Sie sind auf Seite 1von 12

Reliability Engineering and System Safety 138 (2015) 112

Contents lists available at ScienceDirect

Reliability Engineering and System Safety


journal homepage: www.elsevier.com/locate/ress

Integrated risk management of safety and development


on transportation corridors
Shital A. Thekdi a,n, James H. Lambert b
a

Department of Management, Robins School of Business, University of Richmond, 1 Gateway Road, Richmond, VA 23173, USA
Center for Risk Management of Engineering Systems, Research Professor, Department of Systems and Information Engineering, University of Virginia,
PO Box 400747, 112C Olsson Hall, 151 Engineers Way, Charlottesville, VA 22904, USA

art ic l e i nf o

a b s t r a c t

Article history:
Received 23 December 2013
Received in revised form
22 November 2014
Accepted 27 November 2014
Available online 6 January 2015

Prioritization of investments to protect safety and performance of multi-regional transportation networks from
adjacent land development is a key concern for infrastructure agencies, land developers, and other
stakeholders. Despite ample literature describing relationships between transportation and land use, no
evidence-based methods exist for monitoring corridor needs on a large scale. Risk analysis is essential to the
preservation of system safety and capacity, including avoidance of costly retrots, regret, and belated action.
This paper introduces the Corridor Trace Analysis (CTA) for prioritizing corridor segments that are vulnerable
to adjacent land development. The method integrates several components: (i) estimation of likelihood of
adjacent land development, using inuence diagram and rule-based modeling, (ii) characterization of access
point density using geospatial methods, and (iii) plural-model evaluation of corridors, monitoring indices of
land development likelihood, access point densities, and trafc volumes. The results inform deployment of
options that include closing access points, restricting development, and negotiation of agencies and developers.
The CTA method is demonstrated on a region encompassing 6000 centerline miles (about 10,000 km) of
transportation corridors. The method will be of interest to managers investing in safety and performance of
infrastructure systems, balancing safety, nancial, and other criteria of concern for diverse stakeholders.
& 2014 Elsevier Ltd. All rights reserved.

Keywords:
Infrastructure management
Transportation and land use
Sustainable development
Multimodal transportation planning
Risk assessment and management

1. Introduction
A recent study of the US National Cooperative Highway Research
Program has prompted agencies to align investments and policies for
corridor management with consideration of concurrent goals such as
preserving safety on corridors, promoting economic growth, developing sustainable communities, and protecting against incompatible
real estate development [8]. Incompatible real estate development
can compromise the performance of the regional transportation
system and increase future costs to maintain or improve capacities
[15]. Costly belated action to address the needs of vulnerable corridors can be avoided through proactive investment in corridor
segments with highest risk. There is need for agencies to utilize
data-driven decision support methods for forecasting corridor needs
and guiding investments. As unprotected land development involves
stakeholders from state agencies, private developers, infrastructure
users, utility providers, community members, and others, agencies
must use the objective tools developed in this paper to encourage
coordinated protective investments.

Corresponding author.
E-mail addresses: sthekdi@richmond.edu (S.A. Thekdi),
lambert@virginia.edu (J.H. Lambert).
http://dx.doi.org/10.1016/j.ress.2014.11.015
0951-8320/& 2014 Elsevier Ltd. All rights reserved.

Many have studied the role of sustainability in urban transportation [3,21] and the partnering of land use with transportation
infrastructure [28,4,13,29]. Han et al. discuss the need for policy to
address land use change by managing transportation networks
and promoting sustainable city development. Vanka et al. discuss
the role of access management in coordinating land use policies
and transportation infrastructure management. Although existing
literature describes the relationship between land development
and infrastructure risk [26], past efforts have neglected to address
coordination of land use forecasting, corridor management, and
existing infrastructure conditions for large heterogeneous regions,
as described in the this paper.
This paper develops the Corridor Trace Analysis (CTA), a decisionsupport model containing a linear control chart diagram representing
geospatial characteristics to support monitoring and prioritization of
transportation corridors segments that are vulnerable to future land
development. Consistent with mental model approaches for risk, the
likelihood index described herein improves incrementally on a body
of literature [20,15,31] that identies numerous factors inuencing
land development. The method of assembling factors in a mental
model itself is established in the literature [30,23,2].
CTA approach has an aim of addressing economic growth while
avoiding surprise and regret in terms of safety and performance.
Regret resulting from incompatible real estate development may

S.A. Thekdi, J.H. Lambert / Reliability Engineering and System Safety 138 (2015) 112

include insufcient infrastructure to accommodate future demand,


unsafe infrastructure design for future demand, land development
inducing physical barriers to expansion, land development inducing a nancial burden on potential expansion, and others. The
Corridor Trace Analysis describes the inuence of risk scenarios on
agency priorities with recognition that combining heterogeneous
factors could result in opportunities as well as threats
[11,12,25,26]. The innovation of the CTA in this paper is to
integrate established replicable analyses such that no single
element stands alone. The more detailed innovations include:
(i) linearized summary statistics of the several model outputs,
(ii) trace analysis for prioritization of corridor segments, with comparisons of access point densities, trafc volumes, and potential for
land development, and (iii) replicable evidence collection for
performance monitoring and risk-based strategic decision making.
The CTA method can be adapted to the unique needs of communities and regions.
The results will guide investments in detailed corridor studies
that will prompt focused investigations of access-point densities and
other risk indices. Such investigations might include (1) preservation
techniques such as land-use regulations, land acquisition, access
management, and land-use regulations [7], (2) engineering simulation and optimization tools for land-use decisions, (3) growth-management plans in collaboration with local governments and developers, (4) proactive congestion management methods, such as building access roads, restricting left turns, aligning corridor signaling
with promoted development, etc., and (5) other corridor investments, such as re-orienting development such that properties face
away from corridors and imposing transportation impact fees. In
urban and urbanizing areas, it may be cost-effective to compensate
for a degraded level of service by investing in transit improvements.
Trace analysis methods for monitoring or detecting process
extremes are widespread across disciplines such as physics, astronomy,
nance, and others. In addition, the control-chart method for monitoring stability of processes has been established within quality control
literature [19]. Although control charts are traditionally applied to
time-series data, the CTA model expands on the concept by assessing
risk associated with geographic instability. This concept suggests that
risk assessment for transportation infrastructure requires concurrent
evaluation of both infrastructure conditions and the adjacent geographical conditions. This concept recognizes that instability in characteristics, as demonstrated by relatively high levels of relevant characteristics, contribute to system risk. This paper will assess geographic
system stability within a trace analysis for risk classication using
characteristics related to land geography, infrastructure conditions, and
infrastructure investments. The monitoring for coincidences of characteristics beyond natural process limits provides signaling for agencies
to investigate causes and potential remediation for such conditions.
Use of the described methods will make efcient use of agency time,
nancial resources, and other resources by proactively addressing the
most vulnerable corridor segments. As a decision support framework,
the methods of this analysis are intended to provide agencies with
analysis and visualization of relevant data and multi-regional highlevel priorities for resources.
Monitoring of corridor segments at risk for land development
involves predictive modeling and characterization of existing vulnerabilities in order to prioritize and protect regions along the transportation infrastructure [24]. The coincidence of high access point
density, high potential for land development, and high trafc
volumes suggests immediate need for protective investments due
to: (1) correlation between high access point density and accident
risk [9], (2) the potential for land development to directly result in
land value escalation, thereby increasing future right-of-way costs,
and (3) existing high trafc volume indicating the current demand
for the corridor segment. Thus, a discrimination of corridor risk based
on access-point densities [10,17,22] potential for land development,

and trafc volumes is important for the identication of corridor


segments that will encounter regret due to safety, performance, and
nancial consequences after land development occurs.
The CTA as described in this paper could be translated to
protection of infrastructure corridors in other large-scale infrastructure systems, both in developed and developing regions, with various
time horizons and geographic scales. The multi-regional scope addresses geographically expansive areas with multiple administrative
autonomies, terrains, demographic characteristics, political priorities,
economic needs, and social concerns, among others. No single set of
model parameters or control limits may be sufcient to address the
varying regional needs. The CTA method assembles evidence of economic development, safety, and performance that allows decisionmakers to explore changing the weights of the criteria with immediate visualization of the implications. For example, a community or
stakeholder might increase weights for safety and performance
criteria, while decreasing weights for economic development criteria,
and nd the implications for prioritization of risk mitigation investments. The ndings and implications of the CTA model are validated
through expert engagement and elicitation in that it allows visualization of the available evidence across a large-scale system.
The CTA method is suited for municipalities, states, provinces,
regions, and nations. The combination of predicted future development scores with access point densities and trafc volumes will
be useful to diverse stakeholders with varying interests in addition
to transportation system managers. The methods will be of
particular interest to agencies with limited directed resources for
land use and transportation planning. The stakeholders include,
but are not limited to, infrastructure owner/operators, real estate
investors and developers, preservation groups, and environmental
managers.

2. Methods
This section will describe the components of the methodology.
Fig. 1 describes the context of the Corridor Trace Analysis (CTA)
method. The components of an integrated approach that includes
the Corridor Trace Analysis are: (1) Step 1: estimation of likelihood
of adjacent land development, which uses data search, inuence
diagram, the rule-based modeling tools to categorize likelihood of
land development for each corridor segment, (2) Step 2: characterization of access point density, which uses geospatial analysis
to classify the access point density level for each corridor segment,
and (3) Step 3: a plural-model evaluation of risk factors along the
corridors, which concurrently monitors indices of potential land
development, access point densities, and trafc volumes. This
paper describes the implementation of this approach for prioritizing infrastructure needs on 6000 miles (about 10,000 km) of
corridors in a region adjacent to the US national capital region.
2.1. Estimation of likelihood of adjacent land development
This approach builds upon previous efforts to study predictive
modeling for land development [14] and modeling utilizing expert
elicitation methods [16]. The rst step is dening factors that most
inuence adjacent land development. Many of the relevant data
sources have only recently emerged due to advances in satellite
imagery, increased data storage capacities, increased capacity to
process large data les, and increased access to public data
resources. Although this process utilizes abundant big-data
resources, this process recognizes that some data resources may
not adequately contain information on past conditions at a multiregional scale. For example, although public population records
are collected every 10 years, other data resources may not be
frequently collected. Land values may not be centrally owned in a

S.A. Thekdi, J.H. Lambert / Reliability Engineering and System Safety 138 (2015) 112

Fig. 1. Risk-based methodology using Corridor Trace Analysis to focus corridor studies on several factors including access-point densities, potential land development, trafc
volumes, etc.

digital format and may not contain point-in-time records from


previous years. As a result, traditional time series and regression
models are insufcient. Thus, validation with experienced professionals through expert elicitation is necessary to validate results.
As the likelihood estimation process is data-driven, there is need
to identify multi-regional, objective, and digitized data resources. As
described in the case study of this paper, potential factors may
include population data from the national census counts, housing
data from national housing agencies, infrastructure information
from transportation agencies, and others. The determination of
the most relevant factors is performed through expert elicitation,
including expert planners, economic ofcials, and other representatives from related agencies. The expert elicitation process includes
the discussion of relevant factors, costs, and opportunities in the
system and then ltering to determine the most important factors.
The experts should be knowledgeable about substantive issues, but
do not require extensive knowledge of the probability concepts
used in later stages of modeling.
An inuence diagram is used to lter the most inuential factors
for use in probability statements representing the potential for
land development, as described by Thekdi and Lambert [26]. This
diagram provides the ability to graphically represent conditional
interdependencies among random variables such as factors, decisions, uncertain events, and objectives. This process enables experts
to consider the importance of random variables without a reliance
on mathematical notation.
A baseline likelihood of land development for each block of
adjacent land is determined using fuzzy rule-based modeling
methods [1] for factors determined to be most inuential from
the inuence diagraming process. This method denes probability
statements based on values and relationships among contributing
factors. Each inuential factor is assigned an inuence value, such
as low or high. The number of levels used to represent likelihood of
land development are based on decision-maker input and needs
for further analysis. For example, levels may be dened as high,
medium, and low. Then, probability statements are described using

if-then statements, such as the following:


IF population density medium AND employment f orecast
high; THEN likelihood of development
high
As relevant data may not be available for validating this model
using historical land development, validation can consist of
dialogue with transportation and land-use experts, representing
knowledge across heterogeneous geographies.
2.2. Characterization of access point density
A database containing the relevant corridor network data is
prepared by selecting the routes to be counted from a map,
segmenting these routes into one-mile segments, and creating
buffers around the individual segments. Aerial imagery is used to
manually identify and mark corridor access points and consequently use the marked access point data for analysis. Fig. 2 shows
an image of the aerial, roadway, and access point layers overlaid
for counting purposes. To reduce counting errors and missed
segments, access point counting starts at the end of a corridor or
major highway continuing systematically along the roadway. Due
to satellite image quality restrictions, access points are treated
equally: a four-way intersection was marked the same way as a
rural driveway.
Access point density is characterized by access points per onemile (1.6 km) segments. The methods can be generalized to any
length of segment deemed appropriate by the transportation
infrastructure managers. To calculate the access point density
values, the layer containing access points is spatially joined to
the corridor layer segmented into one-mile (1.6 km) parts. Analysis of this layer is used to characterize access point density along
mile segments or specic areas to address as discussed later in this
paper. It is important to minimize through data-quality analysis
that segments are neither double-counted nor omitted.

S.A. Thekdi, J.H. Lambert / Reliability Engineering and System Safety 138 (2015) 112

access point density. Comparison of these histograms within and


across regions allows analysts to benchmark access management
policies and inuence future development to align with successful
strategies.
Corridor-specic linear density maps visually represent access
point density areas along the network. For example, density maps
can be used to highlight the one-mile segments classied as high
density, based on the percentiles for the data set. This approach
can be applied to the entire network or for a specic corridor,
thereby allowing planners to visually identify where high access
point densities are located.

2.3. Plural-model evaluation of corridor risk factors


Fig. 2. Example of an inventory of access points in the risk-based methodology.
Each access point represents a curb cut, a median cut, or an intersection along the
roadway.

An inspection for data quality may be useful for the access-point


dataset. Access points may have been counted multiple times or be
missing. A validation can be performed comparing randomly sampled data with known access points or by focusing comparison of
known access points at corridor segments with atypical characteristics, such as outliers. However, high access point densities could
legitimately occur on corridor segments in rural areas due to
numerous low-trafc driveways. In contrast, corridor segments in
urban areas with multiple four-way intersections quite close to one
another are in jeopardy of being overlooked depending on the
methods of data collection.
With some research suggesting a positive correlation between
access point density and crash rates [9], the following access point
density analysis enables the CTA method to include the identication of corridor segments with the relatively greater safety needs,
as evidenced by a high density level characterization.
Corridor segments with a high access point density can be identied using several established mathematical methods. For example,
high-density segments may be dened in accordance with the
number of standard deviations from the mean, as commonly used
in quality control chart literature [18]. As further study is needed to
evaluate the distribution of access point density, preliminary analysis
may classify corridor segments as high based on the access-pointdensity percentile, such as the 90th or 95th percentile. Given the
large scale of this analysis, there may be further need to classify
access point density on a regional level, such that relatively high
density near large cities may pose less concern versus relatively high
density in smaller cities and rural areas.
In addition, access point volumes categorized by attributes
enable planners to visually identify the types of corridors of
greatest concern. Commonly available attributes include Median
Type, Federal Functional Class, and other region-specic identiers.
The charts serve to identify, for a specic attribute, instances of
that attribute that are high-density segments in the data. Once
these instances are identied, a further in-depth analysis can be
performed to explain why high-density conditions exist.
Histograms are useful to identify the counts of one-mile
segments that fall within a particular bin of Access Points Per
Mile for corridors in the network. They also serve to verify the
inferences drawn from the attribute-specic charts. The shape of
the distribution provides insight into the current status of access
management of the corridors. For example, a positive skew (large
number of segments with low access point density) indicates that
a large proportion of corridor segments have adequately managed
access point densities. A negative skew (large number of segments
with high access point density) indicates that a large proportion of
corridor segments face relatively greater vulnerability due to high

The CTA method then compares access point density, trafc


volume, and likelihood of land development along one-mile segments of corridors to provide a multiple perspective evaluation of
current and future corridor needs. The classication for likelihood of
land development is represented with a gradient color scale on the
map. Trafc volume is measured by the average daily trafc for each
mile segment and is graphed along the linear roadway mileage. The
access point densities measured by the linear-mile are graphed in
the same manner to chart the progression along the route mileage.
As determined through elicitation of experts, several combinations
of perspectives require immediate action by the corridor managers.
Table 1 shows a notional categorization for combinations of perspectives along with the suggested timing for corridor management that
will be used within the case study. Eight scenarios are given, based on
each of land development likelihood, trafc volume, and access point
metrics being assigned either a high or low value. This table can be
adapted to represent more detailed classication. Scenario 1 outlines a
situation in which trafc volume and access point density are high
and additional development is expected to occur. This scenario details
a safety and performance concern that is expected to escalate as the
costs of belated action rise. Conversely, Scenario 7 outlines a situation
in which trafc volume is high, while the land-development likelihood and access point density are low. This situation would not be
of highest priority for protective action due to a low likelihood of
development and adequate access management.
The classication rules for dening high and low levels for
land development likelihood, trafc volume, and access point
density may vary by application, region, and other decisionmaker needs. For example, classication of high levels for trafc
volume and access point density may be dened by the 90th or
95th percentile, as shown in the case study. Classication of high
levels for likelihood of land development may require increased
expert elicited values and modeling using fuzzy rule-based methods or other related mathematical models. In addition, a classication high levels for trafc volume and access point density may
require consideration of regional patterns, such that a high level
may be recognized as a peak within the control chart within each
unique region.
Table 1
Prioritization of corridor segments based on multi-perspective evaluation.
Scenario Land development
likelihood

Trafc
volume

Access point
density

Management
need

1
2
3
4
5
6
7
8

High
Low
High
Low
High
Low
High
Low

High
High
Low
Low
High
High
Low
Low

Immediate
Immediate
Immediate
Mid-term
Mid-term
Mid-term
Long-term
Long-term

High
High
High
High
Low
Low
Low
Low

S.A. Thekdi, J.H. Lambert / Reliability Engineering and System Safety 138 (2015) 112

3. Results
This section demonstrates the CTA method on a 6000-mile
(9000 km) interstate and primary transportation system in the
Commonwealth of Virginia, USA, known as the Statewide Mobility
System (SMS). Virginia is located at the southern edge of an
expanding mega-region that extends from Massachusetts through
New York and south through New Jersey, Delaware, Maryland, the
District of Columbia and Virginia. The multi-regional system
services critical multimodal transportation hubs and includes the
National Highway System, critical evacuation routes, primary
corridors providing regional connectivity, and other vital links
between population and employment centers. This region, with a
population approaching 20 million persons, contains over 197,000
US Census blocks with over 43,000 blocks adjacent to the SMS. The
described results are the product of an iterative vetting process
that updates assumptions and parameters of the model based on
validation with experts across the studied regions.

3.1. Estimation of likelihood of adjacent land development


The expert elicitation process was conducted with regional planners, economic ofcials, and representatives from related state agencies. First, the elicitation process included identication of public data
and reliable data sources. Relevant data sources included the US Bureau
of Economic Analysis, the United States Census, the US Department of
Housing and Urban Development, commercial land and home value
databases, Landsat/National Land Cover Database satellite imagery, and
population and economic development forecasts.
Table 2 shows the relevant predictive factors. The factors included
land values, population and projected population, historically underutilized business zones, historical changes in land use, conservation
easements and parcels economically suitable for development. This
study has been unique in its combination of these data sources for
the region, integrating economic, demographic, and land cover.

The next step is to lter the most inuential factors for use in
probability statements representing the likelihood of land development.
The expert elicitation process consisted of an assessment of relevance
for each factor and discussion of how relevant factors inuence others.
The resulting graphical representation of relevant factors is shown on
the inuence diagram in Fig. 3. Although the shown inuence diagram
contains comprehensive information pertaining to a variety of factors
inuencing the nancial costs, safety, and sustainability of the system, the
relevant component to this analysis is factors contributing to time to
develop. The elicitation process determined this time to develop metric
directly contributes to the likelihood of land development. The diagram
shows the time to develop is inuenced by factors including population,
population forecast, employment centers, land values and undervalued
land. Although unique decision-maker perspectives may consider a
variety of combinations of inuential factors for time to develop, a single
perspective will be used for the case study.
The likelihood of land development was modeled as follows.
While the denition of time to develop and contributing factors has
been modeled in earlier work using several alternative perspectives
[26,27], this paper has focused on a single perspective elicited from
experts. The data for each inuential factor was processed to dene
levels for each factor value. Level 1 corresponds to a high level, Level
2 corresponds to a medium level, and Level 3 corresponds to a low
level. For example, as higher employment forecasts percentages
were assumed to be more inuential, higher employment forecasts
were associated with higher factor levels. Similarly, as higher
population projection percentages were assumed to be the more
inuential, higher population projection percentages were associated with higher factor levels. Any employment or population
forecast with a negative value was assigned to Level 3 (low level);
any value between zero and half of the maximum value was
assigned a value of medium; while any value greater than half of
the maximum value was assigned a value of high.
A single model representing a baseline perspective will be described
in this case study. This perspective was determined to include only
employment forecast (represented by employment centers in the

Table 2
Relevant factors and data sources used to estimate likelihood of land development used for 6000-mile
(9000 km) transportation system demonstration.
Relevant factor
US Census Bureau: population
Virginia Workforce Connection: population forecast
Bureau of Labor Statistics: unemployment
City-data.com: land value
Virginia Department of Transportation: suburban urban rural classication
Virginia Department of Conservation and Recreation: Federal lands
Virginia Department of Conservation and Recreation: protected lands
US Geological Survey National Land Cover Database: slope
Public domain: named Streams and Rivers
Public domain: drainage systems
Public domain: bridges
Public domain: landmarks
Public domain: parks
Public domain: institutions
Public domain: cemeteries
Public domain: railroads
Public domain: utilities
Public domain: legislative districts
Public domain: planning districts
US Census Bureau: farm density
US Census Bureau: urban areas
Bureau of Economic Analysis: personal Income
Department of Housing and Urban Development: residential construction building permits
Department of Housing and Urban Development: American Housing Survey
Great Schools: Elementary, Middle, and High School ratings
Virginia Employment Commission: jobs
US Census Bureau: housing units

S.A. Thekdi, J.H. Lambert / Reliability Engineering and System Safety 138 (2015) 112

Fig. 3. Inuence diagram analysis with actors contributing most to adjacent land development, useful in expert elicitation with transportation ofcials.

Table 3
Level denitions for relevant factors contributing to potential for land development.
Rank

Unit

Level 1 (high)

Level 2 (medium)

Level 3 (low)

Employment forecast
Population density
Population projection
Home value

People
People per Square Km.
%
Dollars

484,500
42500
454
4290,000

o 84,500 and 49801


o 2500 and 4 62
o 53 and 41
o 290,000 and 4138,000

o 9800
o 62
o1
o 138,000

inuence diagram), population density (represented by population in the


inuence diagram), population projection (represented by population
forecast in the inuence diagram), and home value (represented by land
value in the inuence diagram). Experts introduced if-then statements
relating factor levels to probability statements to represent the
potential for land development for corridor segments within each
Census block. Table 3 denes the level denitions for relevant factors
used for the fuzzy rule-based modeling process. For example, if the
population density for a particular area is greater than 2500, the
population density factor for the area was categorized as Level 1 (high
level). For this case study, a corridor segment within a given Census
block is land development likelihood score based on if-then statements
applying equal inuence for each factor.

Fig. 4 shows the baseline likelihood of land development along


the critical transportation corridors. The darker areas have the
highest likelihood of development within one mile of the corridor
while the lighter areas have the lowest likelihood.
3.2. Characterization of access point density
The method for access-point data collection involved locating
70,596 access points over 6000 miles (about 10,000 km) of critical
transportation corridors of the region. Interstate highways were not
counted due to the fact that they are designated as limited access
roads and therefore do not require attention for access management.
Secondary arterial roads (parallel to interstate highways) were of

S.A. Thekdi, J.H. Lambert / Reliability Engineering and System Safety 138 (2015) 112

Fig. 4. Likelihood of land development along 6000 miles (about 10,000 km) of infrastructure corridors in the case study, which in a validation was compared to past
development according to the USGS National Land Cover Database.

Fig. 5. Number of 1-Mile segments in the top 5% access-point density, per US Route. The Other category is an average segment count for other routes that appear in the top
5%. Note the routes are different lengths (1 mile 1.6 km).

Fig. 6. One-mile access point density comparison between two corridors (1 mile 1.6 km), an element of the Corridor Trace Analysis (CTA) for a regional transportation
network.

S.A. Thekdi, J.H. Lambert / Reliability Engineering and System Safety 138 (2015) 112

particular interest to this effort due to placement near areas that


could accommodate future land development and because the
corridors are less restricted in access point placement.
Preliminarily analysis of access point densities can be directed
toward corridor residencies, localities, median types, districts, and
federal functional classes that may require additional study or
attention following the CTA methods. For example, Fig. 5 shows
segment counts per route by attribute for the top 5% of access point
density segments. In this case, routes 11, 60 and 1 have the
greatest number of segments with high (top 5%) access point
densities. In addition, Fig. 6 shows a corridor density chart heavily
weighted towards middle and high-end access point bins. Using
similar charts, supplementary analysis efforts may investigate root
causes for high-density segments, patterns in access management
policies, and the distribution shape for tangential investigation of
land use and corridor performance goals.
Fig. 7 shows an example of a corridor-specic linear density
map, created to visually represent access point density areas along
a transportation corridor. The highlighted mile segments represent

Fig. 7. Access point densities for one-mile segments along US 29 in the Commonwealth of Virginia. Top 10% of access point density segments are indicated in red
(1 mile 1.6 km). (For interpretation of the references to color in this gure legend,
the reader is referred to the web version of this article.)

the top 10% of segments along the specic corridor. This approach
can be applied to the entire network or for a specic corridor.
3.3. Multi-perspective evaluation of corridor performance factors
Figs. 812 demonstrate the CTA method used to compare
access point density, trafc volume, and average priority scores
along one-mile segments of several-hundred mile corridor. Fig. 13
compares four corridors using the access point density, trafc
volume and average priority score metrics. The scores indicating
likelihood of land development are represented by a gradient color
scale with darker areas indicating a higher likelihood. Specically,
red and orange represent a high likelihood, yellow represents a
medium likelihood, and green represents a low likelihood. Trafc
volume is measured by the average daily trafc for each mile
segment and is graphed along the linear roadway mileage. The
linear mile-densities are graphed in the same manner to chart the
progression along the route mileage.
As described in Scenario 1 of Table 1, corridor segments
exhibiting a coincidence of high potential for land development,
high access-point density, and high trafc volume are high priority
segments requiring immediate protective investments. As this
coincidence suggests the corridor segment is already somewhat
developed, yet facing additional future development, this situation
indicates an already costly and increasing access management
implementation. These segments will require further detailed
corridor investigation of potential remediation strategies that
may include land-use regulations, advance right-of-way acquisition, building access roads, restricting left turns, imposing transportation impact fees, and others in collaboration with the diverse
infrastructure stakeholders. Conversely, segments low in trafc
volume and access point density may be more cost-effective to
protect since they represent areas that are expected to develop but
have not yet been developed. Although land with a high likelihood
of development includes highway corridors, the greatest potential
for regret exists on corridors that currently do not have sufcient
access management. Because interstate highways already have
controlled access, they are of less concern for this study.
For example, Fig. 8 shows extremes in trafc volumes observed
near mileposts 22, 58 through 60 and 70 through 80 on US 50. The

Fig. 8. Corridor Trace Analysis (CTA) showing relationships among Access Points Per Mile, likelihood of Land Development Prediction, and Average Daily Trafc on a
transportation corridor (1 mile 1.6 km).

S.A. Thekdi, J.H. Lambert / Reliability Engineering and System Safety 138 (2015) 112

Fig. 9. Corridor Trace Analysis (CTA) showing relationships among Access Points Per Mile, likelihood of Land Development Prediction, and Average Daily Trafc on the US 52
corridor (1 mile 1.6 km).

Fig. 10. Corridor Trace Analysis (CTA) showing relationships among Access Points Per Mile, likelihood of Land Development Prediction, and Average Daily Trafc on the I-81:
Crescent corridor (1 mile 1.6 km).

gure shows peaks in access-point densities, suggesting a relatively high level, observed near Winchester, Upperville, Fairfax,
and Washington, DC (near milepost 90). Based on regional
characteristics, high access point density and trafc volume are
expected as the chart approaches the largely urbanized region of
Washington, DC. There is a need to identify segments classied as
the notional Scenario 1 (described in Table 1), with high likelihood
of land development, high access point density, and high trafc
volume. These conditions may imply that implementation of an
access management protection plan would be costly [24]. The dark
coloring of land parcels near Winchester (milepost 22) suggests
that this area is threatened by future development. However, there
are also signicant amounts of current activity, as demonstrated

by a relatively high access point density and trafc volume for the
region. Conversely, segments classied as Scenario 4, with high
likelihood of development, low in trafc volume, and low access
point density may be more cost-effective to protect in the midterm because access management actions do not require costly
retrotting and can be proactively managed with land developers,
localities, and other stakeholders.

4. Conclusions and future work


This paper has described the Corridor Trace Analysis tool (CTA) to
monitor and prioritize transportation corridor segments that are

10

S.A. Thekdi, J.H. Lambert / Reliability Engineering and System Safety 138 (2015) 112

Fig. 11. Corridor Trace Analysis (CTA) showing relationships among Access Points Per Mile, likelihood of Land Development Prediction, and Average Daily Trafc on the I-64:
East to West corridor (1 mile 1.6 km).

Fig. 12. Corridor Trace Analysis (CTA) showing relationships among Access Points Per Mile, likelihood of Land Development Prediction, and Average Daily Trafc on the US
29: Seminole corridor (1 mile 1.6 km).

vulnerable to future land development of the region. The CTA method


provides agencies with decision support anchored in a quantitative
framework that can be adapted to the unique needs of communities
and regions. As the contribution of this paper is the integration of
established replicable analyses, no single element stands alone. The
method is in need across the nation for the protection of transportation infrastructure [8]. The CTA provides transportation planners with
an approach to perform systematic and strategic monitoring for
resource allocation in access management programs. The approach
is useful for engaging a variety of stakeholders in a transportation
system of the region and is re-scalable to a variety of time horizons
and regions. Although CTA implementation costs for a single state are
estimated to be $50,000, incorporating CTA results into public private
partnerships, corridor management, and protective investment

prioritization processes have potential to result in multi-million dollar


long-term cost savings.
The following limitations exist for this paper. First, data availability may limit the application of a Corridor Trace Analysis. For
example, some geographic areas may not be represented in satellite
imagery with sufcient detail to characterize access point densities,
some satellite imagery may contain cloud cover or foliage that
blocks the view of access points, and types identied within the
expert elicitation process may also be unavailable in certain geographic regions and at the necessary scale. Second, categorization of
numerical risk levels as High, Medium, and Low may lose information. The use of qualitative ratings may add error to results, degrade
resource allocation, and be subjective [5]. Instead, retaining quantitative variable levels could improve robustness of the model. Third,

S.A. Thekdi, J.H. Lambert / Reliability Engineering and System Safety 138 (2015) 112

11

Fig. 13. Corridor Trace Analysis (CTA) making a comparison of four corridors using Access Points Per Mile, Likelihood of Land Development Prediction, and Average Daily
Trafc (1 mile 1.6 km).

if access point identication is performed manually, performing


Corridor Trace Analysis on a multi-state level may be cost prohibitive. However, emergent image processing tools [6] may streamline
the process.
This paper has highlighted several opportunities for future work.
Although the methods of this paper focused on highway corridors
supporting multimodal functions, future work will adapt the methods
to transit alternatives. There is opportunity to assess system characteristics in order to identify control limits for coincidences of access point
density, trafc and future land use potential. There is also need to
investigate how regional conditions for access point density and trafc
volumes should inuence classication of high or extreme values.
For example, a peak in access point density and trafc volume in a
rural region may be of higher concern for decision-makers compared
to a similar peak in an urbanized region. In addition, there is need to
adapt the methods of this paper to contain a continuous risk function,
allowing trafc volume and access point density retain quantitative
properties. As a result, the quantitative output of the risk function can
indicate the urgency and need for protective investment, thereby
increasing prioritization accuracy and addressing the limitations described above. As data availability increases, future work will test the
CTA model robustness to additional risk factors, such as economic
conditions, environmental conditions, regional corridor management
policies, and others.
Acknowledgment
The authors are grateful to the constructive feedback from the
reviewers of the Reliability Engineering System Safety Journal. In
addition, they are grateful to advice Chad Tucker, Ross Hudnall, Robin
Grier, Marsha Fiol, Amy O'Leary, John Miller, Gary Allen, Michael
Perfater, Rick Tambellini, and others of the Virginia Department of
Transportation; Chris Gist of the University of Virginia Scholars Lab;
Qian Zhou of the Department of Systems and Information Engineering

at the University of Virginia; The Virginia Center for Transportation


Innovation and Research; the National Cooperative Highway Research
Program; and the US Federal Highway Administration. University of
Virginia students Juan Aguilar, Matthew R. Giorgis, Cait E. Walcoff,
Michael A. Gude, Preut Thanarat, Andrew Watson, Natasha Hemminger, Philip Rinehart, John Fitzsimmons, John MacKenzie, and Zubin
George made signicant contributions to the data collection and
analysis.

References
[1] Bardossy A, Duckstein L. Fuzzy rule-based modeling with applications to
geophysical, biological and engineering systems. Boca Raton: CRC Press; 1995.
[2] Bier VM. On the state of the art: risk communication to the public. Reliab Eng
Syst Saf 2001;71(2):13950.
[3] Black JA, Paez A, Suthanaya PA. Sustainable urban transportation: performance
indicators and some analytical approaches. J Urban Plan Dev 2009;128(4):184230.
[4] Burchell RW, Listokin D . Land, infrastructure, housing costs and scal impacts
associated with growth: The literature on the impacts of sprawl versus
managed growth (No. Product Code: WP95RB1); 1995.
[5] Cox Jr. LA. What's wrong with risk matrices? Risk Anal 2008;28(2):497512.
[6] Exelis. ENVI, http://www.exelisvis.com/ProductsServices/ENVI/ENVI.aspx;
2014.
[7] Federal Highway Administration. Transportation corridor preservation: a
survey of current practice by local government planning departments,
http://www.fhwa.dot.gov/real_estate/practitioners/right-of-way/corridor_
management/case_studies/cp_local.cfm; 2000.
[8] Fiol M, DeLong MW, Knaster P, Glendening C, Bohard J, Kay C, et al. Best
practices for risk-based forecasts of land volatility for corridor management
and sustainable communities. scan team report. NCHRP Project 2068A, Scan
10-01. National Cooperative Highway Research Program. January 2012. p. 126.
http://onlinepubs.trb.org/onlinepubs/nchrp/docs/NCHRP20-68A_10-01.pdf;
2012.
[9] Frawley WE, Eisele WL. Investigation of access point density and raised
medians: crash analysis and micro-simulation. Texas Transportation Institute.
Project 0-4221. http://d2dtl5nnlpfr0r.cloudfront.net/tti.tamu.edu/documents/
0-4221-P1.pdf; 2004.
[10] Gluck J. Impacts of access management techniques. Washington, DC: National
Academy Press; 1999.

12

S.A. Thekdi, J.H. Lambert / Reliability Engineering and System Safety 138 (2015) 112

[11] Hamilton MC, Thekdi SA, Jenicek EM, Harmon RS, Goodsite ME, Case MP, et al.
Case studies of scenario analysis for adaptive management of natural resource
and infrastructure systems. Environ Syst Decis 2013;33(1):89103.
[12] Hamilton MC, Lambert JH, Keisler JW, Linkov I, Holcomb FM. Research and
development priorities for energy security of military and industrial installations. J Infrastruct Syst 2012;19(3):297305.
[13] Kenworthy J, Laube F. Urban transport patterns in a global sample of cities &
their linkages to transport infrastructure, land use, economics & environment.
World Transp Policy Pract 2002;8:519.
[14] Lambert, JH Linthicum AS, Kim EK, Kincaid LR, Rash SM, Schmidt GW. Riskbased framework using geographic information systems to identify transportation corridors vulnerable to development. VTRC No. 08-CR8. US Federal
Highway Administration and Virginia Transportation Research Council, 2008.
p. 76.
[15] Linthicum A, Lambert JH. Risk management for infrastructure corridors
vulnerable to adjacent land development. J Risk Res 2010;13(8):9831006.
http://dx.doi.org/10.1080/13669877.2010.486076.
[16] Martinez LJ, Joshi NN, Lambert JH. Diagramming qualitative goals for multiobjective project selection in large-scale systems. Syst Eng 2011;14(1):7386.
[17] Maze, T, Access management awareness program phase II report. Ames
Center for Transportation Research and Education, Iowa State University, 1997.
[18] Mitra A. Fundamentals of quality control and improvement. Hoboken, N.J.:
John Wiley & Sons; 2012. p. 275.
[19] Montgomery DC. Statistical quality control. New York: Wiley; 2000.
[20] Moudon AV, Sheikh A, Kraus A, Stewart O, Hallenbeck ME. Land Development
Risks Along State Transportation Corridors, 2013. (No. WA-RD 805.1).
[21] Oswald MR, McNeil S. Rating sustainability: transportation investments in
urban corridors as a case study. J Urban Plan Dev 2009;136(3):17786.

[22] Plazak, D. Long-term impacts of access management on business and land


development along Minnesota Interstate 394. In: Proceedings of the 2005
mid-continent transportation research symposium; 2005.
[23] Rosqvist T. On the validation of risk analysisA commentary. Reliab Eng Syst
Saf 2010;95(11):12615.
[24] Schmidt GW, Rash SM, Kincaid LR, Kim EK, Lambert JH. Incorporating the costs
of access management to the prioritization of transportation corridors vulnerable to land development. In: Proceedings of the IEEE systems and information
engineering design symposium, 2008 (SIEDS 2008); 2525 April 2008. p. 959.
http://dx.doi.org/%2010.1109/SIEDS.2008.4559692 URL:http://ieeexplore.ieee.
org/stamp/stamp.jsp?tp=&arnumber=4559692&isnumber=4559674.
[25] Teng K, Thekdi SA, Lambert JH. Identication and evaluation of priorities in the
business process of a risk or safety organization. Reliab Eng Syst Saf
2012;99:7486.
[26] Thekdi SA, Lambert JH. Decision analysis and risk models for land development affecting infrastructure systems. Risk Anal: Int J 2012;32(7):125369.
[27] Thekdi SA, Lambert JH. Quantication of scenarios and stakeholders inuencing priorities for risk mitigation in infrastructure systems. J Manag Eng
2014;30(1):3240.
[28] Waddell P. UrbanSim: modeling urban development for land use, transportation, and environmental planning. J Am Plan Assoc 2002;68(3):297314.
[29] Williamson IP. Land administration best practice providing the infrastructure for land policy implementation. Land Use Policy 2001;18(4):297307.
[30] Wood MD, Bostrom A, Bridges T, Linkov I. Cognitive mapping tools: review
and risk management needs. Risk Anal 2012;32(8):133348.
[31] Wu F, Webster CJ. Simulation of land development through the integration of
cellular automata and multicriteria evaluation. Env Plan B 1998;25:10326.

Das könnte Ihnen auch gefallen