment Accession No.

3. Re Recipient's Catalog No.

SWUTC/96/465100-1  4. Title and Subtitle

 5. Report Date

Using Acceleration Characteristics in Air Quality and Energy Consumption Analyses

August 1996

 7. Author(s)

 8. Performing Organization Report No.

 William L. Eisele, Shawn M. Turner, and Robert J. Benz

465100-1

9. Performing Organization Name and Address Address

10. Work U Unit nit No.

Texas Transportation Tra nsportation Institute Institute The Texas A&M University System Collegee Station, Texas 77843-3135 Colleg 7784 3-3135

11. Contract or Grant No.

0079

12. Sponsoring A Agency gency Name and A Address ddress

13. Type of Re Report port and Period Covered

Southwest Region University Transportation Center  Texas Transportation Tra nsportation Institute Institute The Texas A&M University System Collegee Station, Texas 77843-3135 Colleg 7784 3-3135

 6. Performing Organization Code

14. Sponsoring Agency Code

15. Supplementary Notes

Supported by a grant from the Office of the Governor of the state of T Texas, exas, Energy Office 16. Abstract

This research investigated investigat ed the effects effects of detailed speed and acceleration characteristics chara cteristics on energy consumption uti utilizing lizing several fuel consumption models. The relationships between spee speed d and acceleration characteristics, geometric characteristics (e.g., number of lanes, signal density, driveway driveway density), and traffic traffic flow variability variabil ity for various roadways were were also investigated. Finally, distributions distribut ions were produced that summarize the operating characteristics of freeway and arterial street facilities in the Houston, Texas T exas area. Data for the study stu dy were collected on a second-by-s second-by-second econd basis on selecte selected d freeways and arterial streets in Houston, Texas using an electronic distance-measuring instrument (DMI) and the floating car technique. The study found that fuel consumption models incorporating detailed speed and acceleration characteristics provide statistically statisti cally different results. Similar results were obtained for both arterial and ffreeway reeway roadways. Low coefficients coefficients of  2 determination (i.e., R   less than 0.35) were found when regressing geometric characteristics with the speed and acceleration characteristics characteristi cs such as average speed speed or acceleration noise. Relationshi Relationships ps between the coefficient coefficient of variation of speed or  2 acceleration noise with average speed provided much higher R   values when investigating the traffic flow variability of the travel time runs. These results were similar similar for peak and off-pe off-peak ak conditions and the different roadway classifications (e.g., arterials and freeways). The distributions distribut ions of operating characteristics for Houston, Texas summarize summari ze the percent percent of time vehicles are operating within a given speed speed and acceleration range. This data is expected to be invaluable for individual individualss desiring the operational characteristics of the Houston roadway system, or similar large urban area, as well as those individuals who can apply this information to future or current mobile source emissions and energy consumption modeling applications. 18. Distribution Statement

17. Key Words

Fuel consumption models, emissions models, acceleration characteristics, travel time variability

19. Security Classif.(of this report)

Unclassified

  Form DOT F 1700.7 (8-72)

 No restrictions restrictions.. This document document is available to the  public through through NTIS:  National Technical Technical Informatio Information n Service Service 5285 Port Royal Road Springfield, Springf ield, Virginia 22161 221 61

20. Security Classif.(of this page)

Unclassified

Reproduction of completed page authorized

21. No. o off Pages

96

22. Price

 

ii

 

USING ACCELERATION CHARACTERISTICS IN AIR QUALITY AND ENERGY CONSUMPTION ANALYSES ANALYSES

by

William L. Eisele Assistant Research Scientist Shawn M. Turner Assistant Research Scientist and Robert J. Benz Assistant Research Scientist

Technical Report 465100-1

Sponsored by The Office of the Governor of the State of Texas, Energy Office Southwest Region University Transportation Center Texas Transportation Institute The Texas A&M University System College Station, TX 77843-3135

August 1996

 

iv

 

ABSTRACT

This research investigated the effects of detailed speed and acceleration characteristics on energy consumption utilizing several fuel consumption models. The relationships between speed and acceleration accelerati on characteristics, geometric characteristics (e.g., number of la lanes, nes, signal signal density, driveway density), and traffic tra ffic flow variability for various roadways were also inves investigated. tigated. Finally, distributions were produced that summarize the operating characteristics of freeways and arterial streets in the Houston, Texas T exas area. Data for the study were collected collected on a second-by-seco second-by-second nd basis on selected selected freeways and arterial streets in Houston, Texas using an electronic distance-measuring instrument (DMI) and the floating car technique. The study found that fuel consumption models incorporating detailed speed and acceleration characteristics chara cteristics provide statistically different results. Similar results were obtained for both arterial and freeway freeway roadways. Low coefficients of determination determinat ion (i.e., R 2 less than 0.35) 0.35 ) were found when regressing geometric characteristics with the speed and acceleration characteristics such as average speed or acceleration acceler ation noise. Relationships Relat ionships between the coefficient of variation varia tion of speed or  2 acceleration noise with average speed provided much higher R   valu  values es when investigating investigating the traffic flow variability of the travel time runs. These results were similar for peak and off-peak off-peak conditions and the different roadway classifications (e.g., arterials and freeways). The distributions of operating characteristics characteristics for Houston, Texas summarize the percent of  time vehicles are operating within a given speed and acceleration range. This data is expected to be invaluable for individuals desiring the operational characteristics of the Houston roadway system, or  similar large urban area, as well as those individuals who can apply this information to future or  current mobile source emissions and energy consumption modeling applications.  

v

 

DISCLAIMER 

The contents of this report reflect the views of the authors who are responsible for the opinions, findings, and conclusions conclusion s presented herein. The contents do not necessarily reflect the official views or policies of the Southwest Region University Transportation Tra nsportation Center (SWUTC). (SWUTC) . This report does not constitute constitute a standard, specification, or regulation. Any reference reference to commercial commercial software packages or hardware is for explanatory purposes only and does not constitute an endorsement.

vi

 

ACKNOWLEDGMENTS

This publication was developed as a part of the University Transportation Centers Program which is funded 50 percent in oil overcharge funds from the Stripper Well Settlement as a s provided by the State of Texas Governor’s Energy Office and approved by the US Department of Energy. Mention of trade names or commercial products does not constitute endorsement or recommendation for use. The authors would like to thank Dr. George Dresser for his insight at critical points in the  project. In addition, the authors would like to acknowledge acknowledge the following following individuals individuals for their  assistance: Luke Albert - data reduction; Brett Baker - travel time runs; Pat Beck - graphics; David Berry - data analyses and computer programming; Monye Brookover - travel time runs; Ryan Christianson - data analyses and computer programming; Ken Clark - travel time runs; Mark Coscio - travel time runs; Jim Cullison - travel time runs and quality control of data collection; Kim Duren - data reduction; David Fenno - travel time runs; Chris Hallin - travel time runs; Monty Poppe - data analyses and computer programming; Jordan Richard - graphics; Troy Rother - data reduction and final report editing; Woodraylyn Smith - travel time runs; John Vaughn - travel time runs; Tony Voight - travel time runs; Kathy Williams - final report editing; and Steve Wohlschlaeger - travel time runs.

vii vii

 

viii

 

EXECUTIVE SUMMARY Introduction

Current mobile source emissions and energy consumption analyses are based on average vehicular speeds over roadway sections that are typically greater than 1 mile (0.6 km) in length. Recent research has ha s indicated that the fluctuation in speed (i.e., acceleration and deceleration) is more important than the average speed in determining mobile source emissions and energy consumption. This fluctuation in speed, known as acceleration noise, has not yet been effectively utilized in vehicle emissions and fuel consumption analyses because of 1) 1 ) the difficulty of collecting or estimating speed data for very short time or distance intervals, and 2) the absence of appropriate computer models to conduct such analyses. Study Objectives and Scope

The primary objective of this study is to characterize the speed and acceleration accelera tion characteristics characteri stics of a wide range of traffic flow. flow. Data were collected with a DMI using short increments of time. Researchers made a preliminary investigation of the effects of detailed speed and acceleration data on existing fuel consumption models. Comparisons of ffuel uel consumption estimates were made using speed and acceleration calculations based upon a segment-wide method (average method) and a second-by-second second-by-s econd method (instantaneous (instantaneous method). Detailed acceleration characteristics characteristics could be incorporated into the next generation of mobile source emissions and energy consumption models. Models that incorporate acceleration characteristi chara cteristics cs are expected to provide more accurate accura te estimates of mobile source emissions and energy consumption and of the changes in the emissions and energy consumption associated associated with various transportation tra nsportation projects and programs. The second objective of this study is to examine relationships between the geometric characteristics, speed and acceleration characteristics, and traffic flow variability for the different roadway functional functional classes. classes. These regressio regression n equations will be based upon data that are disaggregated by functional roadway type (e.g., arterial Class I or II, freeways). The final objective is to compile a reliable data set that describes the speed and acceleration characteristics chara cteristics of various variou s roadways and operating conditions. The data set produced from this project can be supplied to interested individuals or organizations for use in development and/or validation of fuel consumption and/or emissions modeling. Overview of the Study Design

To accomplish these objectives, data were collected with a distance measuring instrument (DMI) on a total of 233 centerline-miles (375 km) of freeway routes and 198 centerline-miles (319 km) of arterial routes. From the sspeed peed information information provided provided by the DMI, further further speed and acceleration characteristics characteristics were calculated. In addition, geometric characteristics characteristics were collected collected along the corridors.

ix

 

Once the data were collected with the DMI, three distinct modules of information were created. The first data set includes the speed and acceleration chara characteristics cteristics computed from the DMI files for each corridor. The second data set includes the results of utilizing both an instanta instantaneous neous and average calculation of several fuel consumption model estimates based upon speed and acceleration rat rate. e. Such analyses ana lyses are imperative to show any differences differences in such models when collection of vehicle vehicle speeds is allowed as often often as every seco second. nd. The last data set that is merged with with the others is the geometric characteristics characteristics that were collected along the the travel time routes. The data were then combined, summarized appropriately, appropria tely, and statistical ana analyses lyses were were performed in order to evaluate evaluat e the objectives of the study. Findings

 Fuel Consumption Model Analyses Since the difference between the average and instantaneous methods of fuel consumption estimation for several models was was desired, a paired t-test was utilized. T-tests were performed on the different arterial classes (e.g., Class I, Class II, and freeways) at the aggregated level (i.e., not disaggregated disaggreg ated by average averag e speed, for example). The null hypothesis for the tests is that there is no difference between the two methods of calculating calcu lating fuel consumption. Therefore, if significance is found, the null hypothesis can be rejected and there is a difference between the two methods of fuel consumption estimation. A critical level of significance of 5 percent was used in the analyses to determine significance. significance. Results of the analyses are shown in Table 3. Some of the findings findings from from the fuel consumption analyses are discussed below. Raus’ model did not yield significant differences differences in fuel consumption estimation for aany ny of the functional classes. cla sses. Although the model is not intended for freeways, freeways, and indeed normality was not found for that condition, the arterial classes cla sses yielded insignificant insignificant results as well. The FREQ10 FREQ1 0 models for freeways and arterials were both found to be insignificant insignificant for the Class I arterials. The final model that demonstrates insignificant results is McGill for for the Class II arterials. arteria ls. It should be noted that insignificance insigni ficance was only found for a few situations. situa tions. Furthermore, Furth ermore, some models demonstrate significance for operating conditions for which the model does not have speed data. It is important to note that the analyses presented here compare only the average and instantaneous instantaneo us methods of fuel consumption consumption estimation for for any one model. It does n not ot compare the models to one another, nor is it possible from this analyses to determine that one model is better than tha n another model. However, one can determine which models demonstrate a significant difference difference that can be attributed a ttributed to the detailed data set produced by perform performing ing travel time runs with the D DMI. MI.  Regression and Correlation Analyses Models relating geometric characteristics and acceleration characteristics were developed  based upon peak and off-peak off-peak conditions, differen differentt functional classific classifications, ations, and stratifications stratifications on some variables (e.g., for driveway driveway densities greater than 20 per mile). The linear regression models models

x

 

generally produced low coefficients of determination (R 2). The highe highest st R 2 values achieved were less than 0.35 for any of the conditions studied. This model was of often ten between the dependent variable varia ble of average speed and the independent variables of signal density and/or driveway density for the arterial sections in either either the peak or off-peak conditions. The addition of independent variables after  signal density and driveway density (i.e., producing graphs with greater than three independent variables) often resulted in increasing the R 2  value only a few hundredths. Analysis of variance (ANOVA) procedures were performed as part of the linear regression using a critical level of  significance signif icance of 5 percent to determine the significance significance of the independent variables in the models. The research team hypothesized that the driveway density and signal density variables would have the most explanatory power in such relationships and provide higher R 2 results than those observed. For nearly all of the models developed, developed, the signal density and driveway density variables were found to be significant significant in the ANOVA procedure. Therefore, these variables were contributing contribu ting to the explanation explanat ion of the variance within the model. model. It is interesting interesting to note that the 24 hour volume and the length of the section, the only independent variables used in the freeway analysis, were not always significant. signifi cant. The 24 hour hou r volume produced significant results results more often, however, than the variable representing the length of the section. This would indicate that the 24 hour volume was critical in many cases in explaining the variance in the freeway segments with the variables available. Another observation observa tion that was made from evaluating evalua ting the resulting models was the ssigns igns on the coefficients of the independent variables. Often times, these signs signs did not make intuitive sense. For  example, as the signal density went down, the coefficient of variation of the speed (CV) would go up. In this example, it does not make sense that the variation vari ation of the traffic speeds, represented by the CV, should go up when there is less interruption in the traffic stream (i.e., a lower signal density). However, Howeve r, it is possible that this indicates along these arterial corridors that the signal timing has been optimized to provide sufficient green time and increased average speeds. Many observations can be made with regard to the traffic flow variability linear regression results. The relationship between average speed and the coefficient coefficient of variation varia tion provided relatively 2 high R    values for for all functional functional roadway classes, classes, peak, and off-peak off-peak conditions. conditions. Further, relationships utilizing average speed to predict the acceleration noise produced relatively lower R 2 results. Although acceleration noise is a better measure of the traff traffic ic variability over a travel time run 2 than average speed, the lower R  values determined for this relationship indicate a significant amount of traffic operation that is unexplained by aggregating the instantaneous readings from the DMI. The next portion of the analysis focused on investigating the value of CV, average speed, relevant geometric characteristics, chara cteristics, and the speed profile at a given speed to realize any possible trends that may ma y exist. The most interesting characteristic of these analyses was realizing the im importance portance of  the location of the travel time run section that is being investigated. Theoretically, one could have sections placed such that the CV could be just about any value (i.e., located loca ted anywhere along the speed  profile).  profil e). Due to the inherent variability in these relationships, relationships, and the geome geometric tric versus speed speed and acceleration accelerat ion characteristics, developing developing estimating regression equations is difficult. Similar results were found for other roadway classes and conditions (e.g., peak and off-peak).

xi

 

Since the geometric characteristics (e.g., number of lanes, signal density) do not change for  a given roadway section, it is possible to aggregate aggrega te the resulting speed and acceleration cha characteristics racteristics together toge ther for these travel time runs. This was performed and regression equations were produced and the results are shown in Table 5. The R 2 values in Table 5 are very similar, or slightly higher, than those produced when each travel time run was plotted. This was expected since since it produces a graph with fewer points that are aggregated closer to the regression line.  Roadway Operating Characteristics: Characteristics: Speed and Acceleration The travel time and speed data collected for this study were summarized to obtain speed and acceleration distributions. These speed and acceleration distributions provide quantitative information about the operating characteristics characteristics of the freeways freeways and arterial streets streets under study. study. These distributions are also very important in designing and validating the next generation of emissions models that are based upon acceleration patterns, not average speeds. The speed distributions for different functional classes cla sses were were markedly markedl y different, with freeways exhibiting higher speeds and arterial arter ial streets exhibiting lower speeds speeds and more idle time. The data for  off-peak period conditions (mid-day) were also examined, and found to be similar to peak period conditions. condition s. Although the researchers had hypothesized that a significant difference difference would would exist  between peak and off-peak period operating characteristics  between characteristics,, the examination of speed speed distributions distributions was unable to confirm the hypothesis. The acceleration distributions for different functional classes where different but not necessarily neces sarily distinctive. distinctive. The floating car method of data collection may have affe affected cted the true acceleration characteristics of different roadway types, thereby smoothing the potential acceleration/decelera accelera tion/deceleration tion differences differences between freeways freeways and arterial arteria l streets. streets. The similarity of the distributions for different functional classes may also al so indicate that, indeed, only small difference exist  between  betwee n acceleration characteristics characteristics for differe different nt functional roadway classes classes.. The three-dimensional speed-acceleration speed-acceleration distribution distribu tion for all freeway freeway and arterial arteria l street routes shows a large “peak” of the data at 60 mph (97 kph), with another smaller “peak” at 0 mph (steadystate). state The acceleration accelera tion and deceleration ranges close to 0 mph/s mph/sec ec can also be seen on the figure as sm small all).“ridges.” The three-dimensional speed-acceleration distribution for freeway routes only show a large  proportion of travel that occurs in the 55 5 5 to 60 6 0 mph (89 (8 9 to 97 9 7 kph) range with a small range in accelerations. accelera tions. Figures 31 and a nd 32 show the speed-acceleration distributions for for Class I and II arterials, arteria ls, respectively. Like the speed distributions distribu tions discussed earlier, earlie r, there is a marked difference between functional classes. Class I and II arterial streets show smaller but comparable compara ble speed “peaks” at 0 mph, or idle time.

xii xii

 

Conclusions and Recommendations

 Fuel Consumption Model Comparisons Significance was determined for many of the functional classes when comparing fuel consumption estimation estimation based upon both the average and instantaneous methods. From these results results it can be concluded that, in general, significant differences can be expected when applying a detailed data set such as that produced by a DMI in a travel time run to the estimation of fuel consumption. It is important to note that when reviewing reviewing the results of Table 3, it is imperative to st study udy Table T able 2 to verify the conditions (e.g., speed range, functional classification) for which a model is valid.  Regression and Correlation Analyses Development of regression equations between speed and acceleration characteristics, geometric characteristics, chara cteristics, and traffic flow variability varia bility was performed in the study. The regression 2 equations did not yield an R   higher than 0.35 when comparing any combination of the geometric characteristics chara cteristics with the speed speed and acceleration characteristics. Signal density and/or driveway dens density ity were found to be significant for most of the conditions evalua evaluated ted with the aid of ANOVA procedures using a critical level of significance of 5 percent. Several factors factors that could account for the findings were considered. considered. The true affe a ffect ct of the driveway density may not be reflected in the travel time data since the floating car method was utilized. It is possible possible that the influence of driveways on the right-mos right-mostt lane may not be included into a travel time run that tha t includes a driver passing as many vehic vehicles les as pass the d driver. river. In addition, travel variability variabili ty induced by traffic signals signals is difficult difficult to quantify. qua ntify. Peak and a nd off-peak off-peak conditions often have different signal signal timings to optimize traffic flow. Average speeds, and motorist delay, will vary depending upon when motorists motorists arrive at the traffic ssignal. ignal. The location of the travel time run section was also found to be of importance when measuring measur ing the coefficient of variation varia tion of the speed. speed. If a travel time run is performed immediately prior to a traffic signal or lane-drop on a freeway, the results willl differ compared to a run performed in an uninterr wil uninterrupted upted flow section. section. Unfortunately, Unfortuna tely, the data  base did not contain contain a variable relating relating to the the section section defini definition tion (e.g., (e.g., before before or after after a traffic traffic signal) signal) of the that, travelalthough time run, but this would interesting element for for further further study. Finally, it was found acceleration noise be is aanbetter measure to determine the operating characteristics of a section than average speed, there is still a significant portion of the instantaneous travel characteristics (e.g., speed, speed, acceleration) that tha t are lost when aggregating over an entire section.  Roadway Operating Characteristics: Characteristics: Speed and Acceleration The travel tra vel time/speed data collected for this study showed a significant difference di fference in the speed distributions for different different ffunctional unctional classes (e.g., freeways, freeways, Class I arterials, Class Cla ss II arterials). The acceleration distributions for different roadway functional classes were less distinctive between functional classes, indicating that acceleration characteristics were similar between freeways and arteriall streets. The floating car data collection technique used in this sstudy arteria tudy may have “smoothe “smoothed” d”

xiii

 

some of the acceleration differences between freeways and arterials streets, so a definitive statement cannot be made. A data collection that obtains a more representative representative sample of the range of  operating characteristics of motorists (e.g., instrumenting random vehicles) would likely provide a more distinct difference between functional classes. The study also produced three-dimensional speed-acceleration distributions distribu tions that were typical of the freeway and arterial street system in Houston, Texas. Texa s. The speed-acceleration distributio distributions ns do exhibit significant differences between freeways and arterial streets, mainly with respect to speed differences diff erences.. The speed and acceleration data set used to produce the these se summary distributions distributions is expected to be useful in validating the next generation of emissions models that are currently in the developmental stages.  DMI Technology for the Data Collection Effort  The distance measuring instrument was found to be an invaluable tool for performing this study. The instantaneous instantaneou s data points provided at every 0.5 second yielded a data set that allows for  detailed speed and acceleration information. From this data, data , the significance significance of the instantaneous data set on estimating fuel consumption estimates could be evaluated, regression equations were evaluated, and traffic operating distributions distributions could be prepared. The ASCII format of the output was easily manipulated for analyses easily analyses and evaluation. Data collection methods methods that produce produce these instantaneous speed and acceleration data will continue to prove to be useful in the transportation community for application to many transportation concerns (e.g., air quality, traffic operations).  Future Research Needs The study identified some areas where additional research is needed. The first is the need for  for  the development of mobile source source emissions models that can incorpor incorporate ate acceleration accelera tion characteristics. chara cteristics. Research of this kind is currently in progress. There is a need for for better characterization characteriza tion of acceleration accelera tion characteristics character istics for for different roadway roadwa y facilities.. Characterizing facilities Chara cterizing acceleration characteristics by percent of time in a particular particula r driving condition (e.g.,that idle,replicate cruise, acceleration, a cceleration, or deceleration) is u usef seful ul for the development development of appropriate a ppropriate driving cycles these conditions. There is much variability both along a travel time run and between travel time runs along sections. Additional research is needed that focuses on determining appropriate appropria te methods to quantify this variability in a consistent consistent and meaningful manner (e.g., separate the driver and a nd traffic influences). influences). In general, the DMI and similar technologies for data collection, allow for larger amounts of  descriptive descrip tive data that has not been possible in the past. Research must now begin to focus on  performance  perform ance measures that are best utilized (e.g., coefficient coefficient of variation) for quantifying the aggregation of this data for transportation-related concerns concerns..

xiv

 

TABLE OF CONTENTS

Page ABSTRACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v DISCLAIMER . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi ACKNOWLEDGMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii EXECUT IVE SUMMARY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix TABLE OF CONTENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvii LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xix CHAPTER I. INT RODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Study Objectives and Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Organization of Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 CHAPT ER II. BACKGROUND . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Acceleration Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Acceleration Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Quality of Flow Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Total To tal Abso Absolute lute Sec Second ond-t -too-Sec Secon ond d Dif Differ eren ence cess iin n Spee Speed d Per Per Mile Mile (TAD (TAD)) . . . . . . 7 Other Acceleration Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Data Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Graphical Representation of Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Driving Cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 C rrceenrtnD rivbionugt C s P. . .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. Cu on sA thyeclFeT Fuel Consumption Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Emissions Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Grap Gr aph hical Rep Repre ressentat atiion of of Fue Fuel Co Consump umptio tion an and Em Emissions Dat Data . . . . . . . . . . . . . . . Summary of Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

xv

13 2 1 13 15 16 17

 

TABLE OF CONTENTS (continued)

Page CHAPT ER III. ST STUDY DESIGN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Overview of Study Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Data Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 Thee Houst Th uston Metropolitan Area and and the Data Collected . . . . . . . . . . . . . . . . . . 21 Roadway Ge Geometric Ch Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 DMI Data Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 Development of Data Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 Speed and Acceleration Characteristics Module . . . . . . . . . . . . . . . . . . . . . . . . 28 Fuel Consumption Module . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 Geometric Characteristics Module . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 Levels of Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 Quality Control Measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 Initial Examination of the Data Set . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 Fuel Consumption Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 Regression and Correlation Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 Data ata Bas asee of Useful Emissions Modeling Information . . . . . . . . . . . . . . . . . . . . 35 CHAPTER IV. FINDINGS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Fuel Consumption Model Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Regression and Correlation Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Roadw adway Operat atiion Char arac actteristics: Speed and and Accelerat atiion . . . . . . . . . . . . . . . . . . . . Speed Di Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acceleration Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-Dimensional Speed-Acceleration Distributions . . . . . . . . . . . . . . . . . . . . . . . .

37 37 39 49 49 54 59

CHAPT ER V. CONCLUSIONS AND RECOMMENDATIONS . . . . . . . . . . . . . . . . . . . . . . 69 Fuel Consumption Model Comparisons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 Regression and Correlation Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 R adw ad ati arac act stoicllse:cStipoened an ati 71 0 DoM Iw Taey chO nopleorgat yiofn orCthhear D atteariCo C Efand fodrt A. c.c.e.le.r.at . i.o.n. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. 7 Future Research Needs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

xvi

 

LIST OF FIGURES

FIGU FIGURE RE 1. FIGURE 2. FIGURE 3. 3. FIGURE 4. FIGURE 5. FIGU FIGURE RE 6. FIGUR GURE 7. 7. FIGURE FIGU RE 8. FIGURE 9. 9. FIGURE 10. 10. FIGURE 11. FIGURE 12. FIGURE FIGU RE 13. FIGURE FIGU RE 14.

Page Exam Examp ple Spe Speed Prof rofile ile for Kat aty y Fre Freeway (I (I--10) 10) in Hous Houstton, Tex Texas . . . . . . . . . . . 9 Vehicle Speed Distributions for Three Data Sets . . . . . . . . . . . . . . . . . . . . . . . . . 9 Cumulative D Diistribution of of Acceleration Va Values . . . . . . . . . . . . . . . . . . . . . . . . 10 Frequency Bar Graph of Acceleration Values . . . . . . . . . . . . . . . . . . . . . . . . . . 10 Graph of Speed and Acceleration Noise about the Mean . . . . . . . . . . . . . . . . . . 11 3-Di 3-Dim men enssiona ionall Grap Graph h Com Compa pari ring ng Spee Speed, d, Acc ccel eler erat atio ion, n, an and d Fr Freequen quency cy . . . . . . . 11 Grap aph hical Representati ation of of Speed Pr Profile an and Em Emission Rat Ratees . . . . . . . . . . . . . 16 Overall Overall Appro Approach ach for for Usin Using g Acce Acceler leration ation Characteri Characterist stics ics in Air Air Quality and Energy Consumption Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 Sample Ro Roadway In Inventory Fi Field Da Data Co Collection Sh Sheet . . . . . . . . . . . . . . . . . . 22 Trav Traveel Tim Time Route utes (Houst uston, Tex Texas as)) Used for the Study udy . . . . . . . . . . . . . . . . 23 Example of DMIREAD Input Screen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 Example of Output from DMIREAD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 Relations Relationship hip Betwee Between n Average Average Speed Speed and CV for for Freeway Freeway Section Sectionss During Pe Peak Pe Periods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 Relations Relationship hip Betwee Between n Average Average Speed Speed and Accelerati Acceleration on Noise Noise for

Freeway Sections During Peak Periods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . FIGURE FIGU RE 15. Relations Relationship hip Betwee Between n Average Average Speed Speed and Accelerati Acceleration on Noise Noise for Class I Arterials During Off-Peak Pe Periods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . FIGURE FIGU RE 16. Relations Relationship hip Betwee Between n Average Average Speed Speed and the the Standard Standard Deviation Deviation of Speed for Class I Ar Arterials During Off-Peak P Peeriods . . . . . . . . . . . . . . . . . . . . . FIGURE FIGU RE 17. Relations Relationship hip Betwee Between n Average Average Speed Speed and CV for for Class II Arter Arterials ials During Pe Peak Pe Periods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . FIGURE FIGU RE 18. Relations Relationship hip Betwee Between n Average Average Speed Speed and Accelerati Acceleration on Noise Noise for Class II Arterials During Peak Periods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . FIGURE FIGU RE 19. Relations Relationship hip Betwee Between n Average Average Speed Speed and CV for for Freeway Freeway Segmen Segments ts Duri uring Peak Periods (Aggregate ated by by Tr Trav aveel Ti Tim me Section) . . . . . . . . . . . . . . . . FIGURE FIGU RE 20. Relations Relationship hip Between Between Average Average Speed Speed and Accelerat Acceleration ion Noise Noise for Freeway Freeway Se Segm Dur Peak Pe ods (Agg Travel Tim S.ecti S pgmen eedents Dts istDurin ribuing tigonPea fork FPeri reriod ew asys(A aggre ndregat Agated rteed rialby StTrav reetsel. Time . . . e. .Sec . tion . on) . . ). .. .. .. .. .. .. .. .. Speed Distribution for Freeways . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Speed Distribution for Class I Arterial Streets . . . . . . . . . . . . . . . . . . . . . . . . . . Speed Distribution for Class II Arterial Streets . . . . . . . . . . . . . . . . . . . . . . . . . Accele lerrat atio ion n Dis Distrib ributio ution n for Fre Freeways ays an and d Art rteeria iall Str treeets . . . . . . . . . . . . . . . Acceleration Distribution for Freeways . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Accelerati ation Distributi ution for Clas asss I Arterial Streets . . . . . . . . . . . . . . . . . . . . . Accelerati ation Distributi ution for Clas asss II Arterial Streets . . . . . . . . . . . . . . . . . . . . 3-Dimensional Speed-Acceleration Distributio Distribution n for Freeways and Arterial Streets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . FIGU FIGURE RE 30. 3-Di 3-Dim men enssiona ionall Sp Spee eedd-A Acc ccel eleera rati tion on Dis Distr trib ibut utio ion n for Fr Freeeway ewayss . . . . . . . . . . . . . .

FIGURE 21. FIGURE 22. FIGURE 23. FIGURE 24. FIGU IGURE 25. 25. FIGURE 26. FIGURE 27 27.. FIGURE 28 28.. FIGURE 29.

xvii

43 44 44 45 45 48

48 50 51 52 53 55 56 57 58 60 62

 

LIST OF FIGURES (continued)

Page FIGURE FIGU RE 31. 3-Dimens 3-Dimension ional al Speed-A Speed-Acce ccelerat leration ion Distribut Distribution ion fo forr Class Class I Arterial Arterial Streets Streets . . . . 64 FIGURE FIGU RE 32. 3-Dimens 3-Dimension ional al Speed Speed-A -Acce ccelerat leration ion Distributi Distribution on fo forr Class Class II Arter Arterial ial Stree Streets ts . . . . 66

xviii

 

LIST OF TABLES

TABLE 1. TAB TABL TA BLE E 2. TABLE TAB LE 3. TABLE TAB LE 4. TABLE 5. TABLE 6. TAB 6. TABLE 7. TABLE 8. TABLE 9. TABLE TAB LE 10. 10.

Page Speed and and Accelerat atiion Distributi ution by Percentage of Tim Time . . . . . . . . . . . . . . . 18 Char Charac acte teri risstic tics of Fuel Fuel Cons Consum umpt ptio ion n Mode Models ls Uti tili lize zed d in th thee Ana naly lysses . . . . . . . . 31 Probab Probabili ilitie tiess Resul Resulti ting ng From From Comp Compari aring ng the the Avera Average ge and and Instan Instantan taneo eous us Methods of Fuel Consumption Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 Traffi Trafficc Oper Operati ating ng Char Charact acter eris istic ticss for for Road Roadway way Class Classes es During During Peak Peak and Off-Peak Periods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 R2  Values Observed Between Average Speed and Operati ating Char arac actteristics for Dif Different Road adw way Cl Clas assses . . . . . . . . . . . . . . . . . 47 Speed-Accelerat atiion Ma Matrix for Fr Freeways ays and Arterial St Streets . . . . . . . . . . . . . . 61 Speed-Acceleration Matrix for Freeways . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 Speed-Acceleration Matrix for Class I Arterial Streets . . . . . . . . . . . . . . . . . . 65 Speed-Acceleration Matrix for Class II Arterial Streets . . . . . . . . . . . . . . . . . . 67 Perce Percent nt of of Time Time S Spe pent nt in in Each Each Operati Operating ng Mod Modee by Roadway Functional Class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

xix

 

1

 

CHAPTER 1. INTRODUCTION

  Since the oil embargo of 1973, there has been an increased concern for energy efficiency and reduction reduct ion of mobile source emissions emissions from vehicles operating on the transportat transportation ion system. system. This concern for energy efficiency and reduced emissions has fostered the development of a new discipline within the transportation transporta tion field. Models for both energy consumption and mobile source source emissions have made up a large facet of this new discipline. As such such models become more rigorous, there has also been an improvement in the ability to collect more detailed operations data. This additional data, da ta, in turn, can be applied to improve the accuracy and level of detail of the energy consumption and mobile source emissions modeling. Mechanical distance measuring instruments (DMIs) that attach to a vehicles’s transmission were used in the late 1950s to collect speed speed and delay data. Reducing the large quantitie quantitiess of data data collected with the mechanical mechanica l equipment to a usable usa ble form proved to be difficult difficult and time-consuming. Electronic DMIs have replaced the mechanical versions, and the advent of portable computers has simplifie simp lified d the collection and reduction of detailed speed data. data . Several transportation tra nsportation agencies across the United States use DMIs and portable computers for travel time studies. Most of these these agencies, however, have few uses for speed profiles other than the identification of geometric bottlenecks and  problem areas. These speed speed profiles profiles are commonly commonly aggregated to provide average speeds between between major cross streets streets [½ to 1 mile intervals (0.8 to 1.6 km)] of the study corridor. Acceleration noise, or the fluctuation of speed along a roadway, is a concept that was first studied in the early early 1960s. 19 60s. Acceleration noise is defined as the standard deviation of changes in vehicular speed speed and has units of miles per hour per second. Detailed Detai led speed data at small intervals interva ls (speeds every second) are required requ ired to accurately calculate calcula te acceleration noise. noise. This type of detailed speed data is readily available using electronic DMIs and portable computers. Two computer models, EMFAC and MOBILE, are used to estimate mobile source emission rat rates es for for California and a nd the remainder of the United States, respectively. Other computer models that use emission rates from EMFAC or MOBILE have been developed to estimate the potential potentia l emission reduction reduct ion benefits of transportation control measures (TCMs). TCMs are required by the Clean Air  Act (CAAA) 1990 forrely areas as severe orof extreme non-attainment areas. The Amendments TCM computer modelsof primarily on designated changes in the number trips, vehicle miles of travel, and average speed to estimate estimate the emission emission reductions of proposed proposed TCMs. The Environmental Protection Agency (EPA), the agency responsible for enforcing the CAAA of 1990, has not issued any standard procedures or methodologies for calculating the potential emission reduction benefits of TCMs. Extensive research continues on new methods of evaluating evalua ting drive cycle changes on vehicular  emissions emis sions and fuel fuel economy. Drive cycle changes are based on the four vehicle operating modes: acceleration, accelera tion, deceleration, cruise, and idle. The changes in speed are more important to estimating emissions and fuel consumption than average speed. These operating modes can be easily characterized chara cterized with the detailed speed data available availa ble through the use of electronic DMIs. Models in

1

 

development would woul d presumably place an a n emphasis on the fluctu fluctuation ation of speeds instead instead of the average speed in computing emissio emission n rates and energy consumption. consumption. The concept of such techniques represents a new approach to calculating emission rates and could become a standard for the next generation of emission rate computer models. Problem Statement

Current mobile source emissions and energy consumption analyses are based on average vehicular speeds over roadway sections that are typically greater than 1 mile (0.6 km) in length. Recent research has ha s indicated that the fluctuation in speed (i.e., acceleration and deceleration) is more important than the average speed in determining mobile source emissions and energy consumption. This fluctuation in speed, known as acceleration noise, has not yet been effectively utilized in vehicle emissions and fuel consumption analyses because of 1) 1 ) the difficulty of collecting or estimating speed data for very short time or distance intervals, and 2) the absence of appropriate computer models to conduct such analyses. Study Objectives and Scope

The primary objective of this study was to characterize the speed and acceleration characteristics chara cteristics of a wide range of traffic flow. Data Dat a were collected with a DMI using short short increments of time. Researchers made a preliminary preliminar y investigation of the effects effects of detailed speed and acceleration accelerat ion data on existing fuel fuel consumption models. Comparisons of fuel fuel consumption estimates were made using speed and acceleration calculations based upon a segment-wide (average method) and a second-by-s second-by-second econd (instantaneous (instantaneous method). Detailed acceleration characteristics could be incorporated into the next generation of mobile source emissions and energy consumption models. Models that incorporate acceleration characteristi chara cteristics cs are expected to provide more accurate accura te estimates of mobile source emissions and energy consumption and of the changes in the emissions and energy consumption associated associated with various transportation tra nsportation projects and programs. The second objective of this study is to examine relationships between the geometric characteristics, speed and acceleration characteristics, and traffic flow variability for the different roadway functional functi classes. classes.roadway Thesetype regressio regre ssion n equations befreeways). based upon data that are disaggregated byonal functional (e.g., arterial Class will I or II, The final objective is to compile a reliable data set that describes the speed and acceleration characteristics chara cteristics of various variou s roadways and operating conditions. The data set produced from this project can be supplied to interested individuals or organizations for use in development and/or validation of fuel consumption and/or emissions modeling.

2

 

Organization of Report

This report is organized into five chapters: Chapter One, Introduction, provides an introduction to the research topic and presents the research objectives and scope. Chapter Two, Background, provides general information about previous studies of  acceleration accelerat ion characteristics and driving cycles. In addition, a summary is provided of previous fuel consumption and emissions modeling. Chapter Three, Study Stu dy Design, contains a summary of the procedures use used d to collect the data and develop the appropriate data bases (e.g., speed and acceleration characteristics, geometric characteristics, chara cteristics, and fuel fuel consumption model estimates). The analysis techniques are also described in this section of the report. Chapt er Four, Findings, presents the major findings for the research study. The findings Chapter include the results of statistical tests to evaluate the significance of utilizing a detailed speed and acceleration data set on fuel fuel consumption estimation. estimation. The relationships between between the speed speed and accelerat ion characteristics, geometric characteristics, acceleration chara cteristics, and fuel consumption model estimates are also discussed in this section. section. Trends in the data base containing operational chara characteristics cteristics of roadways in the Houston, Texas area and its application in emissions and fuel consumption modeling is also addressed. addr essed. This chapter concludes with remarks about the success of using DMI technology for data collection. Chapter Five, Conclusions and Recommendations, presents the conclusions and recommendations recommendati ons based upon the findings described described in Chapter Four. These conclusions begin with a discussion of the use of the detailed data set for evaluation of the differences for fuel consumption model estimation. estimation. The useful regression relationships and the application to transportation planning concerns are addressed. The content and usef u sefulness ulness of the operating characteristics characteristics data base for the Houston, Hou ston, Texas T exas area is also reviewed. The advantages advanta ges of DMI technology in data data collection, and the need future research in several areas a reas encountered in this research study are aaddress ddressed ed at the end of thisfor chapter.

3

 

4

 

CHAPTER II. BACKGROUND

This chapter provides general information about other studies of acceleration characteristics, cha racteristics, related data da ta collection techniques, driving driving cycles, and fuel cons consumption umption and emissions emissions modeling. In addition, sections that discuss the graphical representation of fuel consumption and emis emissions sions data are included. The literature literatu re search identified over over forty references that are summarized topically in the paragraphs that follow. Acceleration Characteristics

Several acceleration characteristics will be evaluated to make comparisons between the Federal Test Procedure (FTP) and the various variou s ffunctional unctional classes upon which which data are obtained. These acceleration characteristics found in the literature are discussed in the following section.  Acceleration Noise Early studies demonstrate that acceleration noise is a useful traffic parameter for evaluating traffic flow flow by investigating investigating it under different different conditions conditions (e.g., hilly, heavy volume) ( 1). Acceleration noise, or standard deviation of the acceleration, is defined as the root mean square of the accelerations, and a nd is also described described by Montroll Montroll and Potts Potts in their car following following study ( 2). The Jones and Potts approximation to acceleration noise, which utilizes three variables of a constant speed change [2 mph (3 kph) k ph) is frequently used], the running time of the vehicle for each speed change, and the total running time of of the vehicle vehicle,, is found in the the literature in various reports reports ( 1,3,4). The T he relationship for acceleration noise, or standard deviation of acceleration, a cceleration, for good level roads ranges from about 0.01 times the acce acceleration leration due due to gravity + 0.002 times times the acceleration acceleration due due to gravity for speeds between 20 mph (32 kph) and 60 mph (97 kph). Furthermore, for speeds speeds greater than 60 mph (97 kph) or less less than 20 mph (32 kph), these these values increase increase ( 2,5,6). One study sugges suggests ts that acceleration noise generally has been observed to decrease with increasing speed, though that may not hold hold true for for very high high speeds speeds [perhaps greater than than 60 mph (97 kph)] ( 7). Furthermore, several studies have indicated that some acceleration noise is “natural” due to ainformation driver’s inability to maintain constantattention speed through all 8). the changes in geometry and other  processing processing tasks thataconsume consume time ( 2,7,8). 2,7, The additional acceleration noise noise is due to vehicle interactions.   Equation 1 is the form of the acceleration noise as presented by Drew and Dudek (3).

5

 

Total Absolute Second-to-Second Differences in Speed Per Mile (TAD) Larsen and Effa introduce the use of a characteristic termed the Total Absolute Second-toSecond Differences in Speed Per Mile (TAD) in research performed in developing real-world drive cycles (11). This topic, as well as this variable, will be discussed discussed in later sections of this report. This variable is calculated in the same same manner as “? “ ? s”, or speed changes per mile, mile, in Greenshields’ Greenshields’ Quality of Flow Index. Other Acceleration Characteristics A study attempting a ttempting to predict accident risk proposed using the following characteristics characteristics:: mean velocity gradient (about the mean and the origin), velocity noise, average velocity, average acceleration, and acceleration acceleration no noise ise ( 7). The mean velocity gradient, defined as the acceleration nois noisee divided by the average velocity, was introduced introduced by Helly and Baker ( 9). The authors point out the fact that acceleration noise is not a good measure when traffic is flowing slowly (e.g., signalized segments)) and that the “ mean velocity segments velocity gradient” is a better measure since it it is a relative measure that can accommodate a ccommodate for the congestion at slower speeds. speeds. Data Collection

The method of data collection is one of the most critical aspects to be considered in any research project. For studies of this type, test vehicle techniques are often utilized. This is the method used for data collection in Houston, Texas for the Houston-Galveston Regional Transportation Transportatio n Study to determi determine ne the the travel travel times times on diff different erent roadways in the the area ( 12). This method operates on the premise that the driver doing the data collection will pass roughly the same number of vehicles that pass him/her. A distance measuring instrument (DMI) can be used to collect speed information at a given time or distance interval. The instrument, instrument, which which is accurate to + 1 foot (0.3 meter) in 1000 100 0 feet (305 meters), is secured to the vehicle’s transmission transmission and sends a pulse to an on-board lap-top computer. The computer then records the appropriate time, speed, speed, and incremental incremental distance. From this information, a speed profile is easily easily constructed. A similar method of data collection has been used in the past past to determine determine a freeway freeway conges congestion tion index index (FCI) ( 13). Another method of data collection is the “chase car” technique in which a vehicle is randomly selected in the traffic tra ffic stream stream (“target” vehicle) and is followed by an equipped data collection vehicle. This was the method method utilized utilized in a study study to develop develop “real-world “real-world”” driving cycles cycles ( 11). Data was collected with the aid of a laser system that can determine the “target” vehicle’s speed from the known change in speed and distance of the vehicle. Larsen has indicated that since emissions emissions are a larger problem at the higher speeds and higher accelerations and decelerations (i.e., non-average driving conditions), conditions), test vehicle me methods thods may not be obtaining the appropriate type of data. It was suggested that perhaps the outliers are the individuals of mos mostt importance and a method to select and obtain data for these individuals is necessary necessary ( 14). Larsen expressed expressed how how data collection collection can also be

7

 

supplemented with simultaneou simultaneouss video taping. Effa and Larsen used such video as an additional quality control measure in their study to ensure that the facility types for each section of the route were characterized and grouped accurately. These route sec sections tions were were aggregated together based upon average speed speed only for for analysis purpose purposess ( 11). Pela expressed expressed a similar similar concern in regard to the use of test vehicle data collection collection ( 15). Pela’s concern was that random vehicle sampling will not be achieved if instrumented vans are used, and he discussed the fact that individuals were actually solicited for vehicle instrumentation in a study in which he was involved (16). Graphical Representation of Data

After the data are collected, there are several diff different erent ways in which they can be presented to make eventual comparisons between, between, for example, the FTP, other drive cycles cycles,, or field data. A very common graphical representation is speed speed versus time, time, or a speed profile profile (Figure 1) . Such a graph enables the reader to see the number and location of starts, stops, and the respective slopes in the graph. The proportion of time throughout the trip that a vehicle is operating within a given speed range is a valuable valu able way of representing representing speed data data (Figure (Figu re 2). This information can also be presented presented with a cumulative cumu lative distribution for either speeds speeds or accelerations of vehicles (Figure (Figure 3). A frequency  bar graph with the acceleration acceleration rate in mph/sec mph/sec is also a lso helpful (Figure 4). This graph could be established with metric units also (e.g., kph/sec). It allows the reader to see the distribution of the acceleration rates ra tes easily. easily. Such a graph was found found in many studies studies while reviewing the literature. A similar frequency frequency bar graph can be constructed with with speed speed ranges of 5 mph (8.0 kph). k ph). Such a graph may not always demonstrate a normal distribution since the distribution may be based upon a limited number of observations. There are several additional graphical representations in the literature literatu re as well. One study demonstrated the use of graphical representations that show sspeed peed on the x-axis and the acceleration noise about the mean along the y-axis (Figure 5). A similar similar graph was shown for for acceleration noise about the origin, origin, and this graph was also created created for for speed speed ( 7). These graphs graphs tend tend to show show a curvilinear relationship with relatively higher standard deviations of acceleration or speed at lower  speeds [7 to 10 ft/sec (2 to 3 m/s)] and relatively rela tively smaller standard deviations of acceleration or speed at higher speeds [49 ft/sec (15 m/s)]. Winzer presents some interesting interesting graphs relating acceleration noisee and combinations nois combinations of two two of the the three three primary macroscopic macroscopic flow parameters ( 19). One such nomograph shows the increase in acceleration noise as the traffic volume increases. Other graphs featuring two curves were required to account for the discontinuity in flow parameters between congested and non-congested non-congested flow.

8

 

Driving Cycles

The FTP is a driving cycle that was developed over twenty years ago to provide emission information for light-duty light-duty vehicles. However, vehicles and driving characteristics have changed sinc sincee the development of this cycle, and this area of study is quickly expanding as more representative driving cycles are studied. Current Driving Cycles The US Environmental Protection Agency (EPA) supplied information about how to obtain the current curr ent driving cycles being considered along with reports reports explaining their development. Below is a summary of these these driving cycles ( 21): developed loped by the California Air Resources Resources Board (CARB) based on ARB02: This cycle was deve data from their Los Angeles chase car study. The purpose pur pose of the cycle is to test vehicles over in-use operation outside of the FTP, including extreme in-use driving events. HL07: This engineered cycle was developed by EPA in coordination with the auto manufacturers. manufactu rers. The purpose of this cycle is to test vehicles on a series series of acceleration events over  a range of speeds. The severity of the accelerations are such that tha t most vehicles vehicles will go into wide open throttle. This cycle has constant power (therefore, (therefore, constant slopes in the tim timee vs. speed profile) profile) with high engine load loa d for the engine wide-open.

developed loped to represent represent in-use driving driving that is outside the boundary of  REP05: This cycle was deve the current FTP FT P driving cycle. The cycle was generated from from a composite data set which equally represented Los Angeles Angeles chase car data and Baltimore 3-parameter instrumented vehicle data. The  primary purpose of the the cycle cycle is is assess assessing ing in-use in-use emission emissions. s. REM01: This cycle was developed to represent represent start driving behavior as well as that portion of in-use driving which which is not represented by REP05. REP05 . Start driving is represented by the first 258 seconds of the cycle. The remainder of the cycle represents in-use driving which was not captured

 by the start or REP05 cycles. cycles. When combined, combined, the REP05 and REM01 are intended intended to characterize characterize the full range of in-use driving. The primary purpose of this cycle cycle is assessing assessing in-use in-use emissions. emissions. The cycle was generated from a composite data set which equally represented represented Los Angeles chase car data and Baltimore 3-parameter instrumented vehicle data. UNIF01: This cycle was developed to represent the full-range of in-use driving in a single cycle. The methodology used in generating the cycle is largely consistent w with ith previous efforts by CARB to develop a unified cycle. The cycle was generated from a composite data set which equally equally represented Los Angeles chase car data and Baltimore 3-parameter instrumented vehicle data.

12

 

US06: This cycle is being proposed to make FTP more “realistic.” However, the proposal proposal is not to replace replace the FTP but to make this an additional additional tes testt to the FTP. Therefore, Therefore, there there would be two (2) tests for for the federal procedure. US06 is based on high speed speed accelerations which are not  present  pres ent in the FTP FTP and is an “aggregation” “aggregation” to some some extent extent of the REP05 and ARB02 ARB02 cycles cycles (which (which are non-FTP conditions). AC866 : This cycle is the part partsecond of the stage UDDS (Urban Dynamometer Driving Schedule--FTP). It is the cycle which represents (i.e., the bag II stage) of the FTP. Schedule--FTP).

which is the same as the first first part of REM01. However, the SC01: This cycle is a start cycle which remainder of it is different. start cycle which is similar similar to SC01. ST02: This is another start Other cycles cycles are being being developed developed in research research by individuals individuals and organization organizationss ( 22). In addition, the Coordinating Research Council has annual workshops that address these and many related issues issues (23). ( 23). Concerns About the FTP  A study by Denis, et al, demonstrates the discrepancies between the “real-world” conditions and what the FTP actually actua lly considers considers ( 20). This report explains, for example, that the FTP does not take into account higher acceleration and a nd deceleration rates that their test vehicle could exhibit while driving. Furthermore, the study explains that, “the FTP has more cruise, percent time stopped, and hard decelerations than the on-road on-road data. The FTP under-repres u nder-represents ents coasting, coasting, and hard and medium accelerations accele rations”” ( 20). A study done by the California Ca lifornia Air Resources Board (CARB) addressed these proble problems ms with the FTP by using “chase” cars in the Los Angeles area to construct seven cycles that are more representative of the on-road on-road performance of motorists motorists ( 11). These cycles are for for both freeway freeway and arterial sections. Perhaps the most most significant significant work in re-evaluating the FTP was performed by the Environmental Environ mental Protect Protection ion Agency Agency ( 18). This work concludes by explaining explaining that current driving cycles are still being developed with the data obtained from this study. Fuel Consumption Modeling

There is extensive literature available on the subject of fuel consumption modeling, and considerable work is being continued in this area. Two documents explain the different different types of fuel fuel consumption models available and the various independent variables that are necessary for their  operation (24,2 ( 24,25). 5). The models are ranked according to their simplicity. simplicity. For example, a simple linear  regression model utilizes u tilizes input such as the section distance and the number of stop/starts to determ determine ine fuel consumption per trip. On the other hand, an instantaneous fuel consumption model uses secondsecond by-seco  bysecond nd speed speed and acceleration acceleration to to obtain obtain fuel consumption consumption.. 13

 

One study of a fuel economy model looks closely at driving cycle cons considerations iderations as well well ( 26). The authors au thors note that their equation for fuel economy depends on five princip principal al summary variables, va riables, and with certain simplifying assumptions, three or four variables may be sufficient: average speed, free-flow free-f low velocity, fraction of time vehicle stopped, and perhaps stops per mile mile.. They also discuss cold start as a factor that they have not yet incorporated incorporated into the equation. They also discuss that in the early 1980s the FTP was estimate estimated d to have an error of about 15 percent percent as compared to actual driving. the authors has estimated beand as high 30 percent 2010, 201 0, and that itFurthermore, is currently estimated es timated atindicate about 20itto 25 2 5been percent error. to An Rossasmention that thatby “certain driving characteristics characteristics are critical for for emissions emissions but not for for fuel use” ( 26). The authors cite velocity times acceleration (a variable variabl e closely closely related to engine power power output) as an example. With that distinction made, they suggest that driving cycles for regulation of emissions should be defined differently from from driving cycles for fuel economy. Emissions modeling will be further further discussed in a later section. One fuel consumption model that is available was developed for FREFLO (the freeway simulation module of the FHWA CORFLO package pack age for macroscopic modeling of ffreeway reeway and arterial ar terial networks) (27). The model includes five terms terms which which address rolling, air, and effective effective inertial resistances, idle fuel consumption, consumption, and effective effective acceleration. The acceleration was approximated from FREFLO’s density density output. Density was used as a surrogate for acceleration noise. It was found that the model model accounted accounted for for “99.5 percent percent of the the variation in the the constant constant speed speed fuel consum consumption ption data and 86.1 percent percent of the variation variation in fuel fuel consumpti consumption on due to acceler acceleration” ation” ( 27). The model was was incorporated into the logical structure of FREFLO, and comparison with results from an INTRAS simulation (microscopic (microscopic model) showed good correlation. Winzer reinforces Rao and Krammes’ use of density density as a surrogate for for acceleration noise noise ( 19). Winzer’s study shows shows that there is indeed a high correlation between the two, in all types of traffic conditions (e.g., low to heavy). Another fuel consumption model called ARFCOM, developed by the Australian Road Research Board (ARRB), was also found found in the literature literature ( 28). The model is is a detailed, detailed, incremental incremental power  power  model for estimating estimating the fuel consumption of fully fully warmed-up vehicles. The inputs to the model include power consumption due to tractive forces, drive-train inefficiencies inefficiencies,, access a ccessories ories and internal engine friction. The author au thor asserts that even though individual components of total power power demand (such as a s internal engine friction) change dramatically over a range of engine speeds, the fuel-to-powe fuel-to-power  r  efficiency efficie ncy of an engine is fairly constant over a range of power and engine speeds. This simplification makes the model much more more practical for use in traffic management sstudies. tudies. The output outpu t of the model model is fuel consumption for the duration of a trip and estimates are within about 5 percent for trip durations of 30 to 60 minutes. The model appears to be well-suited well-suited to estimating estimating the fuelfuelconsumption impacts of geometric and other improvements in roadways. Another model that seems to be suitable for use in estimating the incremental effects of  changes in traffic management schemes was also found in the literature (29). ( 29). The author describes this model as an instantaneous, basic ba sic (detailed and microscopic) mode model, l, and the inputs inpu ts for this model model are instantaneous speed, acceleration, and grade. The output is fuel consumption ffor or the duration of  a trip and the accuracy accur acy is about three percent. The three terms of this model require further 

14

 

explanation. The first term allows for fuel fuel consumption required to maintain engine engine operation. The second term allows for fuel consumption required to provide tractive force to the vehicle in overcoming drag, inertia and a nd gradient forces. forces. The third term uses a product of energy and effective effective acceleration (acceleration including effects of gravity on a grade) to account for increased fuel consumption during du ring hard accelerations. A model developed by Bester in 1981 was also discovered in the literature search. Although it is also a function of acceleration, speed, and gradient, it is of a much simpler form (30). Several models that are only a function of speed were also examined. Five of these models, from Raus (31), Lindley (32), FREQ10 (33), McGill (34), and NETFLO (35,36) utilize equations to determine fuel consumption given only speed. However, two of thes thesee models are presented presented in a table format, as a s opposed opposed to equations, that define fuel consumption consumption as a function of the sspeed. peed. One such modell was develop mode developed ed by McGill McGill in 1985 (34) (3 4) and the other other is FREQ10 (33). In addition, addition, FREQ10 contains values for both arterials and freeways, allowing separate evaluation of both facility types (37). Conversely, NETSIM utilizes a table to determine the fuel consumption as a function of both acceleration accele ration and speed speed ( 35,36,38). Emissions Modeling

There is considerable literature in the area of emissions emissions modeling as well. well. Much of the literature supports the suspicions that the FTP is not representative of real world driving. One such report focuses on the high accelerations that are outside of the envelope of the FTP and are high emitters (39). The authors report that, “a single hard acceleration accelerati on event could produce emissions equivalent to 50 percent to 64 percent of the total FT FTP P emissions emissions for hydrocarbons, and 236 percent to 262 percent percent of the total total for for carbon monoxide” monoxide” ( 39). Further studies investigate doubts about the accuracy of the inputs and, hence, the output of  some emissions models. models. In her report on carbon monoxide modeling, Chapin expresses the concern that microscale dispersion dispersion models depend depend heavily on input variables va riables that have substantia substantiall uncertainties (40). She suggests suggests that traffic volumes, both overall and on local arterials, may be underestimated  by as much as ten to twenty twenty percent. percent. This, in turn, can result result in in uncertaint uncertainties ies inherent inherent in determini determining ng the composite emissions emissions factor that may result in underestimating emissions emissions by as much as 50 5 0 percent. Furthermore, the author indicates that with all these uncertainties built into dispersion models, the models are better used in relative comparisons (e.g., comparing “do project” to “don’t do project” alternatives) than in attempting to determine whether absolute CO concentrations at critical points (e.g., hospitals and parks) exceed exceed the National Ambient Air Quality Qua lity Standards (NAAQS). A study similar to this was performed by Vaughn Vau ghn to describe the impact of Intelligent Transportation Tra nsportation Systems (ITS) on projects with respect to emissions in arterial systems and networks ( 41).

15

 

Another report r eport found in the literature describes graphs which are presented in a sim similar ilar manner  (44). However, this report provides various percentile ranges within within the box plots. For comparison of different driving cycles and their respective emission emission rates for different conditions (e.g., running, cold start, start, or hot hot start), start), bar charts charts can can be utilized utilized ( 45). A useful useful method to provid providee speed speed and acceleration acceleration data in table form form is shown shown in Table 1 ( 18). Such a table shows the frequency at each speed and acceleration in two dimensions rather than with a three-dimensional three-dimensional graph as shown in Figure 6. With this method of prese presentation, ntation, it is easier to  precisely  preci sely read percent percent distribution distribution values than on the three-dim three-dimens ensional ional graph. This aids individuals individuals who desire these results for entry into emissions emissions models. Summary of Literature Review

The preceding literature review has discussed numerous acceleration characteristics, driving cycles, and related fuel consumption and emissions modeling issues. issues. Acceleration noise is found discussed with relation to car following stability as early as 1959 ( 5). The report discusses discusses the interactions of vehicles and its effect effect upon motorists’ speeds. speeds. These alternating speeds, and ensuing acceleration noise, are formulated and special note is made that future research will be performed to attempt to correlate acceleration noise to parameters para meters such as mean speed, number of lanes of traffic, and traffic density. As eviden evidenced ced in the literature review, acceleration noise has been sstudied tudied further, along with additional acceleration characteristics. Additional characteristics include the quality of  traffic flow flow index (Q Index) introduced by Greenshields ( 10). In addition to the study of acceleration characteristics, there is continued study on the issues of fuel consumption and emissions modeling. modeling. Shortcomings in regard to the errors of these mode models ls are apparent throughout throu ghout the literature. A large concern of such such models is the error inherent with some of the input variables (e.g., traffic traffic volumes), volumes), and the subsequent subsequent effec effectt on the results results ( 40). Several other concerns stem from the fact that the cu current rrent driving cycle for testing emissions is not representative of current driving conditions. Hence, much w work ork has been done, and is being continued, to develop additional cycles that will consider the current driving situations that are not represented in the FTP.

17

 

CHAPTER III. STUDY DESIGN

This chapter contains a summary of the procedures used to collect the data and develop the appropriate data bases (e.g., geometric geometric and acceleration characteristics). characteristics). Also contained contained within this chapter are the methodology and analyses ana lyses techniques techniques that were used to quantify qu antify the fuel consumption model estimates, investigate correlations and relationships, and develop the data base of useful emissions modeling information. The overall study approach emissions a pproach is described ffirst irst and is followed followed with a discussion of the data collection effort. Subsequent sections disc discuss uss the pertinent data bases that were created for analyses purposes. purposes. The final sections of this ch chapter apter discuss the statistical statistical analyses a nalyses utilized to study correlations and relationships within the data set and to evaluate several fuel consumption model estimates. Overview of Study Design

Figure 8 illustrates the procedure that was followed to accomplish the objectives of the study. The top of the figure figure begins with the objectives objectives the study study has targeted. These objectives are as follows: • • •

Determine Determine the effects effects of detailed speed speed and acceleration acceleration characteris characteristics tics on fuel fuel consumption; Investigate Investigate relationsh relationships ips between between speed speed and acceleratio acceleration n characteristics characteristics,, geome geometric tric characteristics, and traffic flow variability; and, Establi Establish sh a data base ffor or emiss emission ionss mod modeli eling ng that can can be utilized utilized by by others. others.

To accomplish accomplish these tasks, tasks, data were collecte collected d with a distance distance measuring measuring instrument instrument (DMI) on a total of 233 centerline-miles (375 km) of freeway routes and 198 centerline-miles (319 km) of  arterial routes. Summary speed and acceleration characteristics we were re calculated from the speed speed information provided by the DMI. Geometric characteristics were also collected along the corridors. The following section section entitled, “Data “ Data Collection” discusses the data collection procedures used for  the study. Once the data were collected with the DMI, three distinct modules of information were created. These are shown in Figure 8. The first data set includes the speed and acceleration characteristics computed from the DMI files files for for each corridor. The second data set includes the results of utilizing both an instantaneous and average a verage calcula calculation tion of several several fuel consumption model estimates based upon speeds and acceleration rates. Such analyses are imperative to show any differences differenc es in such models when collection of vehicle vehicle speeds is allowed as often as every second. The last data set that is merged with the others is the geometric characteristics that were collected along the travel time routes. The data were then combined, summarized appropriately, and statistical analyses were  performe  perf ormed d in order to complete complete the objectives objectives of the study. The results results and conclusion conclusionss are contained contained in Chapters IV and V. 19

 

Data Collection

The Houston Metropolitan Area and the Data Collected  Houston is the fourth largest city in the United States, and the metropolitan area ranks as the tenth largest. The population is estimated estimated at nearly 1.8 million within the city limits, limits, and 4 million within the Greater Houston metropolitan area. Geographically, the Houston urbanized area covers approximately 3,000 square miles (7,770 (7,7 70 square kilometers). Due to the large population and geographic area, Houston Hou ston has heavy traffic traffic during peak periods that occur from approximately 6:00 to 9:30 a.m. and 3:30 to 7:00 p.m. Several different different types of data were collected for for the study. Speed and acceleration characteristics were obtained from base travel time data. Roadway geometrics were were collected in conjunction conjunc tion with another another study study being being conducted conducted by TTI ( 12). Driveway Driveway information information was was also collected by recording r ecording the number of “curb cuts” along sections. The speed data were collected collected using the DMI technology discussed in the next section section of this chapter to obtain speed information every half second.  Roadway Geometric Characteristics Characteristics The Houston Hou ston TxDOT District’s Planning Department is unique since the they y have been collecting detailed roadway geometric information since the mid 1960s for long-term transportation planning  purposes.  purpose s. Roadway invento inventory, ry, as the the geomet geometric ric inform information ation data collec collection tion is is called, is is one of the types of data that tha t is collected for for planning purposes. This information is necessary necessary since since Houston has no zoning, with the exception of deed restrictions. restrictions. Geometric roadway information is often used to develop growth trends, estimate existing capacity, and determine projected facility needs. Data for the Roadway Inventory were collected within the Houston Galveston Regional Transportation Tra nsportation Study Stu dy (HGRTS) (HGRT S) (12). (1 2). The study known as Roadway Inventory consists of the collection of roadway information shown shown in Figure 9 (field data collection ssheet). heet). The data collection includes a survey of every FHWA functionally classed roadway segment as well as other selected roadway segments. Data are collected on all segments segments and compiled in a data base. Roadways were divided into segments with limits set where minor arterials or higher class roadways cross the surveyed roadway, roa dway, or when the roadway cross section has a geometric change (e.g., number of lanes, median type). type). Segments typically ranged from 0.10 0.1 0 miles (0.2 km) k m) to 1.5 miles (2.4 km) in urban areas and up to 6.0 miles (9.7 km) in rural area areas. s. With the roadways divided divided into segments, segments, physical inventories were conducted in the field. field. Field inventories involved measurements measurements with a measuring wheel, observations, and recording all other pertinent information. From the roadway segments provided in the Roadway Inventory, the research team determined which travel time routes would be utilized for the data collection effort. The team considered the location and a nd convenience of the roadways with their their data collection cost, while providing providing a random ra ndom sampling of routes. Figure 10 illustrates illu strates the location location of the freeways, freeways, Class I arterials, and Class II

21

 

arterials used for the travel time routes in the study. study. The terminology of Class I and Class II arterial segments,, defined in the Highway segments Highway Capa Capacity city Manual (HCM), (H CM), was used by the research team to classif classify y thesee segments thes segments based based upon the posted posted speed speed limit and geometri geometricc characteristic characteristicss of the arterial arterial ( 46)

Key Map # Sheet # HOUSTON-GALVESTON REGIONAL TRANSPORTATION STUDY HARRIS COUNTY STREET INVENTORY AS OF / / STREET LIMITS SECTION NO. STATE SYSTEM LOCATION

TYPE FEDERAL SYSTEM URBAN AREA

FUNCTIONAL CLASS TEXAS TRUNK SYSTEM MAINTENANCE

LENGTH MI. R.O.W. WIDTH ROAD WIDTH  NO. OF LANES LANES MEDIA MEDIAN N WIDTH MEDIAN DESIGN SIDEWALKS CURBS SHOULDERS SHOULDER S FT. SHOULDER TYPE SURFACE TYPE SURFACE CONDITION ILLUMINATION PARKING MARKING TRAFFIC SIGNALS STOP SIGNS YIELD SIGNS CAUTION LIGHTS CHANNELIZED INTERSECTIONS SPEED LIMIT RAILROAD CROSSINGS: UNPROTECTED CROSSBUCK CROSSBUCKS S FLASHING LIGHTS FLASHING LIGHTS & GATES 24 - HOUR VOLUM VOLUME E

COUNT DATE

ADT

(OVER)

FIELD INITIALS

REMARKS: RECORD #: PAGE #:

1 1

Figure 9. Sample Roadway Roadway Inventory Field Data Data Collection Sheet

22

YEAR

 

 DMI Data Collection Travel time data collection in the Houston area is currently performed with a DMI for the Houston-Galveston Houston -Galveston Regional Regional Transportation Transportation Study (HGRTS) ( 12). The follow following ing discus discussion sion explains the details of DMI data collection and how it was utilized for this study. About the Software:   Computer Aided Transportation Tr ansportation Software Software (CATS), an integrated computer   program,, was developed  program developed by the the Texas Transportation Instit Institute ute (TTI) (TT I) to improve improve the collection collection of  travel time and speed data for traffic studies ( 47). The CATS software software currently consists consists of three modules:: DMIREAD, DMISTAT, and DMIPLOT. modules

The data collection module, DMIREAD, which is used with a laptop computer connected through the serial port to a DMI, collects detailed speed, distance, and clock time data while traveling through a corridor. The operator enters information information about the run from a me menu nu system that includes: Roadway Name, Roadway Type, Travel Direction, User Name, Odometer Reading, Weather  Conditions, etc. etc. (Figure 11). DMIREAD opens the serial serial port, sets the DMI to the correct mode,  places the operator’s operator’s input (run data) data) in the header, header, obtains obtains speed speed and distance distance information information from the DMI, time stamps each data record, formats the data, and writes it to an ASCII file. The information that is provided in the ASCII file are the event number, cumulative and interval distance, speed, a clock time stamp for each reading, and the header (run data) information (Figure 12). The program typically collects data on a half-second interval, but has the capability of collecting the data as frequently as 0.10 of a second. Reliability in the data collection effort results when the travel time run is performed on a  predetermi  prede termined ned route. Prior to beginning beginning the data collection, collection, a driver and an observer observer will use the the DMI to accurately accura tely determine the the distances between between the predetermined predetermined checkpoints. The checkpoint distancess are collected and entered distance entered into a “yar “yard-stick” d-stick” file that will be used by the DMISTAT D MISTAT analysis module to perform perform statistics on each section. The travel time run is then then conducted, taking care to mark the first checkpoint and all subsequent checkpoints very accurately by pressing any key k ey on the laptop computer which will will write a “!!! MARK !!!” to the ASCII file (see Figure 12). 12) . Redundancy is introduced by the operator concurrently marking the checkpoints as the travel time run is being made and comparing this information with the data obtained by using the “yard-stick” to determine the checkpoints. If a conflict is found, found, typically the problem is that the operator is marking the wrong checkpoint location (i.e., the wrong cross street). Of course, each problem requires a case-by-case case-by-case investigation. Extensive quality control was performed performed on the travel time data files prior to analyses to ensure the data were acceptable prior to analyses. analyses. Quality Qua lity control measures measures are discussed discussed in a later  section of this report.

24

 

FREEWAY NAME FREEWAY TYPE FREEWAY DIRECTION DATE TODAY  WEATHER C ONDITION LIGHT CONDITION PAVEMENT CONDITION SCHDEULED TIME DRIVER  MILE STAR T START TIME

: : : : : : : : : : :

KATY FREEWAY MAIN LANES IN BOUND 3/9/1995 CLEAR  NORMAL DAYLIGHT DRY 06:30 atm  15171 Thu Mar 09 06:33:11 1995

!!! MARK !!!   1. 0.001 2. 0.003 3. 0.004 4. 0.007 5. 0.016 6. 0.021 7. 0.026 8. 0.031 9. 0.037 10. 0.042 11. 0.048 12. 0.054 13. 0.059 14. 0.065 15. 0.070

0.001 0.002 0.001 0.004 0.009 0.005 0.006 0.005 0.006 0.006 0.006 0.006 0.006 0.006 0.006

46 46 46 46 47 47 47 47 47 47 47 47 47 46 46

@ @ @ @ @ @ @ @ @ @ @ @ @ @ @

Thu Thu Thu Thu Thu Thu Thu Thu Thu Thu Thu Thu Thu Thu Thu

Mar Mar Mar Mar Mar Mar Mar Mar Mar Mar Mar Mar Mar Mar Mar

09 09 09 09 09 09 09 09 09 09 09 09 09 09 09

06:33:11 06:33:11 06:33:12 06:33:12 06:33:13 06:33:13 06:33:14 06:33:14 06:33:14 06:33:15 06:33:15 06:33:16 06:33:16 06:33:17 06:33:17

1995 1995 1995 1995 1995 1995 1995 1995 1995 1995 1995 1995 1995 1995 1995

16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33.   34.   35.

0.076 0.081 0.086 0.091 0.097 0.102 0.107 0.113 0.118 0.123 0.128 0.133 0.138 0.143 0.148 0.153 0.157 0.162 0.167 0.171

0.006 0.006 0.005 0.006 0.006 0.006 0.005 0.006 0.005 0.006 0.005 0.005 0.006 0.005 0.006 0.005 0.005 0.005 0.005 0.005

47 47 46 46 46 45 45 44 44 44 45 45 44 44 42 42 42 40 40 39

@ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @

Thu Thu Thu Thu Thu Thu Thu Thu Thu Thu Thu Thu Thu Thu Thu Thu Thu Thu Thu Thu

Mar Mar Mar Mar Mar Mar Mar Mar Mar Mar Mar Mar Mar Mar Mar Mar Mar Mar Mar Mar

09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09

06:33:17 06:33:18 06:33:18 06:33:19 06:33:19 06:33:20 06:33:20 06:33:20 06:33:21 06:33:21 06:33:22 06:33:22 06:33:22 06:33:23 06:33:23 06:33:24 06:33:24 06:33:25 06:33:25 06:33:25

1995 1995 1995 1995 1995 1995 1995 1995 1995 1995 1995 1995 1995 1995 1995 1995 1995 1995 1995 1995

  36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50.

0.176 0.180 0.184 0.188 0.193 0.197 0.201 0.206 0.210 0.214 0.219 0.223 0.227 0.232 0.236

0.005 0.005 0.005 0.004 0.005 0.004 0.005 0.005 0.005 0.005 0.005 0.005 0.005 0.005 0.005

39 37 37 37 37 37 36 36 36 37 37 37 37 37 37

@ @ @ @ @ @ @ @ @ @ @ @ @ @ @

Thu Thu Thu Thu Thu Thu Thu Thu Thu Thu Thu Thu Thu Thu Thu

Mar Mar Mar Mar Mar Mar Mar Mar Mar Mar Mar Mar Mar Mar Mar

09 09 09 09 09 09 09 09 09 09 09 09 09 09 09

06:33:26 06:33:26 06:33:27 06:33:27 06:33:28 06:33:28 06:33:28 06:33:29 06:33:29 06:33:30 06:33:30 06:33:30 06:33:31 06:33:31 06:33:32

1995 1995 1995 1995 1995 1995 1995 1995 1995 1995 1995 1995 1995 1995 1995

Figure 12. Example of Output From DMIREAD

26

!!! MARK !!! !!! MARK !!! !!! MARK !!!

 

Either method enables the user to post-analyze the data via the ASCII format as shown in Figure 12. Header informatio information n (run data) provided provided by the operator operator from a menu format is used used to uniquely name each ASCII output data file so that it indicates a roadway name, roadway type, direction of travel, date, and time. This feature eliminates the problem of overwriting previously collected data files. The other two software software modules are Excel macros that are written in Visual Basic. The DMIPLOT module opens the ASCII file, formats the data, plots a speed profile, profile, and uses u ses the header  information for title and labeling purposes. The DMISTAT DMIST AT module opens the ASCII file, formats formats the data, provides interval and cumulative distance and time, calculates average speed, standard deviation, percent time between different speed ranges, and estimates a level of service based on speed. provides a vast amount amou nt of data for each travel tra vel Increase in Quantity of Data: The software system provides time run, as shown shown in Figure 12. The operator enters information information about the travel time run such as route name, direction, type, and weather (see Figure 11), before the program starts, and writes the header data da ta to the ASCII file. The software system system (program and DMI) provides header header information which includes: C C C C C

Route Name Route Direction Route Type Driver Name Run Date

C C C C C

Start Time (computer generated) Weather Conditions Light Conditions (daylight, night, fog) Pavement Status (wet, dry, etc.) Scheduled Start Time

In addition, the program provides an event number, cumulative cu mulative distance, speed, and computer  time stamp at a rate of up to once every 0.1 seconds. This results in much more data than is collected using the manual method. This level of detail provides some some distinct distinct advantages. First, checkpoints can be determined from the travel time data log even if the observer does not mark the location while making the travel time run. Quality Qua lity control checks can be made to determine if the location of the checkpoints are ar e being marked accurately. accu rately. The greatest benefit benefit from this type of data collection technique is the increased volume of data. Instead of an average speed every quarter mile to two miles (0.4 to 3.2 kilometers) the data can be recorded every 0.10 second. From these these data files a far  more detailed analysis can be performed. performed. In addition, the program uses the header data for the automatic file naming system to ensure that no files are overwritten. Ease and Consistency of Data Collection: The software system provides ease and efficiency of  operation, as well as consistency of the data collection process. process. Once the travel time runs are completed, files can be downloaded and the analysis software, DMIPLOT and DMISTAT, can be used to process the data within minutes per output file. With computer generated files, files, there are no  problems  proble ms with interpretation interpretation of the data or data entry errors. errors. More accurate and consiste consistent nt recording recording of time that each run was started and when checkpoints are crossed are provided. The consistent data format allows for automated data reduction, which will be discussed later.

27

 

Another distinct advantage is that the data is in ASCII format, allowing the output files to be viewed or analyzed in almost any software software package. packa ge. Other programs are in various data base formats which are not easily viewed or manipulated. Potential Limitations: As with any type of data collection, there are some drawbacks. One of the limitations associated with computer aided data collection is that the number of runs are limited to

the number of laptops and the number of DMI units that are installed in available vehicles vehicles.. Additional drawbacks are that the size of each data file can get overwhelming and disk storage space is a  potential  poten tial problem problem.. However, However, the increased increased size size of and decrease decreased d cost cost of hard disk disk space, space, along with the use of compression compression utilities for long-term storage, alleviates this problem. A minor problem is getting power to the laptop computers. Battery life can be as long as 4 hours, but if batteries are not charged fully or if they are altered by constant charging and discharging, problems can develop. A/C adapters are available at a modest cost. The human factor (e.g., missing checkpoints and inaccurate calibration) will always be a  problem.  proble m. However, However, these problems problems can be solved solved or isolated with with proper education education and training. These problems can be easily ea sily overcome with modest modest precautions, and a nd the benefits far outweigh any disadvantages. Quality control was a very impo important rtant aspect of of the study. Several checks checks and crosscrosschecks, performe performed in order units, to discover and correct tra travel vel time run files that were incorrect for  differentwere reasons (e.g.,dincorrect unreasonable speeds). Development of Data Matrix

The final data da ta set analyzed ana lyzed for this study was quite extensive and was comprised of data from three sources: 1) speed and acceleration characteristics, 2) fuel consumption calculations, and 3) geometric characteristics. This section will further further explain the development, conten content, t, and usefulness of these three modules. Speed and Acceleration A cceleration Characteristics Module The speed and acceleration accelera tion characteristics portion of the data set is a dir direct ect result of the DMI travel time runs. The travel time run output (see Figure 12), 12 ), contains the speed speed of the vehicle at a specified specif ied distance. A total of 382 travel tra vel time runs are incorporated into the final final data set. There are: 100 Class Cla ss I arterials, 81 Class Cla ss II arterials, 71 with both Class I and Cla Class ss II arterial segmen segments ts together, and 130 freeway segments. These travel time runs are further disaggregated by ssection ection because several of the travel time runs cover sections of roadway that are not geometrically similar and, therefore, should shou ld not be analyzed together. For example, a roadway can change from divided to undivided or from a two-lane section section to a four-lane section. section. Travel time runs that contain portions of arterial Class I and arterial Class II sections were separated.

28

 

After disaggregating the DMI travel tr avel time runs for the different functional functional classes cla sses of of roadway, there were 843 travel time runs for Class I arterials, 1,018 runs for Class II arterials, and 1,087 runs on freeways. As discussed discussed in the preceding section section “ Data Collection,” Collection,” the travel time runs were collected at different times of the day. The reduction of these DMI travel time runs into similar  similar  groups for analyses is discussed in the “Levels of Analyses” section in a later section of this report. The following speed and acceleration characteristics were calculated with the DMI data: • • • • • • •

Average Speed Average Ac Acceleration Standard Deviation of Speed St Stan anda dard rd Devi Deviat atio ion n of of Acce Accele lerat ratio ion n (Ac (Acce cele lerat ratio ion n Noi Noisse) Mea ean n Ve Veloci locity ty Grad Gradie ient nt as Def Define ined by by He Helly lly and and Bake Bakerr (9) Greenshields’ Q Index ( 1 10 0) Total Total Ab Abso solut lutee Seco Second nd-t -too-Se Seco cond nd Diff Differ eren ence cess in Spee Speed d Per Per Mile Mile (TAD) (TAD) ( 11)

While calculations were created to perform the analyses of the acceleration characteristics, variables were also utilized that calculate and save the speed and acceleration distributions by  percentage  perce ntage of of time time (see Table 1). This info information rmation is valuable valuable for for individuals individuals who require data sets sets  broken down Such down by percen percent t of timeisatcritical a specif specified speed speed and acceleration accel for diff dfields ifferent erent conditions condi tions (e.g., facility type). information in ied emissions modeling mode lingeration and related (e.g., determination of drive cycles).  Fuel Consumption Module The literature literatu re search identified several several fuel consumption models that would be beneficial in a comparative analysis. a nalysis. The instantaneous speed and acceleration, and average speed and acceleration model results were were compared using these models. models. The instantaneous method involved calculating calcula ting the speed and acceleration for every other observation (i.e., every 1.0 second) in the data set and calculating calcula ting the instantaneous fuel consumption estimate estimate for the diff different erent models. models. The instantaneous instantaneou s fuel consumption estimate estimate for every other observation was then avera averaged ged over the entire travel time run to determine the average fuel consumption. Hence, the final result is a fuel consumption esti estimate mate  based upon instantaneous instantaneous calculations of fuel fuel consumption consumption.. The fuel consumption estimate based upon average speed values was obtained as follows. follows. The speed for the travel time run was computed as the total distance over the total time. The acceleration was computed the same as for the instantaneou instantaneouss method. This computation remained the same since the true acceleration is required and can be obtained in this fashion. In addition, by keeping the accelerations consistent, the respective differences in fuel consumption will be produced in a consistent manner. After computing speed and acceleration in this manner, the fuel consumption estimate for the average method was performed performed for each travel time run. Analyses in in the “Levels of  Analyses” section of this report discuss the methods used to test for significant differences between the instantaneous and average methods of fuel consumption estimation.

29

 

Models Used for Analyses. Analyses . Several of the models models identified identified in the literature literature search were used in the fuel consumption consumption analyses portion of the study. Ta Table ble 2 displays the models models used in the analyses. In addition to the model name and date, this table also shows the independent variables used in the model to calculate calcula te fuel consumption, the applicable speed range, and the functional cla class ss orientation of the model (e.g., arterial and/or freeway). There are a re some important notes that should be made about abou t the information provided in Table 2. First, most most of the models models are based upon equations, but some some utilize look-up tables to estimate fuel consumption. consumption. The models that estimate estimate fuel consumption with the look-up tables are FREQ10, McGill, and NETSIM. Another important point is that throughout throughou t this study, grade, which is incorporated into two of the models, models, was assumed to be zero. zero. This is a reasonable estimate for the Houston,, Texas area where the data were collected. Houston collected. The “Speed Range” and “Functional “Functional Class Class Orientation Orient ation”” values valu es are based upon what is in the literature, or in the absence, professional professional judgement. ju dgement. For the purposes of this study it was necessary to convert the output units of the fuel consumption models models to a common common unit. The unit 1,000 multiplied by gallons/second, gallons/second, was was utilized. A time-based unit for fuel consumption (e.g., seconds) seconds) was selected over a distance-based distance-based unit (e.g., miles) since it provides reasonable values during idle conditions (i.e., stopped at signals or traffic queues). In order to perform analyses on all the models in the same manner, it was necessary necessary to ensure that all the models were extrapolated from 0 to 75 miles per hour (121 kph) (i.e., the range of   possible  poss ible values encountered encountered in the data set). set). Linear extrapolation extrapolation provided provided a reas reasonable onable assess assessment ment of fuel consumption consumption estimates for undefined ranges. Interpolation was used to estimate fuel fuel consumption rates in the models that use look-up tables since the values are often recorded in increments of 5 miles per hour (8.0 kph) in the tables.

30

 

  s   e   s    l   y   a   n    A   e    h    t   n    i    d   e   z    i    l    i    t    U   s    l   e    d   o    M   n   o    i    t   p   m   u   s   n   o    C    l   e   u    F    f   o   s   c    i    t   s    i   r   e    t   c   a   r   a    h    C  .    2   e    l    b   a    T

   S    S    A   N    L   O    C   I    L   T    A   A    N   T    O   N    I    E    T   I    C   R    N   O    U    F

  y   a   w   e   e   r    F    d   n   a    l   a    i   r   e    t   r    A

  y   a   w   e   e   r    F    d   n   a    l   a    i   r   e    t   r    A

  y   a   w   e   e   r    F    d   n   a    l   a    i   r   e    t   r    A

   E    G    N   )    A   H    R   P    D   (    M    E    E    P    S

   6    7   o    t    9

   6    7   o    t    9

   1    8   o    t    5    2

   T   )    S    N   (    E   E    D   L    N   B    E   A    P   I    E   R    D   A    N   V    I

  n   o    i    t   a   r   e    l   e   c   c    A    d   n   a    d   e   p    S

   d   n   a  ,   n   o    i    t    1   a   e   r    d   e    l   a   e   r   c   c   G    A  ,    d   e   p    S

   d   n   a  ,   n   o    i    t    1   a   e   r    d   e    l   a   e   r   c   c   G    A  ,    d   e   p    S

 ,    k   5    E    i    8    T    l    9   e   c   1    A    k    /    D   s    A  g    /  ,   g   r    E   e   i    M   y   B   w    A   o   d   n    N    B   a  .   o    N    L    E    D    O    M

   1

   l   a    i   r   e    t   r    A

  y   a   w   e   e   r    F    d   n   a    l   a    i   r   e    t   r    A

  y   a   w   e   e   r    F

   l   a    i   r   e    t   r    A

  y   a   w   e   e   r    F

  y   a   w   e   e   r    F

   l   a    i   r   e    t   r    A

   l   a    i   r   e    t   r    A

   5    3   o    t    1

   5    5   o    t    1

   0    7   o    t    0

   0    4   o    t    5

   0    7   o    t    5    1

   5    7   o    t    7

   5    2   o    t    0

   5    7   o    t    0

   d   e   e   p    S

  n   o    i    t   a   r   e    l   e   c   c    A    d   n   a    d   e   p    S

   d   e   e   p    S

  n   o    i    t   a   r   e    l   e   c   c    A    d   n   a    d   e   p    S

   4    8    9    1    /    l    l    i    G   c    M

   4    9    9    1    /    O    L    F    E    R    F

   0    8    9    1     e   r   a   p   t    /   a    O   d    L    F    T    E    N

   6    8    9    1    /    a    t    M   a    I    S   d    T    E    N

   d   e   e   p    S

   d   e   e   p    S

   d   e   e   p    S

   d   e   e   p    S

   6    8    9    1    /   y   e    l    d   n    i    L

   0    9    9    1    /    0    1    Q    E    R    F

   0    9    9    1    /    0    1    Q    E    R    F

   /    k    i    l   e   c    k   6    A   8    d   9   n   a   1   s   g   g    i    B

   0    8    9    1    /   r   e    t   s   e    B

   1    8    9    1    /   s   u   a    R

   2

   3

   4    5    6    7    8    9

   0    1

 .   o   r   e   z   e    b   o    t    d   e   m   u   s   s    1   a    1   e    d   a   r    G  

   1  

31

 

Geometric Characteristics Module The final component of the data set is the geometric characteristics. Many of these these characteristics characte ristics are part of the roadway inventory that is collected for the Houston-Galveston Regional Transportation Transportati on Study (HGRTS) (12). In addition addition to the variables variables available available in HGRTS, the number  number  of driveways along the study routes we were re recorded during the data collection effort. effort. The list below displays the variables contained in the geometric characteristics module along with any necessary description. • • • • • • • • •

Study udy rro oute ute 1s 1stt ccro rosss sstr tree eett 2n 2nd d cro crosss str streeet Roadway Roadway sectio section n number number (numbere (numbered d consecut consecutive ively) ly) Facility type (e.g., (e.g., under under constructi construction, on, divided divided,, undivided, undivided, one-w one-way) ay) Functional Functional classific classification ation (e.g., (e.g., interstate interstate,, freeway, freeway, principal principal arterial, arterial, minor minor arterial) Le Leng ngth th of of the the sec secti tion on Widt Width h of the the roadw roadway ay surf surface ace Num Number o off lan lanees

•• Widt Wi dth hnof ed edia ian n (e.g., Design Desig ofmmedian m edian (e.g., no median, median, curbs, guardrail and/or fence, fence, open and/or and/or drainage ditch, ditch,  painted))  painted • Parking restrictio restrictions ns (e.g., (e.g., none, no restrictio restrictions ns both sides sides,, no parking anytime both sides sides,, a.m. and/or p.m. restrictions both sides) • Num Number ber o off ssig igna nals ls • Numbe Numberr o off sto stop p sig signs ns • Po Possted ted spe speed ed limi limitt • 24-ho 24-hour ur tub tubee vol volum umee • Number of channelization channelizationss per per sectio section n (e.g., (e.g., turn bays) bays) • Number Number of driv drivew eways ays in in traveli traveling ng direc directio tion n From the list above, informative density variables such as signals per mile, stops per mile, and driveways per mile can be computed. Additional variables that were calculated include average lane width and the hourly volume volu me per lane. These density variables will will be used for analyses of geometric characteristics and a nd the speed speed and acceleration characteristics. characteristics. Such analyses are further des described cribed in the sections that follow. Levels of Analyses

The three data da ta modules described above were merged together prior to beginning the statistical analyses. In addition, several quality control measures were used to ensure the data were read  properly  prope rly and contain contain appropriate appropriate information information..

32

 

Quality Control Measures Qua lity control of the data prior to statistical analyses Quality a nalyses was an important element of this study. Since the data were subject to many conversions and disaggregations it was imperative that the data da ta  be closel closely y scrutinize scrutinized d throughout throughout the data reduction reduction proces process. s. The original DMI travel time run files were checked for several items that would create  problems in the computer  problems computer analyses and subsequent subsequent results. results. These checks checks include searching searching for blank  lines or missing variables, formatting problems (i.e., information not located in the correct place), incorrect units, nonconsecutive nonconsecutive observation numbers, la lack ck of consistency, and unreasonable va values. lues. Similar checks were also made after the original DMI travel time run ru n files were were disaggregated into sections. sections. A check was was made to ensure that observations (i.e., lines of the data files files)) were not copied twice or left out while being being disaggregated. Several of the DMI travel time run files were selected for individual scrutiny to ensure that no areas of concern could be found. found. The most effective effective method to search for errors was running the files through the statistical package utilized for analyses. The program alerts the operator to suspect values and syntax problems that may be present in the DMI travel time run files.  Initial Examination of the Data Set  Prior to analyses ana lyses of the data set, it is important to examine the data to determine exactly what trends and a nd patterns may be present. At this preliminary preliminary point of the analyses, it is important important to obtain a basic understanding of the data set and its contents. This information is generally provided by obtaining descriptive statistics for the data set. The following list contains the analyses that were  performe  perf ormed d at this initial level. level. •

Produce frequency frequency tables tables that illustrate illustrate the number of observatio observations ns in in differ different ent groups groups of  data (e.g., at average speeds in a given speed range for freeways).



Ensure that the the number number of obse observations rvations in the different different groups is is ssuffi ufficient cient for statist statistical ical analyses.



Produce histograms histograms and normality normality tests tests to to ensure ensure the data obtained obtained for for each of the the groups groups originates from a normal distribution.

In addition to allowing a llowing the researcher researcher to get an a n understanding for the distributions of th thee data, such analyses ana lyses allow for for checks that ensure the data originate from a normal distribution and whether  the sample sizes are adequate. Since many statistical analyses (e.g., t-tests, t-tests, analysis o off variance) are only applicable when these assumptions are true, it is important to investigate these descriptive statistics.

33

 

After looking at the data, da ta, it was necessary to decide how the roadway functional classes would would  be separated into groups. The Highway Capacity Ca pacity Manual (HCM) shows the level of service service for  different diff erent arterial classe classess based based upon the average travel sp speed eed ( 46). The research research team used used the terminology of Class I and Class II from the HCM to classify the arterial segments based upon the  posted  post ed speed limit and geometric geometric characteristi characteristics cs of the arterial. A typical typical freefree-flo flow w speed speed of 40 mph (64 kph) is reported for Class I arterials in the HCM, 33 mph (53 kph) for Class II arterials, and 27 mph (43 kph) for Class Class III arte arterials rials according according to to the HCM ( 46). Class III arterials arterials tend to be in, in, or  near, the downtown and since travel time data were not being collected in these areas, Class III arterials were not included in the study.  Fuel Consumption Analyses The objective of the fuel consumption analyses is to investigate any significant differences  between  betw een model results results for for the instantane instantaneous ous and average method methodss of estim estimating ating fuel consum consumption. ption. The calculation calcu lation of the instantaneous and average fuel consumption estimates was explained in detail in a previous section entitled, “Fuel Consumption Module ”. The tes tests ts for significance were  performe  perf ormed d utilizing t-tests t-tests within within each roadway functional functional class. class. If signifi significance cance is found, found, analyses analyses of  variance (ANOVA) tests were performed to discover what variables explain where the variance is  being introduced. Example variables include signal signal density, density, number number of of lanes, lanes, or or hourly volume per  lane. introduced. It is important to note that these analyses will not conclude which model is the “best” model model for fuel consumption estimation. These analyses only identify signif significant icant differences b between etween the two methods of fuel fuel consumption estimation. estimation. However, the analyses can provide a relative measure of  the estimation trends of the models (i.e., a group grou p of models give similar results while one or two seem to give “outlier” estimations relative to the group).  Regression and Correlation Analyses  The next step of the analyses was to bring the three modules of data together to investigate any correlations among the variables involved for for a given functional functional roadway roadwa y class. Regression equations were evaluated between the speed and acceleration characteristics, geometric characteristics, and traffic flow variability (e.g., average speed compared to the coefficient of  variation of speed). Coefficient of determination (R 2 value) was evaluated to determine the degree to which the independent independent variables explain the effects on the de dependent pendent variable. There were several relationships that were expected to make intuitive sense to be highly related (e.g., signal density and speed for a given arterial class). These analyses allowed such intuitions to be tested as w well ell as to discover other relationships.

34

 

 Data Base of Useful Emissions E missions Modeling Information The final objective of the study is to establish distributions of the operating characteristics chara cteristics of  the Houston, Texas Texa s area. This distribution will stratify stratify the operating characteristics (e.g., speed and acceleration) by peak and off-peak conditions for for different different roadway classifications. classifications. This information can be utilized by individuals or organizations for use in development and/or validation of fuel consumption and/or emissions emissions modeling. The tables that were produced with this analysis step step are similar to Table 1. The graphs show acceleration rates along the xx-axis axis and speed bins along the yaxis. Each cell contains the percent of operating time at the given acceleration and speed ranges along the travel time run.

35

 

36

 

CHAPTER IV. FINDINGS

Thi s chapter presents the major findings for This for the research study. Presented first are the findings that discuss the significant significant differences found found in compa comparing ring the fuel consumption model estimates from  both the instantaneo instantaneous us and average calculations calculations of both speed speed and and accelerati acceleration on characteri characteristic stics. s. The next section discusses the correlations and regression equations equa tions that were found between the speed and acceleration characteristics and geometric geometric characteristics. The findings with regard to the development of a speed and acceleration data base ba se are then prese presented. nted. Finally, the success of using DMI technology for data collection is discussed. Fuel Consumption Model Analyses

The first step of the fuel consumption analyses was to ensure that the data are from a normal distribution. The normality test was satisfied for for all the variables varia bles representing representing a difference between between the average a verage method and instantaneous method of fuel consumption estimation except for Raus’ model on freeways. Since Raus’ model is only good for velocities velocities ranging from from 1 to 35 mph (1.6 to 56 kph) it makes sense sense that normality may may not be found for freeway conditions. conditions. Once normality was discovered for the other models, statistical analyses a nalyses were performed. Since the difference between the average and instantaneous methods of fuel consumption estimation is of concern, a paired t-test is the appropriate analysis tool. T-tests were were performed performed on the different arterial classes (e.g., Class I, Class II, and freeways) at the aggregated level (i.e., not disaggregated by average speed, for example). example). The null hypothesis for the tests is that there is no differencee between the two methods of calculating fuel consumption. Therefore, if significance is differenc found, the null hypothesis can be rejected and a nd there is a difference difference between the two methods of fuel fuel consumption estimation. A critical level of significance significance of 5 percent was was used in the analyses to determine significance. significance. Results of the analyses are shown shown in Table 3. A cursory review of Table 3 indicates that Raus’ Ra us’ model did not yield significant difference differencess in fuel consumption estimation estimation for any of the functional classes. The FREQ10 FREQ1 0 models for freeways freeways and arterials were both found to be insignificant for the Class I arterials. arteria ls. The final model that demonstrates insignificant results is McGill for the Class II arterials.   It is interesting to note that, aalthough lthough insignificance was only found for a few situations shown in Table 3, 3 , some models demonstrate demonstrate significance significance for which the model does not have spee speed d data. For  example, FREFLO is utilized for modeling freeway freeway conditions, but it is based upon speed data in the range from 7 to 75 mph (11 (1 1 to 121 12 1 kph) (i.e., including inclu ding arterial speeds) and was found found to demonstrate a significant difference difference for for both arterial and freeway conditions. NETFLO NET FLO is also for for modeling arterials and is based upon speeds from 0 to 25 mph (0 to 40 kph), but was found to be significant significant for freeways. Finally, NETSIM, utilized u tilized for for simulating arterials, is based based upon speed data data from 0 to 75 mph (0 to 121 1 21 kph) k ph) and yields significant significant results for both arterial classes as well as freeways. freeways. The authors would like to alert the re reader ader that it is important important to look at Table 2 in union with with Table 3 when interpreting the results to realize what conditions the fuel consumption model is based upon

37

 

(e.g., speed range) and the functional functional class orientation for whic which h the model was developed. This ensures an accurate understanding and interpretation of the results. It is important to note that the analyses presented here compare only the average and instantaneous methods of fuel consumption consumption estimation for any one model. It does not compare the the models to one another, nor is it possible from this analyses to determine that any one model is better  than another model. However, by interpreting Tables 2 and 3 together, one can de determine termine which models mode ls demonstr demonstrate ate a signifi significant cant differenc differencee that can be attributed to the detailed detailed data set produced produced  by perform performing ing travel travel time time runs with with the the DMI.

Table 3. Probabilities Resulting From Comparing the Average and Instantaneous Methods of Fuel Consumption Estimation Probability for Different Functional Classes1 MODEL

Class I Arterials (n=841)

Class II Arterials (n=1020)

Freeways (n=1087)

Bowyer, Akcelik, and Biggs (1985)

0.0001

0.0001

0.0001

Biggs and Akcelik  (1986)

0.0001

0.0001

0.0001

Bester

0.0001

0.0001

0.0001

Raus

0.1000

0.0641

N/A2

Lindley

0.0001

0.0001

0.0001

FREQ10 (Freeways)

0.1257

0.0001

0.0001

FREQ10 (Arterials) McGill

0.2349 0.0001

0.0001 0.0592

0.0001 0.0001

FREFLO

0.0001

0.0001

0.0001

 NET FLO

0.0001

0.000 0.0001 1

0.0001 0.000 1

 NETSIM

0.0001

0.000 0.0001 1

0.0001 0.000 1

1

Shaded cells indicate indicate insignificance insignificance compared to a critical critica l level of significance of 2.5 percent ((twotwo  tailed test). 2 The data for this cell were not found to be from a normal distribution.

38

 

Regression and Correlation Analyses

The next objective of the study was to investigate regression equations between speed and acceleration characteristics, cha racteristics, geometric geometric characteristics, and traffic flow variability. variability. A concern prior  to beginning this evaluation was the time periods of the travel time runs that included the peak period conditions (i.e., relatively lower average speeds). The data were studied, including investigating investigating normal distributions of speed speed by the time of day of the travel time run, to notice reductions in average speeds due to the peak period conditions. conditions. From such analysis, the peak periods for for the travel time runs in Houston, Texas were taken as 6 to 9 a.m. and 4 to 7 p.m. Table 4 provides a summary of the operating characteristics that were discovered in the data set once once these peak and off-peak off-peak times were determined. determined. Summary statistics are shown for speed, speed, acceleration, and the coefficient coefficient of variation varia tion (CV) for for speed. The summary statistics include include the following values: average, maximum, minimum, standard deviation, average minimum, and average maximum. The average maximum and minimum values are the averages ffor or the maximum and minimum that were recorded for each travel time run section. The average avera ge value is the average recorded for each travel time run on a particular roadway segment. cursoryfor review of Table 4 indicates that The thereresearch is not ateam significant difference between thevalues A obtained peak and off-peak off-peak conditions. believes that the peak and a nd offoff  peak condit conditions ions are similar similar for freeways freeways since since there there were were numerous numerous miles miles of freeways freeways included included in the  peak period period that were not congeste congested. d. Thes T hesee freeways freeways were not congeste congested d since since they are further further away from the the urban area, and a nd some some experience experience nearly nearly free-f free-flow low condition conditionss during the peak period. period. The same argument can be made for the high off-peak speeds for arterial Class I and Class II segments as well. Signal timing can also affect the average speeds of Class I and Class II arterial segments along the corridor (e.g., optimized signal signal timing can increase average a verage speeds while poor poor signal timing can reduce the average avera ge speeds). speeds). In addition, the acceleration accelerat ion and CV variables varia bles are lower and the speeds are ar e higher for for freeway sections sections than for for arterials. This would imply that the arterial segments are generally experiencing more stop-and-go stop-and-go conditions w with ith hard brea breaking king (i.e., deceleration) and/or  relatively high accelerations. The stop-and-go conditions conditions are generally due to the signal timing along the corridor. Once the peak period conditions were determined, geometric characteristics were identified identified to use in the linear regression models. models. The following variables were utilized as independent variables: variables: signal density expressed expressed as the number of signals per mile, driveway density expressed as the nu number  mber  of driveways per mile, stop density express expressed ed as the number of stops per mile, length of the section in miles, miles, and the 24 hour volume volu me per lane. Signal density, driveway density, density, and stop density were not applicable for freeway sections. The dependent variables varia bles (i.e., speed and acceleration characteristics) that were used for model development were the sstandard tandard devia deviation tion of the acceleration (i.e., acceleration noise) in mph/sec, standard deviation of the speed in mph, average speed in mph, and the coefficient coefficient of variation expressed expressed as a percentage (standard deviation devia tion of the speed/average speed/average speed).

39

 

Table 4. Traffic Operating Characteristics for Roadway Roadway Classes During Peak and Off-Peak Periods

Traffic Stream Characteristics

All Freeways and Arterials

Arterials  Freeways

Class I1

Class II1

Peak 2

OffPeak 3

Peak 2

OffPeak 3

Peak 2

OffPeak 3

Peak 2

OffPeak 3

Average

42.0

39.6

54.7

57.2

38.1

38.5

28.8

28.8

Avg. Maximum

54.0

51.3

61.1

61.6

53.0

51.9

45.7

43.9

Avg. Minimum

22.6

17.8

46.4

51.2

8.7

8.8

2.9

3.7

Standard Deviation

  9. 9 .7

10.2

3.9

2.8

13.8

12.9

13.9

12.8

Speed (mph)

Acceleration (mph/sec)

Avg. Maximum

1.96

2.06

1.09

0.91

2.36

2.33

2.75

2.57

Avg. Minimum

-2.30

-2.44

-1.32

-1.10

-2.81

-2.68

-3.16

-3.11

Standard Deviation (Acceleration  Noise))  Noise

  0.64

0.66

0.42

0.37

0.74

0.72

0.85

0.80

Coefficient of Variation of Speed 4

     

 

Average

32.5

34.2

10.2

5.9

42.3

38.2

53.5

49.3

Maximum

235.6

125.5

111.9

86.2

132.8

125.5

235.6

120.0

Minimum

0.5

0.7

0.5

0.7

1.1

2.4

6.6

2.6

1

As defined by the 1994 Highway Capacity Manual (HCM). Peak Conditions are from 6 to 9 a.m. and 4 to 7 p.m. 3 Off-Peak Condition is from 10 a.m. to 3 p.m. 4 Coefficient of Variation = Standard Deviation of Speed/Average Speed. 2

40

 

 Normal distributio distribution n graphs of these these variables variables were then evaluated to look for useful useful stratific stratifications ations (e.g., for driveway densities densities greater than 20 per mile) for for model development. development. These stratifications were used when evaluating linear regression model development between the acceleration characteristics and geometric characteristics. characteristics. Models were were developed based based upon peak and a nd off-peak  off-peak  conditions, different functional functional classif cla ssifications, ications, and stratifications on some variables varia bles (e.g., for driveway densities greater than 20 per mile). mile). The linear regression models generally produced low R  2 relationships. In fact, the highest relationships achieved were less th than an 0.3 0.35 5 for any of the conditions studied. This model was often between between the dependent variable of average speed and the independent variables of signal signal density and/or driveway density for the arterial sections in either the peak or off peak condition conditions. s. The addition addition of independe independent nt variables after signal signal density density and drive driveway way density density (i.e.,  producing  producin g graphs with greater than three three independ independent ent variables) variables) often often resulted resulted in increasing increasing the the R  2 value only a few hundredths. Analysis of variance (ANOVA) (ANO VA) procedures were performed performed as part of  the linear regression using a critical level of significance of 5 percent for the independent variables in the relationships. The research team hypothesized that the driveway density and sig signal nal density variables would have the most explanatory power in such relationships and provide higher R  2 results.

For nearly all of the models developed, the signal density and driveway density variables were found to be significant signif in the ANOVA procedure. variables we contributing to the explanation of icant the variance within theprocedure. model. It isTherefore, interesting interestingthese to note that the were 24re hour volume volu me and the length of the section, the only independent variables used in the freeway analysis, were not always significant. The 24 hour volume produced significant results more often, howe however, ver, than the variable representing the length length of the section. section. This would indicate that the 24 hour volume was critical in many cases in explaining the variance in the freeway segments with the variables available. The research team suspected that floating car method did not measure the true affects of the driveway density on travel times. Since the floating car technique is based upon the premise that the driver will pass as many vehicles as pass the driver, it is possible that any significant affects of the driveway density density in the right-most right-most lane are not reflected in the results of the travel time runs. This could influence the results of the model model relationships. It was also important to note that there were not distinct differences between the peak and off-peak periods with respect to the R  2 values and significance of variables. This will be discussed further further in later sections when further analysis are described. Another observation that was made from evaluating the resulting models was the signs on the coefficients coeff icients of the independent variables. Often times, times, these signs signs did not make intuitive sense. sense. For  example, as the signal density went down, the coefficient of variation of speed (CV) would go up. In this example, it does not make mak e sense that the variation of the traffic speeds, speeds, represented by the CV, should go up when there is less interruption in the traffic stream (i.e., a lower signal density). However, this could indicate that along these arterial corridors the signal timing has been optimized optimized to provide sufficient green time and increased average speeds.

41

 

One point that should be discussed is the use of the variable representing the length of the segment. This variable was included into the analysis since it was hypothesized hypothesized that the length of a segment could explain some of the relationship with the acceleration and speed characteristics. Intuition would wou ld suggest that it is possible, given a longer or shorter section, the degree of aggregation incurred would produce significant results for for the variable. However, this variable was found found to be significant in only one or two cases. cases. This result, along with further analysis below that will be described, led the research team to believe that a variable variable explaining the location of the travel time run would possibly be more explanatory. Unfortunately, such a variable did not exis existt in the data set for the study. Closer investigation of the speed speed profiles indicated indicated that depending upon where a tra travel vel time section section was located there could be anticipated anticipa ted results. For example, lower speeds speeds may be incurred prior to a signal location on arterials or upstream of a lane drop on a freeway section. One final point that should be discussed is the the relationship between average speed and the 24 hour  volume. This relationship also produced an R 2 value less than 0.35 which is counter-intuitive. However, this low correlation results since the volume and speed data were not collected simultaneously. Project resources did not allow for for the collection of ssimultaneous imultaneous speed and volume data collection. Therefore, spee speed d data were collected with the DMI while 24 hour volume information was obtained from the Roadway Inventory. After the the geometric characteristics were evalua evaluated, ted, attention was focused focused upon the traffic flow 2 variability. This was studied with the aid of graphs producing R   values between average speed (independent variable) and the dependent variables of acceleration noise, standard deviation of the speed, and the coefficient coefficient of variation of the speed. One point on the figure represents the average speed and average avera ge operating characteristic (e.g., CV) for for a travel time run through a section. Figures 13 through 18 show these these relationship relationshipss for for selected selected conditi conditions ons.. Figures 13 and 14 are for freeway freeway sections (peak conditions), Figures 15 and 16 are for Class I arterials (off-peak conditions), and Figures 17 and 18 are for Class II arterials (peak conditions).

42

 

The relationship between average speed and the coefficie coefficient nt of variation provided relatively high R   values for all functional roadway classes, peak, and a nd off-peak off-peak conditions. Since CV is really a normalization with the average speed for a travel time run, the relationships are more explanatory. This normalization is evidently critical since the R 2 values decrease significantly when investigating the relationship between average speed and the standard deviation of the speed. Further, relationships utilizing average speed to predict the acceleration noise produced relatively lower R  2 results. This is probably due to the inherent averaging that occurs to determine the acceleration noise over a  particular travel travel time time run. 2

Investigation of Figure 13 indicates that there is a cluster of points at higher spe speeds eds with lower  coefficient of variations while at lower speeds the coefficients of variation indicate a much larger  range of values. The next portion of the analysis focused focused on investigating investigating the value of CV, average speed, relevant geometric characteristics, and the sspeed peed profile at a given speed to realize any  possible  poss ible trends trends that may may exist. exist. This evaluation was was performed performed by selecti selecting ng an average speed speed for for a given condition (e.g., peak period freeway sections) and evaluating a point below, near, and above the regression line. After the three points were identified, other runs on the same section were studied to ensure that the points selected were “typical” for the particular travel time run section. As the CV increased for the off-peak off-peak Class Cla ss I arterials, the driveway density and volume increa increased, sed, making intuitive sense. sense. However, signal density did not increase throughout as the CV increased. In fact, the signal density went down to 0.6 signals/mile for the point with the highest CV of 56  percent.  perce nt. In addition to some of the previous previous findings, findings, this also alludes to the fact that there may simply not be any clear relationships between the variables considered. For the peak period Class Cla ss I arterials, it is interesting to note that the point with the highest CV was from from a travel time run that concluded just ju st prior to a traffic traffic signal. Therefore, the section had no signal density and produced a relatively high CV since the vehicle was was slowing as it approached the intersection. This finding reinforces that the location of the section should be evaluated since theoretically one could have sections placed such that the CV could be just about abou t any value valu e (i.e., located anywhere along the speed  profile).  prof ile). Due to the inherent inherent variability variability in these these relations relationships hips,, and the geometric geometric versus versus speed speed and acceleration characteristics, char acteristics, developing developing estimating estimating regression equations is difficult. difficult. Similar results were found for other roadway classes and conditions. After the evaluations evalua tions above, consideration was again given to the relationships shown shown in Figures 13 through 18. 18 . Since the geometric geometric characteristics (e.g., number of lanes, signal density density)) do not change for a given section on which which travel time runs are a re being performed, performed, it is possible to aggregate the resulting speed and acceleration characteristics together for these these runs. This was performed and regression equations equa tions were were produced and the results are shown shown in Ta Table ble 5. The R  2 values in Table 5 are very similar, or slightly higher, than those produced when when each travel time run was plotted. This was expected since it produces a graph with fewer points that ar aree aggregated closer to the regression line. For illustrative purposes, Figures 19 and 20 2 0 show the resulting graphs ffor or average speed versus CV and acceleration noise, respectively, for peak period freeway conditions.  

46

 

Table 5. R 2 Values Observed Between Average Speed and Operating Characteristics for Different Roadway Classes R 2 Value of Linear Regression

Roadway Functional

Dependent Variable Regressed with

Class

Average Speed

Peak Period Conditions

Off-Peak Period Conditions

Coefficient of  Variation

0.87

0.84

Acceleration Noise

0.79

0.72

Standard Deviation of Speed

0.67

0.66

Coefficient of  Variation

0.85

0.73

Acceleration Noise

0.75

0.55

Standard Deviation of Speed

0.61

0.70

Coefficient of  Variation

0.84

0.78

Acceleration Noise

0.54

0.43

Standard Deviation of Speed

0.58

0.46

Coefficient of  Variation

0.68

0.68

Acceleration Noise

0.21

0.13

Standard Deviation of Speed

0.03

0.13

All Roadway Classes

Freeways

Arterials Class I

Arterials Class II

47

 

Roadway Operating Characteristics: Speed and Acceleration

The travel time and speed data collected for this study were summarized to obtain speed and acceleration distributions. These speed and acceleration distributions provide quantitative information about the operating characteristics of the the freeways freeways and arterial streets under study. These distributions are also very important in designing and validating the next generation of emissions models that are based upon acceleration patterns, not average speeds. The objective of this study was to quantify qu antify the speed and acceleration characteristics chara cteristics on a large sample of freeways and arterial arteria l streets in Houston, Texas, ideally representing conditions for other urban areas with similar roadway and development patterns. The following sections contain a description description of the speed, acceleration, and speed-acceleration distributions found on these freeways and arterial streets. Speed Distributions The speed distributions distributions are a quantitative qua ntitative summary of the percentage of time that was sspent pent in different 5 mph (8 kph) speed ranges. Figure 21 illustrates illu strates the speed distribution distribution for all freeway freeway and arterial street routes that were surveyed surveyed in this study. The distribution shows that 15 percent of the total trip time was spent traveling above 55 mph (89 kph), while 7 percent was spent in idle conditions (0 mph). Because this distribution combines combines freeway freeway and arterial street routes, the higher  speeds of freeway travel are mixed mixed with with lower lower speeds speeds typical of arterial streets. By disaggregating the data by functional class, the speed distributions for different functional classes become more distinctive. This study disaggregated the study study routes by freeways, freeways, Class I arterials, arteria ls, and Class II arterials. The speed distribution for the freeway freeway routes is shown in Figure 22. The speed distribution show showss a marked increase in travel speeds speeds above 55 mph (89 kph). For freeways, freeways, approximately 72 percent of the total freeway trip time is relatively free-flow free-flow [i.e., above 55 mph (89 kph)]. k ph)]. Another 14 percent of the trip time is between between 45 mph (72 (7 2 kph) k ph) and 55 mph (89 kph) k ph) (i.e., slightly congested congested conditions), and the remaining 14 percent of the trip time is spent spent in moderate to severe congestion congestion [i.e., below 45 mph (72 kph)]. The freeway speed distributions contrast sharply with the arterial street speed distributions shown shown in Figures 23 and 24. 24 . The arterial arteria l street distributions show speeds patterns more typical of  interrupted flow. The speeds are lower and more distributed over the entire entire speed speed range. For  instance, ins tance, 32 percent of the total trip time on Class I arterials was above 55 mph (89 (8 9 kph), k ph), while only 1 percent of the the total trip trip time on Class II arterials was above above 55 mph (89 kph). The arterial sstreet treet distribution also illustrate the additional idle time experienced at signalized signalized intersections: intersections: 7 percent of total trip time for Class I arterials, and 13 percent for Class II arterials. In summary, the speed distributions for different different functional classes cla sses w were ere markedly different, with freeways exhibiting higher speeds speeds and arterial streets exhibiting exhibiting lower speeds speeds and more idle time. The data shown in Figures 21 through throu gh 24 are for peak period conditions, conditions, or those times when congestion congestion was prevalent on the study routes. The data for off-peak off-peak period conditions (mid-day) (mid-day)

49

 

were also a lso examined, examined, and found to be similar similar to peak period conditions. Although the researchers researchers had hypothesized that a significant difference would exist between peak and off-peak period operating characteristics, the examination of speed distributions was unable to confirm the hypothesis.  Acceleration Distributions Similar to the speed distribution, the acceleration distribution examined in this study is a quantitative quantita tive summary of the acceleration rates [in 1 mph/s mph/sec ec (1.6 k kph/sec) ph/sec) ranges] experienced experienced during a trip. The acceleration distributions were calculated for all freeway freeway and arterial street routes, then disaggregated by functional class. The acceleration distribution for all freeway and arterial street study routes is shown shown in Figure 25. 25 . The figure shows that 60 percent of the total trip time on freeways and arterial streets was at 0 mph/sec, or steady-state conditions. conditions. These steady-state con conditions ditions could have occurred while the vehicle was idling at a traffic signal or while the vehicle was traveling at a constant speed on a freeway or arterial route. Accelerations account for another 21 percent of the trip time, whereas decelerations are a re 19 percent of the total trip. The maximum acceleration range was 2 to 3 mph/s mph/sec ec (3 to 5 kph/sec), and the maximum deceleration range was -4 to -5 -5 mph/sec (-6 to -8 -8 kph/sec). As with the speed distributions, the researchers hypothesized hypothesized that acceleration characteristics cha racteristics would vary  by functional functional class, class, with arterial streets streets exhibiting exhibiting a broader broader range of accelerations accelerations and deceleratio decelerations. ns. Figure 26 shows the the acceleration distribution for freeway routes only. The steady-state conditions now account for 66 percent of the total trip time, with the maximum acceleration range at 1 to 2 mph/sec (2 to 3 kph/sec) and the maximum deceleration range from -2 to -3 mph/sec (-3 to -5 kph/sec). The freeway distribution shows more time spent spent in steady-state steady-state conditions (traveling at constant speed) with less severe acceleration and deceleration rates. Figures 27 and 28 illustrate the acceleration distributions for Class I and II arterial streets, respectively. About 60 percent of total trip time for for Class I arterials is steadysteady-state, state, whereas only 54  percent  perce nt is steady-s steady-state tate for for Class II arterials. arterials. The maximum acceleration acceleration and decele deceleration ration rates are also comparable as well. In summary, the acceleration distributions for different functional classes where different different but not necessarily distinctive. The floating car method of data collection may have affected the true acceleration characteristics of different roadway types, thereby smoothing the potential acceleration/deceleration differences between freeways freeways and arterial streets. streets. The similarity of the distributions for different functional classes may also indicate that, indeed, only small difference difference exist  between  betw een acceleratio acceleration n characteristi characteristics cs for for differe different nt functional functional roadway classe classes. s.

54

 

3-Dimensional Speed-Acceleration Distributions The most useful distribution in the development of emissions and related modeling tools is the speed-accelerati on distribution, which shows the typical acceleration ra rates tes for various speed ranges. The speed-acceleration distribution combines the speed distributions (Figures 21 through 24) 24 ) with the acceleration distributions (Figures 25 through 28) to form form a 3-dimensional distribution. distribution. These distributions provide a quantitative summary of the speeds and accelerations accelera tions on a second-by-s second-by-second econd  basis..  basis Figure 29 illustrates the 3-dimensional speed-acceleration speed-acceleration distribution for all freeway and aarterial rterial street stre et routes in this study. The figure figure shows shows a large “peak” of the data at 60 mph (97 kph), k ph), with another smaller “peak” at a t 0 mph mph (steady-state). (steady-state). The acceleration and deceleration deceleration ranges close close to 0 mph/sec can also be seen on the figure as small “ridges.” Table 6, shown on the ffacing acing page of  Figure 29, 29 , shows the matrix matrix of the percent of total trip time in various speed and acceleration ranges used to create Figure 29. The 3-dimensional speed-acceleration distribution for freeway freeway routes only is shown in Figure 30. 30 . The figure clearly show the large proportion of travel that occurs in the 55 to 60 mph (89 to 97 kph) range with a small range in accelerations. a ccelerations. Ta Table ble 7 shows the matrix of total trip time in various speed and acceleration ranges used to create Figure 30 30.. Figures 31 and 32 show the speed-acceleration speed-acceleration distributions for Class I and II arterials, respectively. Like the speed distributions discussed discussed earlier, there is a marked ma rked difference between between functional classes. classes. Class I and II arteria arteriall streets show show smaller   but com comparab parable le spee speed d “peaks” “peaks” at 0 mph, or idle time. time. Tables 8 and 9 provide provide the percent percent of total trip time in various speed and acceleration ranges for Figures 31 and 32, respectively.

59

 

Table 10 summarizes the percent of time that the data collection vehicles were operating in the modes of idle, steady-state steady-state (cruise), acceleration, aand nd deceleration for each roadway functional class  based upon the informati information on contained contained in Tables 6 through through 9. A review review of Table 10 illustrate illustratess that the arterial streets experienced more time spent in idle, acceleration, and a nd deceleration modes and less time in steady-state conditions than observed on the freeway freeway segments. The research team believes that these characteristics are explained by the stop-and-go traffic flow due to signalized corridors.

Table 10. Percent of Time Spent in Each Each Operating Mode by Roadway Roadway Functional Functional Class

Vehicle Operating Mode

Roadway Roadwa y Functional Class Freeways and Arterial Streets

Freeways

Class I Arterials

Class II Arterials

6.4

0.4

9.2

12.0

Steady-state

53.6

65.7

49.1

41.9

Acceleration

21.3

17.5

22.6

24.9

Deceleration

18.7

16.4

19.1

21.2

Idle

68

 

CHAPTER V. CONCLUSIONS & RECOMM RECOMMENDATIONS ENDATIONS

This chapter presents presents the conclusions and recommendations based upon the findings described in Chapter Four. The discussion begins with conclusions drawn drawn from findings findings of statistical statistical analyses that were applied to the fuel consumption estimation estimation models. The discussion continues w with ith the relationships that were discovered when comparing the speed and acceleration characteristics, geometric characteristics, and fuel consumption consumption estimations. The potential uses for these relationships and prediction equations in planning applications is explained. The content and usefulness of the data base of operating conditions in the Houston, Texas area is then evaluated. The recommendationss conclude with discussion recommendation discussion of the usefulness of the DMI for data collection and the need for future research in the areas encountered throughout this research study. Fuel Consumption Model Comparisons

The findings presented presented in Ta Table ble 3 show the significance that the detailed data set has when applied to fuel consumption analyses. Significance was determined determined for many of the functional functional classes when comparing fuel consumption estimation based upon both the average and instantaneous methods. From these results it can ca n be concluded tha that, t, in general, significant significant differences can be expected when applying a detailed data set such as that produced by a DMI in a travel time run ru n to the estimation of  fuel consumption. consumption. It is important to note that when when reviewing the results results of Table 3, it is imperative to study Table 2 to verify the conditions (e.g., speed range, functional classification) for which a model is valid. Regression and Correlation Analyses

Development of regression equations between speed and acceleration a cceleration characteristics, geometric characteristics, and traffic flow variability was performed in the study. study. The regression equations did 2 not yield an R  higher than 0.35 when comparing any combination of the geometric characteristics with the speed and acceleration characteristics. chara cteristics. Signal density and/or driveway density w were ere found to be significant for most of the conditions evaluated with the aid of ANOVA procedure using a critical level of significance of 5 percent. The evaluation of traffic flow variability performed by regressing average speed with the independent variables of speed and acceleration characteristics resulted in the R  2 values shown in Table 5. These values clearly indicate that the CV of speed speed was a more explanatory indicator of the trip variability when regressed with average speed (i.e., yielded higher R  2 values) than the standard deviation of the speed or the acceleration noise.   Several factors that could account for the findings were considered. The true effect of the driveway density may not be reflected in the travel time data since the floating car method was utilized. It is possible possible that the influence of driveways on the right-most lane may not be included into a travel time run that tha t includes a driver passing as many vehic vehicles les as pass the drive driver. r. In addition, travel

variability induced by traffic signals is difficult difficult to quantify. Peak and a nd off-peak off-peak conditions often have 69

 

different signal timings to optimize traffic traffic flow. Average speeds, speeds, and motorist delay, will vary depending upon when when motorists motorists arrive at the traffic signal. signal. The location of the travel time run was was also found to be of importance when measuring measuring the coefficient of variation of the speed. If a travel time run is performed immediately prior to a traffic signal or lane-drop on a freeway, the results will differ compared to a run performed in an uninterrupted flow section. Unfortunately, the data base did not contain a variable relating to the section definition definition (e.g., before or after a traffic signal) of the travel time run, but this would be an interesting interesting element element for further further study. Finally, it was found found that, although acceleration noise is a better measure to determine the operating characteristics of a section than average a verage speed, there is still a significant portion of the instantaneous travel tra vel characteristics (e.g., speed, acceleration) that are lost when aggregating over an entire section. It was also a lso found that peak and a nd off-peak off-peak conditions seemed to operate very similarly (see Table 4). Although the peak and off-peak conditions were were carefully selected to def define ine these ranges, traffic traffic operation did not appear significantly significantly different different for the two two conditions. conditions. This apparent discrepancy is likely due to the inclusion of freeway routes into the peak-period analyses that are operating near  free-flow free-f low conditions during the peak-period. These freeways are thos thosee that are relatively far from the urban area. A similar similar concern exists for the arterial segments segments as well. In addition, the arterial arteria l sections could have similar operating characteristics cha racteristics due to the signal timings in the peak or off-peak  off-peak   periods.  periods. Roadway Operating Characteristics: Speed and Acceleration

This study examined the operating characteristics of freeway freeway and arteria arteriall street routes in Houston, with a specific specific emphasis on the speed speed and acceleration distributions. The travel time/speed data collected for this study showed a significant difference in the speed distributions for different functional classes (e.g., freeways, Class I arterials, Class II arterials). The result confirmed the obvious but also provided specific quantitative quantita tive evidence of the differences between classes classes.. The acceleration distributions for different roadway functional classes were less distinctive  between  betw een function functional al classes, classes, indicating indicating that acceleration acceleration characterist characteristics ics were similar similar between between freeways freeways and arterial streets. The floating car data collection technique used in this study study may have “smoothed” some of the acceleration differences between freeways and arterials streets, so a definitive statement cannot be made. A data collection method that obtains a representative sample of the range of operating characteristics of motorists (e.g., instrumenting random vehicles) would likely provide a more distinct difference difference between functional classes cla sses.. The study also a lso produced three-dimensional three-dimensional speed-acceleration distributions that were typical of  the freeway and arterial street system system in Houston, Texas. The speed-acceleration distributions do exhibit significant differences between freeways and arterial streets, mainly with respect to speed differences. differenc es. The speed and acceleration data set used to produce these ssummary ummary distributions is expected to be useful in validating the next generation of emissions models that are currently in the developmental stages.

70

 

The distribution data contained in this report was collected during the peak traffic period (6 to 9 a.m. and 4 to 7 p.m.). Data during du ring the off-pe off-peak ak traffic tra ffic period at mid-day mid-day (10 a.m. a .m. to 3 p.m.) were were also collected and analyzed, with the analyses indicating no significant difference between the peak  and off-pe off-peak ak data. Although Although it was expecte expected d that the peak and off-pe off-peak ak data would be distincti distinctive, ve, the study could not determine why these distinctions were not apparent. DMI Technology for the Data Collection Effort

The distance measuring instrument was found to be an a n invaluable invalua ble tool for performing this study. study. The instantaneous data points provided at every ½-second yield a data set that allows for detailed speed and a nd acceleration information. From this data, the significance significance of the instantaneous instantaneous data set on estimating fuel consumption could be evaluated, regression equations were studied, and traffic operating distributions distribu tions could could be prepared. The ASCII format format of the output was easily manipula manipulated ted for analyses and evaluation. Data collection methods, ssuch uch as a DMI or global positioning ssystems ystems,, that produce these instantaneous instantaneous speed and acceleration data will continue to prove to be useful useful in the transportation community for application to many transportation concerns such as air quality qua lity and traffic operations. Future Research Needs

The study identified some some areas where additional research is needed. The first is the need ffor or the development of mobile source emissions models that can incorporate acceleration characteristics. Research of this kind is currently in progress. There is a need for better characterization of acceleration characteristics for different roadway facilities. Characterizing Chara cterizing acceleration chara characteristics cteristics by percent of time in a particular particul ar driving condition (e.g., idle, cruise, acceleration, or deceleration) is useful for the development of appropriate driving cycles that replicate these conditions. There is much variability var iability both along a travel tra vel time run and between travel time runs along sections. Additional research is needed that focuses on determining appropriate methods to quantify this variability in a consistent and meaningful manner (e.g., separate the driver and traffic influences). In general, the DMI and similar technologies for data collection, allow for larger amounts of  descriptive data that has not been possible possible in the past. Research must now begin to focus focus on  performacn  perf ormacnee measures that are best utilized (e.g., coeffici coefficient ent of variation) for quantifying quantifying the aggregation of this data for transortation-related concerns.

71

 

72

 

REFERENCES

1. T. R. Jones and R. B. Potts, Potts, “The Measurement Measurement of of Acceler Acceleration ation Noise-Noise--A A Traffic Traffic Parameter,” Parameter,” Operationss Research Operation Research , Nov.-Dec., Nov.-Dec., 1962. 19 62. 2. E. W. Montroll Montroll and R. B. Potts, Potts, “Car Following Following and Accele Acceleration ration Noise, Noise, An Introduction Introduction to Traffic Flow Flow Theory,” Highway Research Research Board Special Report Report 79 , Washington Washington D.C., 1964. 1964 . 3. D. R. Drew and C. L. Dudek, Dudek, “Inves “Investigation tigation of an Internal Internal Energy Energy Model Model for for Evaluating Freeway Freeway Level of Service,” Research Report No. 24-11, 24-11 , Texas Transportation Institute, Texas A&M University, 1965. 4. T. Ryden, “Measuri “ Measuring ng Levels Levels of Service Service of a City City Street Street by Using Using Energy-Mo Energy-Moment mentum um Techniques,” Transportation Tra nsportation Research Record 567 , Transportation Research Research Board, Washington D.C., 1976. 5. R. Herman, E. W. Montroll, Montroll, R. B. Potts, Potts, and R. W. Rothery, Rothery, “Traffic Dynamics: Analysis Analysis of  of  Stability Stabilit y in Car following,” following,” Operations Operations Research Research , Vol. 7, 1959. 1959 . 6. E. W. Montroll, Montroll, “ Accelerat Acceleration ion Noise Noise and Clusterin Clustering g Tendency Tendency of Vehicular Vehicular Traffic Traffic,” ,” Theory of  Traffic Flow, Edited by R. Herman, Elsevier Publishing Company, 1961. 7. K. Shoarian-Satt Shoarian-Sattari ari and D. Powell, Powell, “ Measured Measured Vehicl Vehiclee Flow Flow Parameters Parameters as Predict Predictors ors in Road Traffic Accident Studies,” Traffic Engineering & Control, June 1987. 8. N. J. Rowan, “ An Invest Investigation igation of Accel Acceleratio eration n Noise Noise as a Measure Measure of the the Quality of Traffic Service on Major Streets, “ Doctoral Dissertation, Civil Engineering, Texas A&M Univers University, ity, 1967. 9. W. Helly Helly and P. G. Baker, “ Acceleratio Acceleration n Noise Noise in a Congested Congested Signaliz Signalized ed Environm Environment,” ent,” Proceedings of the Third International Symposium on the Theory of Traffic Flow, 1965. 10.

B. D. Greens Greenshield hields, s, “The “The Quality Quality of of Traffic Traffic Flow,” Flow,” Qualit Quality y and Theory Theory of of Traffic Traffic Flow Flow ,” Bureau of Highway Traffic, Traffic, Yale, 1961. 19 61.

11.

R. C. Eff Effaa and L. L. C. Lars Larsen, en, “ Develo Developme pment nt of of RealReal-Wor World ld Driv Driving ing Cycles Cycles fo forr Estim Estimatin ating g Facility-Specific Emissions from Light-Duty Vehicles,” Vehicles,” California Environmental Environmental Protection Agency Air Resources Resources Board, Presentation at the Air and Waste Management Association Specialty Conference on the Emission Inventory: Perception and Reality, October 1993. 199 3.

12.

R. Benz, Benz, D. Morris Morris,, and E. Crowe Crowe,, “ Houston Houston-Gal -Galves veston ton Region Regional al Transport Transportatio ation n Study,” Study,” Texas Transportation Transportation Institute, Institute, Report Number 7-1986-5, May 1995. 199 5.

73

 

13.

G. Thurgood Thurgood,, “ Develo Developme pment nt of of a Free Freeway way Conges Congestio tion n Index Index (FCI) (FCI) Usi Using ng an Instrum Instrument ented ed Moving Vehicle,” Vehicle,” Presented at the Transportation Tra nsportation Research Board 74th Annual Meeting, January 1995.

14.

L. C. Larsen, Larsen, Californ California ia Air Resource Resourcess Board, Board, Telepho Telephone ne Convers Conversatio ations ns,, 1995.

15.

A. Pela, Pela, Californ California ia Air Resourc Resources es Board, Board, Telepho Telephone ne Conver Conversat sation ions, s, 1995.

16.

M. Barth, Barth, J. Norbe Norbeck, ck, and and T. Youngl Younglove ove,, “ A Prog Program ram to Coll Collec ectt Mass Massive ive Activ Activity ity Data in in  Northern California: California: Preliminary Preliminary Instrumented Instrumented Vehicle Study,” Californi Californiaa Air Resources Resources Board, January 1995. 1995 .

17.

D. LeBlanc LeBlanc,, F. F. Saunde Saunders, rs, M. Meyer, Meyer, and R. Guens Guensler ler,, “ Driving Driving Pattern Pattern Variabi Variabilit lity y and and Impacts on Vehicle Vehicle Carbon Monoxide Monoxide Emissions Emissions,” ,” Transportation Transportation Research Research Record Record 1472 , Transportation Research Board, Washington D.C., 1995.

18.

U. S. Environmental Environmental Protection Protection Agency, Agency, “Federal “Federal Test Test Proce Procedure dure Review Review Project: Project: Preliminary Preliminary Technical Technic al Report,” Report,” May 1993.

19.

T. Winzer Winzer,, “ Measurem Measurement ent of Accel Accelerat eration ion Noise Noise and and Disc Discuss ussion ion of the Energy Energy Model Model Developed Develo ped by by Drew,” Drew,” Transportation Transportation Research Research , Vol. 15A, no. 6., June 1981 1981..

20.

M. J. J. St. St. Denis, Denis, P. Cicer Ciceroo-Fer Fernand nandez, ez, A. M. Winer, Winer, J. W W.. Butler Butler,, and G. Je Jesi sion, on, “Effe “Effects cts of  In-Use Driving D riving Conditions and Vehicle/Engine Operating Para Parameters meters on ‘Off-Cycle’ ‘Off-Cycle’ Events: Comparison with Federal Test Procedure Conditions,” Air and Waste Management Associatio Ass ociation n Journal, Vol. 44, January 1994. 1994 .

21.

U. S. Environ Environme mental ntal Protect Protection ion Agenc Agency y (EPA), (EPA), EPA Electr Electroni onicc Bulletin Bulletin Board, Board, TTN Technical Technic al Areas: Areas: Light Light Duty, 1995. 1995 .

22.

J. Markey, United United States Environmen Environmental tal Protection Protection Agency, Agency, Telephone Telephone Conversatio Conversations, ns, 1995.

23.

Coordin Coordinatin ating g Research Research Council Council,, Inc. On-Ro On-Road ad Vehicl Vehiclee Emiss Emission ion Worksh Workshop op Procee Proceedin dings gs (1991-1995).

24.

R. Akcel Akcelik, ik, C. Bayley, Bayley, D. P. P. Bowy Bowyer, er, and D. C. Bigg Biggs, s, “A Hierarchy Hierarchy of of Vehicle Vehicle Fuel Consumption Models,” Traffic Tra ffic Engineering & Control , Vol. 24, No. 10, October 1983. 198 3.

25.

D. P. Bowyer Bowyer,, R. Akceli Akcelik, k, and D. C. Biggs Biggs,, “ Guide Guide to Fuel Fuel Cons Consump umpti tion on Anal Analys yses es fo forr Urban Urban Traffic Management,” Management,” Australian Australian Road Research Research Board Special Special Report SR No. 32 , 1986.

74

 

26.

An and and M. Ross, Ross, “ Model Model of of Fuel Fuel Econom Economy y with with Appl Applicat ication ionss to Driv Driving ing Cycl Cycles es and and Traffi Trafficc Management,” Manageme nt,” Transportatio Transportatio n Research Research Record Record 1416 141 6 , Transportation Transportation Research Research Board, Washington Washin gton D.C., 1993. 1993 .

27.

K. S. Rao and and R. A. Kramm Krammes es,, “ An Energ Energyy-Bas Based ed Fuel Fuel Consum Consumpti ption on Model Model for for FREFLO, FREFLO,”” Texas Transportation Tra nsportation Institute, Presented at the 73rd Annual Meeting of the Transportation Research Rese arch Board, January 1994. 1994 .

28.

D. C. Biggs, Biggs, “Fuel Consumpt Consumption ion Estimat Estimation ion Using Using ARFCOM.” ARFCOM.” Proceed Proceeding ingss of 14th Australian Road Research Board Conference, Part 3, 1989.

29.

D. C. Biggs and R. Akcelik, Akcelik, “ An EnergyEnergy-Relate Related d Model Model of Instantan Instantaneous eous Fuel Consumpti Consumption,” on,” Traffic Engineeri Engineering ng & Control, Vol. 27, No. N o. 5, 1986. 19 86.

30.

C. Best Bester, er, “Fuel “Fuel Cons Consumpt umption ion on Con Conges gested ted Freew Freeways ays,” ,” Transpor Transportation tation Researc Research h Rec Record ord 801 , Transportation Research Board, Washington D.C., 1981.

31.

J. Raus, Raus, “ A Metho Method d for for Estim Estimatin ating g Fuel Consum Consumpti ption on and and Vehic Vehicle le Emis Emissi sions ons on U Urban rban Arterials and Networks,” Networks,” U. S. Department of Tra Transportation, nsportation, Report No. FHWA-TS-81-210, FHWA-TS-81-210 , April 1981.

32.

J. A. Lindle Lindley, y, “ A Met Method hodolo ology gy ffor or Quanti Quantify fying ing Ur Urban ban Free Freeway way Conges Congestio tion,” n,” Trans Transpor portati tation on Research Rese arch Record 1132 , Transpo Tra nsportation rtation Research Research Board, Washington Washington D.C., 1987. 19 87.

33.

A. May, May, M. Crott Crotty, y, and M. Morris Morris,, “Texas “Texas Worksh Workshop: op: A Basic Basic Course Course in FR FREQ10,” EQ10,” Inst Institute itute of Transportation Studies at the University of California at Berkeley for the Texas State Department of Highways and Public Transportation, June 1991.

34.

R. McGill McGill,, “Fuel “Fuel Consum Consumpti ption on and and Emis Emissi sion on Values Values fo forr Traffic Traffic Model Models, s,”” U. S. S. Dep Departm artmen entt of Transportation, Tra nsportation, Federal Highway Administration, Administration, National Nationa l Technical Information Service, 1985.

35.

TRAF User User Refe Referen rence ce Guide, Guide, U. U. S. Departm Department ent of Transp Transportat ortation ion,, Federal Federal Highw Highway ay Administration, Office of Safety and Traffic Operations R&D, Version 4.0, April 1993.

36.

H. Lieu, Lieu, TurnerTurner-Fai Fairb rbanks anks,, Inc., Inc., Telep Telepho hone ne Conv Conver ersa sati tion ons, s, 1995.

37.

A. May, May, L. Leiman Leiman,, and B. Ostro Ostrom, m, “FREQ10 “FREQ10 Modif Modificat ication ions: s: Emis Emissio sions ns Facto Factors rs,, Gasoline Gasoline Consumption, and Growth Factors,” Institute of Transportation Studies at the University of  California Califor nia at Berkeley, June 1991. 199 1.

38.

K. Rao, Texas Transpor Transportati tation on Institut Institute, e, Perso Personal nal Interv Interview iews, s, 1995.

75

 

39.

P. Cic Cicer eroo-Fe Fern rnan ande dezz and and J. R. Long Long,,   “Modal Acceleratio Acceleration n Testing on Current Technology Technology Vehicles,” California Air Resources Resources Board, Presentation at the Air and Waste Management Association Asso ciation Specialty Conference on the Emission Emission Inventory: Perception and Reality, October  1993.

4 0.

C. E. Chapin,   “Mobile Source CO Modeling--How Modeling--How Good Is It?” Presentation at the 86th Annual Meeting & Exhibition Exhibition of the Air & Waste Management Ass Association, ociation, June 13-18, 1 3-18, 1993. 199 3.

41.

K. L. L. Vaugh Vaughn, n, “ An Emis Emissi sion onss Mode Modell for for Arte Arteri rial al Str Stree eets ts,, ”  Gradu Graduate ate Student Papers on Advanced Traffic Management Systems, Systems, Texas A&M University, Civil Engineering Engineering Department, Departme nt, August 1992. 199 2.

42.

P. Cicer Ciceroo-Fern Fernande andezz and and J. J. R. Long, Long, “ Grades Grades and Other Other Loads Loads Ef Effe fects cts on On-Road On-Road Emissions:: An On-Board Analyzer Study,” California Air Resources Board. Presentation at Emissions the Fifth Coordinating Research Council, Inc. On-Road Vehicle Emission Workshop, April 1995.

43.

R. Guensler Guensler,, School School of Civil Civil and Envir Environm onment ental al Enginee Engineerin ring, g, Georgi Georgiaa Institute Institute of of Technol Technology ogy,, Telephone Conversations, 1995.

44.

P. Cice Ciceroro-Fer Fernand nandez ez and J. R. Long, Long, “Instantan “Instantaneous eous and Short Short Term Term Emiss Emission ion Rates Observed During Modal Acceleration Acceleration Testing,” Testing,” California Air Resources Board, Presentation at the Fourth Coordinating Research Council, Inc. On-Road Vehicle Emission Workshop, March 1994.

45.

R. T. Gammarie Gammariello llo and and J. R. R. Long, Long, “ An Emi Emiss ssion ionss Co Comp mparis arison on Betw Betwee een n the the Unif Unified ied Cycle Cycle and the Federal Test T est Procedure,” Presentation at the Air and Waste Management Association Association Specialty Conference on the Emission Inventory: Perception and Reality, October 1993.

46.

“ Hi High ghwa way y Capacity Capacity Manual: Manual: Spe Specia ciall Repo Report rt 209.” 209.” Transpor Transportati tation on Res Researc earch h Board, Board, Natio National nal Research Council, Washington, D.C., Third Edition, 1994.

47.

R. J. J. Benz, Benz, J. T. Culliso Cullison, n, and N. Vaidy Vaidya. a. “Compute “Computerr Aide Aided d Transp Transport ortatio ation n Sof Softwar twaree CATS).” Unpublished Computer Software. Software. Texas Transportation Institute. Houston, Texas. 1996.

Sponsor Documents

Or use your account on DocShare.tips

Hide

Forgot your password?

Or register your new account on DocShare.tips

Hide

Lost your password? Please enter your email address. You will receive a link to create a new password.

Back to log-in

Close