Clemson University
TigerPrints
All Theses
Theses
5-2017
Annual Average Daily Traffic (AADT) Estimation
with Regression Using Centrality and Roadway
Characteristic Variables
McKenzie Keehan
Clemson University, mkeehan@clemson.edu
Follow this and additional works at: https://tigerprints.clemson.edu/all_theses
Recommended Citation
Keehan, McKenzie, "Annual Average Daily Traffic (AADT) Estimation with Regression Using Centrality and Roadway Characteristic
Variables" (2017). All Theses. 2644.
https://tigerprints.clemson.edu/all_theses/2644
This Thesis is brought to you for free and open access by the Theses at TigerPrints. It has been accepted for inclusion in All Theses by an authorized
administrator of TigerPrints. For more information, please contact kokeefe@clemson.edu.
ANNUAL AVERAGE DAILY TRAFFIC (AADT)
ESTIMATION WITH REGRESSION USING CENTRALITY
AND ROADWAY CHARACTERISTIC VARIABLES
A Thesis
Presented to
the Graduate School of
Clemson University
In Partial Fulfillment
of the Requirements for the Degree
Master of Science
Civil Engineering
by
McKenzie Keehan
May 2017
Accepted by:
Mashrur Chowdry, Committee Chair
Wayne Sarasua
Eric Morris
ABSTRACT
Accurate estimation of annual average daily traffic (AADT) is critical in nearly every
roadway decision, such as allocations of funding for roadway improvements and maintenance.
While some roadway locations have permanent count stations capable of counting vehicles
24-hours a day throughout the entire year, they are typically only installed at selected
locations on major roadways (i.e., freeways and major arterials) with high traffic volumes.
On lower functional class roads and roadway segments on higher functional class roads
without permanent count stations, short-term coverage counts are collected and adjusted
with data from permanent count stations to estimate AADT. Short-term coverage counts
are essential because they provide data from roadways of all functional classes and lane
configurations, accounting for varying volumes on all roads maintained by an agency.
Although necessary, coverage counts can be expensive and can exhaust resources such as
investment in data collection workforce, equipment and data analysis. This study develops
a strategy for estimating AADT on every roadway within a given jurisdiction using
permanent count stations and short term coverage counts, while limiting the number of
coverage counts needed. The goal of this thesis is to illustrate a noteworthy time and cost
savings using a new centrality based AADT estimation method. A set of new deterministic
variables, based on the theory of centrality, are introduced. This study revealed that
estimated root mean square error (RMSE) for the new centrality based AADT method is
half of the estimated RMSE in the travel demand based AADT model for the same area.
Additionally, it was found that using centrality based AADT estimation model, the number
of coverage count stations necessary can be reduced by more than 60% compared to the
ii
standard factor method for AADT estimation without compromising the AADT estimation
accuracy.
iii
DEDICATION
I would like to dedicate this thesis to my mother and father, for their unconditional
love and support.
iv
ACKNOWLEDGMENTS
I would like to express so much gratitude and appreciation to my advisor, Dr. Mashrur
Chowdhury, for constantly challenging me and aiding in my growth as a researcher and
graduate student. During my last two years under his wing, I have seen myself evolve in ways
I could have never imagined, and it has been due to his relentless support and guidance. He has
helped me to exceed expectations I had previously set for myself, and has helped me to realize
that there are no limitations to our potential.
I would also like to thank Dr. Wayne Sarasua and Dr. Eric Morris for serving as my
thesis committee members, and providing valuable feedback throughout my research. Dr.
Sarasua has also taught me in several courses and has proved to be a passionate professor who
cares deeply about the success of each of his students.
I appreciate the administrative staff members from the Glenn Department of Civil
Engineering, including Kristi Baker, Sandi Priddy, CJ Bolding, Karen Lanning, and Monica
Hughes, for helping me with all of the managerial and organizational aspects of my position
and my needs as a student.
I would like to extend my deep appreciation to Dr. Kakan Dey for being an amazing
mentor through his continuous effort to improve the research quality and for always being
there whenever I needed him. I would like to thank Md Mizanur Rahman, Sakib Kahn, and
Sababa Islam, for being the best mentors a fellow student could ask for. Their kindness and
expertise have done so much for me as I student and I would not have been able to be as
successful in my student career if not for them. Moreover, I would like to thank Md Mhafuzul
v
Islam and Md Zadid khan for helping me with many projects, including class projects,
competitions, and presentations, which were all made better due to their help and guidance.
I would like to thank the South Carolina Department of Transportation (SCDOT)
for funding a lot of the work I’ve done throughout my graduate student career, and
providing me with the data that were necessary for my research.
I cordially thank my parents, my siblings, and my entire family for being the support
system for me during any critical times. They have always been my inspiration to reach my
goals.
vi
TABLE OF CONTENTS
Page
TITLE PAGE .................................................................................................................. i
ABSTRACT .................................................................................................................... ii
DEDICATION............................................................................................................... iv
ACKOWNLEDGEMENTS .......................................................................................... v
TABLE OF CONTENTS ............................................................................................ vii
LIST OF FIGURES ...................................................................................................... ix
LIST OF TABLES ......................................................................................................... x
INTRODUCTION.......................................................................................................... 1
1.1 Problem Statement.............................................................................................. 1
1.2 Objective of the Thesis........................................................................................ 2
1.3 Organization of the Thesis ................................................................................. 2
LITERATURE REVIEW ............................................................................................. 4
2.1 Overview ............................................................................................................... 4
2.2 AADT Estimation Methods................................................................................. 4
2.3 AADT Estimation through Centrality ............................................................... 9
RESEARCH METHOD .............................................................................................. 11
ANALYSIS AND RESULTS ...................................................................................... 21
4.1 AADT Estimation Using Centrality and Linear Regression ......................... 21
4.2 Comparison of AADT Estimation using Centrality based Model and
Traditional Travel Demand Model Output ........................................................... 23
vii
Table of Contents (Continued)
Page
4.3 Possibility of Reduction of Coverage Counts using Centrality based AADT
Method ...................................................................................................................... 26
4.4 Cost Savings Analysis ........................................................................................ 29
CONCLUSIONS AND RECOMMENDATIONS ..................................................... 32
5.1 Conclusions ......................................................................................................... 32
5.2 Recommendations .............................................................................................. 33
REFERENCES............................................................................................................. 36
APPENDIX A .............................................................................................................. A1
viii
LIST OF FIGURES
Page
FIGURE 1 Method of Origin-Destination Centrality Calculation .......................... 11
FIGURE 2 Map of External Gateway Points ............................................................ 14
FIGURE 3 Greenville Parcels with Relative Weights .............................................. 16
FIGURE 4 External-to-External Centrality Map ..................................................... 17
FIGURE 5 Internal-to-External Centrality Map...................................................... 17
FIGURE 6 Internal-to-Internal Centrality Map ...................................................... 18
FIGURE 7 Comparison of AADT Estimation Methods ........................................... 23
FIGURE 8 Method to Determine Coverage Count Reduction Potential................ 24
FIGURE 9 Reduction of Coverage Counts by Using the Centrality Method ........ 26
ix
LIST OF TABLES
Page
TABLE 1 Shapefiles and Attributes Used in the Model Development ................... 12
TABLE 2 Three Stress Centrality Types and Associated Inputs ............................ 15
TABLE 3 Regression Summary Statistics ................................................................. 20
TABLE 4 Comparison of Model Inputs for Centrality Method and TDF Model . 21
TABLE 5 Comparison of Model Steps for Centrality Method and TDF Model ... 21
TABLE 6 Trials for Reduction of Coverage Counts ................................................ 25
TABLE 7 Cost Savings Analysis................................................................................. 27
x
CHAPTER ONE
INTRODUCTION
1.1 Problem Statement
Annual average daily traffic (AADT) is defined as the average daily measure of the
total volume of vehicles on a roadway segment over a year. Traffic volumes are the lead
indication travel demand and utilization of the roadways within a specified network (1, 2).
Therefore, accurate traffic volume estimations are critical in nearly every roadway decision.
While some roadway locations have permanent count stations capable of collecting
vehicle volumes 24-hours a day throughout the entire year, they are very costly to
implement and maintain, and are typically only provided at selected roadway segments.
Therefore, short term coverage counts are taken at thousands of strategically placed
roadway segments of all functional classes and lane configurations, accounting for varying
volumes on the roads maintained by an agency (3). The counts at these locations are usually
only collected once a year or every few years with pneumatic tubes, and the data collection
period can last from 24 hours to 7 days, depending on the responsible transportation
agency’s policy and reporting requirements (3, 4). For short-term count stations, the
permanent count stations serve as control counts, and are used to determine daily, monthly,
and seasonal factors to calibrate the data collected at short-term count stations in estimating
AADT.
However, operating and estimating AADT using coverage counts can be expensive
and can exhaust a significant amount of resources in terms of manpower, equipment, and
1
data analysis. In addition, the data provided by these counts is limited, and not always
sufficient in predicting accurate AADT values. For example, a 24-hour count on one day
throughout the entire year may need to be used to calculate AADT using only the one count
and calculated daily and monthly factors using the factor method, as further explained in
the Chapter 2: Literature Review. Any inconsistencies between the day’s data and the other
elements of that day and month may result in inaccurate AADT estimation.
1.2 Objective of the Thesis
This thesis aims to develop a unique means of estimating AADT on roadways
within a given jurisdiction, while limiting the required amount of implementation time and
effort. All of the data used in development of the AADT estimation method is readily
available. The motivation for this research is to develop an AADT estimation method,
which any jurisdiction can implement with minimal time and effort.
1.3 Organization of the Thesis
Chapter 2 describes the background research and literature on the currently
available AADT estimation methods. Some of these methods are currently being utilized
by state and local transportation agencies, while others are simply being used in academia
for research purposes. Additionally, Chapter 2 discusses the theory of centrality, its
applications, and its potential in AADT estimation. Chapter 3 discusses the method for
applying the theory of centrality in AADT estimation used in this thesis. The steps used in
the application are outlined and explained in detail.
2
Chapter 4 provides a detailed explanation of the linear regression model produced
by this method. Additionally, Chapter 4 compares the newly developed model to an
existing travel demand forecasting model for the city, both conceptually and statistically.
Next in Chapter 4, a procedure is performed in order to determine if this method could
reduce the number of short term count stations that the city of Greenville should use in
order to produce AADT estimates for all current short term count locations. Finally in
Chapter 4, a cost savings analysis is performed to determine the financial competence of
this method. Chapter 5 concludes this thesis, offering conclusions and recommendations
for further research.
3
CHAPTER TWO
LITERATURE REVIEW
2.1 Overview
Because quality AADT estimation on local roads is vital, ample amounts of
research have been carried out in order for models to be developed that can adequately
estimate the AADT of every roadway within a given area (2,5,6), which is discussed further
in this chapter. It is important to note the amount of time and costs associated with these
often inaccurate or inconsistent methods.
In this chapter we have reviewed AADT estimation methods and the theory of
centrality, as well as its feasibility in AADT estimation.
2.2 AADT Estimation Methods
Although collecting traffic count data every day in a year is the most accurate
method of calculating AADT, it is not economically feasible to install and maintain data
collection systems on a widespread scale. Traditionally, for roadways without permanent
count stations, AADT is calculated using the America Association of State Highway and
Transportation Official’s (AASHTO’s) factor method. Using this method, daily and
monthly factors are calculated using permanent counts stations, and typically each
permanent counts station is associated with a group of short term count locations, based on
clustering. 24-hour traffic counts from short term count stations are then multiplied with
the adjustment factors to find AADT, as shown in the following equation:
4
Eqn. 1
where V24ab = 24-hour volume for day a in month b (vehs); DFa = daily adjustment factor
for day a; and MFb = monthly adjustment factor for month b (1). Another AADT
estimation method, used by some states and implemented in programs such as the Traffic
Count Database System (7), used additional factors such as a seasonal factor and an axelcorrection factor, as shown in the following equation:
Eqn. 2
where V24ab = 24-hour volume for day a in month b (vehs); SFab = seasonal adjustment
factor for day a and month b; and AFc = applicable axel-correction factor for c number of
axels (7).
However, these formulas can only be used on roadway segments that have short
term count locations, which are not used to collect data on most publically maintained
roadways within a given state or jurisdiction. Therefore, ample amounts of research have
been carried out in order for models to be developed that can adequately estimate the
AADT of every roadway within a given area (2,5,6).
Several studies have shown that linear regression that utilizes roadway
characteristics and socioeconomic factors at short term count locations can be used to
estimate AADT (2,5,6,8). Doustmohammadi et al. developed a linear regression model to
calculate AADT using a variety of socio-economic factors and roadway data for small and
medium sized urban communities in Alabama (5). The significant variables identified in
the final two models (one for a small city and one for a large city) included the functional
classification (FCLASS) and lane counts of roadway segments (LANE), in addition to
5
population,
retail
employment
(RETAILEMBUFF),
and
all-other
employment
(NONRETAILEMBUFF) all inside a 0.25 mile radius around the traffic count site. These
two models are shown below:
Model 1 (R2 = 0.82): AADT = -5625 + 8493 FCLASS + 219 LANE - 1.16 POPBUFF
-0.58 NONRETAILEMBUFF + 11.55 RETAILEMBUFF
Model 2 (R2 = 0.79): AADT = -12590 + 4479 FCLASS - 1.15 POPBUFF – 0.86
NONRETAILEMBUFF +7.91 RETAILEMBUFF
It was concluded that their AADT estimation models can accurately estimate
AADT on desired roadways in cities of similar populations (5). However, the accuracy and
means of collecting the socio-economic factors used are resource intensive and difficult for
annually updating AADT. For example, the population data was obtained from the Census
Department, where data is only updated every 5 years (9). In this study, the other two socioeconomic factors - retail and all-other employment – were found from case studies that
were completed as a part of long-range transportation plans recently completed by a third
party organization. This socio-economic data can be time consuming to collect, and relies
on sources not maintained by the DOT. Additionally, socio-economic factors determined
by surveys or sampling may not be accurate or capable of validation.
Zhao et al. used regression and further statistical analysis to find factors
supplemental to AADT estimation in a study conducted for Florida. The study used
geographic information system (GIS) technology to investigate various factors that may be
good predictor of AADT on a road (2). A variety of land-use and accessibility
measurements were developed and tested in (2). The four models developed achieved R2
6
values from 0.66 to 0.82. The variables for each are: i) Model 1: Lane count, functional
class, access to employment centers, directness of access to expressways, and employment
inside a 0.25 mile radius (R2= 0.8180); ii) Model 2: Lane count, access to employment
centers, directness of expressway access, and network distance to the mean centers of
population (R2= 0.6607); iii) Model 3: Lane count, access to regional employment centers,
directness of expressway access, network distance to the regional mean centers of
population, and population inside a 0.25 mile radius around a traffic count site (R2=
0.7624); and iv) Model 4: Lane count, access to regional employment centers, directness
of expressway access, network distance to the regional mean centers of population,
employment inside a 0.25 mile radius around a traffic count site, and population inside a
0.25 mile radius around a traffic count site (R2= 0.7648). During their data collection
efforts, the author used employment data, which was purchased from a third party,
containing the number of employees at each business location and the standard industrial
classification code. In this case, the data had been purchased for a prior project, however,
for most DOTs, implementing this method would not only require attaining or purchasing
additional data not readily available to their agency, but also relying on data not operated
and maintained by the DOT and only updated at the third party’s discretion.
Wang et al. presented a means to estimate AADT for roadways through travel
demand forecasting (6). A main factor of applying the travel demand model involved using
land-use data at the parcel level to determine estimated trips produced from or attracted to
each parcel. All-or-nothing trip assignment was conducted using free-flow travel times.
Then, the trips were dispersed through a trip distribution gravity model at the parcel-level.
7
The results show that the proposed model generated 52% MAPE, which is 159% lower
than the MAPE from regression models ran for the same area (6). While travel demand
modeling methods have proven accurate in AADT estimation, they are often time
consuming to develop and require a lot of data collection resources and modeling expertise.
The classic four step travel demand model is complex, time consuming, and costly for
many jurisdictions. This is explained in much further detail in the Chapter 4: Results and
Analysis, where the newly developed model is compared to an existing Travel Demand
Model.
Artificial Neural Networks are also commonly utilized as a means to find AADT
on roadways. Sharma et al. developed two models for estimating AADT using an Artificial
Neural Network – one using the previous 48-hour count data and one using the previous
two 48-hour count data (10). When compared to the traditional factor model, the artificial
neural network models proved to be less accurate. However, an advantage of the neural
network models is that they did not require the ATRs to be grouped, unlike in the factor
method. The study found that the errors of AADT estimation using the factor approach
could be lowered by grouping the ATR sites appropriately and accurately assigning shortterm count stations to each ATR site. For two 48-hour counts, the 95th-percentile error is
between 14.14 to 16.68 percent, as compared with the range of 16.77 to 24.89 percent for
a single 48-hour count. While their results are adequate, many practicing traffic engineers
find the neural network approach to be more complex than the existing factor approach due
to lack of expertize.
8
In emerging Connected Vehicle Technology where vehicles continuously send data
to roadside infrastructure through a wireless communication medium potentially could
reduce the needs of permanent as well as coverage count stations for collecting traffic
volume data (11, 12). Although it will be several more years before enough connected
vehicles on different road segments reduce the need for count stations for volume data
collection, they potentially are considered as potentially an economically beneficial data
collection strategy (13).
2.3 AADT Estimation through Centrality
Land-use characteristics at the parcel level have been incorporated in several
AADT estimation models (8, 14). Centrality models, for example, seek to apply numerical
values to the topological significance of each element in a network using land-use
characteristics. Centrality is used in graph theory and network analysis in order to identify
the level of importance of certain elements of a graph or network. A network in this case
is broadly defined, and can refer to street networks, urban networks, and even social
networks (15).
In 2012, Zhang et al. developed a model based on road network patterns and traffic
analysis zones using three types of centrality - betweenness centrality, degree centrality,
and closeness centrality (14). The betweenness centrality of a node is the count of the times
that the node is intersected by all of the “shortest paths” within the network – when
considering the paths from every node in the network to every other node in the network.
Degree centrality of a node is the count of the number of other nodes that are adjacent to
9
it, and with which it is, therefore, in direct contact. Closeness centrality of a node is based
upon the degree to which a point is close to all other points. The greater the closeness
centrality of a node is, the closer the node is to all others within the network. Zhang et al.’s
study used a centrality property of the whole network, based on the node centralities
calculated, to find the each type of network centrality – betweenness, degree, and closeness.
It was found that network betweenness centrality was the most accurate in distinguishing
and describing various Traffic Analysis Zone (TAZ) road network patterns (14).
Most recent approaches for estimating AADT use a modified form of stress
centrality. Stress centrality is defined as the total count of the times a link would be used if
one were to travel from every node to every other node via the shortest path in a network.
Lowry introduced a new metric called origin-destination (OD) centrality that can be used
in linear regression as a contributing variable to calculate AADT spatially (8). Finding OD
centrality can be executed using a geographic information system (GIS) platform and less
data than other methods require, including land use data and the street network. A case
study of this concept yielded an R2 of 0.95. AADT can vary greatly among roadway
segments of the same functional classification, and this method can demonstrate significant
variation along roadway segments of the same functional class (8).
There are several benefits of using centrality concepts in estimating AADT. First,
the concept is simple and can be realistically applied to any given network, without the use
of expensive, proprietary, and/or subjective data. Also, to derive centrality measures, a GIS
program is essentially the only required tool. Not only can this entire model can be
10
completed within a short time period, but it has the potential to reduce the number of
coverage counts needed to accurately estimate AADT.
In this research, the author was motivated to investigate if the centrality method
utilized on a small city could be applied to a medium size city, and determine if the model
would benefit from the inclusion of additional variables, rather than the centrality variables
alone (16).
11
CHAPTER THREE
RESEARCH METHOD
Opting to use only data that is readily available to South Carolina DOT, this study
attempts to find new deterministic variables to calculate AADT on roadways in a medium
size city (population roughly 65,000) based on the theory of centrality. A similar study was
conducted for the small city of Mascow, Idaho (population of roughly 24,000) (7). Our
study area, the city of Greenville, SC currently has 6 automatic traffic recorders (ATRs)
and 153 short-term count locations within the city limits. The main objective of this
particular research was to apply origin-destination centrality to the city as a means to find
new variables to estimate AADT that are statistically significant at least 95% (pvalue<0.05).
To expand off of the discussion in the Literature Review, origin-destination
centrality is a type of centrality that is derived from stress centrality. Stress centrality is
defined as the total count of the times a link would be used if one were to travel from every
node to every other node via the shortest path in a network. In terms of calculating stress
centrality, the stress centrality equation of a link within a network is given below:
Eqn. 3
∑
∈
where V = the set of all zones in a network, σij = the shortest route from node i to node j,
and σij(e) = 1 if link e is used in this path and σij(e) = 0 if link e is not used in the shortest
path. This stress centrality would be based on three types of travel in or through a city:
12
Internal-to-Internal, Internal to External, and External to External. These will be explained
further in Step 5, later in this chapter.
FIGURE 1 METHOD OF ORIGIN-DESTINATION CENTRALITY
CALCULATION
Unlike stress centrality, the origin-destination centrality of a link uses relative
weights of the TAZs and gateways in each origin-destination combination. Additionally,
the theory of centrality assumes all links are of the same size, and does not take into account
the capacity of each. Therefore, this O-D centrality of each link is actually significant only
per lane. Therefore, we will incorporate the number of lanes into the equation, making it
simply:
13
Eqn. 4
∑
∈
where N = is the number of lanes on link e, V = the set of all zones in a network, σij = the
shortest route from node i to node j, σij(e) = 1 if link e is used in the path and σij(e) = 0 if
link e is not used in the shortest path, Wi = the relative weight of origin I, and Wj = the
relative weight of destination j. A “node” in this instance will refer to either a TAZ or a
gateway.
The majority of the steps (Steps 1-5 and Step 7) included in this research effort were
performed through a GIS software. All of the data used in the new model development are
publically accessible, updated at least once annually. The base data used in the model
development was collected from the City of Greenville, SC’s GIS Data website (17) and
the ITE Trip Generation Manual (18). This method does not require any additional data
collection or cost for the purchasing of data from private organizations. The data used are
geocoded in GIS shapefiles for the data sets shown in Table 1 below.
TABLE 1 Shapefiles and Attributes used in the Model Development
Shapefile
Street Centerlines
Parcels
Zoning
Count Station Locations
Notable Attributes
Street Name, Functional Class, Speed Limit, Number of Lanes
Parcel Number, Number of Buildings, Building SF, Parcel Area
Zone Number, Zoning Code
Station Number, Count Data
Step 1: Develop Street Network
First, the polyline shapefile of the street network, titled “Street Centerlines” was
used to develop a street network for the City of Greenville, SC. The ESRI ArcGIS
extension, “Network Analyst”, has a tool called “Create Network Dataset” that can easily
14
convert a shapefile to a network. When using ArcGIS to make routing decisions, it is
essential to first convert a polyline shapefile to a network of links.
Step 2: Create Traffic Analysis Zones (TAZs)
Much like travel demand models, the stress centrality and origin-destination
centrality models are composed of Traffic Analysis Zones (TAZs). The “Parcels” shapefile
contains land use attributes such as land area (acre), land use type, number of buildings,
etc. Each TAZ is composed of multiple adjacent parcels based on geometry, land use type,
and roadway access. Within the city limits of Greenville, SC, the parcels were sorted into
1,262 zones.
Step 3: Create TAZ Centroids
The area (acre), land use type, number of buildings, building square feet, and others
are all specified in the shapefile’s attributes for each of the parcels. Given these attributes,
combined with trip generation data from the ITE Trip Generation Manual (18), each parcel
is assigned its own relative weight, by calculating daily trip generation rate for each TAZ,
as shown in Figure 2. For example, if a parcel’s land use was identified as a single family
residence, the daily trip generation rate from the Trip Generation Manual, in this case 9.5
trips per dwelling unit, would be multiplied by the number of dwelling units within that
parcel, giving that parcel a relative trip generation rate.
Centroids for each zone are established through a weighted mean analysis based on
the geometry and weight of each parcel within a zone. The weighted mean is essentially
15
the centroid of the zone taken as the mean center of all of the parcels within the TAZ,
except instead of each of the parcels contributing equally to the center, some parcels
contribute more than others based on relative trip generation rates.
Step 4: Create External Entry and Exit Points
Because the utilized methods of centrality use travel that can begin and/or end
outside of the city limits, external points of entry and exit to the city are established by
examining all potential major exit/entry roadways. Just outside of the city limits of
Greenville, 32 major points of entry or exit are identified.
FIGURE 2 Eternal Gateway Points
16
Step 5: Determine the Shortest Route between TAZs and Gateway Points
Once all internal TAZ centroid and external points are established, the shortest
routes between each origin-destination combination are determined. Table 2 below shows
the inputs required for each routing type. Using the previously created Network in ArcGIS,
the shortest distance between all of the points in a given dataset can be found through the
use of the “Closest Facility” tool in the Network Analyst extension. The output of the tool
produces routes between each origin-destination combinations established.
TABLE 2 Three Centrality Types and Associated Inputs
Stress Centrality
Method
Internal ‐ Internal
Internal ‐ External
External ‐ External
Inputs
Origins
Destinations
Zone Centroids
Zone Centroids
Zone Centroids
External Points
External Points
External Points
Travel throughout a city (i.e., jurisdiction under consideration for modeling) is
based on three types of origin-destination combinations:
1) Internal-to-Internal (I-I): Trips from one travel analysis zone (TAZ) to another travel
analysis zone within the city;
2) Internal-to-and-from-External (I-E): Trips from one travel analysis zone to a
destination outside of the city limits, or vice versa;
3) External-to-External (E-E): Trips from an origin outside the city limits to a
destination outside of the city limits, that requires traveling through the city
17
Step 6: Assign Weights to Each TAZ and External Point
As in determining the stress centrality for each type of origin-destination
combination, finding the origin-destination centrality began with creating a street network
dataset, TAZs, and external exit/entry points. In this method, each zone’s weight is
determined by summing the weights of all of the parcels within that zone, which were
established in Step 3. The relative weight of each external point (E) is then taken as the
nearest AADT value available on that roadway.
FIGURE 3 Greenville Parcels with Relative Weights
18
Step 7: Create Three Origin-Destination Centrality Models
The three models are created using the origin-destination travel combinations: I-I,
I-E, and E-E. Their outputs incorporate the weights created in Step 5 above. The three new
origin-destination centrality maps are shown in Figure 4, Figure 5, and Figure 6. The
findings of the analysis are explained in the Chapter 4: Results and Analysis.
FIGURE 4 EXTERNAL-TO-EXTERNAL CENTRALITY MAP
19
FIGURE 5 Internal-to-External Centrality Map
20
FIGURE 6 Internal-to-Internal Centrality Map
Step 8: Create Regression Model to Estimate AADT
Using the outputs of the stress centrality formula above for each link and each type
of stress centrality, the three new variables were calculated as potential independent
variables in AADT multiple linear regression and non-linear regression model
development. Other variables were incorporated in an attempt to produce the most accurate
model are speed, number of lanes, and functional class. These methods and the findings
of the analysis are explained in Chapter 4: Results and Analysis.
21
CHAPTER FOUR
ANALYSIS AND RESULTS
4.1 AADT Estimation Using Centrality and Linear Regression
Once the three new origin-destination centrality were calculated following the steps
presented in Chapter 3: Research Method, multiple linear regression analysis was
performed with these three and three roadway characteristic variables, in order to develop
the most accurate centrality based AADT estimation model. The six independent variables
considered are:
i.
I-I Origin-Destination Centrality (I-I OD)
ii.
I-E Origin-Destination Centrality (I-E OD)
iii.
E-E Origin-Destination Centrality (E-E OD)
iv.
Functional Classification (FC)
5 - Interstate
4 - Major Arterial Freeway/Expressway
3 - Major Arterial
2 - Minor Arterial, Major Collector
1 - Minor Collector
v.
Speed Limit (SL)
The centrality variables were combined with roadway characteristic variables (i.e.,
speed, functional class and number of lanes) with the goal to derive new variables that
could prove to be independent variables for AADT estimation model. The final regression
model is shown in Table 3 below.
22
TABLE 3 Regression Analysis Summary Statistics
Regression Statistics
Multiple R
0.910628468
R Square
0.829244206
Adjusted R Square
0.825851707
Standard Error
4828.530729
Observations
155
ANOVA
df
Regression
Residual
Total
Intercept
I-I OD
E-E OD
Speed
3
151
154
Coefficients
-18267.96215
3.12073E-06
0.001284391
712.4749964
SS
17096765045
3520521060
20617286105
Standard
Error
2496.893055
4.95513E-07
0.000124769
77.96341866
MS
5698921682
23314709
F
244.4346
Significance F
9.9506E-58
P-value
1.39986E-11
3.12489E-09
3.68774E-19
3.94468E-16
Due to these results, the formula below was selected as the best formula:
AADT = 3.12073E-06* I-I OD + 1.284391E-04 *E-E OD+ 712.4749964 SL –
18267.96215
As shown, I-E OD centrality was not used in this final model. This is because, due
to the nature of the three models, I-E OD centrality was too similar to E-E centrality and
I-I OD centrality to provide statistical significance in the final model. The coefficient of
I-I Centrality is much lower than that of E-E Centrality, because the values used for I-I
centrality were much larger than those used for E-E Centrality. The coefficients for all
variables used are positive, showing a positive correlation with AADT. Additionally, the
p-values for each of the variables used are all significant at greater than 99% (p-value <
23
0.001). Finally, the intercept is a very large negative value, which is also not surprising,
due to the large values of these variables.
4.2 Comparison of AADT Estimation using Centrality based Model and Traditional
Travel Demand Model Output
In order to validate the accuracy of the centrality based AADT estimation model,
the author compared the model’s predictive ability with that of the city’s existing travel
demand model. First, it is important to carefully specify the differences between the steps
and data necessary to use the two models.
Travel demand forecasting models consists of four major steps: Trip generation,
trip distribution, mode choice, and trip assignment (19). Time-of-day and directional
factoring is a very important step as well, but is not explicitly mentioned in four steps of
the “four-step” model. For example, in the four step model, ample amount of data inputs
are necessary prior to the initiation of the four steps, which can include, but are not limited
to, employees, students, automobiles, and households by TAZ. In addition, there are two
ends of any trip generation – trip productions and trip attractions – and all trips must have
a given purpose – Home-based Work (HBW), Home-based Other (HBO), and Non Homebased (NHB). Considering both trip productions and attractions, in addition to each
purpose, there are six types of formulas that must be used for each TAZ in order to calculate
the overall trip production and attraction of each TAZ. Then, once the formulas are
estimated and the trip productions and attractions are calculated, the number of attractions
must be balanced to the number of productions (19). The trips for Internal-to-External and
External-to-External travel are then calculated separately. Table 4 and Table 5 below
compare more of the inputs and steps of each of the models.
24
TABLE 4 Comparison of Model Inputs for Centrality Method and TDF Model
Model Inputs
Street Network
Street Data (Name, Number of Lanes, Speed, etc.)
Employees by TAZ
Students by TAZ
Automobile by TAZ
Households by TAZ
Centrality
Method
x
x
x
Travel Demand
Forecasting
Model
x
x
x
x
x
x
TABLE 5 Comparison of Model Steps for Centrality Method and TDF Model
Model Steps
Develop Street Network
Create Traffic Analysis Zones (TAZs)
Create TAZ Centroids
Create External Entry and Exit Points
Assign Weights to TAZs and External Points
Generate Trip Productions by Purpose and TAZ
Generate Trip Productions by Purpose and TAZ
Balance Number of Attractions to Productions
Generate Trips for Internal‐to‐External Travel
Generate Trips by Gateway for External‐to‐External Travel
Generate Production & Attraction Trip Tables by Purpose
Transpose Table to Create Origin‐Destination Tables by
Purpose
Create a QuickSum Matrix
Direct Traffic between all TAZ pairs using Trip Assignment
Direct Traffic between all gateway and TAZ pairs using trip
assignment
Create Three Origin‐Designation Centrality Models
Create Regression Model to Estimate AADT
Centrality
Method
x
x
x
x
x
x
x
x
x
Travel Demand
Forecasting
Model
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
The Base AADT estimates used in Figure 7 are based on observations from 149
short-term count locations in the City of Greenville. Base AADT data was calculated using
25
adjustment factors and short term count data collected by SCDOT. The travel demand
model was created by a third party for the Greenville Metropolitan Planning Organization
(MPO) for year 2010. Average growth rates from the 6 ATRs located in the City of
Greenville were used to estimate the 2015 volumes from the travel demand model’s
calibrated volumes for 2010. The comparison of the AADT estimates for short term count
stations using two models (i.e., centrality based AADT method, and AADT from travel
demand model) is shown in Figure 7. It is important to note that both the origin-destination
centrality model and the travel demand model were calibrated using the estimated AADT
using the factor approach and 24-hour count for each short-term count station.
The travel demand model achieved lower goodness-of-fit (i.e., R2 value of 0.61),
while the new centrality based AADT method developed in this research has higher
goodness-of-fit (i.e., R2 value of 0.77). Additionally, the root mean square error (RMSE)
for the centrality-based AADT linear regression model and the travel demand based AADT
model are 7352.54 and 14073.19 respectively. It could be concluded that centrality based
AADT method performs better compared to AADT estimate from travel demand model in
terms of RMSE and R2.
26
120000
Travel Demand
Model
Origin‐Destination
Centrality Model
BASE AADT VALUES
100000
80000
y = 1.4132x ‐ 3035.8
R² = 0.6147
60000
y = x ‐ 3E‐12
R² = 0.8292
40000
20000
0
0
10000
20000
30000
40000
50000
60000
PREDICTED AADT VALUES
FIGURE 7 Comparison of AADT Estimation Methods
4.3 Possibility of Reduction of Coverage Counts using centrality based AADT Method
In an attempt to explore if the number of short-term count locations maintained
throughout the city could be reduced applying centrality based AADT estimation method,
random sub-sets of short term count were used to determine how many short-term stations
were necessary to achieve same level of AADT estimation for all roads in the city. Five
random sub-sets of each scenario were used, and the scenarios consisted of 100%, 80%,
60%, 40% and 20% of the total short term count stations in the city of Greenville. In each
random sub-set, two thirds of the data were used for model calibration and the other third
was used for model validation. This is shown in further detail in Figure 8 below, for
clarification.
27
FIGURE 8 Method to Determine Coverage Count Reduction Potential
After the model of the calibration set was developed, the validation set was ran using the
same linear regression model. Then the Median Absolute Percent Error (MdAPE) of the
calibration and validation data sets were calculated. This process was repeated 5 times for
each scenario. The results of each trial in each scenario are shown in Table 6 below. As
demonstrated in Figure 9, after using more than 60 short-term count stations (40% of
28
current short-term count stations), little AADT estimation accuracy is gained in terms of
MdAPE, suggesting that the number of coverage counts used in the modeling can be
reduced by 60%, which could substantially reduce data collection cost for the SCDOT. The
author believes this method can further save DOTs resources without compromising model
accuracy. A cost savings analysis is performed later in this chapter.
TABLE 6 Trials for Reduction of Coverage Counts
Percentage of
Count Stations
100
80
60
40
20
Trial
Number
Validation
Data MdAPE
1
2
3
4
Calibration
Data MdAPE
55.264
57.391
59.166
59.331
5
1
2
3
4
64.716
56.670
59.967
58.667
64.045
62.194
5
1
2
3
4
58.784
53.213
58.629
45.903
63.355
70.419
5
1
2
3
4
5
1
2
3
4
5
65.522
54.727
38.788
62.132
59.158
60.729
68.901
64.788
45.558
38.878
67.126
29
92.253
97.678
64.559
66.062
60.501
56.790
64.027
47.361
61.825
54.766
72.636
53.099
64.652
79.689
69.292
58.174
51.264
111.861
96.816
59.709
195.504
90.297
57.824
120
Calibration Data (67%)
Median Absolute Percent Error
100
Validation Data (33%)
80
60
40
20
0
100
80
60
40
20
Number of Short Term Count Stations
FIGURE 9 Reduction of Coverage Counts Applying the Centrality Method
4.4 Cost Savings Analysis
Because the centrality based AADT method has the potential to reduce the
number of short-term count stations, a cost savings analysis was performed to estimate
annual savings for the South Carolina Department of Transportation by using this method
throughout the state.
SCDOT performs 12,000 short-term counts per year, each collecting 24 hour
volume data. Due to the estimated reduction of over 60% short-term counts in Greenville
applying centrality based AADT method cost savings analyses was performed for 40%,
50%, 60%, and 70% short-term count reduction statewide, as presented in Table 7 below.
30
Each time SCDOT performs a short-term count, they use a PEEK-ADR 2000
counter and two pneumatic tubes. The counter costs an average of approximately $1000
(depending on features, etc.) and was assumed to have an average lifespan of
approximately 200 uses. The tubes used have an approximate cost of $200 per pneumatic
tube, with two tubes needed in every count. The average lifespan per tube is estimated at
approximately 20 uses (due to tearing, breakage, and wearing out). It was assumed that
every short term count requires 3 SCDOT employees for a total of 5 hours at $30/hour,
which includes labor, travel, etc. The initial cost to develop the model is not included in
this analysis. Additionally, according to the author’s estimates, it will cost approximately
$10,000 per year to pay an in-house traffic engineer to maintain the centrality model vs.
maintaining the factor method model. Then, the total cost of the centrality based method at
each of the four levels of count location reduction was calculated in order to calculate a
percent cost savings. Therefore, the cost savings of each reduction level is simply:
Eqn. 6 Cost savings =
Cost of factor method
Cost of centrality method with x% reduction
The cost of each method is calculated using the final equation below:
Eqn. 5
Total Cost = Cost of traffic counter
Cost to run and update the model
Cost of tubing
Cost of manpower
Using this equation, the cost of the factor method was determined to be $5,700,000
annually.
31
TABLE 7 Cost Savings Analysis
Percent
Reduction
of Counts
Counts
Needed
40%
7,200
50%
6,000
60%
4,800
70%
3,600
Total
Savings
Total Cost
∗
∗$ ,
∗
∗$ ,
∗
∗$ ,
∗
∗$ ,
∗$
∗$
∗
$
,
∗$
∗$
∗
$
,
∗$
∗$
∗
$
,
∗$
∗$
∗
$
,
$3,430,000
$2,860,000
$2,290,000
$1,720,000
The costs illustrated in Table 7 (Column 4) are the costs of performing the number
of count locations specified at each level of reduction. It is estimated that 12,000 counts
are reduced by 4,800, 6,000, 7,200, and 8,400 counts per year for 40%, 50%, 60%, and
70% short-term count reduction scenarios, respectively. Using the Equation 5 shown, the
final savings range from $1,720,000 to $3,430,000. Clearly, there is a financial benefit of
utilizing this method for SCDOT. The monetary benefits could save DOTs hundreds of
thousands of dollars, which could be allocated to other valuable projects, such as
roadway and infrastructure maintenance.
32
CHAPTER FIVE
CONCLUSION AND RECOMMENDATIONS
5.1 Conclusions
Affordably estimating AADT on all of the roadways within a given street network
has been a challenge that many transportation agencies face, especially in smaller cities
and municipalities where resources are limited. As AADT estimation on local roads is vital,
ample amounts of research have been conducted to develop models that can estimate the
AADT of every roadway within any jurisdiction. For example, linear regression, travel
demand modeling, and Artificial Neural Network models have all been used in an attempt
to estimated AADT. The new centrality-based AADT estimation model, utilizing roadway
characteristics and new derived origin-destination centrality variables based on the theory
of centrality, illustrates a method that is capable of minimizing cost and effort when
compared to deploying dozens of short term count locations or using models such as the
travel demand model. Not only was a new variable to estimate AADT created, but the
model has a lower RMSE than the city’s travel demand model, and can be developed using
a GIS software alone. This thesis establishes a means of estimating AADT on every
roadway within a given jurisdiction, while reducing the number of coverage counts in
Greenville, SC by over 60%.
This method has the potential to be used in cities that do not have the time, funding
or resources to use coverage counts at many roads throughout their jurisdictions. In
addition, this method uses existing and publically available data to quickly and accurately
estimate AADT within a reasonable estimation threshold.
33
While this method was successfully applied to Greenville, SC, it has not yet been
tested for transferability among other cities within and outside SC. Future research should
include a study of transferability. Another limitation of this study is that the “base AADT”
counts that were used and compared to this regression model were not perfectly accurate
AADT values. At 142 of the 148 locations utilized were calculated counts, the AADT
values compared with the ones calculated using this model were also calculated counts,
and may not be entirely accurate. The base data was calculated using 24-hour short-term
traffic counts and the factor method.
5.2 Recommendations
The following recommendations are made based on this research:
The method presented in this study should be examined for other cities before
a decision can be made in its broader adoption. This should include cities of
various sizes to determine its applicability in various size cities.
This method should be tested at the state level. In addition to its use at the state
level, this method can be used at local level as well for low-cost AADT
estimation strategy.
This method should be tested against AADT estimations using short-term
counts and the factor method at the lowest functional classifications. Currently,
there is no short-term count data available for local roads in Greenville, SC.
This method should be tested in an area with a much larger number of
permanent count stations for validation. While AADT estimations using short-
34
term counts and the factor method are generally very accurate, they are not
perfect data, and this model should be tested against actual ground truth data.
Cost savings estimates could be conducted for a specific area based on
regional costs, which can vary based on location-based manpower and
equipment costs, in order to determine an exact savings value for their
particular agency prior to making a decision.
35
REFERENCES
1. Roess, R. P., E. S. Prassas, and W. R. McShane. Traffic Engineering. Upper
Saddle River, NJ, 2011.
2. Zhao, F., and S. Chung. Contributing factors of annual average daily traffic in a
Florida county: exploration with geographic information system and regression
models. Transp. Res. Rec. 1769, 2001, pp. 113–122.
3. Traffic Forecasting & Analysis. Minnesota Department of Transportation.
http://www.dot.state.mn.us/traffic/data/coll-methods.html. Accessed Aug 1, 2016.
4. Guide to LTPP Traffic Data Collection and Processing. McLean, VA: Federal
Highway Administration, Office of Infrastructure Research, Development and
Technology, Turner-Fairbank Highway Research Center, 2001.
http://www.fhwa.dot.gov/publications/research/infrastructure/pavements/ltpp/trfc
ol/trfcol.pdf Accessed Jul 2016.
5. Doustmohammadi, M., and M. Anderson. Developing Direct Demand AADT
Forecasting Models for Small and Medium Sized Urban Communities. In
International Journal of Traffic and Transportation Engineering 5.2, 2016, pp.
27-31.
6. Wang, T., G. Albert, and P. Alluri. Estimating annual average daily traffic for
local roads for highway safety analysis. In Transportation Research Record:
Journal of the Transportation Research Board, No. 2398, 2013, pp 60-66.
7. Katter, Greg. "INDOT Traffic Data Sources and Using Traffic Count Data System
Provided by MS2." (2015).
36
8. Lowry, Michael. Spatial interpolation of traffic counts based on origin–
destination centrality. In Journal of Transport Geography 36, 2014, pp. 98-105.
9. United States Census Bureau. When is the census of governments conducted?
https://www.census.gov/history/www/faqs/governments_faqs/when_is_the_censu
s_of_governments_conducted.html. Accessed on Jul 2016.
10. Sharma, S., P. Lingras, F. Xu, and G. Liu. Neural networks as alternative to
traditional factor approach of annual average daily traffic estimation from traffic
counts. In Transportation Research Record: Journal of the Transportation
Research Board, No. 1660, Transportation Research Board of the National
Academies, Washington, D.C., 1999, pp. 24-31.
11. Khan, S. M., Dey, K., and Chowdhury, M., “Real-time Traffic State Estimation
with Connected Vehicles,” IEEE Transactions on Intelligent Transportation
Systems, doi: 10.1109/TITS.2017.2658664, (2017).
12. Ma, Y., Chowdhury, M., Sadek, A., and Jeihani, M., “Real-Time Highway Traffic
Condition Assessment Framework Using Vehicle-Infrastructure Integration (VII)
with Artificial Intelligence (AI),” IEEE Transactions on Intelligent
Transportation Systems, Vol. 10, No. 4, pp 615-627, (2009).
13. He, Y., Chowdhury, M., Ma, Y., and Pisu, P., “Merging Mobility and Energy
Vision with Hybrid Electric Vehicles and Vehicle Infrastructure Integration,”
Energy Policy, Modeling Transport (Energy) Demand and Policies, Vol. 41, pp
599-609, (2012).
37
14. Zhang, Yuanyuan, X. Wang, P. Zeng and X. Chen. Centrality characteristics of
road network patterns of traffic analysis zones. In Transportation Research
15. Freeman, Linton C. "Centrality in social networks conceptual clarification."
Social networks 1.3 (1978): 215-239.
16. Keehan, M., Chowdhury, M., and Dey, K., “AADT Estimation with Regression
Using Centrality and Roadway Characteristic Variables,” Accepted for
presentation in the 96th Annual Meeting of the Transportation Research
Board, Washington, D.C., (2017).
17. City of Greenville, South Carolina. GIS Data.
http://www.greenvillesc.gov/364/Access-GIS-Data. Accessed Jul 2016.
18. ITE (Institute of Transportation Engineers), 2012. Trip Generation Manual, 9th
Edition, ITE, Washington, DC.
19. U.S. Department of Transportation, Federal Highway Administration,
Introduction to Urban Travel Demand Forecasting, Instructor Guide, NHI Course
No. 152054, Publication No. FHWA-NHI-08-061, May 2008.
38
APPENDIX A
SAMPLE DATA
Streets _Shapefile
FID
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
OBJECTID
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
RECNUM
3
4
5
6
7
8
9
10
11
12
13
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
ST_LABEL
RUTHERFORD RD
RAYFORD LA
N PLEASANTBURG DR
TILBURY WY
WEDGEWOOD DR
BROUGHTON DR
INGLEWOOD DR
TAMBURLAINE CT
WEDGEWOOD DR
MEADOW CREST CIR
WEDGEWOOD DR
SUMMIT DR
SUMMIT DR
DEARSLEY CT
BRENTWOOD DR
RUTHERFORD RD
GREEN MEADOW LA
BRENTWOOD DR
BRENTWOOD DR
INGLEWOOD DR
SUMMIT DR
TILBURY WY
WEDGEWOOD DR
BROUGHTON DR
STONE LAKE DR
N PLEASANTBURG DR
SUMMIT DR
GOBLET CT
VENNING CT
RUTHERFORD RD
COOL SPRINGS DR
A1
ROADNUM
S-23-21
SC-291
S-23-21
SC-291
S-23-21
LADD1
1241
1
2001
1
401
201
13
1
305
1
301
1001
1011
1
31
1221
1
0
13
1
927
51
201
101
1
1901
1043
1
1
1201
1
LADD2
1299
99
2099
11
499
299
99
99
399
99
303
1009
1041
99
41
1239
99
0
29
11
999
99
299
199
99
1999
1051
99
99
1219
99
RADD1
1240
2
2000
2
400
200
12
2
304
2
300
1000
1010
2
30
1220
2
0
12
2
926
50
200
100
2
1900
1042
2
2
1202
2
A2
RADD2
1298
98
2098
12
498
298
98
98
398
98
302
1008
1040
98
40
1238
98
0
28
10
998
98
298
198
98
1998
1050
98
98
1218
98
PREFIX
N
N
NAME
RUTHERFORD
RAYFORD
PLEASANTBURG
TILBURY
WEDGEWOOD
BROUGHTON
INGLEWOOD
TAMBURLAINE
WEDGEWOOD
MEADOW CREST
WEDGEWOOD
SUMMIT
SUMMIT
DEARSLEY
BRENTWOOD
RUTHERFORD
GREEN MEADOW
BRENTWOOD
BRENTWOOD
INGLEWOOD
SUMMIT
TILBURY
WEDGEWOOD
BROUGHTON
STONE LAKE
PLEASANTBURG
SUMMIT
GOBLET
VENNING
RUTHERFORD
COOL SPRINGS
TYPE
RD
LA
DR
WY
DR
DR
DR
CT
DR
CIR
DR
DR
DR
CT
DR
RD
LA
DR
DR
DR
DR
WY
DR
DR
DR
DR
DR
CT
CT
RD
DR
SUFFIX
A3
ALTNAME
RDCLASS
3
1
2
1
1
1
1
1
1
1
1
1
1
1
1
3
1
1
1
1
1
1
1
1
1
2
1
1
1
3
1
MAINTENANC
2
1
2
1
1
1
1
1
1
1
1
1
1
1
1
2
1
1
1
1
1
1
1
1
1
2
1
1
1
2
1
SPEED
A4
40
30
40
30
30
30
30
30
25
30
30
35
35
30
30
40
30
30
30
30
35
30
30
25
30
40
35
30
30
40
30
MAJRDS
Y
N
Y
N
N
N
N
N
N
N
N
N
N
N
N
Y
N
N
N
N
N
N
N
N
N
Y
N
N
N
Y
N
LOW_STREET
1240
1
2000
1
400
200
12
1
304
1
300
1000
1010
1
30
1220
1
0
12
1
926
50
200
100
1
1900
1042
1
1
1201
1
HIGH_STREE
1299
99
2099
12
499
299
99
99
399
99
303
1009
1041
99
41
1239
99
0
29
11
999
99
299
199
99
1999
1051
99
99
1219
99
STREET_ID
5600
2644
2574
5251
3709
446
1629
5252
3709
2117
3709
3105
3105
5563
390
5600
1349
390
390
1629
3105
5251
3709
446
3081
2574
3105
5401
5288
5600
775
SECT_NO
2000000002
0000167600
2000000004
0000167700
0000165800
0000164400
0000164700
0000168000
0000165700
0000167500
0000165600
0000163500
0000163400
0000164200
0000164800
2000000018
0000168100
0000165100
0000164900
0000164600
0000163600
0000167900
0000165500
0000164500
0000168200
2000000028
0000163300
0000164100
0000164300
2000000032
0000167400
A5
EDITORNAME
LASTUPDATE
CSOURCE
LIB
LIB
LIB
LIB
LIB
LIB
LIB
LIB
LIB
LIB
LIB
LIB
LIB
LIB
LIB
LIB
LIB
LIB
LIB
LIB
LIB
LIB
LIB
LIB
LIB
LIB
LIB
LIB
LIB
LIB
LIB
8954
8955
8956
8957
8958
8959
8960
8961
8962
8963
8964
8965
8966
8967
8968
8969
8970
8971
8972
8973
8974
8975
8976
8977
8978
8979
8980
8981
8982
8983
8984
FACILITYID
Shape_STLe
527.1766826
1150.473647
614.3874936
415.4401001
186.6661026
735.1952526
308.0473068
235.2168267
263.642934
645.0751654
165.4496045
202.4614411
697.1660385
140.9242893
297.5336242
752.6313822
404.4736069
396.117093
68.30224875
412.1011087
287.5025614
387.9861478
548.1792152
1409.610659
295.3921772
883.2929854
317.2860422
118.7954023
416.1107824
400.3369536
586.6512895
A6