Traffic Microsimulation of Illegal On-street Parking in Downtown Toronto
by
Ahmed Ramadan
A thesis submitted in conformity with the requirementsfor the degree of Masters of Applied Science
Department of Civil EngineeringUniversity of Toronto
© Copyright by Ahmed Ramadan 2016
ii
Traffic Microsimulation of Illegal On-street Parking in Downtown
Toronto
Ahmed Ramadan
Master of Applied Science
Department of Civil EngineeringUniversity of Toronto
2016
Abstract
This research investigates the effect of vehicles parked illegally on-street on the performance of
city streets during the AM rush hour in Downtown Toronto. Illegally parked vehicles effectively
block a lane on a street, forcing traffic to merge onto a single lane. This problem might be
inconvenient at uncongested times of day, but it creates significant delays to drivers, buses,
streetcars and cyclists as roadways are already operating near or at capacity during the rush hour.
A traffic microsimulation model that integrates illegal on-street parking into the Toronto
Waterfront Network Paramics Model was generated to measure the effect of illegally parked
vehicles on the flow of traffic. The study concluded that illegal on-street parking significantly
increases link delays, link travel times, and reduces link flows and link speeds. The results imply
that illegal on-street parking reduces the level of service of Downtown Toronto streets during the
AM rush hour, and that existing traffic microsimulation models underestimate the vehicle travel
times in the network for that period.
iii
Acknowledgments This research would not have been possible without the opportunity provided to me by Professor
Matthew Roorda, whose guidance, patience and support kept me going through turbulent times
of the research. Matt was a great source of motivation, and his always positive attitude also
served as a life lesson to never give up.
I would also like to of course dedicate this thesis to my family, the best and kindest family
anyone could ever wish for. I will start by thanking my wonderful parents, whose success in life
motivated me to constantly push my limit in order to be just a little bit like them. They have
managed to give me all the care and support I need while leading very busy and hectic lives. My
parents have dedicated their life to us, and for that I will forever be grateful. My sister, Sara, is
one of my best friends, and she always the first one to congratulate me when I succeed, and also
scold me when I screw up.
My thanks also go to the friends and colleagues that I gained in the ITS lab at the University of
Toronto. You have made my journey easier with your companionship. Sami and Toka in
particular were my un-biological siblings, who have always stood by my side and jumped at the
chance to help me whenever they could.
iv
Table of Contents
Table of Contents
ACKNOWLEDGMENTS..........................................................................................................................III
TABLE OF CONTENTS ...........................................................................................................................IV
LIST OF TABLES....................................................................................................................................VI
LIST OF FIGURES .................................................................................................................................VII
1 INTRODUCTION ............................................................................................................................1
2 LITERATURE REVIEW ....................................................................................................................2
2.1 PARKING POLICIES............................................................................................................................2
2.2 CRUISING FOR PARKING ....................................................................................................................6
2.3 PARKING MICROSIMULATION MODELS ................................................................................................6
2.4 BEHAVIORAL MODELLING OF DRIVERS’ PARKING CHOICES ......................................................................9
3 TRAFFIC MICROSIMULATION ......................................................................................................13
3.1 WHAT IS TRAFFIC MICROSIMULATION ...............................................................................................13
3.2 USES OF MICROSIMULATION ...........................................................................................................13
3.3 ILLEGAL PARKING & MICROSIMULATION STUDIES................................................................................14
3.4 SHORTCOMINGS OF MICROSIMULATION STUDIES ................................................................................14
4 ILLEGAL PARKING IN THE CITY OF TORONTO...............................................................................16
4.1 PARKING SUPPLY AND DEMAND .......................................................................................................16
4.2 TORONTO’S PEAK PERIOD ON-STREET PARKING POLICY........................................................................17
4.3 PARKING VIOLATIONS IN THE CITY OF TORONTO..................................................................................17
4.4 PARKING ENFORCEMENT IS EFFECTIVE BUT LIMITED.............................................................................17
5 DATA..........................................................................................................................................19
5.1 TRAVEL DEMAND MATRICES............................................................................................................19
5.2 STUDY AREA .................................................................................................................................19
5.3 MICROSIMULATION FRAMEWORK.....................................................................................................20
5.3.1 Quadstone Paramics .............................................................................................................20
5.3.2 Toronto’s Waterfront Paramics Network..............................................................................21
5.3.3 Microsimulation Model: The Base Case ................................................................................22
v
5.3.4 Microsimulation Model: The Illegal Parking Component......................................................22
5.4 MODEL CALIBRATION .....................................................................................................................23
5.4.1 Number of Model Runs Required ..........................................................................................24
5.4.2 Model Development..............................................................................................................25
5.4.3 Model Calibration .................................................................................................................25
5.5 TORONTO PARKING CITATIONS RECORD ............................................................................................26
5.5.1 Overview ...............................................................................................................................26
5.5.2 A Measure of the Non-compliance Rate ...............................................................................28
6 METHODOLOGY OVERVIEW........................................................................................................29
6.1 DATA FILTERING ............................................................................................................................29
6.2 GEOCODING INFRACTIONS’ ADDRESSES .............................................................................................30
6.3 CODING ILLEGAL ON-STREET PARKING INTO PARAMICS ........................................................................31
7 RESULTS .....................................................................................................................................33
7.1 SCENARIOS ...................................................................................................................................33
7.2 SUMMARY OF RESEARCH SCOPE .......................................................................................................33
7.2.1 Simulation Runs ....................................................................................................................33
7.2.2 Simulated Links .....................................................................................................................34
7.2.3 Performance Metrics.............................................................................................................34
7.3 SUMMARY OF RESULTS ...................................................................................................................35
7.3.1 Individual Simulation Days....................................................................................................35
7.3.2 Overall Summary...................................................................................................................35
7.4 T-STATISTIC TEST ...........................................................................................................................36
7.5 DISCUSSION OF RESULTS .................................................................................................................38
8 CONCLUSION & FUTURE WORK ..................................................................................................39
8.1 POLICY IMPLICATIONS OF SIMULATION MODEL ...................................................................................39
8.2 CONCLUSION ................................................................................................................................40
8.3 FUTURE WORK..............................................................................................................................40
REFERENCES .......................................................................................................................................42
APPENDIX...........................................................................................................................................47
vi
List of Tables Table 1 Three broad paradigms or mindsets on parking based on two criteria (Barter, 2015) .......2
Table 2 Paramics Model Calibration .............................................................................................25
Table 3 Breakdown of Parking Citations, 2011.............................................................................27
Table 5 Adjacent Links - Results Summary ..................................................................................35
Table 6 t-Test: Two-Sample Assuming Unequal Variances - Link Delay ....................................36
Table 7 t-Test: Two-Sample Assuming Unequal Variances - Link Flow ....................................36
Table 8 t-Test: Two-Sample Assuming Unequal Variances - Link Speed....................................37
Table 9 t-Test: Two-Sample Assuming Unequal Variances - Link Travel Time..........................37
vii
List of Figures Figure 1 Toronto Waterfront Network (Amirjamshidi, Mostafa, Misra, & Roorda, 2013) ..........19
Figure 2 A Portion of the Quadstone Paramics Simulated Toronto Waterfront Network.............21
Figure 3 Distribution of Infractions by Period of Day ..................................................................27
Figure 4 Toronto Parking Citations Record...................................................................................28
1
1 IntroductionThe City of Toronto Central Business District (CBD) experiences the highest volumes of traffic
during the A.M. and P.M. peak periods, when travel demand is at its maximum value for the day.
During these peak periods, congestion resulting from high traffic volume arises, causing
significant delays to passenger vehicles, commercial vehicles, streetcars and buses. In an effort to
alleviate these congestion levels, the City of Toronto, like many major cities around the world,
restricts on-street parking on most major streets during the peak periods in the CBD. This policy
ensures that the streets’ full capacity is utilized since on-street parking effectively blocks the
right-most lane. A vehicle parked on-street forces the vehicles behind it to merge into the next
lane, causing a bottle-neck at that location.
However, as with any parking policy, the compliance rate is not 100%. Between the years 2008
and 2014, 2.7 million parking infractions per year on average were recorded in the City of
Toronto (City of Toronto, 2015). The offenders that do not comply to the rush hour parking
restrictions, either by standing, stopping or parking on prohibited streets, exacerbate the already
critical traffic situation in the CBD. In addition to the delays caused by illegal parking, the
conflict resulting from vehicles switching lanes and cyclists exiting the bike lanes can pose a
safety concern. In an effort to try to discourage this phenomenon, a parking enforcement blitz
was launched in January 2015 and again in October of that year. Extra parking enforcement
officers were dispatched during morning and afternoon rush hours. Offenders were ticketed then
towed. The cost of the ticket is $150 and towing costs $200, in addition to the inconvenience
encountered by drivers to recover their towed vehicles. Between January and October, more than
61,000 vehicles were ticketed and more than 12,000 towed (Shum, 2015).
This research uses traffic microsimulation to study the impact of illegal parking on congestion
during the A.M. peak period in Toronto’s CBD. Although simulation models for Toronto’s road
network exist, these models omit illegal parking and therefore do not account for their adverse
effects on network travel times and delays. This research builds on an existing microsimulation
model and tries to improve its accuracy and realism by incorporating illegal parking into the
model.
2
2 Literature ReviewParking remains an under-studied area in the field of transportation. However, more and more
studies are emerging recently as a result of parking shortages across major cities and as
successors to previous studies that proved parking as a significant component of a vehicle trip.
2.1 Parking PoliciesBarter (2015) proposes a new approach to classifying parking policies. He argues that the
research field lacks a common classification approach, and as result confusion arises on the
distinguishing factors between policy alternatives and the key assumptions behind them. Barter
proposes a new taxonomy that classifies parking polices into three main paradigms that are
identified based on two criteria. These two criteria are:
a) Whether parking is provided on every site or it is provided to serve many sites within the
surrounding area.
b) Whether parking is seen as infrastructure that is planned based on certain guidelines or
whether it is seen as a market good where prices, supply, and demand interact through
market mechanisms.
Table 1 Three broad paradigms or mindsets on parking based on two criteria (Barter,
2015)
Parking facilities serve their
district
Every site should be fully
served by on-site parking
Parking is a market good
(real-estate based service)
“Responsive” approaches No cases
Parking is “infrastructure” “Area Management”
approaches
“Conventional site-focused”
approaches
3
The systematic classification approach proposed by Barter was then used to identify and lay out
the distinctions between three main parking policy reform streams. The first reform type is
known as right-sizing reforms, where parking supply is reduced from what can be described as
plentiful to a level that matches demand. Then, the second reform stream involves a shift from
site-focused approaches to area management approaches. Finally, a third reform stream redefines
parking as a market good, a departure from a more conventional definition that classifies parking
as an “infrastructure” item, where minimum parking requirements needs to be mandated by law.
Simićević et al. (2013) use stated preference and logistic regression to predict drivers’ response
to changes in parking policies. More specifically, the effects of changing parking price and time
limitation were investigated.
Several hypothetical scenarios with different combinations of parking price and time limitation
levels were presented to drivers in a short interview. For each scenario presented, respondents
were asked to choose one of the following five responses: park on-street, park off-street, park at
the fringe of the zone, switch to public transport, or other.
A multinomial logit model (MNL) was used to represent the gathered data. The model was able
to predict that as the parking price increases, the probability of parking in the zone decreases, and
the share of visitors that give up driving to the zone increases. On the other hand, time limitation
was shown to have no significant impact on the amount of drivers driving to the area. This was
attributed to the fact that a 1-hour time restriction is already being imposed on on-street parking
in the study area, and that visitors needing to park for durations that exceed the allowable
duration have the alternative of parking off-street. Furthermore, visitors that work in the area
were the most sensitive to parking policy changes, and a negligible amount of respondents
indicated that they would give up going to the zone all together as a result of policy changes.
Therefore, parking pricing can be used as a tool to influence demand to an area, while time
restriction enables the re-allocation of existing demand between different parking types (on-
street and off-street).
4
Millard-Ball et al. (2014) evaluate the relationship between parking occupancy rules and metrics
of direct policy interest. More specifically, the relationship between average occupancy and the
probability that a driver finds a block full, as well as the relationship between average occupancy
and the number of blocks cruised were evaluated. Then, the authors presented a methodology of
simulating cruising and arrival rates from hourly occupancy data. Finally, the degree to which
the SFpark met its objectives was evaluated.
SFpark is a smart-parking initiative launched in San Francisco, California with the aim of
improving the management of on-street and off-street parking. The pilot project was launched in
2011 and included seven pilot areas and two control areas in the San Francisco downtown area.
The SFpark program consists of three main elements:
a) Demand-responsive pricing to reduce cruising for on-street parking.
b) Smart payment methods to offer alternatives to the conventional payment-by-coins
method
c) Better parking information by providing drivers with real-time and static information
about availability and location of parking.
SFpark’s target average block occupancy rate is between 60-80%, which is in line with the
accepted occupancy rate in literature that is thought to eliminate cruising for parking. Parking
rate would go up on busy blocks during peak periods and go down when parking is in low
demand.
Millard-Ball et al. (2014) conduct both descriptive and regression analyses to evaluate the
effectiveness of the SFpark program. The results of the descriptive analysis show that despite
several rate changes, the hourly average occupancy of pilot areas did not change significantly,
and intuitively there was also little effect on the amount of cruising and the probability of a block
being full. However, it should be noted that the control area was worse off in that regard. Parking
availability was good in large in the study area, and the majority of the system operated below
80% occupancy. As for the regression analysis, the impacts of the first ten rate changes were
evaluated. It was concluded that the effect of each individual rate change on its own is small, but
when combined together, their collective impact is significant. On average, a rate change brings
occupancy 0.1-0.2% closer to the 60-80% range, and a reduction in cruising distance by 0.007-
0.017 blocks. This suggests an occupancy reduction of 1-2% and a cruising distance reduction of
5
0.07-0.17 blocks (50% reduction when compared to the mean cruising distance in the area), both
substantial outcomes.
Finally, the authors conclude that the SFpark program is slowly succeeding in bringing the
average occupancy rate within the target 60-80%, thus implying that the benefits of demand-
responsive pricing are realized in the long term. On the other hand, the lack of notable change in
average occupancy rates of the pilot area is attributed to the rebounding economy, which is likely
to have caused an increase in demand.
Arnott & Inci ( 2006), on the other hand, examine the issue of on-street parking and traffic
congestion from an economic perspective by presenting an integrated model. The model’s
underlying hypothesis is that the demand for parking in downtown areas is dependent on both the
money and time costs of the trip, with cruising for parking being a significant contributor to the
latter cost component. The model was used to develop two strategies that could eliminate
cruising for parking in a saturated parking environment.
The optimum solution was to increase the on-street parking fee to a level that ensures cruising is
eliminated without leaving the parking unsaturated. The second-optimum solution suggested
increasing the amount of curbside parking if the parking fee is to be kept at a suboptimal level.
Kobus et al. (2013) estimate several probit models to study the effect of parking prices on
drivers’ choice between on-street and off-street parking. The study focuses on the city of Almere,
which is the fastest growing city in the Netherlands. The study concludes that when the average
distance to the final destination is larger for off-street parking, drivers are willing to pay a
premium for on-street parking. Furthermore, the price elasticity of demand for on-street parking
for a duration of one hour is -5.5. The price elasticity for on-street parking is much smaller for
shorter parking durations. As a result, a parking policy that imposes a price premium on on-street
parking is deemed as welfare improving.
Arnott, Inci & Rowse (2015) acknowledge the abundance of literature focusing on parking
policies that increase the price of on-street parking in an attempt to reduce the externalities of
cruising. The study focuses on a different policy aspect: given a fixed on-street parking price,
what is the optimal quantity of on-street parking that should be supplied?
6
The optimum solution favours on-street parking when demand is low relative to on-street
capacity, which ensures the elimination of cruising for parking. For intermediate demand, both
on-street and off-street parking are provided. Only off-street parking is provided when demand is
high. Furthermore, the findings confirm the hypotheses of previous studies that as the price
difference between on-street parking and off-street parking increases, cruising for parking also
increases.
2.2 Cruising for ParkingVan Ommeren et al. (2012) use a random sample of car trips in the Netherlands to provide
empirical evidence that supports the largely theoretical literature about cruising for parking. The
prices of on-street and off-street parking are largely the same in the Netherlands, and the average
cruising time of only 36 s obtained from the sample data supported the notion that increasing the
price of on-street parking to match that of off-street parking reduces cruising time dramatically.
The conclusion that increasing the price of on-street parking reduces cruising can also be
supported by comparing the Netherlands average cruising time of 36 s with the average cruising
time of eight minutes computed by Shoup (2006) using data from 22 US cities where on-street
parking tend to be underpriced.
Van Ommeren et al. (2012) also use the sample data to examine the determinants of cruising.
The results indicate that there is a negative relationship between income and cruising time, which
was attributed to the higher value of travel time associated with higher income drivers. The
results also indicate that cruising times are substantially higher for leisure trips, which was
attributed to the assumption that parking is supplied at below peak values. Furthermore, cruising
time was found to peak in the morning when parking demand is higher.
2.3 Parking Microsimulation ModelsSeveral models have been developed over the years to form a basis for a systematic
microsimulation of agents’ parking behavior in an urban environment. The models attempt to
capture the interaction between the supply and demand of parking within a spatial-temporal
context.
When developing a parking simulation model, one has to represent four key decisions made by
the driver in the model: Parking search start point, search route, parking type choice, parking lot
7
choice. Horni et al. (2013) develop an agent-based simulation model that uses cellular automaton
(CA) to simulate cruising for parking. The search start point was dependent on the linear distance
from destination and is uniformly sampled from a distance range specified in a configuration file
to represent different search tactics made by different types of drivers.
The parking search route, on the other hand, was dependent on the agent’s mental map of the
area. As an agent approaches an intersection, it decides on the next link to use based on two
criteria:
• Destination Approaching Efficiency – a link has a higher probability of being
chosen if it leads the agent closer to the final destination.
• Memorized Free Parking Spaces- a link has a higher probability of being chosen
if it leads to parking lots with the most free spaces as predicted by the agent based
on the mental map created from previous parking search experiences around the
same destination.
As for the parking lot choice, a probabilistic choice model was used. The model was depended
on elapsed search time and distance to destination. Intuitively, the probability of accepting a
parking spot increases when distance to destination decreases or when search time increases.
Benenson et al. (2008) present an agent-based, spatially explicit model of city parking, named
PARKAGENT. Similar to the cellular automaton approach, the search start point is dependent on
the linear distance to destination.
However, when it comes to the parking lot choice, PARKAGENT adapts a unique approach.
Firstly, the model assumes that drivers always prefer to look for on-street parking at first, and
they only park off-street when the parking search time exceeds 10 minutes. The parking search
algorithm is as follows:
• Let x be distance from actual destination.
o Parking search area starts when x<dawareness
o dawareness is measured as 250 m from destination
o dparking measured as 100 m from destination
When dawareness<x<dparking, then:
• Agent reduces speed to 25 km/h
8
• Agent estimates fraction of unoccupied on-street parking places using:
𝑃𝑢𝑛𝑜𝑐𝑐𝑢𝑝𝑖𝑒𝑑 = 𝑁𝑢𝑛𝑜𝑐𝑐𝑢𝑝𝑖𝑒𝑑
𝑁𝑢𝑛𝑜𝑐𝑐𝑢𝑝𝑖𝑒𝑑 + 𝑁𝑜𝑐𝑐𝑢𝑝𝑖𝑒𝑑
When x<dparking, then:
• Agent reduces speed to 12 km/h
• Agent estimates the expected number of vacant spots to destination using:
• 𝐹𝑒𝑥𝑝𝑒𝑐𝑡𝑒𝑑 = 𝑃𝑢𝑛𝑜𝑐𝑐𝑢𝑝𝑖𝑒𝑑 ∗ 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝑡𝑜 𝐷𝑒𝑠𝑡𝑖𝑛𝑎𝑡𝑖𝑜𝑛𝐿𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑃𝑎𝑟𝑘𝑖𝑛𝑔 𝑃𝑙𝑎𝑐𝑒
• When the agent arrives at a vacant spot, it has to decide whether to park or to proceed
further towards the destination. For sufficiently high Fexpected (3-5), the driver will decide
to proceed towards destination and not to park in the next vacant spot. However, for low
Fexpected (eg. 0.5), the driver will choose to park in the next vacant spot
• Model continuously reestimates Punoccupied and consequently Fexpected along the
dparking interval to adjust the expectations based on the supply within the interval
When driver passes destination without parking (dparking>100 m):
• dparking starts accumulating as agent moves away from destination. dparking starts at
100 m from destination.
• dparking increases at a rate of 30 m/min
• Agent parks at the first available spot within the interval 100 m<dpakring<400 m
• If the total search time exceeds 10 minutes, the agent chooses to park in the nearest off-
street lot (Assuming vacancy is always available)
Moini et al. (2013a) study the effect of parking guidance systems (PGSs) on network mobility
and vehicular greenhouse gas (GHG) emissions generated by vehicles in a central business
district area. Only on-street parking was considered.
The study employs a simulation framework using the PARAMICS software. In the model, the
variable message signs (VMSs) feature of PARAMICS was used to present drivers with the most-
recent on-street parking availability information, thus mimicking the functionality of a parking
guidance system. The authors compared the existing base scenario where a PGS is not used with
several other scenarios that depict the utilization of PGSs with varying levels of congestion and
9
parking demand. Furthermore, market saturation was manipulated in the scenarios to determine
the optimal percentage of users of a PGS needed to achieve meaningful improvements to
mobility and emissions.
Moini et al. (2013b) conclude that PGSs have the potential to result in substantial improvements
to mobility and emissions, even with market saturation values as low as 25%. It was also
concluded that the coordination between a PGS and an Advanced Traveler Information System
(ATIS) such as GPS could amplify the benefits of PGS.
2.4 Behavioral Modelling of Drivers’ Parking ChoicesCools et al. (2013) study drivers’ mental knowledge of the parking facilities surrounding their
destinations. More specifically, the familiarity of drivers with the geographical location of
available parking lots is investigated. Lack of mental knowledge of the parking supply can have
a negative impact on local roads and parking lots, such as overcrowded “famous” lots, and
drivers having to circle or cruise around the area in search of parking facilities.
A sample of drivers was surveyed about their spatial knowledge of parking facilities in the
vicinity of the central shopping area of Hasselt, Belgium. Several factors that can potentially
impact drivers’ mental knowledge of the parking supply, including socio-demographic and
cognitive variables, were collected.
Only age, education, and the frequency of using the car when making shopping trips to Hasselt
were found to have a significant impact on drivers’ familiarity values. Other factors investigated
such as income and perceived mental knowledge were not proven to be of significance.
Shoup (2006) introduces a model of how drivers choose between on-street and off-street parking.
If drivers choose curbside parking, they may have to cruise for a vacant spot for an unknown
period of time if no spaces are available. Shoup states that “cruising creates a mobile queue of
cars that are waiting for curb vacancies, but no one can see how may cars are in the queue
because cruisers are mixed in with other cars that are actually going somewhere.”
The decision of whether to cruise for on-street parking or choose off-street parking where no
cruising is required was modelled using a simple equation quantifying the benefits and costs of
cruising. The benefit of parking on-street arises from the cost differential between the two
10
parking types when parking for short durations, as off-street parking is usually considerably
pricier. However, these savings are diminished by the cost of fuel burnt while circling & the cost
of time spent circling.
Two pricing strategies have been suggested to reduce cruising. The first is to increase the price
of on-street parking to match that of off-street lots. By doing so, the incentive to cruise is
eliminated. The second strategy suggested reducing the off-street parking price to match that of
on-street, again discouraging cruising. On the other hand, the author acknowledges the
simplifying assumptions that affect the realism of his model. For example, the model assumes
that the value of time for all drivers is constant, which is not the case as different drivers will
value time differently based on their attributes as well their trip’s. In addition, drivers in a real
life parking situation do not know in advance their parking search time, and therefore cannot
make an informed decision on the rational parking type choice.
Walking distance, although an important component of parking, was not explicitly represented in
Shoup’s model. Kobus et al. (2013) state that their study reveals that the average walking
distance to the final destination is minimized when drivers with longer parking durations park
further away from their destination, and consequently on-street parking should be left to drivers
with shorter parking durations as it is usually ubiquitous where off-street lots are available in
limited locations. This outcome can be achieved by making on-street parking pricier per unit
time compared to off-street parking.
Gallo et al. (2011) propose an assignment model to simulate parking choices. The model can be
used to simulate the impact of cruising on congestion. The model presented consists of a demand
model and a supply model. The demand model is taken as an hour of transportation demand,
obtained from an origin destination matrix. The demand is assumed to be constant and in steady
state. The supply model is described as a multilayer network supply model, where each network
layer simulates a trip phase. Gallo et al. (2011) identify three trip phases:
a) On-car trip between origin and destination zones (trip layer)
b) On-car cruising for parking at the destination zone (cruising layer)
c) Walking egress trip (walking layer)
11
Since the model’s objective is to explicitly represent cruising time in the network, a network
link called the parking link was specifically created for that purpose, where the cruising time
is represented by the cost function of that link. The parking link had the task of simulating
three main aspects of parking, namely the parking fare, the time needed to search for a free
parking spot, and the decision of drivers to search for parking at other facilities if parking at
the current facility is fully occupied. The proposed parking link cost function is:
𝐶𝑝 = 𝐶𝑝0 (1 + 𝛼 ( 𝑓𝑝𝐶𝑎𝑝𝑝 )𝛽)
Where is the generalized parking cost when the lot is empty, is the flow on the parking 𝐶𝑝0 𝑓𝑝
link, representing the amount of vehicles searching for parking in a specific lot, is the 𝐶𝑎𝑝𝑝
residual capacity of the parking facility in the simulation hour, and and are parameters to be 𝛼 𝛽
calibrated. Parking fare can be included by adding a fare component to . 𝐶𝑝
The proposed model is then tested on a trial network and it is determined that the model can be
utilized when the average parking saturation degree exceeds 0.7. Gallo et al. (2011)also note that
the model requires small traffic zones that accurately capture vehicles’ destinations to produce an
accurate simulation of cruising in the area.
Montini et al. (2012) analyze raw person-based GPS data from Zurich, Switzerland to obtain
parking search characteristics. A parking search analysis module is developed and then added to
the POSDAP framework to extend the framework’s capability to include parking search analysis.
POSDAP is an open source GPS data analysis framework.
The parking search analysis uses a longitudinal data set of three-dimensional, person-based GPS
positions. As reported by many parking researches before, the parking search start point cannot
be accurately determined for each driver because it is not possible to record the exact moment
when the driver starts to consciously initiate the search process. Therefore, the parking search
start point is assumed to be the point at which a driver is 800 meters from the parking space that
would be selected for parking. The parking search route would likely be captured within the 800
meters radius.
12
The parking search analysis returns results consistent with previous literature. The parking
search time is found to be higher in city centers. The parking search distance however is lower in
the centers. The lower search distance is attributed to the lower speeds associated with busy city
centers, as well as the possibility of drivers undertaking different search strategies. The analysis
of the GPS data indicates that the time driven within the 800 meters radius around the parking
space is less than four minutes for 80% of the cases, which lead the authors to conclude that the
parking search in Zurich does not pose an issue to mobility.
Montini et al. (2012) conclude that the extraction of parking data can be improved using GPS
data analysis, since conventional survey methods that asks drivers to recall their searching times,
walking distances, etc. can be inaccurate as drivers report these parameters from their memories
which are not always reliable.
13
3 Traffic MicrosimulationTraffic Microsimulation is the main component of this body of research. It is used as a
representation of real-life traffic conditions of an existing transportation network. Quadstone
Paramics, a popular microsimulation suite, is used in this study.
3.1 What is Traffic MicrosimulationMicrosimulation is “the dynamic and stochastic modeling of individual vehicle movements
within a system of transportation facilities (Dowling, Holland, & Huang, 2002).” Individual
vehicles, or agents, are explicitly and individually deployed into a simulated transportation
infrastructure, where that infrastructure can be a representation of existing transportation
facilities or future ones. The development of microscopic traffic simulation models started after
the introduction of car following models in the 1950s by Reuschel (1950) and Pipes (1953). The
car following model was based on the premise that driver’s keep a safe distance to the leading
vehicle in a magnitude proportional to the travelling speed.
Microsimulation takes into account the physical characteristics of vehicles and transportation
facilities, and with the help of empirically derived models such as car following, gap acceptance
and lane changing models, agents traverse the simulated environment in a manner that mimics
real life traffic conditions. The interaction of the individual agents with each other as well as
with the surrounding built environment creates a simulation that is consistent with observed
driving behavior. The simulation model results can then be displayed visually or as values of
performance measures of the network.
3.2 Uses of MicrosimulationTraffic simulation models are used to evaluate the performance of existing transportation
networks, identify congestion hotspots, and test different policy and infrastructure alternatives to
determine the most effective solution to improve traffic flow, safety, emissions, etc. Simulation
models can also be used to evaluate the performance of future or proposed transportation systems
and study the impact of future developments on existing road networks. A well calibrated model
would generate reliable performance metrics for a network without the need of conducting
expensive and time consuming field studies.
14
3.3 Illegal Parking & Microsimulation Studies
A microsimulation based assessment was conducted by Kladeftiras and Antoniou to study the
impact of double parking, a form of illegal parking, on traffic conditions in the city of Athens,
Greece (Kladeftiras & Antoniou, 2013). The sensitivity analysis concluded that limiting double
parking by means of increasing enforcement or adding strategically placed cones to prevent
double parking may result in an increase in vehicle speeds by 10-15% and a 15-20% decrease in
delay and stopped time. The study also concluded that eliminating double parking completely
may result in a 44% increase in vehicle speeds and a 33% decrease in delay as well as a 47%
decrease in stopped time.
Another microsimulation study explains that some drivers in Lisbon, Portugal choose to park
illegally when the designated roadside parking lot is fully occupied during high demand periods
(Lu & Viegas, 2007). The authors discovered that the practical scenarios to be simulated are
complex, as a result of the varying parking duration as well as the varying proximity of an illegal
parking incident from the upstream and downstream intersections of a link. The study concluded
that the effect of illegal parking is increased with higher traffic flows and that illegal parking
results in increased conflicts leading to decreased safety and higher chances of accidents.
Jia et al. (2003) use the cellular automata traffic flow model, a form of traffic microsimulation, to
study the effect of bottlenecks, illegal parking being one of its types, on traffic flow (Jia, Jiang,
& Wu, 2003). The simulation results reveal that the capacity of the bottleneck is slightly lower
than the maximum flow rate of a single-lane road.
3.4 Shortcomings of Microsimulation Studies
The main deficiency in traffic microsimulation research is that it doesn’t account for parking
most of the time. Parking activities, legal and illegal, are often omitted altogether from a traffic
simulation study. That omission implicitly implies that parking activities have no effect on the
flow of traffic. However, studies have shown that parking significantly influences the
performance of a transportation network, especially in downtown areas. Shoup (2006), using
data from 22 US cities, determined that the average cruising time for on-street parking in these
cities is eight minutes. Therefore, on average, a vehicle remains on the network eight minutes
15
past their arrival time to their destination. Montini et al. (2012) analyzed raw person-based GPS
data from Zurich to study parking search characteristics. The data revealed that parking search
times are higher in city centers. Shoup (2006) and Arnott & Inci (2006) argue that the main cause
of cruising for on-street parking is the fact that it is underpriced, making it worthwhile to spend
more time looking for that cheap parking spot rather than park in off-street facilities. On the
other hand, Van Ommeren et al. ( 2012) examined cruising for parking in the Netherlands and
found it to be negligible as a result of pricing on-street parking to resemble prices of off-street
parking. Therefore, as a result of most North American cities underpricing their on-street
parking, it is important to account for the significant parking component of passenger vehicle
trips to downtown areas in microsimulation models.
16
4 Illegal Parking in the City of TorontoThe Central Business District (CBD) of the City of Toronto received a large influx of vehicles in
the morning rush hour. In addition to the regular commuters that commute to work every day,
visitors also arrive during the morning peak period to run errands, access government and
commercial services, and shop. The large number of vehicles arriving into a dense urban area
creates a competition for the limited parking supply. This is when the problem of illegal parking
arises.
4.1 Parking Supply and DemandAs with any commodity, a shortage arises when the demand exceeds supply. Parking in the
downtown core is an example of a scarce commodity. The large number of arriving vehicles in
the morning rush hour is met with a limited supply of parking. Further exacerbating the issue of
parking in the downtown core is the competition between commercial and passenger vehicles for
curb space, forcing both types of vehicles to cruise for parking or park illegally when the legal
spots have been consumed by other vehicles. Haider (2009) reveals that the number of parking
tickets in Toronto increased by 70% between 2006 and 2009, and that major couriers paid around
$2.5 million in fines. Simply increasing the parking supply is challenging. First of all, space is at
a premium in the downtown area, and land is very expensive. Nowadays, most vacant land is
allocated to commercial or residential developments, rarely to parking facilities. Also, increasing
parking supply contradicts the attempts of the current urban planning policy that favors public
transit and discourages driving in every way possible to reduce emissions and improve mobility.
“Parking space which is not completely controlled by parking management is an over-saturated
system: that means parking demand exceeds parking supply or—to put it another way—
additional spaces attract additional cars. That is even true with illegal parking (Topp, 1993).” A
study conducted in New York also concluded that the minimum parking requirement, which
imposes a minimum number of spots that need to be provided in an area, encourage driving as
drivers are more likely to drive if parking is “guaranteed” (Weinberger, Seaman, & Johnson,
2008).
17
4.2 Toronto’s Peak Period On-street Parking PolicyThe City of Toronto Central Business District (CBD) experiences the highest volumes of traffic
during the A.M. and P.M. peak periods, when travel demand is at its maximum value for the day.
During these peak periods, congestion resulting from high traffic volume arises, causing
significant delays to passenger vehicles, commercial vehicles, streetcars and buses. In an effort to
alleviate these congestion levels, the City of Toronto, like many major cities around the world,
restricts on-street parking on most major streets during the peak periods in the CBD. This policy
ensures that the streets’ full capacity is utilized since on- street parking effectively blocks the
right-most lane. A vehicle parked on-street forces the vehicles behind it to merge into the next
lane, causing a bottle-neck at that location.
4.3 Parking Violations in the City of Toronto
The compliance rate to any parking policy is never 100%. Between the years 2008 and 2014, 2.7
million parking infractions per year on average were recorded in the City of Toronto (City of
Toronto, 2015). The offenders that do not comply with rush hour parking restrictions, either by
standing, stopping or parking on prohibited streets, exacerbate the already critical traffic situation
in the CBD. In addition to the delays caused by illegal parking, the conflict resulting from
vehicles switching lanes and cyclists exiting the bike lanes can pose a safety concern. In an effort
to try to discourage this phenomenon, a parking enforcement blitz was launched in January 2015
and again in October of that year. Extra parking enforcement officers were dispatched during
morning and afternoon rush hours. Offenders were ticketed then towed. The cost of the ticket is
$150 and towing costs $200, in addition to the inconvenience encountered by drivers to recover
their towed vehicles. Between January and October, more than 61,000 vehicles were ticketed and
more than 12,000 towed (Shum, 2015).
4.4 Parking Enforcement is Effective but Limited
Cullinane & Polak (1992) suggest that evidence of the existence of an illegal parking problem
motivates and justifies further study. The ever increasing number of tickets issued by the
enforcement officers in Toronto clearly proves the existence of a problem. As a response to this
phenomenon, The City of Toronto is constantly pressured to increase enforcement in an attempt
18
to deter illegal parking. Parking enforcement campaigns (also known as the parking blitz) is
often one of those responses.
However, parking enforcement, especially in its current form of deploying enforcement officers
to visually inspect an area, is expensive. Parking enforcement officers cannot be available
everywhere, all the time. Their efforts and resources need to be concentrated in problem areas.
Toronto’s CBD as a whole can be considered a problem area, due to its high demand for parking
and its low supply. Therefore, most of the parking studies in the City of Toronto need to examine
that area.
19
5 DataThis chapter describes the data that is used as an input to the illegal parking simulation model.
5.1 Travel Demand MatricesTravel demand within the simulated network is obtained from the Origin-Destination matrices
from the 2011 Transportation Tomorrow Survey (TTS). The TTS is a comprehensive travel
survey that is conducted every five years in the Greater Toronto and Hamilton Area (GTHA) by
the data management group at the University of Toronto (Data Management Group, 2011).
5.2 Study AreaThe study area is the Toronto Waterfront network. The area is bordered by Dundas Street in the North,
Woodbine Avenue in the east and Parkside Drive in the west.
Figure 1 Toronto Waterfront Network (Amirjamshidi, Mostafa, Misra, & Roorda, 2013)
49,691 vehicles traverse the Toronto Waterfront network in the morning rush hour (8 a.m. to 9
a.m.), according to the data released in the 2011 TTS survey. The highest number of vehicles
entering the network comes from the western portion of the Gardiner Expressway, which
releases 5,877 vehicles into the network, followed by the southbound Don Valley Parkway,
which adds 4,899 vehicles into the network. The zone bordered by Queen Street in the north,
King Street in the south, Bay Street in the east, and University Avenue in the west receives the
largest number of vehicles in the network during the morning rush hour, where it receives 5,253
vehicles, which is expected since this corridor is one of the densest in the downtown core.
20
The Waterfront Network has been chosen for this study since it the largest employment center in
the Greater Toronto Area and it receives a large number of vehicle trips relative to its area in the
morning rush hour. Furthermore, the limited parking supply in that network further exacerbates
the problem of illegal parking.
5.3 Microsimulation FrameworkAn integrated microsimulation approach that incorporates illegal on-street parking into
Quadstone Paramics is developed to create a more realistic microscopic traffic model. A
microsimulation traffic model consists of the various physical attributes of a road network,
including roads, intersections, signals and signal timings, speed limit on each road link and
turning restrictions. In conventional traffic microsimulation models, only one set of data is used
as an input to the network: travel demand. Travel demand is usually in the form of an origin
destination (OD) matrix that defines the number of vehicles travelling from one zone in the
network to the other at a given time interval. Once the travel demand is added to the
microsimulation model, the model’s algorithms create a traffic flow along the various links of
the network, in accordance with empirical traffic flow models such as car following, gap
acceptance and lane changing models (Quadstone Paramics, 2016). These models define how a
vehicle interacts with the various physical attributes of the road network as well as with other
vehicles in proximity.
5.3.1 Quadstone Paramics
In order to simulate the impact of illegal parking on congestion in Toronto effectively, a detailed
and accurate model of the transportation network is needed. Furthermore, the microsimulation
suite to be used needs to allow for a representation of parking within the model, either directly or
indirectly. At the time of writing this thesis, there have been no studies examining the effects of
illegal parking in Toronto specifically or in Canada as a whole.
Quadstone Paramics is used in this illegal parking study. This choice has been influenced by two
main factors:
1) The availability of a Toronto Waterfront Paramics Network (developed by IntelliCan
Transportation Systems)
21
2) Paramic’s API that allows for the use of the software for research topics that can
sometimes depart from the conventional framework of traffic microsimulation studies
Figure 2 A Portion of the Quadstone Paramics Simulated Toronto Waterfront Network
5.3.2 Toronto’s Waterfront Paramics Network
The microsimulation model replicates the physical characteristics of the actual Toronto
Waterfront Network, as attributes such as link lengths and widths, speed limits, turning
restrictions and signal timings have been surveyed by the model developers and subsequently
added into the Paramics software.
Quadstone Paramics deploys vehicles into the network based on an origin-destination (OD)
matrix specified by the user. This OD matrix reflects a realistic travel pattern that is observed
through a travel survey. Travel demand within the simulated waterfront network is obtained from
the Origin-Destination matrices from the 2011 Transportation Tomorrow Survey (TTS). The
TTS is a comprehensive travel survey that is conducted every five years in the Greater Toronto
and Hamilton Area (GTHA) by the Data Management Group at the University of Toronto (Data
Management Group, 2011).
22
5.3.3 Microsimulation Model: The Base Case
For reference, a simulation that does not involve illegal parking needs to be conducted. These
simulations serve as the benchmark to which the network’s performance under illegal parking
scenarios is to be compared.
Intuitively, the network’s performance metrics would deteriorate once illegal parking is added.
Furthermore, a base case simulation would generate the results that are currently being used by
planners in the City of Toronto to predict current traffic conditions and identify problem areas in
the network.
It should be noted however that all parameters aside from adding illegally parked vehicles should
be kept constant between base case simulations and illegal parking simulations. This ensures a
fair and accurate comparison between the two cases which will result in capturing the effect of
the illegal parking activity on traffic conditions without interference from other variables.
5.3.4 Microsimulation Model: The Illegal Parking Component
The main contribution of this study is the addition of the illegal parking activity during the
morning rush hour in the Toronto Downtown core to the simulated network.
The key output of this simulation model is a measure of the degradation of the network’s level of
service (LOS) attributed to illegal parking.
During the scope setting phase of this study, authors were faced with two possible approaches to
simulating illegal parking:
1) Creating and subsequently simulating hypothetical scenarios of randomly placed illegally
parked vehicles on the network
2) Simulating illegal parking activities that actually occurred in the network, at defined
times and locations
The advantage of the first approach is that one can perform a more well-rounded sensitivity
analysis of the effect of the location and time of illegal parking on congestion. By having control
23
over when and where to initiate an illegal parking event, it would be possible to manipulate the
two variables to study their effect on network performance.
However, it was determined that the study would be of more value if it were able to quantify
existing travel behavior rather than hypothesized ones. Putting a value on the current problem of
illegal parking in the City of Toronto can show the extent to which this problem is affecting
commute times and roadway capacities. If this problem is deemed as significant enough,
reflected in a significant difference between the performance metrics of the base case scenario
and the illegal parking scenario, it can serve as a motive for planners to always consider and
incorporate parking into their traffic studies, whether their studies involve microsimulation or
not. In additions, planners and policymakers in the City of Toronto would be provided with
scientific evidence that proves the negative impacts of illegal parking and encourages further
parking policy and infrastructure developments.
The Toronto parking citation record is used as a representation of the current state of drivers’
non-compliance to the AM-peak period parking restrictions. The parking citation record is a list
of all parking tickets issued by the Toronto parking enforcement officers in a year. The parking
citation record is explained in detail in the next chapter.
5.4 Model Calibration Several parameters can be calibrated to reflect different driving behaviors and conditions, to
account for the variability in driver behavior, weather conditions, traffic laws, etc. One of the
most effective ways of calibrating these parameters is to compare the simulation results with
observed data (goodness-of-fit test). This ensures that the combination of parameters used is able
to capture the behavior of vehicles in a transportation environment accurately. Barceló (2010)
describes the validation procedure as an iterative process that calibrates the model parameters
and then compares the model to actual system behavior, and then recalibrate the parameters if
necessary to minimize any discrepancies. Several measures of goodness-of-fit such as percent
error, mean error, route mean squared error and exponential mean absolute normalized error are
used to reflect the difference between performance measures from observed data and from the
simulation model. (Park & Qi, 2005) (Merritt, 2004) (Toledo, Ben-Akiva, Darda, Jha, &
Koutsopoulos, 2004) (Chu, Liu, Oh, & Recker, 2003) (Ma & Abdulhai, 2002) (Hourdakis,
24
Michalopoulos, & Kottommannil, 2003) (Barcelo & Casas, 2004) (Brockfeld, Kühne, &
Wagner, 2004).
Model parameters, such as driver aggressiveness and vehicle type distribution were calibrated by
(Amirjamshidi et al., 2013). A feedback period of three minutes is used as a representation of the
frequency at which the model recalculates the routes for its vehicles. This feedback period value
is a result of the calibration study by (Amirjamshidi et al., 2013).
5.4.1 Number of Model Runs Required
It should also be noted that microsimulation models, using a random number generator, or a
seed, outputs different results depending on the seed selected. The seed generates random
numbers that determine parameters such as the destination of each vehicle, the behavior of its
driver, and route assignment. Hollander and Lui (2008), after reviewing several methodologies
used to calibrate parameters, agree that it is insufficient to examine the results of a single run of a
simulation. The following formula can be used to determine the required number of runs to
achieve a certain level confidence:
(Shaaban & Radwan, 2005)
where:
is the mean of the performance measure generated by different runs𝑥
s is the standard deviation of the performance measure
is the allowable error specified as a fraction of the mean𝜀
is the critical value of t-distribution at significance level𝑡𝛼/2
25
5.4.2 Model Development
The Paramics Waterfront Network was developed by Intellican Transportation Systems and
refined over the years in several studies, most recently by Amirjamshidi et al. (2013). The
Waterfront Network is divided into 111 zones in the simulation model. It includes all the links of
the actual waterfront network, with their accurate speed limits and turning restrictions.
5.4.3 Model Calibration
Amirjamshidi et al. (2013) used freeway ramp counts, road counts and highway counts as a
respresentation of observed data in the Toronto Waterfront network. They calibrated the model
paramters using a simple genetic algorithm.
The calibration effort had three objective functions:
1) C Model – Calibrates to road counts only
2) CS Model- Calibrtes to road counts and link average speeds
3) CSA Model- Calibrates to road counts, link average speeds, and link standard devation of
acceleration
The calibrated network parameters for each model are summarized in Table 2.
Table 2 Paramics Model Calibration
Parameter C Model CS Model CSA Model
Reaction time (sec) 0.85 0.86 0.63
Headway (sec) 0.83 0.8 1.94
Timestep per second 3 4 2
Feedback int (min) 3 3 4
Familiarity (%) 86% 86% 89%
26
Perturbation (%) 8% 8% 14%
Maximum passenger
car acc (m/s2)
2.71 2.53 1.88
Maximum passenger
car dec(m/s2)
-3.68 -3.73 -2.62
The CS model parameters were used in the simulations, since it provides good accuracy by
calibrating to both road counts and link average speeds. The CSA model is intended to account
for driver aggressiveness, which isn’t applicable in this research.
5.5 Toronto Parking Citations Record
5.5.1 Overview
Parking citation data is published by the City of Toronto in its open data website (City of
Toronto, 2015). The citation data is published on a yearly basis and contains a list of all the
parking tickets issued in the City of Toronto for that year. Parking citations for the year 2011 are
used in this research to be consistent with the 2011 travel demands obtained from the TTS. The
parking citations record contains the following details about a parking citation:
• Date of infraction
• Time of infraction
• Type of infraction
• Location of infraction
• Fine amount
The total number of infractions recorded in the year 2011 was 2,805,492 infractions. Out of these
infractions, 882,956 were for vehicles parked or stopped on a street at a prohibited time of day
(220,763 in the waterfront area). The distribution of infractions over the different periods of the
day (AM Peak, Mid-day, PM Peak, Off-peak and overnight) is shown in Figure 3.
27
67,933
441,550
222,757
126,849
23,867
13,769
118,704
53,106
28,820
6,364
0 100,000 200,000 300,000 400,000 500,000
AM Peak (6 am- 9 am)
Mid-day (9 am- 3 pm)
PM Peak (3 pm- 7 pm)
Off-peak (7 pm- 12 am)
Overnight (12 am- 6 am)
Waterfront Area Overall Toronto
Distibution of Parking Infractions by Period of Day
Figure 3 Distribution of Infractions by Period of Day
The breakdown of parking citations by type, time and location is shown in Table 3.
Table 3 Breakdown of Parking Citations, 2011
Total number of citation records for 2011 2,805,492
Number of vehicles parked or stopped during
prohibited time of day on restricted highway,
all times of day
882,956
Number of vehicles parked or stopped during
prohibited time of day on restricted highway,
between 8 a.m. and 9 a.m.
37,152
Number of vehicles parked or stopped during
prohibited time of day on restricted highway,
between 8 a.m. and 9 a.m., in the Toronto
waterfront area
6892
28
Number of vehicles parked or stopped during
prohibited time of day on restricted highway,
between 8 a.m. and 9 a.m., in the Toronto
waterfront area, after omitting one-lane links
4704
Figure 4 Toronto Parking Citations Record
5.5.2 A Measure of the Non-compliance Rate
The parking citations record can be used as an indicator of the current level of non-compliance to
the on-street parking restrictions during the AM peak hour. Since these tickets are a
documentation of real illegal parking events that took place at a given time and location, the
parking citations can be used as an input to the illegal parking microsimulation model. The time
and location at which an infraction occurred can be extracted and then simulated in the
simulation model.
29
6 Methodology OverviewThis chapter describes the data that are used as an input to the illegal parking simulation model.
The illegal on-street parking model setup can be divided into three main components: Data
filtering, geocoding infraction addresses, and coding illegal on-street parking into Quadstone
Paramics.
6.1 Data FilteringThe parking citation database obtained from the City of Toronto is a record of all the parking
tickets issued by the city’s parking enforcement officers in 2011. Parking tickets are issued for a
variety of different parking offences, such as not paying for parking, exceeding the meter
duration and parking on a restricted street. The parking tickets record contains citations occurring
at all times of day, and at various locations throughout the City of Toronto.
However, not all parking infraction types are relevant to this research. The intended infractions
are those of vehicles parked or stopped at a prohibited time of the day. Moreover, since this
study evaluates the impact of illegal parking on AM rush hour traffic, only citations recorded
between 8 a.m. and 9 a.m. are needed. Then, there is the location of the citations. The study area
only examines citations occurring within the Toronto Waterfront area boundaries. All in all,
three main filters are applied to the parking citations record.
1) Infraction type filter- only parking infractions involving vehicles parked or stopped on-
street on roadways that prohibit such parking activity during the AM peak period are
extracted.
2) Infraction time of day filter- only citations recorded between 8 a.m. and 9 a.m. are
considered since the simulation lies within this hour.
3) Infraction location filter- only tickets issued within the Toronto Waterfront boundaries
are considered.
In addition, links with only one lane per travel direction are omitted from the study. This is due
to a limitation of the microsimulation algorithm of Quadstone Paramics, as it does not instruct
vehicles to move onto the lane with the opposing direction of travel to maneuver around an
obstacle ahead, which is what would drivers do in a real-life scenario. Therefore, if a parked
30
vehicle was to be added on to a one-lane link, vehicles would queue up behind that vehicle
without the ability of clearing that vehicle, creating an unrealistic traffic condition.
After applying the filters, the remaining infractions would be those of the intended type, time of
day and location for the study. The next step was to geocode the infraction addresses (see the
next section). But before geocoding these addresses, several omissions and changes to the text of
these addresses were applied in order to align the way the addresses are recorded with the syntax
of the geocoding software (ArcGIS), eliminating errors in the geocoding process. In summary,
the following changes were made to the parking infractions record:
1) Add a municipality column for all entries, set municipality as Toronto
2) Delete entries with empty address field (1138 entries affected)
3) Deleting addresses beginning with 0 or special characters (520 entries affected)
4) Create a column that contains the intersection closest to the address recorded 131,499
entries affected)
5) Delete entries with no number address and no intersection data (2098 entries affected)
6) Delete spaces between street numbers (188 entries affected)
7) Set province to ON for all entries
8) Delete entries with no time of day recorded (2009 entries affected)
6.2 Geocoding Infractions’ Addresses
The location of infractions in the parking citations record obtained from the City of Toronto is
the address of the closest building to where the vehicle was cited (eg. 1 Yonge St. Toronto, ON
Canada). However, Quadstone Paramics, the microsimulation suite used in this study, requires
the distance between the infraction and its closest upstream intersection as means of adding the
illegally parking vehicle into the network.
In order to obtain these distances for all the infractions to be simulated, a geocoding software can
be used to perform this measurement collectively. ArcGIS is used in this research. Geocoding is
“the process of transforming a description of a location—such as a pair of coordinates, an
address, or a name of a place—to a location on the earth's surface” (Esri, 2010). Once the
addresses are geocoded, ArcGIS calculates the distance between each infraction and its upstream
31
intersection, providing the parameter needed to code the infractions into the microsimulation
model (see next section).
6.3 Coding Illegal On-street Parking into Paramics
Quadstone Paramics, the microsimulation software used in this research, requires the following
pieces of data to be incorporated into the code describing the infractions to be simulated:
1. The name of the link at which the incident occurred
2. The distance of the infraction from the upstream intersection
3. Infraction type
4. Infraction duration
The names of the links as well as the distance of the infraction from its upstream location have
been obtained previously in the steps described above. The infraction type instructs Paramics
whether to create parking incidents at random times and locations in the simulation or whether it
should create parking incidents with times and locations specified by the user. Since a dataset of
defined times and locations is used, the second option is selected.
As for the duration of the illegal parking activities, an assumption has to be made. Since the
parking citations record only contains the time at which the ticket was issued, there is no way of
knowing when the vehicle’s parking activity began and when it ended. And to the best
knowledge of the authors, no studies have examined these durations through surveys. The
following durations are assumed:
1. 10-minute duration for parked vehicles
2. 5-minute duration for stopped/standing vehicles
A separate file for each simulation day, which includes all the infractions recorded for a given
day, is created. This ensures that the effects of infractions recorded for a day are captured
without being influenced by infractions recorded on other days.
Another assumption was made for the time at which a vehicle starts its illegal parking activity in
the simulation network. It is assumed in this study to be the time the ticket was issued, rounded
down to the nearest 10 minutes. For example, if the citations record shows that a ticket was
32
issued at 8:46 a.m., that corresponding simulated parking activity would begin at 8:40 a.m. in the
simulated network. This assumption was made for the following reasons:
• The exact time at which the vehicle stopped at the link is unknown
• To capture as much of the effect of the infraction as possible within the 10-minute
interval reporting period, where the reporting period is how often the model reports the
performance metrics of the network
• To avoid an infraction occurring within more than 1 reporting period.
Each simulation day is run with 10 different seeds, as this many runs was deemed to be sufficient
to smooth out the simulation randomness.
33
7 Results
7.1 ScenariosThe two distinct scenarios that have been simulated independently of each other are:
A) Base Case Scenario- The Toronto’s Waterfront network is simulated without the addition
of illegal parking incidents into the network. Therefore, this base case serves as a
benchmark to which the second scenario is compared. The base case generates the
outputs that are currently being generated by users of the conventional Paramics
Waterfront network that has been used in the past and is still being currently used in
various traffic studies of Toronto’s Downtown area. In summary, the base case scenario
does not account for the effect of illegally parked vehicles on the flow of traffic.
B) Illegal Parking Scenario- In this scenario, illegally parked vehicles are added into the
network used in Scenario A. The only difference between the two scenarios is the
presence of illegally parked vehicles. All of the model’s parameters are kept constant to
maintain a fair comparison between the two scenarios and to ensure that any difference in
the performance metrics of the two networks can be solely attributed to illegal parking.
7.2 Summary of Research Scope
7.2.1 Simulation Runs
In total, 310 simulation runs have been conducted, these runs can be categorized as follows:
1) Ten Base Case Simulations: Upon obtaining the relevant illegal parking infractions from
the Toronto Parking Citations Database, the Toronto Waterfront Paramics network was
simulated ten times, each time with a different seed (Seeds 1-10 are used in the
simulation).
2) 300 Illegal Parking Simulations: The 30 days that encountered the highest number of
illegal on-street parking during the AM rush hour according to the Parking Citations
Database are simulated. Each day is simulated separately with its unique set of illegal
parking times and locations that are obtained from the citations database, creating a
34
“simulation day”. Each simulation day is simulated 10 different times with each
simulation having a different seed number ranging from seed 1 to seed 10.
7.2.2 Simulated Links
The links that experienced illegal on-street parking or stopping on the worst 30 days of the year
2011 in terms of the number of illegal on-street parking/stopping incidents amounted to:
• 117 links that experienced illegal on-street parking
• 499 links that experienced illegal on-street stopping
It is observed that the illegal stopping incidents clearly outnumber illegal parking incidents. This
may be attributed to the perceived risk to drivers as they approach their destination. In general,
drivers may perceive illegal parking as a riskier activity compared to illegal stopping in terms of
the probability of being caught and ticketed by a parking enforcement officer.
In addition to the links simulated above, their adjacent links have also been examined. These
adjacent links are examined to ensure that any congestion that propagates beyond the links that
actually experience the illegal parking/stopping incident and to the surrounding areas is captured.
All in all, there were 1778 adjacent links examined.
7.2.3 Performance Metrics
The following performance metrics are used in this simulation to gauge the level of congestion
of the network in both the base case scenario as well as the illegal parking scenarios:
1) Link Delay: Defined as the average difference between the actual travel time and the free
flow travel time for all vehicles on the link for a given time interval
2) Link Flow: Defined as the flow of vehicles transferring from the downstream end of the
link for a given interval
3) Link Speed: Defined as the average speed of a vehicle traversing the link for a given
interval
4) Link Travel Time: Defined as the average time taken by all vehicles to traverse the link
for a given time interval
35
7.3 Summary of Results
7.3.1 Individual Simulation Days
The results of the model runs for the 30 simulation days are summarized in the appendix.
7.3.2 Overall Summary
Overall, with the 30 simulation days combined, the performance metrics are as follows:
Table 4 Links that experience Illegal Parking - Results Summary
Overall Summary Delay (Sec) Flow (pcus/hr) Speed (km/hr) Travel Time
(Sec)
Base Case 7 656 31.1 16.6
Illegal Parking
Added
10.5 609 26.8 20.3
% Increase 50.3 -7.1 -13.9 22.3
Table 5 Adjacent Links - Results Summary
Overall Summary Delay (Sec) Flow (pcus/hr) Speed (km/hr) Travel Time
(Sec)
Base Case 9.2 523 25.8 18.3
Illegal Parking
Added
14.8 353 22.4 20.3
% Increase 60 -32.4 -13.1 11.3
36
7.4 T-statistic TestThe t-statistic test was performed for every performance metric measured in the simulations in
order to determine if the difference in these performance metrics between the base case and the
illegal parking case is statistically significant. The t-statistic test for each performance metric is
shown below.
Table 6 t-Test: Two-Sample Assuming Unequal Variances - Link Delay
Variable 1 Variable 2Mean 6.968298505 10.47597289Variance 59.74967269 117.1637357Observations 616 616Hypothesized Mean Difference 0df 1113t Stat -6.54529542P(T<=t) one-tail 4.52154E-11t Critical one-tail 1.646223839P(T<=t) two-tail 9.04309E-11t Critical two-tail 1.962097686
Table 7 t-Test: Two-Sample Assuming Unequal Variances - Link Flow
Variable 1 Variable 2Mean 655.7210045 609.2045455Variance 198824.4718 160839.2034Observations 616 616Hypothesized Mean Difference 0df 1216t Stat 1.925079704P(T<=t) one-tail 0.027226211t Critical one-tail 1.646107688P(T<=t) two-tail 0.054452421t Critical two-tail 1.961916776
37
Table 8 t-Test: Two-Sample Assuming Unequal Variances - Link Speed
Variable 1 Variable 2Mean 31.09060984 26.76820406Variance 81.74971931 71.65398328Observations 616 616Hypothesized Mean Difference 0df 1225t Stat 8.661596939P(T<=t) one-tail 7.22476E-18t Critical one-tail 1.646098467P(T<=t) two-tail 1.44495E-17t Critical two-tail 1.961902415
Table 9 t-Test: Two-Sample Assuming Unequal Variances - Link Travel Time
Variable 1 Variable 2Mean 16.60836358 20.31080369Variance 89.75943189 148.067211Observations 616 616Hypothesized Mean Difference 0df 1160t Stat -5.95865369P(T<=t) one-tail 1.68449E-09t Critical one-tail 1.646168277P(T<=t) two-tail 3.36898E-09t Critical two-tail 1.962011145
38
7.5 Discussion of ResultsBy examining the comparison between the base case scenario and the illegal parking scenario for
each simulation day separately as well as for all the simulation days grouped together, the
following observations can be made:
A) Link Delay increases with the addition of illegal parking into the network
B) Link Flow reduces with the addition of illegal parking into the network. However, this
reduction is not statistically significant at the 95% level of confidence, according to the t-
statistic test.
C) Link Speed reduces with the addition of illegal parking into the network
D) Link Travel Time increases with the addition of illegal parking into the network
The following can be deduced from the change in the above performance metrics:
A) Illegal parking causes significant delays to all drivers on the link that experiences illegal
parking as well as the surrounding links
B) Illegal parking reduces the capacity of the links significantly.
C) The increased travel times and reduced speeds indicate a reduced LOS caused by illegal
parking activities
D) Network performance metrics are not being realistically measured by existing
microsimulation models that do not account for illegal parking
Comparing the base case to the scenarios where illegal parking is added onto the network
consistently and reliably reveals an increase in link delays and travel time and a reduction in link
flow and speed. Therefore, it can be deduced that illegal parking has a detrimental effect on the
flow of traffic during the morning rush hour, which already experiences higher than normal
travel times for most drivers. Existing traffic microsimulation models that omit illegal parking
underestimate the level of congestion and the trip lengths of vehicles, thus revealing an inability
to reliably and accurately reflect real life traffic conditions, which is the main objective of traffic
microsimulation.
39
It should be also noted that the increase in travel time caused by illegally parked vehicles can be
compounded as a vehicle drives along several links that encounter illegal parking along its trip,
so the total increase in the trip length of a single vehicle can be very significant.
It is also observed that an illegally parked vehicle not only affects the link on which it is stopped,
but the congestion it causes extends to the adjacent links as well. Thus, congestion is not
localized to the location of the bottleneck but extends to the surrounding area, increasing delays
for a larger set of drivers.
8 Conclusion & Future Work
8.1 Policy Implications of Simulation ModelBased on the results generated from the microsimulation model, it can be concluded that a
reduction in the demand for illegal parking is needed in order to alleviate their negative impact
on traffic flow in Toronto’s CBD. A reduction in the number of illegally parked vehicles means
less bottlenecks for vehicles to squeeze through during the morning rush hour, and less delay to
many drivers.
Three types of strategy may be used to cause a reduction in the demand for illegal parking:
A) Increase Parking Enforcement- Drivers are less likely to park illegally if they perceive the
risk of being caught by an enforcement officer and ticketed as high. However, the
benefits of enforcement should be weighed against its substantial cost as the primary
method of parking enforcement is currently human enforcement agents which is
expensive.
B) Increase Supply of Parking- By ensuring that drivers have a legal parking spot to use
close to their destination, the need to park illegally is eliminated, since the main reason
why drivers choose to park illegally is due to the lack of parking at their destinations.
However, space in downtown cores of major cities is at a premium, and the benefits of
adding more parking infrastructure should be weighed against utilizing the space for
other uses.
40
C) Induce a Mode Shift- The need for parking is eliminated when the need for driving is
eliminated. By investing in alternative modes of transportation, drivers will have a more
convenient alternative to driving to their destinations in the busy downtown core,
reducing congestion as a result of the reduced number of vehicles, as well as the reduced
number of illegal parking incidents. This strategy is a longer term strategy since
infrastructure investments require time and money, and shift in the population’s
perception is needed in order to push them to replace their habit of driving with other
modes.
8.2 ConclusionThe results generated by the proposed integrated microsimulation model indicate that illegally
parked vehicles on Toronto Downtown streets during the AM rush hour cause significant delays
to all drivers in the area, and the flow on links that experience illegal parking significantly
reduces, thus reducing the capacity of downtown streets during a critical time of day that is
already experiencing congestion without the introduction of the bottlenecks created by illegally
parked vehicles. Since the delay is compounded by the many drivers that experience it in the area
over the hour, the illegal parking becomes more prominent and costly. A need for reducing the
illegal parking phenomenon arises as the main recommendation from this body of research.
By comparing the performance metric of the Toronto Waterfront Network which does not
account for illegal parking with the one that does, it can be seen that existing microsimulation
models underestimate the level of congestion in the downtown core. The proposed parking-
sensitive microsimulation model enables planners to better represent real life traffic conditions
and extract more accurate travel times for vehicles traversing the network.
8.3 Future WorkA continuation of this research is recommended to improve its accuracy and extend its
applications. Some of the possible future streams of research include:
• Obtain empirical values for the distribution of the duration of the illegal parking/stopping
activity. Since the values used in this research are assumed values due to the lack of
studies that measure these values, a field study that measures the duration of the illegal
41
parking activities in the downtown core can provide a more accurate duration of parking
that can be used in the future to improve the accuracy of the microsimulation model
• The parking-sensitive microsimulation model can be used to identify the most critical
links which, if they were to encounter an illegal parking incident, would cause the most
delay on a network level. This can be performed by assigning illegally parked vehicles
onto different links and measuring their effect on travel times
• Network performance metrics can be correlated with the number of lanes of a link as well
as other attributes such as link type, link length, link speed limit, etc. to determine the
relationship between these attributes and the extent to which an illegally parked vehicle
can cause delays on these links based on their attributes
• The parking-sensitive microsimulation model can be used to test the impact of different
parking policies on reducing congestion, therefore helping policy makers identify the
most effective strategies for improving traffic conditions in the downtown core,
especially during the AM and PM peak periods
42
ReferencesAmirjamshidi, G., Mostafa, T. S., Misra, A., & Roorda, M. J. (2013). Integrated model for
microsimulating vehicle emissions, pollutant dispersion and population exposure.
Transportation Research Part D: Transport and Environment, 18, 16–24.
Arnott, R., & Inci, E. (2006). An integrated model of downtown parking and traffic congestion.
Journal of Urban Economics, 60(3), 418–442.
Arnott, R., Inci, E., & Rowse, J. (2015). Downtown curbside parking capacity. Journal of Urban
Economics, 86, 83–97.
Barceló, J. (2010). Fundamentals of traffic simulation (Vol. 145). Springer.
Barcelo, J., & Casas, J. (2004). Methodological notes on the calibration and validation of
microscopic traffic simulation models. Presented at the Proceedings of the 83rd TRB
annual meeting, Washington, DC.
Barter, P. A. (2015). A parking policy typology for clearer thinking on parking reform.
International Journal of Urban Sciences, 19(2), 136–156.
Benenson, I., Martens, K., & Birfir, S. (2008). PARKAGENT: An agent-based model of parking
in the city. Computers, Environment and Urban Systems, 32(6), 431–439.
Brockfeld, E., Kühne, R., & Wagner, P. (2004). Calibration and validation of microscopic traffic
flow models. Transportation Research Record: Journal of the Transportation Research
Board, (1876), 62–70.
43
Chu, L., Liu, H. X., Oh, J.-S., & Recker, W. (2003). A calibration procedure for microscopic
traffic simulation (Vol. 2, pp. 1574–1579). Presented at the Intelligent Transportation
Systems, 2003. Proceedings. 2003 IEEE, IEEE.
City of Toronto. (2015). Parking Tickets. Retrieved February 4, 2016, from
http://www1.toronto.ca/wps/portal/contentonly?vgnextoid=ca20256c54ea4310VgnVCM
1000003dd60f89RCRD&vgnextchannel=7807e03bb8d1e310VgnVCM10000071d60f89
RCRD
Cools, M., van der Waerden, P., & Janssens, D. (2013). Investigation of the Determinants of
Travelers’ Mental Knowledge of Public Parking Facilities. Presented at the 92nd Annual
Meeting of the Transportation Research Board, Transportation Research Board of the
National Academies.
Cullinane, K., & Polak, J. (1992). Illegal parking and the enforcement of parking regulations:
causes, effects and interactions. Transport Reviews, 12(1), 49–75.
Data Management Group. (2011). Transportation Tomorrow Survey Area Summary: 2011, 2006,
2001, 1996 & 1986 ORIGIN-DESTINATION MATRICES. Department of Civil
Engineering: University of Toronto.
Dowling, R., Holland, P. J., & Huang, P. A. (2002). California Department of Transportation
Guidelines for Applying Traffic Microsimulation Modeling Software. Way, 3, 3–2.
Esri. (2010). Introducing Geocoding. Retrieved from
http://desktop.arcgis.com/en/arcmap/10.3/guide-books/geocoding/what-is-geocoding.htm
44
Gallo, M., D’Acierno, L., & Montella, B. (2011). A multilayer model to simulate cruising for
parking in urban areas. Transport Policy, 18(5), 735–744.
Haider, M. (2009). Challenges Facing Express Delivery Services in Canada’s Urban Centres.
Ryerson University, Toronto.
Horni, A., Montini, L., Waraich, R. A., & Axhausen, K. W. (2013). An agent-based cellular
automaton cruising-for-parking simulation. Transportation Letters, 5(4), 167–175.
Hourdakis, J., Michalopoulos, P., & Kottommannil, J. (2003). Practical procedure for calibrating
microscopic traffic simulation models. Transportation Research Record: Journal of the
Transportation Research Board, (1852), 130–139.
Jia, B., Jiang, R., & Wu, Q.-S. (2003). The traffic bottleneck effects caused by the lane closing in
the cellular automata model. International Journal of Modern Physics C, 14(10), 1295–
1303. http://doi.org/10.1142/S012918310300542X
Kladeftiras, M., & Antoniou, C. (2013). Simulation-based assessment of double-parking impacts
on traffic and environmental conditions. Transportation Research Record: Journal of the
Transportation Research Board, (2390), 121–130.
Kobus, M. B., Gutiérrez-i-Puigarnau, E., Rietveld, P., & Van Ommeren, J. N. (2013). The on-
street parking premium and car drivers’ choice between street and garage parking.
Regional Science and Urban Economics, 43(2), 395–403.
Lu, B., & Viegas, J. (2007). The Analysis of the Influences of the Double Parking Vehicles to
the General Traffic Flow (pp. 3121–3126). Presented at the International Conference on
Transportation Engineering 2007, ASCE.
45
Ma, T., & Abdulhai, B. (2002). Genetic algorithm-based optimization approach and generic tool
for calibrating traffic microscopic simulation parameters. Transportation Research
Record: Journal of the Transportation Research Board, (1800), 6–15.
Merritt, E. (2004). Calibration and validation of CORSIM for Swedish road traffic conditions.
Presented at the Proceedings of the 83rd TRB annual meeting, Washington, DC.
Millard-Ball, A., Weinberger, R. R., & Hampshire, R. C. (2014). Is the curb 80% full or 20%
empty? Assessing the impacts of San Francisco’s parking pricing experiment.
Transportation Research Part A: Policy and Practice, 63, 76–92.
Moini Ph D, N., Hill Ph D, D., Shabihkhani, R., Homami, P., & Rezaei, H. (2013a). Impact
assessments of on-street parking guidance system on mobility and environment.
Presented at the Transportation Research Board 92nd Annual Meeting.
Montini, L., Horni, A., Rieser-Schüssler, N., & Axhausen, K. W. (2012). Searching for parking
in GPS data. Eidgenössische Technische Hochschule Zürich, IVT, Institute for Transport
Planning and Systems.
Park, B., & Qi, H. (2005). Development and Evaluation of a Procedure for the Calibration of
Simulation Models. Transportation Research Record: Journal of the Transportation
Research Board, (1934), 208–217.
Pipes, L. A. (1953). An operational analysis of traffic dynamics. Journal of Applied Physics,
24(3), 274–281.
Quadstone Paramics. (2016, February 17). What is Traffic Microsimulation? Retrieved from
http://www.paramics-online.com/what-is-microsimulation.php
46
Reuschel, A. (1950). Vehicle movements in a platoon. Oesterreichisches Ingenieur-Archir, 4,
193–215.
Shaaban, K. S., & Radwan, E. (2005). A calibration and validation procedure for microscopic
simulation model: a case study of sim traffic arterial streets. Presented at the Proceedings
of the 84rd TRB Annual Meeting, Washington, DC.
Shoup, D. C. (2006). Cruising for parking. Transport Policy, 13(6), 479–486.
Shum, D. (2015, October 5). Toronto’s tag-and-tow traffic blitz is back. Retrieved from
http://globalnews.ca/news/2258391/torontos-rush-hour-tag-and-tow-blitz-begins-oct-5/
Simićević, J., Vukanović, S., & Milosavljević, N. (2013). The effect of parking charges and time
limit to car usage and parking behaviour. Transport Policy, 30, 125–131.
Toledo, T., Ben-Akiva, M., Darda, D., Jha, M., & Koutsopoulos, H. (2004). Calibration of
microscopic traffic simulation models with aggregate data. Transportation Research
Record: Journal of the Transportation Research Board, (1876), 10–19.
Topp, H. H. (1993). Parking policies to reduce car traffic in German cities. Transport Reviews,
13(1), 83–95.
Van Ommeren, J. N., Wentink, D., & Rietveld, P. (2012). Empirical evidence on cruising for
parking. Transportation Research Part A: Policy and Practice, 46(1), 123–130.
Weinberger, R., Seaman, M., & Johnson, C. (2008). Suburbanizing the city: how New York City
parking requirements lead to more driving.
47
Appendix
Summary of Individual Simulation Runs
Day 1 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 9.6 704.6 27.7 18.2
Illegal Parking Added 13.4 657.4 25.2 22.0
% Increase 39.7 -6.7 -9.0 21.1
Day 2 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 5.9 553.1 32.6 15.7
Illegal Parking Added 8.9 567.8 26.9 18.8
% Increase 51.8 2.7 -17.6 19.9
Day 3 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 5.6 491.7 30.8 13.6
Illegal Parking Added 8.5 460.8 26.8 16.9
% Increase 53.2 -6.3 -13.1 24.6
48
Day 4 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 5.6 579.7 30.3 17.3
Illegal Parking Added 9.9 518.7 26.5 21.8
% Increase 76.1 -10.5 -12.6 26.5
Day 5 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 4.8 696.5 34.6 15.8
Illegal Parking Added 8.8 598.6 29.8 19.8
% Increase 85.2 -14.1 -13.9 25.2
Day 6 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 5.8 690.9 32.2 15.4
Illegal Parking Added 12.5 554.0 26.1 22.2
% Increase 116.6 -19.8 -19.1 43.6
49
Day 7 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 7.8 637.3 28.2 17.0
Illegal Parking Added 8.4 599.4 26.7 17.6
% Increase 7.9 -5.9 -5.1 3.6
50
Day 8 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 5.19 679.09 32.84 16.00
Illegal Parking Added 10.97 634.25 27.40 21.76
% Increase 111.53 -6.60 -16.55 35.96
Day 9 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 6.0 762.1 31.5 16.0
Illegal Parking Added 10.5 702.4 26.4 20.6
% Increase 75.6 -7.8 -16.0 28.4
51
Day 10 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 4.3 550.6 36.3 13.7
Illegal Parking Added 5.6 546.0 31.0 17.5
% Increase 30.2 -0.8 -14.7 28.1
Day 11 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 6.6 571.7 31.6 17.1
Illegal Parking Added 10.6 522.2 27.0 21.2
% Increase 62.0 -8.6 -14.5 24.0
Day 12 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 6.2 847.6 33.0 16.3
Illegal Parking Added 10.1 724.0 28.4 20.3
% Increase 63.7 -14.6 -14.0 25.2
52
Day 13 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 6.6 591.9 29.8 16.5
Illegal Parking Added 10.9 555.5 25.0 20.9
% Increase 65.4 -6.1 -16.2 26.2
Day 14 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 5.1 706.8 31.3 14.6
Illegal Parking Added 7.6 590.2 27.5 17.1
% Increase 47.7 -16.5 -12.1 17.0
Day 15 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 7.3 752.9 30.2 16.7
Illegal Parking Added 9.3 676.8 26.2 18.8
% Increase 28.4 -10.1 -13.5 12.8
53
Day 16 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 6.2 604.4 31.9 16.8
Illegal Parking Added 13.1 503.2 27.1 23.6
% Increase 111.0 -16.7 -15.0 40.6
Day 17 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 7.9 522.3 29.3 16.2
Illegal Parking Added 12.4 456.7 23.6 20.8
% Increase 57.4 -12.6 -19.5 28.0
Day 18 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 8.5 577.1 28.6 17.2
Illegal Parking Added 10.3 509.8 24.8 19.0
% Increase 21.9 -11.7 -13.3 10.8
54
Day 19 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 13.9 814.2 28.7 24.4
Illegal Parking Added 15.9 832.6 27.2 26.4
% Increase 15.0 2.3 -5.3 8.2
Day 20 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 9.8 726.8 30.6 19.1
Illegal Parking Added 10.1 669.3 26.4 19.5
% Increase 3.9 -7.9 -13.8 2.0
Day 21 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 6.0 774.7 32.2 15.0
Illegal Parking Added 10.8 728.1 25.9 19.9
% Increase 79.9 -6.0 -19.4 32.3
55
Day 22 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 6.8 581.2 31.5 19.5
Illegal Parking Added 10.0 535.6 27.9 22.8
% Increase 46.2 -7.9 -11.3 16.8
Day 23 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 5.0 536.8 34.4 15.9
Illegal Parking Added 8.4 520.5 29.8 19.4
% Increase 68.7 -3.0 -13.3 22.0
Day 24 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 4.26 622.06 34.03 15.37
Illegal Parking Added 8.40 597.92 27.82 19.61
% Increase 97.05 -3.88 -18.25 27.57
56
Day 25 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 8.8 624.0 27.8 13.1
Illegal Parking Added 13.9 672.5 25.2 23.3
% Increase 57.6 7.8 -9.5 78.2
Day 26 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 12.0 563.0 30.2 22.1
Illegal Parking Added 17.0 499.1 24.5 27.3
% Increase 42.5 -11.3 -18.9 23.4
Day 27 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 5.6 646.1 30.8 15.5
Illegal Parking Added 8.6 649.0 27.1 18.5
% Increase 52.6 0.5 -11.9 19.2
57
Day 28 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 8.4 692.3 30.7 17.5
Illegal Parking Added 10.2 692.5 26.9 19.4
% Increase 21.0 0.0 -12.3 11.0
Day 29 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 9.2 825.6 31.3 18.6
Illegal Parking Added 11.4 774.3 27.0 20.7
% Increase 23.6 -6.2 -14.0 11.7
Day 30 Delay (Sec)
Flow
(pcus/hr)
Speed
(km/hr) Travel Time (sec)
Base Case 5.5 507.0 29.1 15.3
Illegal Parking Added 9.8 498.2 22.9 18.4
% Increase 78.8 -1.7 -21.2 20.0
58